Namespace(accumulate=1, batch_norm=False, batch_size=8, clip_grad=0, crop_ratio=0.875, data_dir='/home/ubuntu/yizhu/data/kinetics400/kinetics400/train_256', dataset='kinetics400', dtype='float32', eval=False, fast_temporal_stride=2, freeze_bn=False, hard_weight=0.5, hashtag='', input_5d=False, input_size=112, kvstore=None, label_smoothing=False, last_gamma=False, log_interval=50, logging_file='r2plus1d_res34_seg1_k400_b8_g8_f16s2_cosine_epoch196_lr1_warm34_warmlr001_run1.log', lr=0.1, lr_decay=0.1, lr_decay_epoch='40,60', lr_decay_period=0, lr_mode='cosine', mixup=False, mixup_alpha=0.2, mixup_off_epoch=0, mode='hybrid', model='r2plus1d_resnet34_kinetics400', momentum=0.9, new_height=128, new_length=16, new_step=2, new_width=171, no_wd=False, num_classes=400, num_crop=1, num_epochs=196, num_gpus=8, num_segments=1, num_workers=32, partial_bn=False, prefetch_ratio=1.0, resume_epoch=0, resume_params='', resume_states='', save_dir='/home/ubuntu/yizhu/logs/mxnet/kinetics400/r2plus1d/r2plus1d_res34_seg1_k400_b8_g8_f16s2_cosine_epoch196_lr1_warm34_warmlr001_run1', save_frequency=10, scale_ratios='1.0,0.8', slow_temporal_stride=16, slowfast=False, teacher=None, temperature=20, train_list='/home/ubuntu/yizhu/data/kinetics400/kinetics400/k400_train_240618.txt', use_amp=False, use_decord=True, use_gn=False, use_pretrained=False, use_se=False, use_tsn=False, val_data_dir='/home/ubuntu/yizhu/data/kinetics400/kinetics400/val_256', val_list='/home/ubuntu/yizhu/code/gluon-cv/extra/kinetics400/k400_val_19761_cleanv3.txt', video_loader=True, warmup_epochs=34, warmup_lr=0.001, wd=0.0001)
Total batch size is set to 64 on 8 GPUs
R2Plus1D(
  (conv1): Conv3D(3 -> 45, kernel_size=(1, 7, 7), stride=(1, 2, 2), padding=(0, 3, 3), bias=False)
  (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=45)
  (relu): Activation(relu)
  (conv2): Conv3D(45 -> 64, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
  (bn2): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
  (layer1): HybridSequential(
    (0): BasicBlock(
      (conv1): Conv2Plus1D(
        (conv1): Conv3D(64 -> 144, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=144)
        (relu): Activation(relu)
        (conv2): Conv3D(144 -> 64, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
      (relu): Activation(relu)
      (conv2): Conv2Plus1D(
        (conv1): Conv3D(64 -> 144, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=144)
        (relu): Activation(relu)
        (conv2): Conv3D(144 -> 64, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn2): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
    )
    (1): BasicBlock(
      (conv1): Conv2Plus1D(
        (conv1): Conv3D(64 -> 144, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=144)
        (relu): Activation(relu)
        (conv2): Conv3D(144 -> 64, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
      (relu): Activation(relu)
      (conv2): Conv2Plus1D(
        (conv1): Conv3D(64 -> 144, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=144)
        (relu): Activation(relu)
        (conv2): Conv3D(144 -> 64, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn2): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
    )
    (2): BasicBlock(
      (conv1): Conv2Plus1D(
        (conv1): Conv3D(64 -> 144, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=144)
        (relu): Activation(relu)
        (conv2): Conv3D(144 -> 64, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
      (relu): Activation(relu)
      (conv2): Conv2Plus1D(
        (conv1): Conv3D(64 -> 144, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=144)
        (relu): Activation(relu)
        (conv2): Conv3D(144 -> 64, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn2): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=64)
    )
  )
  (layer2): HybridSequential(
    (0): BasicBlock(
      (downsample): HybridSequential(
        (0): Conv3D(64 -> 128, kernel_size=(1, 1, 1), stride=(2, 2, 2), bias=False)
        (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=128)
      )
      (conv1): Conv2Plus1D(
        (conv1): Conv3D(64 -> 230, kernel_size=(1, 3, 3), stride=(1, 2, 2), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=230)
        (relu): Activation(relu)
        (conv2): Conv3D(230 -> 128, kernel_size=(3, 1, 1), stride=(2, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=128)
      (relu): Activation(relu)
      (conv2): Conv2Plus1D(
        (conv1): Conv3D(128 -> 230, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=230)
        (relu): Activation(relu)
        (conv2): Conv3D(230 -> 128, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn2): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=128)
    )
    (1): BasicBlock(
      (conv1): Conv2Plus1D(
        (conv1): Conv3D(128 -> 288, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=288)
        (relu): Activation(relu)
        (conv2): Conv3D(288 -> 128, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=128)
      (relu): Activation(relu)
      (conv2): Conv2Plus1D(
        (conv1): Conv3D(128 -> 288, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=288)
        (relu): Activation(relu)
        (conv2): Conv3D(288 -> 128, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn2): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=128)
    )
    (2): BasicBlock(
      (conv1): Conv2Plus1D(
        (conv1): Conv3D(128 -> 288, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=288)
        (relu): Activation(relu)
        (conv2): Conv3D(288 -> 128, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=128)
      (relu): Activation(relu)
      (conv2): Conv2Plus1D(
        (conv1): Conv3D(128 -> 288, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=288)
        (relu): Activation(relu)
        (conv2): Conv3D(288 -> 128, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn2): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=128)
    )
    (3): BasicBlock(
      (conv1): Conv2Plus1D(
        (conv1): Conv3D(128 -> 288, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=288)
        (relu): Activation(relu)
        (conv2): Conv3D(288 -> 128, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=128)
      (relu): Activation(relu)
      (conv2): Conv2Plus1D(
        (conv1): Conv3D(128 -> 288, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=288)
        (relu): Activation(relu)
        (conv2): Conv3D(288 -> 128, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn2): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=128)
    )
  )
  (layer3): HybridSequential(
    (0): BasicBlock(
      (downsample): HybridSequential(
        (0): Conv3D(128 -> 256, kernel_size=(1, 1, 1), stride=(2, 2, 2), bias=False)
        (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=256)
      )
      (conv1): Conv2Plus1D(
        (conv1): Conv3D(128 -> 460, kernel_size=(1, 3, 3), stride=(1, 2, 2), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=460)
        (relu): Activation(relu)
        (conv2): Conv3D(460 -> 256, kernel_size=(3, 1, 1), stride=(2, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=256)
      (relu): Activation(relu)
      (conv2): Conv2Plus1D(
        (conv1): Conv3D(256 -> 460, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=460)
        (relu): Activation(relu)
        (conv2): Conv3D(460 -> 256, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn2): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=256)
    )
    (1): BasicBlock(
      (conv1): Conv2Plus1D(
        (conv1): Conv3D(256 -> 576, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=576)
        (relu): Activation(relu)
        (conv2): Conv3D(576 -> 256, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=256)
      (relu): Activation(relu)
      (conv2): Conv2Plus1D(
        (conv1): Conv3D(256 -> 576, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=576)
        (relu): Activation(relu)
        (conv2): Conv3D(576 -> 256, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn2): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=256)
    )
    (2): BasicBlock(
      (conv1): Conv2Plus1D(
        (conv1): Conv3D(256 -> 576, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=576)
        (relu): Activation(relu)
        (conv2): Conv3D(576 -> 256, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=256)
      (relu): Activation(relu)
      (conv2): Conv2Plus1D(
        (conv1): Conv3D(256 -> 576, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=576)
        (relu): Activation(relu)
        (conv2): Conv3D(576 -> 256, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn2): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=256)
    )
    (3): BasicBlock(
      (conv1): Conv2Plus1D(
        (conv1): Conv3D(256 -> 576, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=576)
        (relu): Activation(relu)
        (conv2): Conv3D(576 -> 256, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=256)
      (relu): Activation(relu)
      (conv2): Conv2Plus1D(
        (conv1): Conv3D(256 -> 576, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=576)
        (relu): Activation(relu)
        (conv2): Conv3D(576 -> 256, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn2): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=256)
    )
    (4): BasicBlock(
      (conv1): Conv2Plus1D(
        (conv1): Conv3D(256 -> 576, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=576)
        (relu): Activation(relu)
        (conv2): Conv3D(576 -> 256, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=256)
      (relu): Activation(relu)
      (conv2): Conv2Plus1D(
        (conv1): Conv3D(256 -> 576, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=576)
        (relu): Activation(relu)
        (conv2): Conv3D(576 -> 256, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn2): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=256)
    )
    (5): BasicBlock(
      (conv1): Conv2Plus1D(
        (conv1): Conv3D(256 -> 576, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=576)
        (relu): Activation(relu)
        (conv2): Conv3D(576 -> 256, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=256)
      (relu): Activation(relu)
      (conv2): Conv2Plus1D(
        (conv1): Conv3D(256 -> 576, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=576)
        (relu): Activation(relu)
        (conv2): Conv3D(576 -> 256, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn2): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=256)
    )
  )
  (layer4): HybridSequential(
    (0): BasicBlock(
      (downsample): HybridSequential(
        (0): Conv3D(256 -> 512, kernel_size=(1, 1, 1), stride=(2, 2, 2), bias=False)
        (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=512)
      )
      (conv1): Conv2Plus1D(
        (conv1): Conv3D(256 -> 921, kernel_size=(1, 3, 3), stride=(1, 2, 2), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=921)
        (relu): Activation(relu)
        (conv2): Conv3D(921 -> 512, kernel_size=(3, 1, 1), stride=(2, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=512)
      (relu): Activation(relu)
      (conv2): Conv2Plus1D(
        (conv1): Conv3D(512 -> 921, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=921)
        (relu): Activation(relu)
        (conv2): Conv3D(921 -> 512, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn2): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=512)
    )
    (1): BasicBlock(
      (conv1): Conv2Plus1D(
        (conv1): Conv3D(512 -> 1152, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=1152)
        (relu): Activation(relu)
        (conv2): Conv3D(1152 -> 512, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=512)
      (relu): Activation(relu)
      (conv2): Conv2Plus1D(
        (conv1): Conv3D(512 -> 1152, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=1152)
        (relu): Activation(relu)
        (conv2): Conv3D(1152 -> 512, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn2): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=512)
    )
    (2): BasicBlock(
      (conv1): Conv2Plus1D(
        (conv1): Conv3D(512 -> 1152, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=1152)
        (relu): Activation(relu)
        (conv2): Conv3D(1152 -> 512, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=512)
      (relu): Activation(relu)
      (conv2): Conv2Plus1D(
        (conv1): Conv3D(512 -> 1152, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
        (bn1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=1152)
        (relu): Activation(relu)
        (conv2): Conv3D(1152 -> 512, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
      )
      (bn2): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=512)
    )
  )
  (avgpool): GlobalAvgPool3D(size=(1, 1, 1), stride=(1, 1, 1), padding=(0, 0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCDHW)
  (dropout): Dropout(p = 0.5, axes=())
  (fc): Dense(512 -> 400, linear)
)
Load 240618 training samples and 19404 validation samples.
Epoch[000] Batch [0049]/[3759]	Speed: 61.959320 samples/sec	 accuracy=0.406250	 loss=5.976696	 lr=0.001039
Epoch[000] Batch [0099]/[3759]	Speed: 117.980245 samples/sec	 accuracy=0.390625	 loss=5.959900	 lr=0.001077
Epoch[000] Batch [0149]/[3759]	Speed: 117.308676 samples/sec	 accuracy=0.458333	 loss=5.954040	 lr=0.001116
Epoch[000] Batch [0199]/[3759]	Speed: 117.008965 samples/sec	 accuracy=0.460938	 loss=5.948361	 lr=0.001155
Epoch[000] Batch [0249]/[3759]	Speed: 115.705042 samples/sec	 accuracy=0.437500	 loss=5.945132	 lr=0.001194
Epoch[000] Batch [0299]/[3759]	Speed: 116.629479 samples/sec	 accuracy=0.411458	 loss=5.942214	 lr=0.001232
Epoch[000] Batch [0349]/[3759]	Speed: 116.152158 samples/sec	 accuracy=0.419643	 loss=5.939321	 lr=0.001271
Epoch[000] Batch [0399]/[3759]	Speed: 116.698532 samples/sec	 accuracy=0.421875	 loss=5.938177	 lr=0.001310
Epoch[000] Batch [0449]/[3759]	Speed: 117.352974 samples/sec	 accuracy=0.409722	 loss=5.936171	 lr=0.001349
Epoch[000] Batch [0499]/[3759]	Speed: 115.975938 samples/sec	 accuracy=0.425000	 loss=5.934347	 lr=0.001387
Epoch[000] Batch [0549]/[3759]	Speed: 115.766812 samples/sec	 accuracy=0.434659	 loss=5.931836	 lr=0.001426
Epoch[000] Batch [0599]/[3759]	Speed: 116.778081 samples/sec	 accuracy=0.481771	 loss=5.930033	 lr=0.001465
Epoch[000] Batch [0649]/[3759]	Speed: 117.082732 samples/sec	 accuracy=0.485577	 loss=5.927710	 lr=0.001504
Epoch[000] Batch [0699]/[3759]	Speed: 116.775769 samples/sec	 accuracy=0.470982	 loss=5.925415	 lr=0.001542
Epoch[000] Batch [0749]/[3759]	Speed: 116.654291 samples/sec	 accuracy=0.485417	 loss=5.923685	 lr=0.001581
Epoch[000] Batch [0799]/[3759]	Speed: 116.334295 samples/sec	 accuracy=0.484375	 loss=5.920331	 lr=0.001620
Epoch[000] Batch [0849]/[3759]	Speed: 116.303777 samples/sec	 accuracy=0.490809	 loss=5.917923	 lr=0.001658
Epoch[000] Batch [0899]/[3759]	Speed: 115.973531 samples/sec	 accuracy=0.494792	 loss=5.915342	 lr=0.001697
Epoch[000] Batch [0949]/[3759]	Speed: 116.610405 samples/sec	 accuracy=0.493421	 loss=5.912566	 lr=0.001736
Epoch[000] Batch [0999]/[3759]	Speed: 116.808575 samples/sec	 accuracy=0.509375	 loss=5.908570	 lr=0.001775
Epoch[000] Batch [1049]/[3759]	Speed: 117.209168 samples/sec	 accuracy=0.531250	 loss=5.905473	 lr=0.001813
Epoch[000] Batch [1099]/[3759]	Speed: 116.892217 samples/sec	 accuracy=0.536932	 loss=5.901534	 lr=0.001852
Epoch[000] Batch [1149]/[3759]	Speed: 117.099602 samples/sec	 accuracy=0.540761	 loss=5.897675	 lr=0.001891
Epoch[000] Batch [1199]/[3759]	Speed: 117.167102 samples/sec	 accuracy=0.555990	 loss=5.892779	 lr=0.001930
Epoch[000] Batch [1249]/[3759]	Speed: 115.543662 samples/sec	 accuracy=0.560000	 loss=5.889012	 lr=0.001968
Epoch[000] Batch [1299]/[3759]	Speed: 116.685394 samples/sec	 accuracy=0.558894	 loss=5.885290	 lr=0.002007
Epoch[000] Batch [1349]/[3759]	Speed: 116.635380 samples/sec	 accuracy=0.559028	 loss=5.881994	 lr=0.002046
Epoch[000] Batch [1399]/[3759]	Speed: 117.079907 samples/sec	 accuracy=0.563616	 loss=5.878654	 lr=0.002084
Epoch[000] Batch [1449]/[3759]	Speed: 116.410856 samples/sec	 accuracy=0.573276	 loss=5.875298	 lr=0.002123
Epoch[000] Batch [1499]/[3759]	Speed: 116.042890 samples/sec	 accuracy=0.586458	 loss=5.872053	 lr=0.002162
Epoch[000] Batch [1549]/[3759]	Speed: 117.717204 samples/sec	 accuracy=0.583669	 loss=5.868778	 lr=0.002201
Epoch[000] Batch [1599]/[3759]	Speed: 117.519231 samples/sec	 accuracy=0.585938	 loss=5.866474	 lr=0.002239
Epoch[000] Batch [1649]/[3759]	Speed: 118.064717 samples/sec	 accuracy=0.593750	 loss=5.863530	 lr=0.002278
Epoch[000] Batch [1699]/[3759]	Speed: 116.189684 samples/sec	 accuracy=0.602022	 loss=5.860173	 lr=0.002317
Epoch[000] Batch [1749]/[3759]	Speed: 116.410762 samples/sec	 accuracy=0.608929	 loss=5.858211	 lr=0.002356
Epoch[000] Batch [1799]/[3759]	Speed: 116.193956 samples/sec	 accuracy=0.615451	 loss=5.855319	 lr=0.002394
Epoch[000] Batch [1849]/[3759]	Speed: 116.574553 samples/sec	 accuracy=0.618243	 loss=5.853001	 lr=0.002433
Epoch[000] Batch [1899]/[3759]	Speed: 117.383812 samples/sec	 accuracy=0.620066	 loss=5.850176	 lr=0.002472
Epoch[000] Batch [1949]/[3759]	Speed: 116.765345 samples/sec	 accuracy=0.625000	 loss=5.847577	 lr=0.002511
Epoch[000] Batch [1999]/[3759]	Speed: 117.354033 samples/sec	 accuracy=0.632812	 loss=5.845607	 lr=0.002549
Epoch[000] Batch [2049]/[3759]	Speed: 116.929330 samples/sec	 accuracy=0.636433	 loss=5.843180	 lr=0.002588
Epoch[000] Batch [2099]/[3759]	Speed: 115.856665 samples/sec	 accuracy=0.636161	 loss=5.841485	 lr=0.002627
Epoch[000] Batch [2149]/[3759]	Speed: 115.544759 samples/sec	 accuracy=0.640262	 loss=5.839215	 lr=0.002665
Epoch[000] Batch [2199]/[3759]	Speed: 117.864898 samples/sec	 accuracy=0.639915	 loss=5.837533	 lr=0.002704
Epoch[000] Batch [2249]/[3759]	Speed: 117.265278 samples/sec	 accuracy=0.645139	 loss=5.835451	 lr=0.002743
Epoch[000] Batch [2299]/[3759]	Speed: 116.204708 samples/sec	 accuracy=0.647418	 loss=5.833303	 lr=0.002782
Epoch[000] Batch [2349]/[3759]	Speed: 116.816594 samples/sec	 accuracy=0.648271	 loss=5.831560	 lr=0.002820
Epoch[000] Batch [2399]/[3759]	Speed: 117.333347 samples/sec	 accuracy=0.654297	 loss=5.829260	 lr=0.002859
Epoch[000] Batch [2449]/[3759]	Speed: 117.676077 samples/sec	 accuracy=0.671556	 loss=5.827170	 lr=0.002898
Epoch[000] Batch [2499]/[3759]	Speed: 117.661496 samples/sec	 accuracy=0.680000	 loss=5.825605	 lr=0.002937
Epoch[000] Batch [2549]/[3759]	Speed: 116.524346 samples/sec	 accuracy=0.696691	 loss=5.823375	 lr=0.002975
Epoch[000] Batch [2599]/[3759]	Speed: 117.803395 samples/sec	 accuracy=0.706130	 loss=5.821151	 lr=0.003014
Epoch[000] Batch [2649]/[3759]	Speed: 116.421523 samples/sec	 accuracy=0.707547	 loss=5.819290	 lr=0.003053
Epoch[000] Batch [2699]/[3759]	Speed: 117.093830 samples/sec	 accuracy=0.714699	 loss=5.817135	 lr=0.003091
Epoch[000] Batch [2749]/[3759]	Speed: 117.295635 samples/sec	 accuracy=0.719318	 loss=5.815126	 lr=0.003130
Epoch[000] Batch [2799]/[3759]	Speed: 116.766265 samples/sec	 accuracy=0.723214	 loss=5.812806	 lr=0.003169
Epoch[000] Batch [2849]/[3759]	Speed: 117.868465 samples/sec	 accuracy=0.729715	 loss=5.810484	 lr=0.003208
Epoch[000] Batch [2899]/[3759]	Speed: 117.500055 samples/sec	 accuracy=0.741379	 loss=5.807717	 lr=0.003246
Epoch[000] Batch [2949]/[3759]	Speed: 117.581897 samples/sec	 accuracy=0.742055	 loss=5.805619	 lr=0.003285
Epoch[000] Batch [2999]/[3759]	Speed: 115.962114 samples/sec	 accuracy=0.748437	 loss=5.803269	 lr=0.003324
Epoch[000] Batch [3049]/[3759]	Speed: 116.197213 samples/sec	 accuracy=0.749488	 loss=5.800648	 lr=0.003363
Epoch[000] Batch [3099]/[3759]	Speed: 116.084568 samples/sec	 accuracy=0.758065	 loss=5.798489	 lr=0.003401
Epoch[000] Batch [3149]/[3759]	Speed: 115.967308 samples/sec	 accuracy=0.767361	 loss=5.796191	 lr=0.003440
Epoch[000] Batch [3199]/[3759]	Speed: 116.032333 samples/sec	 accuracy=0.772461	 loss=5.793349	 lr=0.003479
Epoch[000] Batch [3249]/[3759]	Speed: 115.799545 samples/sec	 accuracy=0.776923	 loss=5.790765	 lr=0.003518
Epoch[000] Batch [3299]/[3759]	Speed: 116.354725 samples/sec	 accuracy=0.789299	 loss=5.788387	 lr=0.003556
Epoch[000] Batch [3349]/[3759]	Speed: 116.500498 samples/sec	 accuracy=0.801772	 loss=5.785903	 lr=0.003595
Epoch[000] Batch [3399]/[3759]	Speed: 114.856887 samples/sec	 accuracy=0.808824	 loss=5.782987	 lr=0.003634
Epoch[000] Batch [3449]/[3759]	Speed: 115.661315 samples/sec	 accuracy=0.821558	 loss=5.780110	 lr=0.003672
Epoch[000] Batch [3499]/[3759]	Speed: 115.669687 samples/sec	 accuracy=0.832143	 loss=5.777544	 lr=0.003711
Epoch[000] Batch [3549]/[3759]	Speed: 116.463117 samples/sec	 accuracy=0.842870	 loss=5.775103	 lr=0.003750
Epoch[000] Batch [3599]/[3759]	Speed: 116.692473 samples/sec	 accuracy=0.853733	 loss=5.772497	 lr=0.003789
Epoch[000] Batch [3649]/[3759]	Speed: 116.656257 samples/sec	 accuracy=0.866010	 loss=5.769668	 lr=0.003827
Epoch[000] Batch [3699]/[3759]	Speed: 116.022408 samples/sec	 accuracy=0.874155	 loss=5.767232	 lr=0.003866
Epoch[000] Batch [3749]/[3759]	Speed: 120.862598 samples/sec	 accuracy=0.885000	 loss=5.764900	 lr=0.003905
Batch [0049]/[0303]: acc-top1=0.968750 acc-top5=4.312500
Batch [0099]/[0303]: acc-top1=1.187500 acc-top5=4.609375
Batch [0149]/[0303]: acc-top1=1.468750 acc-top5=5.906250
Batch [0199]/[0303]: acc-top1=1.320312 acc-top5=6.039062
Batch [0249]/[0303]: acc-top1=1.531250 acc-top5=6.268750
Batch [0299]/[0303]: acc-top1=1.395833 acc-top5=5.812500
[Epoch 000] training: accuracy=0.888700	 loss=5.764405
[Epoch 000] speed: 115 samples/sec	time cost: 2210.315872
[Epoch 000] validation: acc-top1=1.382013 acc-top5=5.760107 loss=5.676277
Epoch[001] Batch [0049]/[3760]	Speed: 86.712388 samples/sec	 accuracy=2.062500	 loss=5.562659	 lr=0.003951
Epoch[001] Batch [0099]/[3760]	Speed: 117.275511 samples/sec	 accuracy=1.765625	 loss=5.559884	 lr=0.003989
Epoch[001] Batch [0149]/[3760]	Speed: 117.528748 samples/sec	 accuracy=1.687500	 loss=5.555867	 lr=0.004028
Epoch[001] Batch [0199]/[3760]	Speed: 117.964276 samples/sec	 accuracy=1.648438	 loss=5.556314	 lr=0.004067
Epoch[001] Batch [0249]/[3760]	Speed: 117.402467 samples/sec	 accuracy=1.662500	 loss=5.551628	 lr=0.004105
Epoch[001] Batch [0299]/[3760]	Speed: 117.169990 samples/sec	 accuracy=1.619792	 loss=5.551170	 lr=0.004144
Epoch[001] Batch [0349]/[3760]	Speed: 118.337803 samples/sec	 accuracy=1.678571	 loss=5.549119	 lr=0.004183
Epoch[001] Batch [0399]/[3760]	Speed: 118.251873 samples/sec	 accuracy=1.722656	 loss=5.545899	 lr=0.004222
Epoch[001] Batch [0449]/[3760]	Speed: 117.726796 samples/sec	 accuracy=1.736111	 loss=5.543590	 lr=0.004260
Epoch[001] Batch [0499]/[3760]	Speed: 117.396219 samples/sec	 accuracy=1.759375	 loss=5.537340	 lr=0.004299
Epoch[001] Batch [0549]/[3760]	Speed: 117.159335 samples/sec	 accuracy=1.752841	 loss=5.536679	 lr=0.004338
Epoch[001] Batch [0599]/[3760]	Speed: 117.433870 samples/sec	 accuracy=1.752604	 loss=5.534479	 lr=0.004377
Epoch[001] Batch [0649]/[3760]	Speed: 116.568415 samples/sec	 accuracy=1.762019	 loss=5.531582	 lr=0.004415
Epoch[001] Batch [0699]/[3760]	Speed: 116.848251 samples/sec	 accuracy=1.808036	 loss=5.529835	 lr=0.004454
Epoch[001] Batch [0749]/[3760]	Speed: 117.476934 samples/sec	 accuracy=1.816667	 loss=5.525795	 lr=0.004493
Epoch[001] Batch [0799]/[3760]	Speed: 116.983936 samples/sec	 accuracy=1.845703	 loss=5.523106	 lr=0.004531
Epoch[001] Batch [0849]/[3760]	Speed: 116.973097 samples/sec	 accuracy=1.854779	 loss=5.519386	 lr=0.004570
Epoch[001] Batch [0899]/[3760]	Speed: 117.050038 samples/sec	 accuracy=1.876736	 loss=5.516836	 lr=0.004609
Epoch[001] Batch [0949]/[3760]	Speed: 117.429952 samples/sec	 accuracy=1.889803	 loss=5.514171	 lr=0.004648
Epoch[001] Batch [0999]/[3760]	Speed: 117.432680 samples/sec	 accuracy=1.910938	 loss=5.511578	 lr=0.004686
Epoch[001] Batch [1049]/[3760]	Speed: 116.984782 samples/sec	 accuracy=1.925595	 loss=5.509047	 lr=0.004725
Epoch[001] Batch [1099]/[3760]	Speed: 116.921148 samples/sec	 accuracy=1.928977	 loss=5.507006	 lr=0.004764
Epoch[001] Batch [1149]/[3760]	Speed: 117.131056 samples/sec	 accuracy=1.942935	 loss=5.503775	 lr=0.004803
Epoch[001] Batch [1199]/[3760]	Speed: 117.756794 samples/sec	 accuracy=1.942708	 loss=5.501885	 lr=0.004841
Epoch[001] Batch [1249]/[3760]	Speed: 117.213116 samples/sec	 accuracy=1.966250	 loss=5.497984	 lr=0.004880
Epoch[001] Batch [1299]/[3760]	Speed: 117.333987 samples/sec	 accuracy=1.985577	 loss=5.495581	 lr=0.004919
Epoch[001] Batch [1349]/[3760]	Speed: 117.640515 samples/sec	 accuracy=1.993056	 loss=5.493168	 lr=0.004958
Epoch[001] Batch [1399]/[3760]	Speed: 117.435861 samples/sec	 accuracy=2.012277	 loss=5.490034	 lr=0.004996
Epoch[001] Batch [1449]/[3760]	Speed: 117.377566 samples/sec	 accuracy=2.032328	 loss=5.486957	 lr=0.005035
Epoch[001] Batch [1499]/[3760]	Speed: 117.362208 samples/sec	 accuracy=2.032292	 loss=5.485068	 lr=0.005074
Epoch[001] Batch [1549]/[3760]	Speed: 117.240427 samples/sec	 accuracy=2.047379	 loss=5.482468	 lr=0.005112
Epoch[001] Batch [1599]/[3760]	Speed: 116.652835 samples/sec	 accuracy=2.063477	 loss=5.479667	 lr=0.005151
Epoch[001] Batch [1649]/[3760]	Speed: 117.630837 samples/sec	 accuracy=2.073864	 loss=5.477303	 lr=0.005190
Epoch[001] Batch [1699]/[3760]	Speed: 117.086805 samples/sec	 accuracy=2.101103	 loss=5.474048	 lr=0.005229
Epoch[001] Batch [1749]/[3760]	Speed: 116.799223 samples/sec	 accuracy=2.121429	 loss=5.471150	 lr=0.005267
Epoch[001] Batch [1799]/[3760]	Speed: 117.054172 samples/sec	 accuracy=2.138021	 loss=5.468501	 lr=0.005306
Epoch[001] Batch [1849]/[3760]	Speed: 117.183612 samples/sec	 accuracy=2.164696	 loss=5.465626	 lr=0.005345
Epoch[001] Batch [1899]/[3760]	Speed: 117.136188 samples/sec	 accuracy=2.173520	 loss=5.463499	 lr=0.005384
Epoch[001] Batch [1949]/[3760]	Speed: 117.991360 samples/sec	 accuracy=2.194712	 loss=5.460384	 lr=0.005422
Epoch[001] Batch [1999]/[3760]	Speed: 116.746253 samples/sec	 accuracy=2.214844	 loss=5.458162	 lr=0.005461
Epoch[001] Batch [2049]/[3760]	Speed: 117.739755 samples/sec	 accuracy=2.226372	 loss=5.455970	 lr=0.005500
Epoch[001] Batch [2099]/[3760]	Speed: 117.335221 samples/sec	 accuracy=2.236607	 loss=5.453183	 lr=0.005538
Epoch[001] Batch [2149]/[3760]	Speed: 116.527037 samples/sec	 accuracy=2.254360	 loss=5.450096	 lr=0.005577
Epoch[001] Batch [2199]/[3760]	Speed: 117.694495 samples/sec	 accuracy=2.271307	 loss=5.447725	 lr=0.005616
Epoch[001] Batch [2249]/[3760]	Speed: 117.478834 samples/sec	 accuracy=2.284028	 loss=5.445715	 lr=0.005655
Epoch[001] Batch [2299]/[3760]	Speed: 116.601031 samples/sec	 accuracy=2.317935	 loss=5.442895	 lr=0.005693
Epoch[001] Batch [2349]/[3760]	Speed: 117.499342 samples/sec	 accuracy=2.335106	 loss=5.440498	 lr=0.005732
Epoch[001] Batch [2399]/[3760]	Speed: 117.226083 samples/sec	 accuracy=2.347005	 loss=5.437899	 lr=0.005771
Epoch[001] Batch [2449]/[3760]	Speed: 117.450735 samples/sec	 accuracy=2.360969	 loss=5.436081	 lr=0.005810
Epoch[001] Batch [2499]/[3760]	Speed: 117.194498 samples/sec	 accuracy=2.368125	 loss=5.433745	 lr=0.005848
Epoch[001] Batch [2549]/[3760]	Speed: 116.599857 samples/sec	 accuracy=2.381740	 loss=5.432015	 lr=0.005887
Epoch[001] Batch [2599]/[3760]	Speed: 117.506228 samples/sec	 accuracy=2.390625	 loss=5.430652	 lr=0.005926
Epoch[001] Batch [2649]/[3760]	Speed: 117.166835 samples/sec	 accuracy=2.397406	 loss=5.428060	 lr=0.005965
Epoch[001] Batch [2699]/[3760]	Speed: 116.788592 samples/sec	 accuracy=2.416667	 loss=5.425458	 lr=0.006003
Epoch[001] Batch [2749]/[3760]	Speed: 117.246219 samples/sec	 accuracy=2.440909	 loss=5.422780	 lr=0.006042
Epoch[001] Batch [2799]/[3760]	Speed: 117.283048 samples/sec	 accuracy=2.455357	 loss=5.420415	 lr=0.006081
Epoch[001] Batch [2849]/[3760]	Speed: 117.496394 samples/sec	 accuracy=2.463816	 loss=5.417780	 lr=0.006119
Epoch[001] Batch [2899]/[3760]	Speed: 118.385977 samples/sec	 accuracy=2.470905	 loss=5.415730	 lr=0.006158
Epoch[001] Batch [2949]/[3760]	Speed: 117.507763 samples/sec	 accuracy=2.495763	 loss=5.413769	 lr=0.006197
Epoch[001] Batch [2999]/[3760]	Speed: 117.125525 samples/sec	 accuracy=2.508854	 loss=5.411752	 lr=0.006236
Epoch[001] Batch [3049]/[3760]	Speed: 116.842815 samples/sec	 accuracy=2.522029	 loss=5.409395	 lr=0.006274
Epoch[001] Batch [3099]/[3760]	Speed: 116.680313 samples/sec	 accuracy=2.536290	 loss=5.407160	 lr=0.006313
Epoch[001] Batch [3149]/[3760]	Speed: 116.476419 samples/sec	 accuracy=2.549107	 loss=5.404841	 lr=0.006352
Epoch[001] Batch [3199]/[3760]	Speed: 117.313633 samples/sec	 accuracy=2.570801	 loss=5.402384	 lr=0.006391
Epoch[001] Batch [3249]/[3760]	Speed: 116.754757 samples/sec	 accuracy=2.580288	 loss=5.400199	 lr=0.006429
Epoch[001] Batch [3299]/[3760]	Speed: 117.600707 samples/sec	 accuracy=2.597064	 loss=5.397866	 lr=0.006468
Epoch[001] Batch [3349]/[3760]	Speed: 116.937925 samples/sec	 accuracy=2.607743	 loss=5.395749	 lr=0.006507
Epoch[001] Batch [3399]/[3760]	Speed: 117.660927 samples/sec	 accuracy=2.630055	 loss=5.393180	 lr=0.006545
Epoch[001] Batch [3449]/[3760]	Speed: 117.205998 samples/sec	 accuracy=2.636322	 loss=5.390915	 lr=0.006584
Epoch[001] Batch [3499]/[3760]	Speed: 117.672357 samples/sec	 accuracy=2.646429	 loss=5.388271	 lr=0.006623
Epoch[001] Batch [3549]/[3760]	Speed: 116.925850 samples/sec	 accuracy=2.662412	 loss=5.385960	 lr=0.006662
Epoch[001] Batch [3599]/[3760]	Speed: 115.799738 samples/sec	 accuracy=2.677951	 loss=5.383490	 lr=0.006700
Epoch[001] Batch [3649]/[3760]	Speed: 117.380906 samples/sec	 accuracy=2.690925	 loss=5.381587	 lr=0.006739
Epoch[001] Batch [3699]/[3760]	Speed: 116.916445 samples/sec	 accuracy=2.708615	 loss=5.378907	 lr=0.006778
Epoch[001] Batch [3749]/[3760]	Speed: 120.366684 samples/sec	 accuracy=2.722500	 loss=5.376446	 lr=0.006817
Batch [0049]/[0303]: acc-top1=0.937500 acc-top5=6.562500
Batch [0099]/[0303]: acc-top1=1.796875 acc-top5=8.203125
Batch [0149]/[0303]: acc-top1=2.593750 acc-top5=10.687500
Batch [0199]/[0303]: acc-top1=3.054688 acc-top5=11.171875
Batch [0249]/[0303]: acc-top1=3.456250 acc-top5=11.962500
Batch [0299]/[0303]: acc-top1=3.390625 acc-top5=11.958333
[Epoch 001] training: accuracy=2.724402	 loss=5.376211
[Epoch 001] speed: 116 samples/sec	time cost: 2182.038572
[Epoch 001] validation: acc-top1=3.403465 acc-top5=11.984323 loss=5.320988
Epoch[002] Batch [0049]/[3759]	Speed: 86.044882 samples/sec	 accuracy=4.156250	 loss=5.200511	 lr=0.006863
Epoch[002] Batch [0099]/[3759]	Speed: 117.061781 samples/sec	 accuracy=4.218750	 loss=5.186996	 lr=0.006902
Epoch[002] Batch [0149]/[3759]	Speed: 117.386692 samples/sec	 accuracy=4.093750	 loss=5.188627	 lr=0.006941
Epoch[002] Batch [0199]/[3759]	Speed: 116.141689 samples/sec	 accuracy=3.992188	 loss=5.191032	 lr=0.006979
Epoch[002] Batch [0249]/[3759]	Speed: 117.309325 samples/sec	 accuracy=4.012500	 loss=5.190513	 lr=0.007018
Epoch[002] Batch [0299]/[3759]	Speed: 116.869153 samples/sec	 accuracy=4.031250	 loss=5.186269	 lr=0.007057
Epoch[002] Batch [0349]/[3759]	Speed: 116.577518 samples/sec	 accuracy=4.071429	 loss=5.188073	 lr=0.007095
Epoch[002] Batch [0399]/[3759]	Speed: 116.853345 samples/sec	 accuracy=4.054688	 loss=5.186807	 lr=0.007134
Epoch[002] Batch [0449]/[3759]	Speed: 115.687978 samples/sec	 accuracy=4.072917	 loss=5.185521	 lr=0.007173
Epoch[002] Batch [0499]/[3759]	Speed: 116.766457 samples/sec	 accuracy=4.056250	 loss=5.185209	 lr=0.007212
Epoch[002] Batch [0549]/[3759]	Speed: 116.819528 samples/sec	 accuracy=4.110795	 loss=5.181608	 lr=0.007250
Epoch[002] Batch [0599]/[3759]	Speed: 118.270254 samples/sec	 accuracy=4.080729	 loss=5.180424	 lr=0.007289
Epoch[002] Batch [0649]/[3759]	Speed: 116.864926 samples/sec	 accuracy=4.093750	 loss=5.178214	 lr=0.007328
Epoch[002] Batch [0699]/[3759]	Speed: 116.932534 samples/sec	 accuracy=4.080357	 loss=5.178546	 lr=0.007367
Epoch[002] Batch [0749]/[3759]	Speed: 117.984808 samples/sec	 accuracy=4.158333	 loss=5.175483	 lr=0.007405
Epoch[002] Batch [0799]/[3759]	Speed: 117.017224 samples/sec	 accuracy=4.171875	 loss=5.171361	 lr=0.007444
Epoch[002] Batch [0849]/[3759]	Speed: 116.826754 samples/sec	 accuracy=4.224265	 loss=5.169997	 lr=0.007483
Epoch[002] Batch [0899]/[3759]	Speed: 117.294509 samples/sec	 accuracy=4.262153	 loss=5.165831	 lr=0.007522
Epoch[002] Batch [0949]/[3759]	Speed: 116.718609 samples/sec	 accuracy=4.282895	 loss=5.165138	 lr=0.007560
Epoch[002] Batch [0999]/[3759]	Speed: 117.470421 samples/sec	 accuracy=4.312500	 loss=5.163838	 lr=0.007599
Epoch[002] Batch [1049]/[3759]	Speed: 117.632519 samples/sec	 accuracy=4.322917	 loss=5.161732	 lr=0.007638
Epoch[002] Batch [1099]/[3759]	Speed: 117.231754 samples/sec	 accuracy=4.359375	 loss=5.159514	 lr=0.007676
Epoch[002] Batch [1149]/[3759]	Speed: 116.639184 samples/sec	 accuracy=4.381793	 loss=5.157793	 lr=0.007715
Epoch[002] Batch [1199]/[3759]	Speed: 117.104370 samples/sec	 accuracy=4.393229	 loss=5.157106	 lr=0.007754
Epoch[002] Batch [1249]/[3759]	Speed: 116.949088 samples/sec	 accuracy=4.430000	 loss=5.153839	 lr=0.007793
Epoch[002] Batch [1299]/[3759]	Speed: 116.374369 samples/sec	 accuracy=4.451923	 loss=5.151082	 lr=0.007831
Epoch[002] Batch [1349]/[3759]	Speed: 117.375807 samples/sec	 accuracy=4.469907	 loss=5.148640	 lr=0.007870
Epoch[002] Batch [1399]/[3759]	Speed: 117.023488 samples/sec	 accuracy=4.485491	 loss=5.145430	 lr=0.007909
Epoch[002] Batch [1449]/[3759]	Speed: 116.416077 samples/sec	 accuracy=4.495690	 loss=5.145203	 lr=0.007948
Epoch[002] Batch [1499]/[3759]	Speed: 116.498128 samples/sec	 accuracy=4.501042	 loss=5.141572	 lr=0.007986
Epoch[002] Batch [1549]/[3759]	Speed: 117.165577 samples/sec	 accuracy=4.482863	 loss=5.139376	 lr=0.008025
Epoch[002] Batch [1599]/[3759]	Speed: 116.579114 samples/sec	 accuracy=4.507812	 loss=5.136895	 lr=0.008064
Epoch[002] Batch [1649]/[3759]	Speed: 116.100225 samples/sec	 accuracy=4.526515	 loss=5.134935	 lr=0.008102
Epoch[002] Batch [1699]/[3759]	Speed: 115.879867 samples/sec	 accuracy=4.522059	 loss=5.133019	 lr=0.008141
Epoch[002] Batch [1749]/[3759]	Speed: 117.204112 samples/sec	 accuracy=4.530357	 loss=5.131719	 lr=0.008180
Epoch[002] Batch [1799]/[3759]	Speed: 116.596002 samples/sec	 accuracy=4.531250	 loss=5.130160	 lr=0.008219
Epoch[002] Batch [1849]/[3759]	Speed: 117.575088 samples/sec	 accuracy=4.546453	 loss=5.127899	 lr=0.008257
Epoch[002] Batch [1899]/[3759]	Speed: 116.948707 samples/sec	 accuracy=4.549342	 loss=5.126279	 lr=0.008296
Epoch[002] Batch [1949]/[3759]	Speed: 117.814369 samples/sec	 accuracy=4.560897	 loss=5.124253	 lr=0.008335
Epoch[002] Batch [1999]/[3759]	Speed: 116.834049 samples/sec	 accuracy=4.570312	 loss=5.123016	 lr=0.008374
Epoch[002] Batch [2049]/[3759]	Speed: 117.068555 samples/sec	 accuracy=4.594512	 loss=5.121697	 lr=0.008412
Epoch[002] Batch [2099]/[3759]	Speed: 117.672911 samples/sec	 accuracy=4.626488	 loss=5.119081	 lr=0.008451
Epoch[002] Batch [2149]/[3759]	Speed: 117.881860 samples/sec	 accuracy=4.655523	 loss=5.117576	 lr=0.008490
Epoch[002] Batch [2199]/[3759]	Speed: 116.514250 samples/sec	 accuracy=4.674716	 loss=5.115558	 lr=0.008529
Epoch[002] Batch [2249]/[3759]	Speed: 116.898812 samples/sec	 accuracy=4.704167	 loss=5.113047	 lr=0.008567
Epoch[002] Batch [2299]/[3759]	Speed: 117.306961 samples/sec	 accuracy=4.730299	 loss=5.110707	 lr=0.008606
Epoch[002] Batch [2349]/[3759]	Speed: 116.970140 samples/sec	 accuracy=4.744016	 loss=5.109275	 lr=0.008645
Epoch[002] Batch [2399]/[3759]	Speed: 116.424332 samples/sec	 accuracy=4.763021	 loss=5.107381	 lr=0.008683
Epoch[002] Batch [2449]/[3759]	Speed: 117.353335 samples/sec	 accuracy=4.761480	 loss=5.105105	 lr=0.008722
Epoch[002] Batch [2499]/[3759]	Speed: 117.108090 samples/sec	 accuracy=4.790000	 loss=5.102478	 lr=0.008761
Epoch[002] Batch [2549]/[3759]	Speed: 117.668597 samples/sec	 accuracy=4.813725	 loss=5.100386	 lr=0.008800
Epoch[002] Batch [2599]/[3759]	Speed: 117.106706 samples/sec	 accuracy=4.828726	 loss=5.098819	 lr=0.008838
Epoch[002] Batch [2649]/[3759]	Speed: 117.159871 samples/sec	 accuracy=4.854953	 loss=5.096738	 lr=0.008877
Epoch[002] Batch [2699]/[3759]	Speed: 116.344443 samples/sec	 accuracy=4.866319	 loss=5.095256	 lr=0.008916
Epoch[002] Batch [2749]/[3759]	Speed: 117.507452 samples/sec	 accuracy=4.885227	 loss=5.093265	 lr=0.008955
Epoch[002] Batch [2799]/[3759]	Speed: 116.807750 samples/sec	 accuracy=4.904576	 loss=5.091204	 lr=0.008993
Epoch[002] Batch [2849]/[3759]	Speed: 116.479561 samples/sec	 accuracy=4.919408	 loss=5.089544	 lr=0.009032
Epoch[002] Batch [2899]/[3759]	Speed: 117.011893 samples/sec	 accuracy=4.946121	 loss=5.087305	 lr=0.009071
Epoch[002] Batch [2949]/[3759]	Speed: 117.019931 samples/sec	 accuracy=4.951271	 loss=5.085429	 lr=0.009109
Epoch[002] Batch [2999]/[3759]	Speed: 116.380306 samples/sec	 accuracy=4.965104	 loss=5.083429	 lr=0.009148
Epoch[002] Batch [3049]/[3759]	Speed: 116.944686 samples/sec	 accuracy=4.986680	 loss=5.081488	 lr=0.009187
Epoch[002] Batch [3099]/[3759]	Speed: 117.402805 samples/sec	 accuracy=5.011089	 loss=5.079256	 lr=0.009226
Epoch[002] Batch [3149]/[3759]	Speed: 116.710630 samples/sec	 accuracy=5.029762	 loss=5.076363	 lr=0.009264
Epoch[002] Batch [3199]/[3759]	Speed: 117.263732 samples/sec	 accuracy=5.044434	 loss=5.074327	 lr=0.009303
Epoch[002] Batch [3249]/[3759]	Speed: 116.859704 samples/sec	 accuracy=5.051923	 loss=5.072690	 lr=0.009342
Epoch[002] Batch [3299]/[3759]	Speed: 116.521737 samples/sec	 accuracy=5.062027	 loss=5.070579	 lr=0.009381
Epoch[002] Batch [3349]/[3759]	Speed: 117.807847 samples/sec	 accuracy=5.072295	 loss=5.068563	 lr=0.009419
Epoch[002] Batch [3399]/[3759]	Speed: 116.893773 samples/sec	 accuracy=5.089154	 loss=5.066239	 lr=0.009458
Epoch[002] Batch [3449]/[3759]	Speed: 116.232377 samples/sec	 accuracy=5.105072	 loss=5.064021	 lr=0.009497
Epoch[002] Batch [3499]/[3759]	Speed: 116.884932 samples/sec	 accuracy=5.127679	 loss=5.062171	 lr=0.009536
Epoch[002] Batch [3549]/[3759]	Speed: 116.601612 samples/sec	 accuracy=5.147447	 loss=5.059827	 lr=0.009574
Epoch[002] Batch [3599]/[3759]	Speed: 116.377209 samples/sec	 accuracy=5.163194	 loss=5.057577	 lr=0.009613
Epoch[002] Batch [3649]/[3759]	Speed: 116.532743 samples/sec	 accuracy=5.176798	 loss=5.055958	 lr=0.009652
Epoch[002] Batch [3699]/[3759]	Speed: 117.342040 samples/sec	 accuracy=5.185811	 loss=5.054688	 lr=0.009690
Epoch[002] Batch [3749]/[3759]	Speed: 119.734941 samples/sec	 accuracy=5.208333	 loss=5.052454	 lr=0.009729
Batch [0049]/[0303]: acc-top1=3.656250 acc-top5=13.937500
Batch [0099]/[0303]: acc-top1=4.343750 acc-top5=14.437500
Batch [0149]/[0303]: acc-top1=5.750000 acc-top5=17.781250
Batch [0199]/[0303]: acc-top1=6.695312 acc-top5=18.703125
Batch [0249]/[0303]: acc-top1=6.725000 acc-top5=19.037500
Batch [0299]/[0303]: acc-top1=6.369792 acc-top5=18.843750
[Epoch 002] training: accuracy=5.207086	 loss=5.052032
[Epoch 002] speed: 116 samples/sec	time cost: 2187.283793
[Epoch 002] validation: acc-top1=6.363449 acc-top5=18.915017 loss=5.048796
Epoch[003] Batch [0049]/[3760]	Speed: 85.511095 samples/sec	 accuracy=6.375000	 loss=4.910545	 lr=0.009775
Epoch[003] Batch [0099]/[3760]	Speed: 117.079953 samples/sec	 accuracy=6.937500	 loss=4.897018	 lr=0.009814
Epoch[003] Batch [0149]/[3760]	Speed: 116.585776 samples/sec	 accuracy=6.656250	 loss=4.895017	 lr=0.009852
Epoch[003] Batch [0199]/[3760]	Speed: 116.508740 samples/sec	 accuracy=6.625000	 loss=4.893402	 lr=0.009891
Epoch[003] Batch [0249]/[3760]	Speed: 117.927642 samples/sec	 accuracy=6.468750	 loss=4.897829	 lr=0.009930
Epoch[003] Batch [0299]/[3760]	Speed: 117.687425 samples/sec	 accuracy=6.510417	 loss=4.888481	 lr=0.009969
Epoch[003] Batch [0349]/[3760]	Speed: 117.254755 samples/sec	 accuracy=6.575893	 loss=4.889239	 lr=0.010007
Epoch[003] Batch [0399]/[3760]	Speed: 117.710719 samples/sec	 accuracy=6.597656	 loss=4.886757	 lr=0.010046
Epoch[003] Batch [0449]/[3760]	Speed: 116.955370 samples/sec	 accuracy=6.625000	 loss=4.882353	 lr=0.010085
Epoch[003] Batch [0499]/[3760]	Speed: 117.135477 samples/sec	 accuracy=6.653125	 loss=4.882338	 lr=0.010123
Epoch[003] Batch [0549]/[3760]	Speed: 116.815506 samples/sec	 accuracy=6.659091	 loss=4.881000	 lr=0.010162
Epoch[003] Batch [0599]/[3760]	Speed: 117.072678 samples/sec	 accuracy=6.734375	 loss=4.876068	 lr=0.010201
Epoch[003] Batch [0649]/[3760]	Speed: 116.234437 samples/sec	 accuracy=6.718750	 loss=4.873919	 lr=0.010240
Epoch[003] Batch [0699]/[3760]	Speed: 117.287615 samples/sec	 accuracy=6.738839	 loss=4.872610	 lr=0.010278
Epoch[003] Batch [0749]/[3760]	Speed: 115.871817 samples/sec	 accuracy=6.733333	 loss=4.873365	 lr=0.010317
Epoch[003] Batch [0799]/[3760]	Speed: 116.338806 samples/sec	 accuracy=6.753906	 loss=4.870349	 lr=0.010356
Epoch[003] Batch [0849]/[3760]	Speed: 116.338573 samples/sec	 accuracy=6.818015	 loss=4.865314	 lr=0.010395
Epoch[003] Batch [0899]/[3760]	Speed: 116.647278 samples/sec	 accuracy=6.842014	 loss=4.862619	 lr=0.010433
Epoch[003] Batch [0949]/[3760]	Speed: 116.672338 samples/sec	 accuracy=6.837171	 loss=4.861193	 lr=0.010472
Epoch[003] Batch [0999]/[3760]	Speed: 116.782207 samples/sec	 accuracy=6.812500	 loss=4.860368	 lr=0.010511
Epoch[003] Batch [1049]/[3760]	Speed: 117.214850 samples/sec	 accuracy=6.860119	 loss=4.854569	 lr=0.010549
Epoch[003] Batch [1099]/[3760]	Speed: 116.126281 samples/sec	 accuracy=6.867898	 loss=4.851305	 lr=0.010588
Epoch[003] Batch [1149]/[3760]	Speed: 116.750232 samples/sec	 accuracy=6.845109	 loss=4.850784	 lr=0.010627
Epoch[003] Batch [1199]/[3760]	Speed: 117.129082 samples/sec	 accuracy=6.876302	 loss=4.849385	 lr=0.010666
Epoch[003] Batch [1249]/[3760]	Speed: 115.956723 samples/sec	 accuracy=6.910000	 loss=4.846203	 lr=0.010704
Epoch[003] Batch [1299]/[3760]	Speed: 118.058788 samples/sec	 accuracy=6.923077	 loss=4.845198	 lr=0.010743
Epoch[003] Batch [1349]/[3760]	Speed: 116.839201 samples/sec	 accuracy=6.936343	 loss=4.842090	 lr=0.010782
Epoch[003] Batch [1399]/[3760]	Speed: 117.257970 samples/sec	 accuracy=6.967634	 loss=4.841783	 lr=0.010821
Epoch[003] Batch [1449]/[3760]	Speed: 116.227623 samples/sec	 accuracy=6.960129	 loss=4.838625	 lr=0.010859
Epoch[003] Batch [1499]/[3760]	Speed: 117.879108 samples/sec	 accuracy=6.979167	 loss=4.836377	 lr=0.010898
Epoch[003] Batch [1549]/[3760]	Speed: 116.691669 samples/sec	 accuracy=6.993952	 loss=4.834734	 lr=0.010937
Epoch[003] Batch [1599]/[3760]	Speed: 116.767345 samples/sec	 accuracy=7.034180	 loss=4.832561	 lr=0.010976
Epoch[003] Batch [1649]/[3760]	Speed: 116.994929 samples/sec	 accuracy=7.043561	 loss=4.830160	 lr=0.011014
Epoch[003] Batch [1699]/[3760]	Speed: 117.418529 samples/sec	 accuracy=7.083640	 loss=4.828098	 lr=0.011053
Epoch[003] Batch [1749]/[3760]	Speed: 117.762208 samples/sec	 accuracy=7.114286	 loss=4.826731	 lr=0.011092
Epoch[003] Batch [1799]/[3760]	Speed: 117.050580 samples/sec	 accuracy=7.118924	 loss=4.826357	 lr=0.011130
Epoch[003] Batch [1849]/[3760]	Speed: 117.868059 samples/sec	 accuracy=7.127534	 loss=4.824769	 lr=0.011169
Epoch[003] Batch [1899]/[3760]	Speed: 116.533778 samples/sec	 accuracy=7.151316	 loss=4.823094	 lr=0.011208
Epoch[003] Batch [1949]/[3760]	Speed: 116.687846 samples/sec	 accuracy=7.168269	 loss=4.821694	 lr=0.011247
Epoch[003] Batch [1999]/[3760]	Speed: 116.865518 samples/sec	 accuracy=7.188281	 loss=4.818916	 lr=0.011285
Epoch[003] Batch [2049]/[3760]	Speed: 116.535504 samples/sec	 accuracy=7.183689	 loss=4.818038	 lr=0.011324
Epoch[003] Batch [2099]/[3760]	Speed: 116.497925 samples/sec	 accuracy=7.200149	 loss=4.815788	 lr=0.011363
Epoch[003] Batch [2149]/[3760]	Speed: 116.200176 samples/sec	 accuracy=7.235465	 loss=4.813689	 lr=0.011402
Epoch[003] Batch [2199]/[3760]	Speed: 117.023347 samples/sec	 accuracy=7.259233	 loss=4.812081	 lr=0.011440
Epoch[003] Batch [2249]/[3760]	Speed: 116.955079 samples/sec	 accuracy=7.263194	 loss=4.810665	 lr=0.011479
Epoch[003] Batch [2299]/[3760]	Speed: 116.456887 samples/sec	 accuracy=7.272418	 loss=4.808834	 lr=0.011518
Epoch[003] Batch [2349]/[3760]	Speed: 117.229798 samples/sec	 accuracy=7.287234	 loss=4.806676	 lr=0.011556
Epoch[003] Batch [2399]/[3760]	Speed: 117.789477 samples/sec	 accuracy=7.305339	 loss=4.804538	 lr=0.011595
Epoch[003] Batch [2449]/[3760]	Speed: 116.621296 samples/sec	 accuracy=7.309949	 loss=4.802759	 lr=0.011634
Epoch[003] Batch [2499]/[3760]	Speed: 117.414948 samples/sec	 accuracy=7.329375	 loss=4.800581	 lr=0.011673
Epoch[003] Batch [2549]/[3760]	Speed: 117.862548 samples/sec	 accuracy=7.359681	 loss=4.797682	 lr=0.011711
Epoch[003] Batch [2599]/[3760]	Speed: 117.034963 samples/sec	 accuracy=7.372596	 loss=4.795600	 lr=0.011750
Epoch[003] Batch [2649]/[3760]	Speed: 117.419297 samples/sec	 accuracy=7.378538	 loss=4.793636	 lr=0.011789
Epoch[003] Batch [2699]/[3760]	Speed: 117.056008 samples/sec	 accuracy=7.403356	 loss=4.791620	 lr=0.011828
Epoch[003] Batch [2749]/[3760]	Speed: 116.735438 samples/sec	 accuracy=7.419886	 loss=4.789555	 lr=0.011866
Epoch[003] Batch [2799]/[3760]	Speed: 117.394680 samples/sec	 accuracy=7.436942	 loss=4.788166	 lr=0.011905
Epoch[003] Batch [2849]/[3760]	Speed: 116.988835 samples/sec	 accuracy=7.453947	 loss=4.786492	 lr=0.011944
Epoch[003] Batch [2899]/[3760]	Speed: 117.566381 samples/sec	 accuracy=7.474677	 loss=4.783767	 lr=0.011983
Epoch[003] Batch [2949]/[3760]	Speed: 117.696304 samples/sec	 accuracy=7.478814	 loss=4.782016	 lr=0.012021
Epoch[003] Batch [2999]/[3760]	Speed: 116.735821 samples/sec	 accuracy=7.492188	 loss=4.779994	 lr=0.012060
Epoch[003] Batch [3049]/[3760]	Speed: 117.309019 samples/sec	 accuracy=7.508709	 loss=4.778891	 lr=0.012099
Epoch[003] Batch [3099]/[3760]	Speed: 117.116131 samples/sec	 accuracy=7.526714	 loss=4.777004	 lr=0.012137
Epoch[003] Batch [3149]/[3760]	Speed: 117.296876 samples/sec	 accuracy=7.534722	 loss=4.774918	 lr=0.012176
Epoch[003] Batch [3199]/[3760]	Speed: 117.426316 samples/sec	 accuracy=7.556641	 loss=4.773295	 lr=0.012215
Epoch[003] Batch [3249]/[3760]	Speed: 116.719264 samples/sec	 accuracy=7.569231	 loss=4.771864	 lr=0.012254
Epoch[003] Batch [3299]/[3760]	Speed: 116.684141 samples/sec	 accuracy=7.580966	 loss=4.770049	 lr=0.012292
Epoch[003] Batch [3349]/[3760]	Speed: 116.869431 samples/sec	 accuracy=7.590951	 loss=4.768387	 lr=0.012331
Epoch[003] Batch [3399]/[3760]	Speed: 116.265770 samples/sec	 accuracy=7.613051	 loss=4.765684	 lr=0.012370
Epoch[003] Batch [3449]/[3760]	Speed: 116.718620 samples/sec	 accuracy=7.626359	 loss=4.764076	 lr=0.012409
Epoch[003] Batch [3499]/[3760]	Speed: 116.758958 samples/sec	 accuracy=7.647321	 loss=4.761828	 lr=0.012447
Epoch[003] Batch [3549]/[3760]	Speed: 117.517570 samples/sec	 accuracy=7.674736	 loss=4.759727	 lr=0.012486
Epoch[003] Batch [3599]/[3760]	Speed: 116.853767 samples/sec	 accuracy=7.696181	 loss=4.757577	 lr=0.012525
Epoch[003] Batch [3649]/[3760]	Speed: 117.492216 samples/sec	 accuracy=7.718322	 loss=4.755102	 lr=0.012563
Epoch[003] Batch [3699]/[3760]	Speed: 116.648638 samples/sec	 accuracy=7.744088	 loss=4.753268	 lr=0.012602
Epoch[003] Batch [3749]/[3760]	Speed: 120.326868 samples/sec	 accuracy=7.757917	 loss=4.751702	 lr=0.012641
Batch [0049]/[0303]: acc-top1=6.906250 acc-top5=22.593750
Batch [0099]/[0303]: acc-top1=6.484375 acc-top5=21.343750
Batch [0149]/[0303]: acc-top1=7.958333 acc-top5=24.312500
Batch [0199]/[0303]: acc-top1=9.039062 acc-top5=25.546875
Batch [0249]/[0303]: acc-top1=8.937500 acc-top5=25.550000
Batch [0299]/[0303]: acc-top1=8.802083 acc-top5=25.619792
[Epoch 003] training: accuracy=7.761802	 loss=4.751092
[Epoch 003] speed: 116 samples/sec	time cost: 2187.395957
[Epoch 003] validation: acc-top1=8.828383 acc-top5=25.670380 loss=4.666686
Epoch[004] Batch [0049]/[3760]	Speed: 87.652529 samples/sec	 accuracy=9.468750	 loss=4.560918	 lr=0.012687
Epoch[004] Batch [0099]/[3760]	Speed: 117.156288 samples/sec	 accuracy=9.609375	 loss=4.570287	 lr=0.012726
Epoch[004] Batch [0149]/[3760]	Speed: 116.776857 samples/sec	 accuracy=9.666667	 loss=4.555873	 lr=0.012765
Epoch[004] Batch [0199]/[3760]	Speed: 116.719775 samples/sec	 accuracy=9.609375	 loss=4.568153	 lr=0.012804
Epoch[004] Batch [0249]/[3760]	Speed: 116.712275 samples/sec	 accuracy=9.393750	 loss=4.577092	 lr=0.012842
Epoch[004] Batch [0299]/[3760]	Speed: 117.427693 samples/sec	 accuracy=9.375000	 loss=4.582658	 lr=0.012881
Epoch[004] Batch [0349]/[3760]	Speed: 116.804915 samples/sec	 accuracy=9.424107	 loss=4.585356	 lr=0.012920
Epoch[004] Batch [0399]/[3760]	Speed: 116.964907 samples/sec	 accuracy=9.347656	 loss=4.586307	 lr=0.012959
Epoch[004] Batch [0449]/[3760]	Speed: 116.281597 samples/sec	 accuracy=9.364583	 loss=4.581477	 lr=0.012997
Epoch[004] Batch [0499]/[3760]	Speed: 117.529521 samples/sec	 accuracy=9.350000	 loss=4.580889	 lr=0.013036
Epoch[004] Batch [0549]/[3760]	Speed: 116.892770 samples/sec	 accuracy=9.355114	 loss=4.582293	 lr=0.013075
Epoch[004] Batch [0599]/[3760]	Speed: 116.552792 samples/sec	 accuracy=9.320312	 loss=4.580724	 lr=0.013113
Epoch[004] Batch [0649]/[3760]	Speed: 118.075207 samples/sec	 accuracy=9.324519	 loss=4.582131	 lr=0.013152
Epoch[004] Batch [0699]/[3760]	Speed: 116.406502 samples/sec	 accuracy=9.390625	 loss=4.578896	 lr=0.013191
Epoch[004] Batch [0749]/[3760]	Speed: 117.754776 samples/sec	 accuracy=9.422917	 loss=4.577271	 lr=0.013230
Epoch[004] Batch [0799]/[3760]	Speed: 117.231408 samples/sec	 accuracy=9.470703	 loss=4.573576	 lr=0.013268
Epoch[004] Batch [0849]/[3760]	Speed: 116.656962 samples/sec	 accuracy=9.417279	 loss=4.573947	 lr=0.013307
Epoch[004] Batch [0899]/[3760]	Speed: 117.346550 samples/sec	 accuracy=9.368056	 loss=4.575180	 lr=0.013346
Epoch[004] Batch [0949]/[3760]	Speed: 116.715963 samples/sec	 accuracy=9.353618	 loss=4.575777	 lr=0.013385
Epoch[004] Batch [0999]/[3760]	Speed: 117.031922 samples/sec	 accuracy=9.428125	 loss=4.573489	 lr=0.013423
Epoch[004] Batch [1049]/[3760]	Speed: 116.487645 samples/sec	 accuracy=9.438988	 loss=4.572307	 lr=0.013462
Epoch[004] Batch [1099]/[3760]	Speed: 117.070281 samples/sec	 accuracy=9.492898	 loss=4.570243	 lr=0.013501
Epoch[004] Batch [1149]/[3760]	Speed: 116.983841 samples/sec	 accuracy=9.536685	 loss=4.565868	 lr=0.013540
Epoch[004] Batch [1199]/[3760]	Speed: 117.243908 samples/sec	 accuracy=9.555990	 loss=4.564858	 lr=0.013578
Epoch[004] Batch [1249]/[3760]	Speed: 117.403073 samples/sec	 accuracy=9.562500	 loss=4.562899	 lr=0.013617
Epoch[004] Batch [1299]/[3760]	Speed: 116.903900 samples/sec	 accuracy=9.585337	 loss=4.560606	 lr=0.013656
Epoch[004] Batch [1349]/[3760]	Speed: 117.314322 samples/sec	 accuracy=9.607639	 loss=4.559126	 lr=0.013694
Epoch[004] Batch [1399]/[3760]	Speed: 117.319142 samples/sec	 accuracy=9.655134	 loss=4.556668	 lr=0.013733
Epoch[004] Batch [1449]/[3760]	Speed: 117.050525 samples/sec	 accuracy=9.660560	 loss=4.555582	 lr=0.013772
Epoch[004] Batch [1499]/[3760]	Speed: 117.583056 samples/sec	 accuracy=9.677083	 loss=4.554253	 lr=0.013811
Epoch[004] Batch [1549]/[3760]	Speed: 116.840507 samples/sec	 accuracy=9.664315	 loss=4.552926	 lr=0.013849
Epoch[004] Batch [1599]/[3760]	Speed: 117.678962 samples/sec	 accuracy=9.697266	 loss=4.551352	 lr=0.013888
Epoch[004] Batch [1649]/[3760]	Speed: 116.727326 samples/sec	 accuracy=9.731061	 loss=4.549433	 lr=0.013927
Epoch[004] Batch [1699]/[3760]	Speed: 116.621206 samples/sec	 accuracy=9.768382	 loss=4.546532	 lr=0.013966
Epoch[004] Batch [1749]/[3760]	Speed: 117.041120 samples/sec	 accuracy=9.795536	 loss=4.544602	 lr=0.014004
Epoch[004] Batch [1799]/[3760]	Speed: 117.140016 samples/sec	 accuracy=9.828125	 loss=4.542561	 lr=0.014043
Epoch[004] Batch [1849]/[3760]	Speed: 116.530499 samples/sec	 accuracy=9.831926	 loss=4.541422	 lr=0.014082
Epoch[004] Batch [1899]/[3760]	Speed: 117.314092 samples/sec	 accuracy=9.865132	 loss=4.539370	 lr=0.014120
Epoch[004] Batch [1949]/[3760]	Speed: 116.056132 samples/sec	 accuracy=9.885417	 loss=4.538064	 lr=0.014159
Epoch[004] Batch [1999]/[3760]	Speed: 116.882338 samples/sec	 accuracy=9.910156	 loss=4.537030	 lr=0.014198
Epoch[004] Batch [2049]/[3760]	Speed: 116.445576 samples/sec	 accuracy=9.926829	 loss=4.535144	 lr=0.014237
Epoch[004] Batch [2099]/[3760]	Speed: 116.682325 samples/sec	 accuracy=9.956845	 loss=4.533435	 lr=0.014275
Epoch[004] Batch [2149]/[3760]	Speed: 116.944985 samples/sec	 accuracy=9.993459	 loss=4.531625	 lr=0.014314
Epoch[004] Batch [2199]/[3760]	Speed: 117.130352 samples/sec	 accuracy=9.983665	 loss=4.530680	 lr=0.014353
Epoch[004] Batch [2249]/[3760]	Speed: 116.563993 samples/sec	 accuracy=9.979861	 loss=4.530609	 lr=0.014392
Epoch[004] Batch [2299]/[3760]	Speed: 117.260767 samples/sec	 accuracy=9.975543	 loss=4.530022	 lr=0.014430
Epoch[004] Batch [2349]/[3760]	Speed: 117.639422 samples/sec	 accuracy=9.964761	 loss=4.529707	 lr=0.014469
Epoch[004] Batch [2399]/[3760]	Speed: 117.129716 samples/sec	 accuracy=9.971354	 loss=4.528644	 lr=0.014508
Epoch[004] Batch [2449]/[3760]	Speed: 117.100481 samples/sec	 accuracy=9.989796	 loss=4.526704	 lr=0.014547
Epoch[004] Batch [2499]/[3760]	Speed: 117.039182 samples/sec	 accuracy=9.990625	 loss=4.525767	 lr=0.014585
Epoch[004] Batch [2549]/[3760]	Speed: 117.845845 samples/sec	 accuracy=10.022059	 loss=4.523435	 lr=0.014624
Epoch[004] Batch [2599]/[3760]	Speed: 116.877898 samples/sec	 accuracy=10.039663	 loss=4.521607	 lr=0.014663
Epoch[004] Batch [2649]/[3760]	Speed: 117.212677 samples/sec	 accuracy=10.061321	 loss=4.519631	 lr=0.014701
Epoch[004] Batch [2699]/[3760]	Speed: 117.794145 samples/sec	 accuracy=10.078704	 loss=4.517792	 lr=0.014740
Epoch[004] Batch [2749]/[3760]	Speed: 117.067528 samples/sec	 accuracy=10.098295	 loss=4.515963	 lr=0.014779
Epoch[004] Batch [2799]/[3760]	Speed: 116.673016 samples/sec	 accuracy=10.130580	 loss=4.514009	 lr=0.014818
Epoch[004] Batch [2849]/[3760]	Speed: 117.239269 samples/sec	 accuracy=10.139254	 loss=4.511963	 lr=0.014856
Epoch[004] Batch [2899]/[3760]	Speed: 117.035029 samples/sec	 accuracy=10.149246	 loss=4.510858	 lr=0.014895
Epoch[004] Batch [2949]/[3760]	Speed: 118.036971 samples/sec	 accuracy=10.172140	 loss=4.509372	 lr=0.014934
Epoch[004] Batch [2999]/[3760]	Speed: 117.621576 samples/sec	 accuracy=10.185937	 loss=4.508511	 lr=0.014973
Epoch[004] Batch [3049]/[3760]	Speed: 116.541316 samples/sec	 accuracy=10.221311	 loss=4.506894	 lr=0.015011
Epoch[004] Batch [3099]/[3760]	Speed: 116.830071 samples/sec	 accuracy=10.243952	 loss=4.504829	 lr=0.015050
Epoch[004] Batch [3149]/[3760]	Speed: 117.441260 samples/sec	 accuracy=10.261905	 loss=4.502997	 lr=0.015089
Epoch[004] Batch [3199]/[3760]	Speed: 116.933452 samples/sec	 accuracy=10.284668	 loss=4.501680	 lr=0.015127
Epoch[004] Batch [3249]/[3760]	Speed: 117.294737 samples/sec	 accuracy=10.323077	 loss=4.499726	 lr=0.015166
Epoch[004] Batch [3299]/[3760]	Speed: 117.020307 samples/sec	 accuracy=10.341383	 loss=4.497758	 lr=0.015205
Epoch[004] Batch [3349]/[3760]	Speed: 117.127387 samples/sec	 accuracy=10.364739	 loss=4.495577	 lr=0.015244
Epoch[004] Batch [3399]/[3760]	Speed: 117.028564 samples/sec	 accuracy=10.394761	 loss=4.493515	 lr=0.015282
Epoch[004] Batch [3449]/[3760]	Speed: 117.030431 samples/sec	 accuracy=10.392663	 loss=4.492642	 lr=0.015321
Epoch[004] Batch [3499]/[3760]	Speed: 117.444669 samples/sec	 accuracy=10.402679	 loss=4.491206	 lr=0.015360
Epoch[004] Batch [3549]/[3760]	Speed: 116.897919 samples/sec	 accuracy=10.420775	 loss=4.489631	 lr=0.015399
Epoch[004] Batch [3599]/[3760]	Speed: 116.942241 samples/sec	 accuracy=10.440538	 loss=4.488342	 lr=0.015437
Epoch[004] Batch [3649]/[3760]	Speed: 117.713752 samples/sec	 accuracy=10.464469	 loss=4.487352	 lr=0.015476
Epoch[004] Batch [3699]/[3760]	Speed: 116.897088 samples/sec	 accuracy=10.480152	 loss=4.485909	 lr=0.015515
Epoch[004] Batch [3749]/[3760]	Speed: 121.056801 samples/sec	 accuracy=10.498333	 loss=4.484260	 lr=0.015554
Batch [0049]/[0303]: acc-top1=7.656250 acc-top5=26.250000
Batch [0099]/[0303]: acc-top1=9.562500 acc-top5=26.875000
Batch [0149]/[0303]: acc-top1=10.677083 acc-top5=29.166667
Batch [0199]/[0303]: acc-top1=11.726562 acc-top5=29.835938
Batch [0249]/[0303]: acc-top1=11.787500 acc-top5=29.787500
Batch [0299]/[0303]: acc-top1=11.458333 acc-top5=29.718750
[Epoch 004] training: accuracy=10.504488	 loss=4.483943
[Epoch 004] speed: 116 samples/sec	time cost: 2184.262303
[Epoch 004] validation: acc-top1=11.458333 acc-top5=29.677186 loss=4.490511
Epoch[005] Batch [0049]/[3759]	Speed: 86.002720 samples/sec	 accuracy=12.375000	 loss=4.430110	 lr=0.015600
Epoch[005] Batch [0099]/[3759]	Speed: 117.509686 samples/sec	 accuracy=12.468750	 loss=4.369273	 lr=0.015639
Epoch[005] Batch [0149]/[3759]	Speed: 117.161436 samples/sec	 accuracy=12.312500	 loss=4.362971	 lr=0.015677
Epoch[005] Batch [0199]/[3759]	Speed: 116.763716 samples/sec	 accuracy=12.179688	 loss=4.347680	 lr=0.015716
Epoch[005] Batch [0249]/[3759]	Speed: 117.562073 samples/sec	 accuracy=12.012500	 loss=4.352484	 lr=0.015755
Epoch[005] Batch [0299]/[3759]	Speed: 116.834972 samples/sec	 accuracy=12.031250	 loss=4.347519	 lr=0.015794
Epoch[005] Batch [0349]/[3759]	Speed: 116.313556 samples/sec	 accuracy=12.040179	 loss=4.339740	 lr=0.015832
Epoch[005] Batch [0399]/[3759]	Speed: 116.370231 samples/sec	 accuracy=12.070312	 loss=4.336586	 lr=0.015871
Epoch[005] Batch [0449]/[3759]	Speed: 116.304585 samples/sec	 accuracy=12.020833	 loss=4.338408	 lr=0.015910
Epoch[005] Batch [0499]/[3759]	Speed: 117.058272 samples/sec	 accuracy=12.128125	 loss=4.336280	 lr=0.015949
Epoch[005] Batch [0549]/[3759]	Speed: 116.547783 samples/sec	 accuracy=12.142045	 loss=4.337410	 lr=0.015987
Epoch[005] Batch [0599]/[3759]	Speed: 116.001781 samples/sec	 accuracy=12.161458	 loss=4.337290	 lr=0.016026
Epoch[005] Batch [0649]/[3759]	Speed: 115.716762 samples/sec	 accuracy=12.211538	 loss=4.335390	 lr=0.016065
Epoch[005] Batch [0699]/[3759]	Speed: 117.334997 samples/sec	 accuracy=12.207589	 loss=4.336818	 lr=0.016103
Epoch[005] Batch [0749]/[3759]	Speed: 116.053497 samples/sec	 accuracy=12.210417	 loss=4.339211	 lr=0.016142
Epoch[005] Batch [0799]/[3759]	Speed: 117.138374 samples/sec	 accuracy=12.251953	 loss=4.336702	 lr=0.016181
Epoch[005] Batch [0849]/[3759]	Speed: 116.088302 samples/sec	 accuracy=12.257353	 loss=4.333620	 lr=0.016220
Epoch[005] Batch [0899]/[3759]	Speed: 116.558942 samples/sec	 accuracy=12.239583	 loss=4.334427	 lr=0.016258
Epoch[005] Batch [0949]/[3759]	Speed: 117.219004 samples/sec	 accuracy=12.261513	 loss=4.331481	 lr=0.016297
Epoch[005] Batch [0999]/[3759]	Speed: 117.118762 samples/sec	 accuracy=12.235938	 loss=4.331663	 lr=0.016336
Epoch[005] Batch [1049]/[3759]	Speed: 117.283100 samples/sec	 accuracy=12.263393	 loss=4.332359	 lr=0.016375
Epoch[005] Batch [1099]/[3759]	Speed: 116.432505 samples/sec	 accuracy=12.248580	 loss=4.333225	 lr=0.016413
Epoch[005] Batch [1149]/[3759]	Speed: 117.136326 samples/sec	 accuracy=12.274457	 loss=4.331109	 lr=0.016452
Epoch[005] Batch [1199]/[3759]	Speed: 115.695554 samples/sec	 accuracy=12.287760	 loss=4.330393	 lr=0.016491
Epoch[005] Batch [1249]/[3759]	Speed: 116.646987 samples/sec	 accuracy=12.300000	 loss=4.328593	 lr=0.016530
Epoch[005] Batch [1299]/[3759]	Speed: 117.247705 samples/sec	 accuracy=12.332933	 loss=4.328123	 lr=0.016568
Epoch[005] Batch [1349]/[3759]	Speed: 117.088526 samples/sec	 accuracy=12.362269	 loss=4.327915	 lr=0.016607
Epoch[005] Batch [1399]/[3759]	Speed: 116.181997 samples/sec	 accuracy=12.389509	 loss=4.325356	 lr=0.016646
Epoch[005] Batch [1449]/[3759]	Speed: 116.581950 samples/sec	 accuracy=12.387931	 loss=4.326266	 lr=0.016684
Epoch[005] Batch [1499]/[3759]	Speed: 117.290466 samples/sec	 accuracy=12.396875	 loss=4.325545	 lr=0.016723
Epoch[005] Batch [1549]/[3759]	Speed: 116.863312 samples/sec	 accuracy=12.420363	 loss=4.323794	 lr=0.016762
Epoch[005] Batch [1599]/[3759]	Speed: 116.686803 samples/sec	 accuracy=12.417969	 loss=4.322178	 lr=0.016801
Epoch[005] Batch [1649]/[3759]	Speed: 117.079734 samples/sec	 accuracy=12.432765	 loss=4.321091	 lr=0.016839
Epoch[005] Batch [1699]/[3759]	Speed: 116.943917 samples/sec	 accuracy=12.442096	 loss=4.319463	 lr=0.016878
Epoch[005] Batch [1749]/[3759]	Speed: 117.066605 samples/sec	 accuracy=12.454464	 loss=4.319476	 lr=0.016917
Epoch[005] Batch [1799]/[3759]	Speed: 116.956209 samples/sec	 accuracy=12.474826	 loss=4.318089	 lr=0.016956
Epoch[005] Batch [1849]/[3759]	Speed: 116.703651 samples/sec	 accuracy=12.494088	 loss=4.316717	 lr=0.016994
Epoch[005] Batch [1899]/[3759]	Speed: 117.401400 samples/sec	 accuracy=12.513980	 loss=4.315384	 lr=0.017033
Epoch[005] Batch [1949]/[3759]	Speed: 116.088657 samples/sec	 accuracy=12.549679	 loss=4.313844	 lr=0.017072
Epoch[005] Batch [1999]/[3759]	Speed: 116.276164 samples/sec	 accuracy=12.549219	 loss=4.313029	 lr=0.017110
Epoch[005] Batch [2049]/[3759]	Speed: 116.733067 samples/sec	 accuracy=12.575457	 loss=4.310697	 lr=0.017149
Epoch[005] Batch [2099]/[3759]	Speed: 116.878327 samples/sec	 accuracy=12.598958	 loss=4.308872	 lr=0.017188
Epoch[005] Batch [2149]/[3759]	Speed: 116.869040 samples/sec	 accuracy=12.615552	 loss=4.307551	 lr=0.017227
Epoch[005] Batch [2199]/[3759]	Speed: 116.378728 samples/sec	 accuracy=12.657670	 loss=4.305156	 lr=0.017265
Epoch[005] Batch [2249]/[3759]	Speed: 116.262057 samples/sec	 accuracy=12.704167	 loss=4.302920	 lr=0.017304
Epoch[005] Batch [2299]/[3759]	Speed: 116.777086 samples/sec	 accuracy=12.726223	 loss=4.300435	 lr=0.017343
Epoch[005] Batch [2349]/[3759]	Speed: 116.298508 samples/sec	 accuracy=12.744016	 loss=4.299126	 lr=0.017382
Epoch[005] Batch [2399]/[3759]	Speed: 116.441279 samples/sec	 accuracy=12.762370	 loss=4.297757	 lr=0.017420
Epoch[005] Batch [2449]/[3759]	Speed: 116.565728 samples/sec	 accuracy=12.769770	 loss=4.296628	 lr=0.017459
Epoch[005] Batch [2499]/[3759]	Speed: 117.405935 samples/sec	 accuracy=12.805000	 loss=4.295380	 lr=0.017498
Epoch[005] Batch [2549]/[3759]	Speed: 116.733153 samples/sec	 accuracy=12.813113	 loss=4.294326	 lr=0.017537
Epoch[005] Batch [2599]/[3759]	Speed: 117.196066 samples/sec	 accuracy=12.859375	 loss=4.292161	 lr=0.017575
Epoch[005] Batch [2649]/[3759]	Speed: 117.490957 samples/sec	 accuracy=12.867925	 loss=4.290498	 lr=0.017614
Epoch[005] Batch [2699]/[3759]	Speed: 117.300817 samples/sec	 accuracy=12.873843	 loss=4.289084	 lr=0.017653
Epoch[005] Batch [2749]/[3759]	Speed: 116.778117 samples/sec	 accuracy=12.883523	 loss=4.288877	 lr=0.017691
Epoch[005] Batch [2799]/[3759]	Speed: 117.823920 samples/sec	 accuracy=12.888951	 loss=4.287853	 lr=0.017730
Epoch[005] Batch [2849]/[3759]	Speed: 116.828033 samples/sec	 accuracy=12.891447	 loss=4.287071	 lr=0.017769
Epoch[005] Batch [2899]/[3759]	Speed: 117.470748 samples/sec	 accuracy=12.897091	 loss=4.286975	 lr=0.017808
Epoch[005] Batch [2949]/[3759]	Speed: 118.215178 samples/sec	 accuracy=12.905720	 loss=4.285692	 lr=0.017846
Epoch[005] Batch [2999]/[3759]	Speed: 116.978182 samples/sec	 accuracy=12.930208	 loss=4.283905	 lr=0.017885
Epoch[005] Batch [3049]/[3759]	Speed: 116.968650 samples/sec	 accuracy=12.943135	 loss=4.281718	 lr=0.017924
Epoch[005] Batch [3099]/[3759]	Speed: 116.475620 samples/sec	 accuracy=12.962702	 loss=4.280797	 lr=0.017963
Epoch[005] Batch [3149]/[3759]	Speed: 117.096923 samples/sec	 accuracy=12.975694	 loss=4.279895	 lr=0.018001
Epoch[005] Batch [3199]/[3759]	Speed: 117.543236 samples/sec	 accuracy=12.992676	 loss=4.279003	 lr=0.018040
Epoch[005] Batch [3249]/[3759]	Speed: 117.170842 samples/sec	 accuracy=13.012019	 loss=4.277954	 lr=0.018079
Epoch[005] Batch [3299]/[3759]	Speed: 116.637332 samples/sec	 accuracy=13.026042	 loss=4.276604	 lr=0.018117
Epoch[005] Batch [3349]/[3759]	Speed: 116.835947 samples/sec	 accuracy=13.035914	 loss=4.275553	 lr=0.018156
Epoch[005] Batch [3399]/[3759]	Speed: 116.611055 samples/sec	 accuracy=13.051930	 loss=4.273900	 lr=0.018195
Epoch[005] Batch [3449]/[3759]	Speed: 117.420941 samples/sec	 accuracy=13.060688	 loss=4.272333	 lr=0.018234
Epoch[005] Batch [3499]/[3759]	Speed: 117.792641 samples/sec	 accuracy=13.071875	 loss=4.271096	 lr=0.018272
Epoch[005] Batch [3549]/[3759]	Speed: 117.819635 samples/sec	 accuracy=13.074824	 loss=4.269873	 lr=0.018311
Epoch[005] Batch [3599]/[3759]	Speed: 117.249871 samples/sec	 accuracy=13.102431	 loss=4.268811	 lr=0.018350
Epoch[005] Batch [3649]/[3759]	Speed: 117.445503 samples/sec	 accuracy=13.125428	 loss=4.267234	 lr=0.018389
Epoch[005] Batch [3699]/[3759]	Speed: 117.910534 samples/sec	 accuracy=13.138514	 loss=4.266006	 lr=0.018427
Epoch[005] Batch [3749]/[3759]	Speed: 120.309777 samples/sec	 accuracy=13.148333	 loss=4.264477	 lr=0.018466
Batch [0049]/[0303]: acc-top1=11.437500 acc-top5=32.562500
Batch [0099]/[0303]: acc-top1=12.234375 acc-top5=33.515625
Batch [0149]/[0303]: acc-top1=13.239583 acc-top5=35.052083
Batch [0199]/[0303]: acc-top1=14.656250 acc-top5=36.445312
Batch [0249]/[0303]: acc-top1=14.387500 acc-top5=36.162500
Batch [0299]/[0303]: acc-top1=14.229167 acc-top5=36.182292
[Epoch 005] training: accuracy=13.149275	 loss=4.264532
[Epoch 005] speed: 116 samples/sec	time cost: 2187.873753
[Epoch 005] validation: acc-top1=14.315182 acc-top5=36.200495 loss=4.107787
Epoch[006] Batch [0049]/[3760]	Speed: 87.595468 samples/sec	 accuracy=14.312500	 loss=4.166405	 lr=0.018512
Epoch[006] Batch [0099]/[3760]	Speed: 116.770367 samples/sec	 accuracy=14.562500	 loss=4.148279	 lr=0.018551
Epoch[006] Batch [0149]/[3760]	Speed: 118.064185 samples/sec	 accuracy=14.614583	 loss=4.142893	 lr=0.018589
Epoch[006] Batch [0199]/[3760]	Speed: 116.866192 samples/sec	 accuracy=14.648438	 loss=4.133728	 lr=0.018628
Epoch[006] Batch [0249]/[3760]	Speed: 117.782189 samples/sec	 accuracy=14.462500	 loss=4.142129	 lr=0.018667
Epoch[006] Batch [0299]/[3760]	Speed: 117.380340 samples/sec	 accuracy=14.338542	 loss=4.148500	 lr=0.018705
Epoch[006] Batch [0349]/[3760]	Speed: 117.099661 samples/sec	 accuracy=14.459821	 loss=4.148651	 lr=0.018744
Epoch[006] Batch [0399]/[3760]	Speed: 117.473774 samples/sec	 accuracy=14.511719	 loss=4.146814	 lr=0.018783
Epoch[006] Batch [0449]/[3760]	Speed: 117.937960 samples/sec	 accuracy=14.454861	 loss=4.147383	 lr=0.018822
Epoch[006] Batch [0499]/[3760]	Speed: 117.129011 samples/sec	 accuracy=14.500000	 loss=4.147288	 lr=0.018860
Epoch[006] Batch [0549]/[3760]	Speed: 117.451721 samples/sec	 accuracy=14.536932	 loss=4.144740	 lr=0.018899
Epoch[006] Batch [0599]/[3760]	Speed: 117.289050 samples/sec	 accuracy=14.648438	 loss=4.140534	 lr=0.018938
Epoch[006] Batch [0649]/[3760]	Speed: 116.388588 samples/sec	 accuracy=14.682692	 loss=4.140790	 lr=0.018977
Epoch[006] Batch [0699]/[3760]	Speed: 117.343823 samples/sec	 accuracy=14.758929	 loss=4.138157	 lr=0.019015
Epoch[006] Batch [0749]/[3760]	Speed: 116.923935 samples/sec	 accuracy=14.772917	 loss=4.137889	 lr=0.019054
Epoch[006] Batch [0799]/[3760]	Speed: 117.543758 samples/sec	 accuracy=14.828125	 loss=4.134481	 lr=0.019093
Epoch[006] Batch [0849]/[3760]	Speed: 116.625472 samples/sec	 accuracy=14.860294	 loss=4.135590	 lr=0.019131
Epoch[006] Batch [0899]/[3760]	Speed: 117.855030 samples/sec	 accuracy=14.875000	 loss=4.133789	 lr=0.019170
Epoch[006] Batch [0949]/[3760]	Speed: 117.026730 samples/sec	 accuracy=14.911184	 loss=4.132775	 lr=0.019209
Epoch[006] Batch [0999]/[3760]	Speed: 116.789889 samples/sec	 accuracy=14.948438	 loss=4.131114	 lr=0.019248
Epoch[006] Batch [1049]/[3760]	Speed: 116.087837 samples/sec	 accuracy=14.968750	 loss=4.129899	 lr=0.019286
Epoch[006] Batch [1099]/[3760]	Speed: 116.636269 samples/sec	 accuracy=14.984375	 loss=4.127819	 lr=0.019325
Epoch[006] Batch [1149]/[3760]	Speed: 117.125028 samples/sec	 accuracy=15.004076	 loss=4.126737	 lr=0.019364
Epoch[006] Batch [1199]/[3760]	Speed: 117.078265 samples/sec	 accuracy=15.016927	 loss=4.126021	 lr=0.019403
Epoch[006] Batch [1249]/[3760]	Speed: 116.514954 samples/sec	 accuracy=15.036250	 loss=4.123738	 lr=0.019441
Epoch[006] Batch [1299]/[3760]	Speed: 116.726562 samples/sec	 accuracy=14.993990	 loss=4.123308	 lr=0.019480
Epoch[006] Batch [1349]/[3760]	Speed: 117.346556 samples/sec	 accuracy=14.988426	 loss=4.121619	 lr=0.019519
Epoch[006] Batch [1399]/[3760]	Speed: 117.172245 samples/sec	 accuracy=14.982143	 loss=4.121268	 lr=0.019558
Epoch[006] Batch [1449]/[3760]	Speed: 117.285913 samples/sec	 accuracy=14.963362	 loss=4.122721	 lr=0.019596
Epoch[006] Batch [1499]/[3760]	Speed: 116.760855 samples/sec	 accuracy=14.950000	 loss=4.122697	 lr=0.019635
Epoch[006] Batch [1549]/[3760]	Speed: 117.048960 samples/sec	 accuracy=14.939516	 loss=4.123065	 lr=0.019674
Epoch[006] Batch [1599]/[3760]	Speed: 117.171961 samples/sec	 accuracy=14.935547	 loss=4.122083	 lr=0.019712
Epoch[006] Batch [1649]/[3760]	Speed: 117.192058 samples/sec	 accuracy=14.941288	 loss=4.121958	 lr=0.019751
Epoch[006] Batch [1699]/[3760]	Speed: 117.088100 samples/sec	 accuracy=14.941176	 loss=4.120747	 lr=0.019790
Epoch[006] Batch [1749]/[3760]	Speed: 117.180260 samples/sec	 accuracy=14.958036	 loss=4.119913	 lr=0.019829
Epoch[006] Batch [1799]/[3760]	Speed: 117.121207 samples/sec	 accuracy=14.968750	 loss=4.118754	 lr=0.019867
Epoch[006] Batch [1849]/[3760]	Speed: 117.458456 samples/sec	 accuracy=14.987331	 loss=4.117721	 lr=0.019906
Epoch[006] Batch [1899]/[3760]	Speed: 117.166539 samples/sec	 accuracy=15.009046	 loss=4.116816	 lr=0.019945
Epoch[006] Batch [1949]/[3760]	Speed: 116.511560 samples/sec	 accuracy=15.005609	 loss=4.115926	 lr=0.019984
Epoch[006] Batch [1999]/[3760]	Speed: 117.056413 samples/sec	 accuracy=15.026563	 loss=4.114818	 lr=0.020022
Epoch[006] Batch [2049]/[3760]	Speed: 117.210873 samples/sec	 accuracy=15.057165	 loss=4.114226	 lr=0.020061
Epoch[006] Batch [2099]/[3760]	Speed: 117.118879 samples/sec	 accuracy=15.058780	 loss=4.113324	 lr=0.020100
Epoch[006] Batch [2149]/[3760]	Speed: 116.931238 samples/sec	 accuracy=15.078488	 loss=4.111309	 lr=0.020138
Epoch[006] Batch [2199]/[3760]	Speed: 116.878038 samples/sec	 accuracy=15.090199	 loss=4.110325	 lr=0.020177
Epoch[006] Batch [2249]/[3760]	Speed: 116.884262 samples/sec	 accuracy=15.113889	 loss=4.107631	 lr=0.020216
Epoch[006] Batch [2299]/[3760]	Speed: 117.164089 samples/sec	 accuracy=15.108696	 loss=4.107717	 lr=0.020255
Epoch[006] Batch [2349]/[3760]	Speed: 117.070881 samples/sec	 accuracy=15.100399	 loss=4.107000	 lr=0.020293
Epoch[006] Batch [2399]/[3760]	Speed: 117.384543 samples/sec	 accuracy=15.120443	 loss=4.105448	 lr=0.020332
Epoch[006] Batch [2449]/[3760]	Speed: 117.339489 samples/sec	 accuracy=15.145408	 loss=4.103994	 lr=0.020371
Epoch[006] Batch [2499]/[3760]	Speed: 117.238739 samples/sec	 accuracy=15.168750	 loss=4.102021	 lr=0.020410
Epoch[006] Batch [2549]/[3760]	Speed: 117.311624 samples/sec	 accuracy=15.222426	 loss=4.099368	 lr=0.020448
Epoch[006] Batch [2599]/[3760]	Speed: 116.882056 samples/sec	 accuracy=15.213341	 loss=4.098715	 lr=0.020487
Epoch[006] Batch [2649]/[3760]	Speed: 117.562655 samples/sec	 accuracy=15.227005	 loss=4.098034	 lr=0.020526
Epoch[006] Batch [2699]/[3760]	Speed: 116.849434 samples/sec	 accuracy=15.263889	 loss=4.095866	 lr=0.020565
Epoch[006] Batch [2749]/[3760]	Speed: 117.487028 samples/sec	 accuracy=15.268182	 loss=4.094819	 lr=0.020603
Epoch[006] Batch [2799]/[3760]	Speed: 116.611471 samples/sec	 accuracy=15.300223	 loss=4.093211	 lr=0.020642
Epoch[006] Batch [2849]/[3760]	Speed: 116.495122 samples/sec	 accuracy=15.313048	 loss=4.092807	 lr=0.020681
Epoch[006] Batch [2899]/[3760]	Speed: 117.196294 samples/sec	 accuracy=15.320043	 loss=4.092351	 lr=0.020719
Epoch[006] Batch [2949]/[3760]	Speed: 117.380700 samples/sec	 accuracy=15.342691	 loss=4.090930	 lr=0.020758
Epoch[006] Batch [2999]/[3760]	Speed: 117.797219 samples/sec	 accuracy=15.342708	 loss=4.090601	 lr=0.020797
Epoch[006] Batch [3049]/[3760]	Speed: 116.943988 samples/sec	 accuracy=15.355020	 loss=4.090139	 lr=0.020836
Epoch[006] Batch [3099]/[3760]	Speed: 116.659957 samples/sec	 accuracy=15.375000	 loss=4.088536	 lr=0.020874
Epoch[006] Batch [3149]/[3760]	Speed: 117.797477 samples/sec	 accuracy=15.405258	 loss=4.086408	 lr=0.020913
Epoch[006] Batch [3199]/[3760]	Speed: 117.086627 samples/sec	 accuracy=15.425293	 loss=4.085634	 lr=0.020952
Epoch[006] Batch [3249]/[3760]	Speed: 116.904784 samples/sec	 accuracy=15.448077	 loss=4.084784	 lr=0.020991
Epoch[006] Batch [3299]/[3760]	Speed: 116.656206 samples/sec	 accuracy=15.461648	 loss=4.083310	 lr=0.021029
Epoch[006] Batch [3349]/[3760]	Speed: 116.959172 samples/sec	 accuracy=15.466418	 loss=4.082754	 lr=0.021068
Epoch[006] Batch [3399]/[3760]	Speed: 117.041546 samples/sec	 accuracy=15.458180	 loss=4.082679	 lr=0.021107
Epoch[006] Batch [3449]/[3760]	Speed: 117.194082 samples/sec	 accuracy=15.459239	 loss=4.082364	 lr=0.021145
Epoch[006] Batch [3499]/[3760]	Speed: 117.528218 samples/sec	 accuracy=15.481696	 loss=4.081104	 lr=0.021184
Epoch[006] Batch [3549]/[3760]	Speed: 117.166249 samples/sec	 accuracy=15.495599	 loss=4.080413	 lr=0.021223
Epoch[006] Batch [3599]/[3760]	Speed: 116.570787 samples/sec	 accuracy=15.518663	 loss=4.079416	 lr=0.021262
Epoch[006] Batch [3649]/[3760]	Speed: 117.143826 samples/sec	 accuracy=15.537243	 loss=4.077974	 lr=0.021300
Epoch[006] Batch [3699]/[3760]	Speed: 116.457175 samples/sec	 accuracy=15.559122	 loss=4.077769	 lr=0.021339
Epoch[006] Batch [3749]/[3760]	Speed: 120.161772 samples/sec	 accuracy=15.588333	 loss=4.075746	 lr=0.021378
Batch [0049]/[0303]: acc-top1=15.062500 acc-top5=38.625000
Batch [0099]/[0303]: acc-top1=14.546875 acc-top5=37.375000
Batch [0149]/[0303]: acc-top1=16.468750 acc-top5=39.302083
Batch [0199]/[0303]: acc-top1=17.929688 acc-top5=41.007812
Batch [0249]/[0303]: acc-top1=17.850000 acc-top5=40.718750
Batch [0299]/[0303]: acc-top1=17.447917 acc-top5=41.057292
[Epoch 006] training: accuracy=15.590093	 loss=4.075337
[Epoch 006] speed: 116 samples/sec	time cost: 2185.435828
[Epoch 006] validation: acc-top1=17.491749 acc-top5=41.073639 loss=3.930839
Epoch[007] Batch [0049]/[3760]	Speed: 88.233150 samples/sec	 accuracy=16.437500	 loss=3.999438	 lr=0.021424
Epoch[007] Batch [0099]/[3760]	Speed: 116.854927 samples/sec	 accuracy=17.265625	 loss=3.966712	 lr=0.021463
Epoch[007] Batch [0149]/[3760]	Speed: 116.900769 samples/sec	 accuracy=17.583333	 loss=3.942652	 lr=0.021502
Epoch[007] Batch [0199]/[3760]	Speed: 117.329220 samples/sec	 accuracy=17.585938	 loss=3.942517	 lr=0.021541
Epoch[007] Batch [0249]/[3760]	Speed: 116.491004 samples/sec	 accuracy=17.443750	 loss=3.946841	 lr=0.021579
Epoch[007] Batch [0299]/[3760]	Speed: 117.991866 samples/sec	 accuracy=17.442708	 loss=3.951632	 lr=0.021618
Epoch[007] Batch [0349]/[3760]	Speed: 117.611879 samples/sec	 accuracy=17.464286	 loss=3.952760	 lr=0.021657
Epoch[007] Batch [0399]/[3760]	Speed: 117.505254 samples/sec	 accuracy=17.460938	 loss=3.955668	 lr=0.021695
Epoch[007] Batch [0449]/[3760]	Speed: 117.175616 samples/sec	 accuracy=17.378472	 loss=3.964205	 lr=0.021734
Epoch[007] Batch [0499]/[3760]	Speed: 117.646102 samples/sec	 accuracy=17.340625	 loss=3.963395	 lr=0.021773
Epoch[007] Batch [0549]/[3760]	Speed: 117.138696 samples/sec	 accuracy=17.264205	 loss=3.963059	 lr=0.021812
Epoch[007] Batch [0599]/[3760]	Speed: 117.836220 samples/sec	 accuracy=17.229167	 loss=3.962997	 lr=0.021850
Epoch[007] Batch [0649]/[3760]	Speed: 116.605573 samples/sec	 accuracy=17.213942	 loss=3.963470	 lr=0.021889
Epoch[007] Batch [0699]/[3760]	Speed: 117.056199 samples/sec	 accuracy=17.223214	 loss=3.962810	 lr=0.021928
Epoch[007] Batch [0749]/[3760]	Speed: 116.949430 samples/sec	 accuracy=17.129167	 loss=3.967566	 lr=0.021967
Epoch[007] Batch [0799]/[3760]	Speed: 117.181283 samples/sec	 accuracy=17.117188	 loss=3.964845	 lr=0.022005
Epoch[007] Batch [0849]/[3760]	Speed: 117.441024 samples/sec	 accuracy=17.132353	 loss=3.965254	 lr=0.022044
Epoch[007] Batch [0899]/[3760]	Speed: 117.407174 samples/sec	 accuracy=17.138889	 loss=3.963079	 lr=0.022083
Epoch[007] Batch [0949]/[3760]	Speed: 116.435412 samples/sec	 accuracy=17.151316	 loss=3.960299	 lr=0.022121
Epoch[007] Batch [0999]/[3760]	Speed: 116.573465 samples/sec	 accuracy=17.145313	 loss=3.959211	 lr=0.022160
Epoch[007] Batch [1049]/[3760]	Speed: 116.400014 samples/sec	 accuracy=17.104167	 loss=3.961107	 lr=0.022199
Epoch[007] Batch [1099]/[3760]	Speed: 117.623132 samples/sec	 accuracy=17.107955	 loss=3.960179	 lr=0.022238
Epoch[007] Batch [1149]/[3760]	Speed: 116.872792 samples/sec	 accuracy=17.116848	 loss=3.962447	 lr=0.022276
Epoch[007] Batch [1199]/[3760]	Speed: 116.567489 samples/sec	 accuracy=17.138021	 loss=3.960877	 lr=0.022315
Epoch[007] Batch [1249]/[3760]	Speed: 117.377999 samples/sec	 accuracy=17.152500	 loss=3.960283	 lr=0.022354
Epoch[007] Batch [1299]/[3760]	Speed: 116.867851 samples/sec	 accuracy=17.195913	 loss=3.958372	 lr=0.022393
Epoch[007] Batch [1349]/[3760]	Speed: 116.899982 samples/sec	 accuracy=17.209491	 loss=3.957967	 lr=0.022431
Epoch[007] Batch [1399]/[3760]	Speed: 117.017048 samples/sec	 accuracy=17.226562	 loss=3.956694	 lr=0.022470
Epoch[007] Batch [1449]/[3760]	Speed: 116.967994 samples/sec	 accuracy=17.224138	 loss=3.956547	 lr=0.022509
Epoch[007] Batch [1499]/[3760]	Speed: 116.688702 samples/sec	 accuracy=17.235417	 loss=3.955331	 lr=0.022548
Epoch[007] Batch [1549]/[3760]	Speed: 117.094976 samples/sec	 accuracy=17.234879	 loss=3.955596	 lr=0.022586
Epoch[007] Batch [1599]/[3760]	Speed: 117.178263 samples/sec	 accuracy=17.247070	 loss=3.954660	 lr=0.022625
Epoch[007] Batch [1649]/[3760]	Speed: 117.117199 samples/sec	 accuracy=17.283144	 loss=3.952967	 lr=0.022664
Epoch[007] Batch [1699]/[3760]	Speed: 116.444322 samples/sec	 accuracy=17.284926	 loss=3.952973	 lr=0.022702
Epoch[007] Batch [1749]/[3760]	Speed: 117.273713 samples/sec	 accuracy=17.313393	 loss=3.950594	 lr=0.022741
Epoch[007] Batch [1799]/[3760]	Speed: 116.552085 samples/sec	 accuracy=17.315972	 loss=3.949613	 lr=0.022780
Epoch[007] Batch [1849]/[3760]	Speed: 116.740555 samples/sec	 accuracy=17.321791	 loss=3.949311	 lr=0.022819
Epoch[007] Batch [1899]/[3760]	Speed: 117.182126 samples/sec	 accuracy=17.352796	 loss=3.948047	 lr=0.022857
Epoch[007] Batch [1949]/[3760]	Speed: 116.862776 samples/sec	 accuracy=17.306891	 loss=3.949344	 lr=0.022896
Epoch[007] Batch [1999]/[3760]	Speed: 116.825589 samples/sec	 accuracy=17.321875	 loss=3.948690	 lr=0.022935
Epoch[007] Batch [2049]/[3760]	Speed: 117.264475 samples/sec	 accuracy=17.318598	 loss=3.949021	 lr=0.022974
Epoch[007] Batch [2099]/[3760]	Speed: 117.749515 samples/sec	 accuracy=17.321429	 loss=3.949752	 lr=0.023012
Epoch[007] Batch [2149]/[3760]	Speed: 117.707613 samples/sec	 accuracy=17.321948	 loss=3.948949	 lr=0.023051
Epoch[007] Batch [2199]/[3760]	Speed: 116.405571 samples/sec	 accuracy=17.335227	 loss=3.947902	 lr=0.023090
Epoch[007] Batch [2249]/[3760]	Speed: 117.005670 samples/sec	 accuracy=17.320139	 loss=3.947573	 lr=0.023129
Epoch[007] Batch [2299]/[3760]	Speed: 116.710834 samples/sec	 accuracy=17.362092	 loss=3.946039	 lr=0.023167
Epoch[007] Batch [2349]/[3760]	Speed: 116.969690 samples/sec	 accuracy=17.342420	 loss=3.945982	 lr=0.023206
Epoch[007] Batch [2399]/[3760]	Speed: 117.513265 samples/sec	 accuracy=17.366536	 loss=3.944591	 lr=0.023245
Epoch[007] Batch [2449]/[3760]	Speed: 117.448004 samples/sec	 accuracy=17.389668	 loss=3.943152	 lr=0.023283
Epoch[007] Batch [2499]/[3760]	Speed: 117.463834 samples/sec	 accuracy=17.429375	 loss=3.942075	 lr=0.023322
Epoch[007] Batch [2549]/[3760]	Speed: 116.955515 samples/sec	 accuracy=17.432598	 loss=3.941860	 lr=0.023361
Epoch[007] Batch [2599]/[3760]	Speed: 117.740212 samples/sec	 accuracy=17.446514	 loss=3.940778	 lr=0.023400
Epoch[007] Batch [2649]/[3760]	Speed: 117.196094 samples/sec	 accuracy=17.446934	 loss=3.940475	 lr=0.023438
Epoch[007] Batch [2699]/[3760]	Speed: 116.445958 samples/sec	 accuracy=17.471065	 loss=3.938309	 lr=0.023477
Epoch[007] Batch [2749]/[3760]	Speed: 117.080811 samples/sec	 accuracy=17.471023	 loss=3.938029	 lr=0.023516
Epoch[007] Batch [2799]/[3760]	Speed: 117.885691 samples/sec	 accuracy=17.492746	 loss=3.937803	 lr=0.023555
Epoch[007] Batch [2849]/[3760]	Speed: 116.822205 samples/sec	 accuracy=17.496711	 loss=3.937254	 lr=0.023593
Epoch[007] Batch [2899]/[3760]	Speed: 117.123853 samples/sec	 accuracy=17.523707	 loss=3.935635	 lr=0.023632
Epoch[007] Batch [2949]/[3760]	Speed: 116.618471 samples/sec	 accuracy=17.531250	 loss=3.935581	 lr=0.023671
Epoch[007] Batch [2999]/[3760]	Speed: 116.925845 samples/sec	 accuracy=17.543229	 loss=3.935482	 lr=0.023709
Epoch[007] Batch [3049]/[3760]	Speed: 117.952227 samples/sec	 accuracy=17.563525	 loss=3.934932	 lr=0.023748
Epoch[007] Batch [3099]/[3760]	Speed: 117.372030 samples/sec	 accuracy=17.567036	 loss=3.934743	 lr=0.023787
Epoch[007] Batch [3149]/[3760]	Speed: 117.482246 samples/sec	 accuracy=17.588294	 loss=3.933626	 lr=0.023826
Epoch[007] Batch [3199]/[3760]	Speed: 117.560164 samples/sec	 accuracy=17.613770	 loss=3.932663	 lr=0.023864
Epoch[007] Batch [3249]/[3760]	Speed: 117.812820 samples/sec	 accuracy=17.624519	 loss=3.932250	 lr=0.023903
Epoch[007] Batch [3299]/[3760]	Speed: 118.076694 samples/sec	 accuracy=17.649621	 loss=3.930207	 lr=0.023942
Epoch[007] Batch [3349]/[3760]	Speed: 117.334924 samples/sec	 accuracy=17.663246	 loss=3.929295	 lr=0.023981
Epoch[007] Batch [3399]/[3760]	Speed: 117.202491 samples/sec	 accuracy=17.655790	 loss=3.928680	 lr=0.024019
Epoch[007] Batch [3449]/[3760]	Speed: 117.146740 samples/sec	 accuracy=17.686594	 loss=3.927674	 lr=0.024058
Epoch[007] Batch [3499]/[3760]	Speed: 116.806611 samples/sec	 accuracy=17.726786	 loss=3.925900	 lr=0.024097
Epoch[007] Batch [3549]/[3760]	Speed: 116.953080 samples/sec	 accuracy=17.726232	 loss=3.925476	 lr=0.024136
Epoch[007] Batch [3599]/[3760]	Speed: 117.995545 samples/sec	 accuracy=17.736979	 loss=3.924597	 lr=0.024174
Epoch[007] Batch [3649]/[3760]	Speed: 117.047122 samples/sec	 accuracy=17.739298	 loss=3.923965	 lr=0.024213
Epoch[007] Batch [3699]/[3760]	Speed: 117.413274 samples/sec	 accuracy=17.758024	 loss=3.923427	 lr=0.024252
Epoch[007] Batch [3749]/[3760]	Speed: 120.688345 samples/sec	 accuracy=17.776667	 loss=3.922583	 lr=0.024290
Batch [0049]/[0303]: acc-top1=17.687500 acc-top5=43.750000
Batch [0099]/[0303]: acc-top1=16.781250 acc-top5=40.812500
Batch [0149]/[0303]: acc-top1=18.562500 acc-top5=42.718750
Batch [0199]/[0303]: acc-top1=19.968750 acc-top5=44.398438
Batch [0249]/[0303]: acc-top1=20.156250 acc-top5=44.706250
Batch [0299]/[0303]: acc-top1=20.119792 acc-top5=44.911458
[Epoch 007] training: accuracy=17.786735	 loss=3.922376
[Epoch 007] speed: 116 samples/sec	time cost: 2183.198476
[Epoch 007] validation: acc-top1=20.162954 acc-top5=44.889645 loss=3.745860
Epoch[008] Batch [0049]/[3759]	Speed: 86.336875 samples/sec	 accuracy=19.375000	 loss=3.835204	 lr=0.024337
Epoch[008] Batch [0099]/[3759]	Speed: 117.977769 samples/sec	 accuracy=18.562500	 loss=3.858016	 lr=0.024376
Epoch[008] Batch [0149]/[3759]	Speed: 116.601118 samples/sec	 accuracy=18.645833	 loss=3.841221	 lr=0.024414
Epoch[008] Batch [0199]/[3759]	Speed: 116.887268 samples/sec	 accuracy=18.484375	 loss=3.846615	 lr=0.024453
Epoch[008] Batch [0249]/[3759]	Speed: 117.225922 samples/sec	 accuracy=18.581250	 loss=3.843380	 lr=0.024492
Epoch[008] Batch [0299]/[3759]	Speed: 116.833941 samples/sec	 accuracy=18.760417	 loss=3.833855	 lr=0.024531
Epoch[008] Batch [0349]/[3759]	Speed: 116.914195 samples/sec	 accuracy=18.754464	 loss=3.834910	 lr=0.024569
Epoch[008] Batch [0399]/[3759]	Speed: 117.407290 samples/sec	 accuracy=19.015625	 loss=3.833723	 lr=0.024608
Epoch[008] Batch [0449]/[3759]	Speed: 116.548577 samples/sec	 accuracy=19.045139	 loss=3.836480	 lr=0.024647
Epoch[008] Batch [0499]/[3759]	Speed: 117.715896 samples/sec	 accuracy=19.131250	 loss=3.837260	 lr=0.024685
Epoch[008] Batch [0549]/[3759]	Speed: 116.559949 samples/sec	 accuracy=19.159091	 loss=3.837270	 lr=0.024724
Epoch[008] Batch [0599]/[3759]	Speed: 117.331346 samples/sec	 accuracy=19.255208	 loss=3.832756	 lr=0.024763
Epoch[008] Batch [0649]/[3759]	Speed: 117.071200 samples/sec	 accuracy=19.278846	 loss=3.830588	 lr=0.024802
Epoch[008] Batch [0699]/[3759]	Speed: 117.512346 samples/sec	 accuracy=19.245536	 loss=3.828974	 lr=0.024840
Epoch[008] Batch [0749]/[3759]	Speed: 117.170999 samples/sec	 accuracy=19.254167	 loss=3.828458	 lr=0.024879
Epoch[008] Batch [0799]/[3759]	Speed: 117.920668 samples/sec	 accuracy=19.205078	 loss=3.831707	 lr=0.024918
Epoch[008] Batch [0849]/[3759]	Speed: 116.975824 samples/sec	 accuracy=19.240809	 loss=3.831764	 lr=0.024957
Epoch[008] Batch [0899]/[3759]	Speed: 117.311019 samples/sec	 accuracy=19.208333	 loss=3.830663	 lr=0.024995
Epoch[008] Batch [0949]/[3759]	Speed: 117.324809 samples/sec	 accuracy=19.171053	 loss=3.830518	 lr=0.025034
Epoch[008] Batch [0999]/[3759]	Speed: 117.106654 samples/sec	 accuracy=19.145312	 loss=3.831611	 lr=0.025073
Epoch[008] Batch [1049]/[3759]	Speed: 117.214037 samples/sec	 accuracy=19.151786	 loss=3.830457	 lr=0.025112
Epoch[008] Batch [1099]/[3759]	Speed: 116.977219 samples/sec	 accuracy=19.176136	 loss=3.827802	 lr=0.025150
Epoch[008] Batch [1149]/[3759]	Speed: 117.816839 samples/sec	 accuracy=19.183424	 loss=3.826621	 lr=0.025189
Epoch[008] Batch [1199]/[3759]	Speed: 117.373405 samples/sec	 accuracy=19.173177	 loss=3.826409	 lr=0.025228
Epoch[008] Batch [1249]/[3759]	Speed: 117.619785 samples/sec	 accuracy=19.153750	 loss=3.827668	 lr=0.025266
Epoch[008] Batch [1299]/[3759]	Speed: 116.779995 samples/sec	 accuracy=19.168269	 loss=3.826208	 lr=0.025305
Epoch[008] Batch [1349]/[3759]	Speed: 117.397634 samples/sec	 accuracy=19.165509	 loss=3.827631	 lr=0.025344
Epoch[008] Batch [1399]/[3759]	Speed: 116.703314 samples/sec	 accuracy=19.207589	 loss=3.826124	 lr=0.025383
Epoch[008] Batch [1449]/[3759]	Speed: 116.763331 samples/sec	 accuracy=19.200431	 loss=3.826711	 lr=0.025421
Epoch[008] Batch [1499]/[3759]	Speed: 116.488594 samples/sec	 accuracy=19.178125	 loss=3.826157	 lr=0.025460
Epoch[008] Batch [1549]/[3759]	Speed: 116.986457 samples/sec	 accuracy=19.161290	 loss=3.826390	 lr=0.025499
Epoch[008] Batch [1599]/[3759]	Speed: 116.999148 samples/sec	 accuracy=19.160156	 loss=3.825394	 lr=0.025538
Epoch[008] Batch [1649]/[3759]	Speed: 117.168403 samples/sec	 accuracy=19.172348	 loss=3.825959	 lr=0.025576
Epoch[008] Batch [1699]/[3759]	Speed: 117.584846 samples/sec	 accuracy=19.181066	 loss=3.827560	 lr=0.025615
Epoch[008] Batch [1749]/[3759]	Speed: 117.040000 samples/sec	 accuracy=19.152679	 loss=3.828142	 lr=0.025654
Epoch[008] Batch [1799]/[3759]	Speed: 116.612529 samples/sec	 accuracy=19.201389	 loss=3.826568	 lr=0.025692
Epoch[008] Batch [1849]/[3759]	Speed: 117.328603 samples/sec	 accuracy=19.195101	 loss=3.825925	 lr=0.025731
Epoch[008] Batch [1899]/[3759]	Speed: 116.576858 samples/sec	 accuracy=19.183388	 loss=3.825354	 lr=0.025770
Epoch[008] Batch [1949]/[3759]	Speed: 117.941463 samples/sec	 accuracy=19.224359	 loss=3.823860	 lr=0.025809
Epoch[008] Batch [1999]/[3759]	Speed: 116.940773 samples/sec	 accuracy=19.233594	 loss=3.823332	 lr=0.025847
Epoch[008] Batch [2049]/[3759]	Speed: 117.342859 samples/sec	 accuracy=19.256860	 loss=3.821413	 lr=0.025886
Epoch[008] Batch [2099]/[3759]	Speed: 117.867561 samples/sec	 accuracy=19.271577	 loss=3.820487	 lr=0.025925
Epoch[008] Batch [2149]/[3759]	Speed: 117.229891 samples/sec	 accuracy=19.273256	 loss=3.819845	 lr=0.025964
Epoch[008] Batch [2199]/[3759]	Speed: 117.960204 samples/sec	 accuracy=19.299006	 loss=3.818167	 lr=0.026002
Epoch[008] Batch [2249]/[3759]	Speed: 117.383358 samples/sec	 accuracy=19.303472	 loss=3.817625	 lr=0.026041
Epoch[008] Batch [2299]/[3759]	Speed: 116.818303 samples/sec	 accuracy=19.338315	 loss=3.817548	 lr=0.026080
Epoch[008] Batch [2349]/[3759]	Speed: 117.103343 samples/sec	 accuracy=19.344415	 loss=3.816363	 lr=0.026119
Epoch[008] Batch [2399]/[3759]	Speed: 117.655985 samples/sec	 accuracy=19.365885	 loss=3.815373	 lr=0.026157
Epoch[008] Batch [2449]/[3759]	Speed: 117.430592 samples/sec	 accuracy=19.353316	 loss=3.815412	 lr=0.026196
Epoch[008] Batch [2499]/[3759]	Speed: 117.415723 samples/sec	 accuracy=19.370000	 loss=3.814603	 lr=0.026235
Epoch[008] Batch [2549]/[3759]	Speed: 117.058339 samples/sec	 accuracy=19.388480	 loss=3.814960	 lr=0.026273
Epoch[008] Batch [2599]/[3759]	Speed: 117.805429 samples/sec	 accuracy=19.406250	 loss=3.814826	 lr=0.026312
Epoch[008] Batch [2649]/[3759]	Speed: 117.069111 samples/sec	 accuracy=19.410967	 loss=3.813171	 lr=0.026351
Epoch[008] Batch [2699]/[3759]	Speed: 117.293829 samples/sec	 accuracy=19.449653	 loss=3.811513	 lr=0.026390
Epoch[008] Batch [2749]/[3759]	Speed: 117.964214 samples/sec	 accuracy=19.488636	 loss=3.809976	 lr=0.026428
Epoch[008] Batch [2799]/[3759]	Speed: 117.083737 samples/sec	 accuracy=19.483817	 loss=3.810014	 lr=0.026467
Epoch[008] Batch [2849]/[3759]	Speed: 117.467568 samples/sec	 accuracy=19.514254	 loss=3.808287	 lr=0.026506
Epoch[008] Batch [2899]/[3759]	Speed: 117.555509 samples/sec	 accuracy=19.508082	 loss=3.808088	 lr=0.026545
Epoch[008] Batch [2949]/[3759]	Speed: 117.597926 samples/sec	 accuracy=19.514301	 loss=3.807311	 lr=0.026583
Epoch[008] Batch [2999]/[3759]	Speed: 117.025193 samples/sec	 accuracy=19.516667	 loss=3.806962	 lr=0.026622
Epoch[008] Batch [3049]/[3759]	Speed: 117.686592 samples/sec	 accuracy=19.542520	 loss=3.805561	 lr=0.026661
Epoch[008] Batch [3099]/[3759]	Speed: 117.349126 samples/sec	 accuracy=19.552923	 loss=3.805455	 lr=0.026699
Epoch[008] Batch [3149]/[3759]	Speed: 116.509808 samples/sec	 accuracy=19.563492	 loss=3.804653	 lr=0.026738
Epoch[008] Batch [3199]/[3759]	Speed: 116.776972 samples/sec	 accuracy=19.589844	 loss=3.803449	 lr=0.026777
Epoch[008] Batch [3249]/[3759]	Speed: 117.373086 samples/sec	 accuracy=19.595192	 loss=3.803212	 lr=0.026816
Epoch[008] Batch [3299]/[3759]	Speed: 117.531961 samples/sec	 accuracy=19.599905	 loss=3.801983	 lr=0.026854
Epoch[008] Batch [3349]/[3759]	Speed: 117.576849 samples/sec	 accuracy=19.616604	 loss=3.800578	 lr=0.026893
Epoch[008] Batch [3399]/[3759]	Speed: 116.737825 samples/sec	 accuracy=19.630515	 loss=3.800451	 lr=0.026932
Epoch[008] Batch [3449]/[3759]	Speed: 117.211721 samples/sec	 accuracy=19.665761	 loss=3.799604	 lr=0.026971
Epoch[008] Batch [3499]/[3759]	Speed: 116.959025 samples/sec	 accuracy=19.686161	 loss=3.798542	 lr=0.027009
Epoch[008] Batch [3549]/[3759]	Speed: 117.365820 samples/sec	 accuracy=19.702905	 loss=3.797325	 lr=0.027048
Epoch[008] Batch [3599]/[3759]	Speed: 116.795292 samples/sec	 accuracy=19.683594	 loss=3.797338	 lr=0.027087
Epoch[008] Batch [3649]/[3759]	Speed: 117.085782 samples/sec	 accuracy=19.692637	 loss=3.796456	 lr=0.027126
Epoch[008] Batch [3699]/[3759]	Speed: 117.053334 samples/sec	 accuracy=19.703970	 loss=3.795371	 lr=0.027164
Epoch[008] Batch [3749]/[3759]	Speed: 120.508538 samples/sec	 accuracy=19.707083	 loss=3.794427	 lr=0.027203
Batch [0049]/[0303]: acc-top1=20.906250 acc-top5=46.687500
Batch [0099]/[0303]: acc-top1=19.390625 acc-top5=44.578125
Batch [0149]/[0303]: acc-top1=21.072917 acc-top5=46.166667
Batch [0199]/[0303]: acc-top1=23.132812 acc-top5=48.148438
Batch [0249]/[0303]: acc-top1=22.637500 acc-top5=47.881250
Batch [0299]/[0303]: acc-top1=22.421875 acc-top5=48.213542
[Epoch 008] training: accuracy=19.707702	 loss=3.794364
[Epoch 008] speed: 116 samples/sec	time cost: 2182.576039
[Epoch 008] validation: acc-top1=22.452558 acc-top5=48.189975 loss=3.565409
Epoch[009] Batch [0049]/[3760]	Speed: 87.734904 samples/sec	 accuracy=21.718750	 loss=3.716986	 lr=0.027249
Epoch[009] Batch [0099]/[3760]	Speed: 117.225913 samples/sec	 accuracy=21.781250	 loss=3.706853	 lr=0.027287
Epoch[009] Batch [0149]/[3760]	Speed: 117.925901 samples/sec	 accuracy=21.802083	 loss=3.692149	 lr=0.027326
Epoch[009] Batch [0199]/[3760]	Speed: 116.468701 samples/sec	 accuracy=21.500000	 loss=3.692096	 lr=0.027365
Epoch[009] Batch [0249]/[3760]	Speed: 117.188551 samples/sec	 accuracy=21.337500	 loss=3.701187	 lr=0.027404
Epoch[009] Batch [0299]/[3760]	Speed: 117.152910 samples/sec	 accuracy=21.187500	 loss=3.702341	 lr=0.027442
Epoch[009] Batch [0349]/[3760]	Speed: 117.398470 samples/sec	 accuracy=21.040179	 loss=3.706257	 lr=0.027481
Epoch[009] Batch [0399]/[3760]	Speed: 117.111719 samples/sec	 accuracy=21.070312	 loss=3.702765	 lr=0.027520
Epoch[009] Batch [0449]/[3760]	Speed: 117.422150 samples/sec	 accuracy=21.041667	 loss=3.703444	 lr=0.027559
Epoch[009] Batch [0499]/[3760]	Speed: 117.084963 samples/sec	 accuracy=20.984375	 loss=3.705270	 lr=0.027597
Epoch[009] Batch [0549]/[3760]	Speed: 116.985110 samples/sec	 accuracy=20.988636	 loss=3.705936	 lr=0.027636
Epoch[009] Batch [0599]/[3760]	Speed: 117.276997 samples/sec	 accuracy=21.075521	 loss=3.703212	 lr=0.027675
Epoch[009] Batch [0649]/[3760]	Speed: 116.599403 samples/sec	 accuracy=21.086538	 loss=3.704287	 lr=0.027713
Epoch[009] Batch [0699]/[3760]	Speed: 118.301200 samples/sec	 accuracy=21.040179	 loss=3.708320	 lr=0.027752
Epoch[009] Batch [0749]/[3760]	Speed: 118.013348 samples/sec	 accuracy=20.995833	 loss=3.707648	 lr=0.027791
Epoch[009] Batch [0799]/[3760]	Speed: 117.378595 samples/sec	 accuracy=21.023438	 loss=3.707082	 lr=0.027830
Epoch[009] Batch [0849]/[3760]	Speed: 117.373437 samples/sec	 accuracy=21.001838	 loss=3.708597	 lr=0.027868
Epoch[009] Batch [0899]/[3760]	Speed: 116.896042 samples/sec	 accuracy=21.015625	 loss=3.706075	 lr=0.027907
Epoch[009] Batch [0949]/[3760]	Speed: 117.353928 samples/sec	 accuracy=21.054276	 loss=3.703086	 lr=0.027946
Epoch[009] Batch [0999]/[3760]	Speed: 117.179822 samples/sec	 accuracy=21.048437	 loss=3.701220	 lr=0.027985
Epoch[009] Batch [1049]/[3760]	Speed: 117.680577 samples/sec	 accuracy=21.059524	 loss=3.699146	 lr=0.028023
Epoch[009] Batch [1099]/[3760]	Speed: 116.727572 samples/sec	 accuracy=21.090909	 loss=3.699992	 lr=0.028062
Epoch[009] Batch [1149]/[3760]	Speed: 117.994927 samples/sec	 accuracy=21.108696	 loss=3.697979	 lr=0.028101
Epoch[009] Batch [1199]/[3760]	Speed: 117.453935 samples/sec	 accuracy=21.082031	 loss=3.698555	 lr=0.028140
Epoch[009] Batch [1249]/[3760]	Speed: 116.724086 samples/sec	 accuracy=21.050000	 loss=3.700919	 lr=0.028178
Epoch[009] Batch [1299]/[3760]	Speed: 117.254448 samples/sec	 accuracy=21.039663	 loss=3.701963	 lr=0.028217
Epoch[009] Batch [1349]/[3760]	Speed: 117.165927 samples/sec	 accuracy=21.063657	 loss=3.701457	 lr=0.028256
Epoch[009] Batch [1399]/[3760]	Speed: 117.820339 samples/sec	 accuracy=21.104911	 loss=3.700822	 lr=0.028294
Epoch[009] Batch [1449]/[3760]	Speed: 116.838600 samples/sec	 accuracy=21.200431	 loss=3.699147	 lr=0.028333
Epoch[009] Batch [1499]/[3760]	Speed: 117.213825 samples/sec	 accuracy=21.188542	 loss=3.700246	 lr=0.028372
Epoch[009] Batch [1549]/[3760]	Speed: 117.234929 samples/sec	 accuracy=21.195565	 loss=3.699921	 lr=0.028411
Epoch[009] Batch [1599]/[3760]	Speed: 116.873403 samples/sec	 accuracy=21.154297	 loss=3.700818	 lr=0.028449
Epoch[009] Batch [1649]/[3760]	Speed: 117.185944 samples/sec	 accuracy=21.163826	 loss=3.700416	 lr=0.028488
Epoch[009] Batch [1699]/[3760]	Speed: 116.806705 samples/sec	 accuracy=21.151654	 loss=3.699573	 lr=0.028527
Epoch[009] Batch [1749]/[3760]	Speed: 117.494921 samples/sec	 accuracy=21.138393	 loss=3.699272	 lr=0.028566
Epoch[009] Batch [1799]/[3760]	Speed: 117.469730 samples/sec	 accuracy=21.156250	 loss=3.697445	 lr=0.028604
Epoch[009] Batch [1849]/[3760]	Speed: 117.599852 samples/sec	 accuracy=21.208615	 loss=3.696546	 lr=0.028643
Epoch[009] Batch [1899]/[3760]	Speed: 117.529414 samples/sec	 accuracy=21.235197	 loss=3.696011	 lr=0.028682
Epoch[009] Batch [1949]/[3760]	Speed: 117.007834 samples/sec	 accuracy=21.276442	 loss=3.694811	 lr=0.028720
Epoch[009] Batch [1999]/[3760]	Speed: 117.391497 samples/sec	 accuracy=21.271094	 loss=3.694289	 lr=0.028759
Epoch[009] Batch [2049]/[3760]	Speed: 117.767222 samples/sec	 accuracy=21.286585	 loss=3.693812	 lr=0.028798
Epoch[009] Batch [2099]/[3760]	Speed: 117.679925 samples/sec	 accuracy=21.299851	 loss=3.693541	 lr=0.028837
Epoch[009] Batch [2149]/[3760]	Speed: 117.306826 samples/sec	 accuracy=21.310320	 loss=3.693527	 lr=0.028875
Epoch[009] Batch [2199]/[3760]	Speed: 117.417434 samples/sec	 accuracy=21.345170	 loss=3.692500	 lr=0.028914
Epoch[009] Batch [2249]/[3760]	Speed: 116.972002 samples/sec	 accuracy=21.360417	 loss=3.692447	 lr=0.028953
Epoch[009] Batch [2299]/[3760]	Speed: 117.015860 samples/sec	 accuracy=21.366168	 loss=3.693055	 lr=0.028992
Epoch[009] Batch [2349]/[3760]	Speed: 117.414961 samples/sec	 accuracy=21.386968	 loss=3.691928	 lr=0.029030
Epoch[009] Batch [2399]/[3760]	Speed: 117.805271 samples/sec	 accuracy=21.376953	 loss=3.693017	 lr=0.029069
Epoch[009] Batch [2449]/[3760]	Speed: 116.939714 samples/sec	 accuracy=21.357143	 loss=3.693489	 lr=0.029108
Epoch[009] Batch [2499]/[3760]	Speed: 117.957226 samples/sec	 accuracy=21.383125	 loss=3.692451	 lr=0.029147
Epoch[009] Batch [2549]/[3760]	Speed: 117.983366 samples/sec	 accuracy=21.399510	 loss=3.692160	 lr=0.029185
Epoch[009] Batch [2599]/[3760]	Speed: 116.775506 samples/sec	 accuracy=21.378005	 loss=3.692661	 lr=0.029224
Epoch[009] Batch [2649]/[3760]	Speed: 118.239406 samples/sec	 accuracy=21.396816	 loss=3.691471	 lr=0.029263
Epoch[009] Batch [2699]/[3760]	Speed: 117.236377 samples/sec	 accuracy=21.386574	 loss=3.692478	 lr=0.029301
Epoch[009] Batch [2749]/[3760]	Speed: 117.579955 samples/sec	 accuracy=21.379545	 loss=3.692224	 lr=0.029340
Epoch[009] Batch [2799]/[3760]	Speed: 117.011724 samples/sec	 accuracy=21.390625	 loss=3.692002	 lr=0.029379
Epoch[009] Batch [2849]/[3760]	Speed: 117.323024 samples/sec	 accuracy=21.385965	 loss=3.691152	 lr=0.029418
Epoch[009] Batch [2899]/[3760]	Speed: 117.132033 samples/sec	 accuracy=21.391703	 loss=3.690800	 lr=0.029456
Epoch[009] Batch [2949]/[3760]	Speed: 117.904104 samples/sec	 accuracy=21.398305	 loss=3.690447	 lr=0.029495
Epoch[009] Batch [2999]/[3760]	Speed: 117.219462 samples/sec	 accuracy=21.420833	 loss=3.689333	 lr=0.029534
Epoch[009] Batch [3049]/[3760]	Speed: 116.929733 samples/sec	 accuracy=21.429816	 loss=3.689453	 lr=0.029573
Epoch[009] Batch [3099]/[3760]	Speed: 117.616268 samples/sec	 accuracy=21.422883	 loss=3.689396	 lr=0.029611
Epoch[009] Batch [3149]/[3760]	Speed: 117.751592 samples/sec	 accuracy=21.426587	 loss=3.689095	 lr=0.029650
Epoch[009] Batch [3199]/[3760]	Speed: 117.873626 samples/sec	 accuracy=21.419922	 loss=3.689940	 lr=0.029689
Epoch[009] Batch [3249]/[3760]	Speed: 118.135072 samples/sec	 accuracy=21.430288	 loss=3.689200	 lr=0.029727
Epoch[009] Batch [3299]/[3760]	Speed: 117.273746 samples/sec	 accuracy=21.452178	 loss=3.688029	 lr=0.029766
Epoch[009] Batch [3349]/[3760]	Speed: 116.807973 samples/sec	 accuracy=21.458955	 loss=3.687305	 lr=0.029805
Epoch[009] Batch [3399]/[3760]	Speed: 117.341040 samples/sec	 accuracy=21.454963	 loss=3.687328	 lr=0.029844
Epoch[009] Batch [3449]/[3760]	Speed: 117.289925 samples/sec	 accuracy=21.456522	 loss=3.687014	 lr=0.029882
Epoch[009] Batch [3499]/[3760]	Speed: 116.945211 samples/sec	 accuracy=21.467857	 loss=3.687102	 lr=0.029921
Epoch[009] Batch [3549]/[3760]	Speed: 117.711219 samples/sec	 accuracy=21.493838	 loss=3.686895	 lr=0.029960
Epoch[009] Batch [3599]/[3760]	Speed: 116.972204 samples/sec	 accuracy=21.510417	 loss=3.686416	 lr=0.029999
Epoch[009] Batch [3649]/[3760]	Speed: 117.479114 samples/sec	 accuracy=21.500428	 loss=3.686395	 lr=0.030037
Epoch[009] Batch [3699]/[3760]	Speed: 117.272752 samples/sec	 accuracy=21.503378	 loss=3.685832	 lr=0.030076
Epoch[009] Batch [3749]/[3760]	Speed: 120.836756 samples/sec	 accuracy=21.508750	 loss=3.685026	 lr=0.030115
Batch [0049]/[0303]: acc-top1=20.718750 acc-top5=46.187500
Batch [0099]/[0303]: acc-top1=20.953125 acc-top5=45.734375
Batch [0149]/[0303]: acc-top1=22.312500 acc-top5=46.968750
Batch [0199]/[0303]: acc-top1=23.953125 acc-top5=48.875000
Batch [0249]/[0303]: acc-top1=23.037500 acc-top5=48.556250
Batch [0299]/[0303]: acc-top1=22.484375 acc-top5=48.343750
[Epoch 009] training: accuracy=21.513880	 loss=3.684906
[Epoch 009] speed: 116 samples/sec	time cost: 2180.014832
[Epoch 009] validation: acc-top1=22.493812 acc-top5=48.267327 loss=3.658664
Epoch[010] Batch [0049]/[3760]	Speed: 86.655798 samples/sec	 accuracy=22.625000	 loss=3.558947	 lr=0.030161
Epoch[010] Batch [0099]/[3760]	Speed: 117.133636 samples/sec	 accuracy=22.640625	 loss=3.584831	 lr=0.030200
Epoch[010] Batch [0149]/[3760]	Speed: 117.484134 samples/sec	 accuracy=22.322917	 loss=3.596851	 lr=0.030239
Epoch[010] Batch [0199]/[3760]	Speed: 117.971418 samples/sec	 accuracy=22.375000	 loss=3.598238	 lr=0.030277
Epoch[010] Batch [0249]/[3760]	Speed: 116.993625 samples/sec	 accuracy=22.375000	 loss=3.605619	 lr=0.030316
Epoch[010] Batch [0299]/[3760]	Speed: 116.936309 samples/sec	 accuracy=22.432292	 loss=3.598376	 lr=0.030355
Epoch[010] Batch [0349]/[3760]	Speed: 116.861097 samples/sec	 accuracy=22.370536	 loss=3.596995	 lr=0.030394
Epoch[010] Batch [0399]/[3760]	Speed: 117.565793 samples/sec	 accuracy=22.351562	 loss=3.600617	 lr=0.030432
Epoch[010] Batch [0449]/[3760]	Speed: 116.847544 samples/sec	 accuracy=22.340278	 loss=3.604301	 lr=0.030471
Epoch[010] Batch [0499]/[3760]	Speed: 117.430317 samples/sec	 accuracy=22.415625	 loss=3.602235	 lr=0.030510
Epoch[010] Batch [0549]/[3760]	Speed: 116.781051 samples/sec	 accuracy=22.306818	 loss=3.608222	 lr=0.030549
Epoch[010] Batch [0599]/[3760]	Speed: 117.680266 samples/sec	 accuracy=22.411458	 loss=3.606490	 lr=0.030587
Epoch[010] Batch [0649]/[3760]	Speed: 117.985743 samples/sec	 accuracy=22.362981	 loss=3.607920	 lr=0.030626
Epoch[010] Batch [0699]/[3760]	Speed: 116.654464 samples/sec	 accuracy=22.330357	 loss=3.606663	 lr=0.030665
Epoch[010] Batch [0749]/[3760]	Speed: 117.336383 samples/sec	 accuracy=22.347917	 loss=3.609246	 lr=0.030703
Epoch[010] Batch [0799]/[3760]	Speed: 117.198973 samples/sec	 accuracy=22.375000	 loss=3.611399	 lr=0.030742
Epoch[010] Batch [0849]/[3760]	Speed: 117.857324 samples/sec	 accuracy=22.362132	 loss=3.611738	 lr=0.030781
Epoch[010] Batch [0899]/[3760]	Speed: 116.679558 samples/sec	 accuracy=22.331597	 loss=3.612902	 lr=0.030820
Epoch[010] Batch [0949]/[3760]	Speed: 116.744996 samples/sec	 accuracy=22.350329	 loss=3.614855	 lr=0.030858
Epoch[010] Batch [0999]/[3760]	Speed: 117.572855 samples/sec	 accuracy=22.398438	 loss=3.611766	 lr=0.030897
Epoch[010] Batch [1049]/[3760]	Speed: 117.144039 samples/sec	 accuracy=22.397321	 loss=3.614595	 lr=0.030936
Epoch[010] Batch [1099]/[3760]	Speed: 117.409211 samples/sec	 accuracy=22.451705	 loss=3.612230	 lr=0.030975
Epoch[010] Batch [1149]/[3760]	Speed: 117.613300 samples/sec	 accuracy=22.483696	 loss=3.608490	 lr=0.031013
Epoch[010] Batch [1199]/[3760]	Speed: 116.872969 samples/sec	 accuracy=22.494792	 loss=3.608650	 lr=0.031052
Epoch[010] Batch [1249]/[3760]	Speed: 117.288918 samples/sec	 accuracy=22.502500	 loss=3.608760	 lr=0.031091
Epoch[010] Batch [1299]/[3760]	Speed: 116.937813 samples/sec	 accuracy=22.543269	 loss=3.608065	 lr=0.031130
Epoch[010] Batch [1349]/[3760]	Speed: 117.081748 samples/sec	 accuracy=22.531250	 loss=3.608742	 lr=0.031168
Epoch[010] Batch [1399]/[3760]	Speed: 117.482812 samples/sec	 accuracy=22.508929	 loss=3.609653	 lr=0.031207
Epoch[010] Batch [1449]/[3760]	Speed: 117.148601 samples/sec	 accuracy=22.481681	 loss=3.611190	 lr=0.031246
Epoch[010] Batch [1499]/[3760]	Speed: 118.293854 samples/sec	 accuracy=22.490625	 loss=3.609651	 lr=0.031284
Epoch[010] Batch [1549]/[3760]	Speed: 117.015977 samples/sec	 accuracy=22.505040	 loss=3.609506	 lr=0.031323
Epoch[010] Batch [1599]/[3760]	Speed: 116.696552 samples/sec	 accuracy=22.511719	 loss=3.609160	 lr=0.031362
Epoch[010] Batch [1649]/[3760]	Speed: 116.922416 samples/sec	 accuracy=22.490530	 loss=3.608268	 lr=0.031401
Epoch[010] Batch [1699]/[3760]	Speed: 117.661140 samples/sec	 accuracy=22.476103	 loss=3.608102	 lr=0.031439
Epoch[010] Batch [1749]/[3760]	Speed: 117.322045 samples/sec	 accuracy=22.517857	 loss=3.606226	 lr=0.031478
Epoch[010] Batch [1799]/[3760]	Speed: 116.810044 samples/sec	 accuracy=22.536458	 loss=3.604273	 lr=0.031517
Epoch[010] Batch [1849]/[3760]	Speed: 117.510846 samples/sec	 accuracy=22.584459	 loss=3.603746	 lr=0.031556
Epoch[010] Batch [1899]/[3760]	Speed: 118.062296 samples/sec	 accuracy=22.595395	 loss=3.603567	 lr=0.031594
Epoch[010] Batch [1949]/[3760]	Speed: 117.635983 samples/sec	 accuracy=22.626603	 loss=3.603098	 lr=0.031633
Epoch[010] Batch [1999]/[3760]	Speed: 117.419025 samples/sec	 accuracy=22.622656	 loss=3.603713	 lr=0.031672
Epoch[010] Batch [2049]/[3760]	Speed: 117.827322 samples/sec	 accuracy=22.671494	 loss=3.603163	 lr=0.031710
Epoch[010] Batch [2099]/[3760]	Speed: 117.800607 samples/sec	 accuracy=22.683780	 loss=3.602946	 lr=0.031749
Epoch[010] Batch [2149]/[3760]	Speed: 117.990505 samples/sec	 accuracy=22.718750	 loss=3.602122	 lr=0.031788
Epoch[010] Batch [2199]/[3760]	Speed: 118.743060 samples/sec	 accuracy=22.708097	 loss=3.602183	 lr=0.031827
Epoch[010] Batch [2249]/[3760]	Speed: 117.083158 samples/sec	 accuracy=22.728472	 loss=3.601256	 lr=0.031865
Epoch[010] Batch [2299]/[3760]	Speed: 117.916884 samples/sec	 accuracy=22.745924	 loss=3.601344	 lr=0.031904
Epoch[010] Batch [2349]/[3760]	Speed: 118.008530 samples/sec	 accuracy=22.767287	 loss=3.599985	 lr=0.031943
Epoch[010] Batch [2399]/[3760]	Speed: 117.639806 samples/sec	 accuracy=22.777344	 loss=3.600529	 lr=0.031982
Epoch[010] Batch [2449]/[3760]	Speed: 118.279420 samples/sec	 accuracy=22.816327	 loss=3.597949	 lr=0.032020
Epoch[010] Batch [2499]/[3760]	Speed: 117.477448 samples/sec	 accuracy=22.800000	 loss=3.598318	 lr=0.032059
Epoch[010] Batch [2549]/[3760]	Speed: 117.154516 samples/sec	 accuracy=22.814338	 loss=3.597586	 lr=0.032098
Epoch[010] Batch [2599]/[3760]	Speed: 117.608082 samples/sec	 accuracy=22.833534	 loss=3.595714	 lr=0.032137
Epoch[010] Batch [2649]/[3760]	Speed: 117.970964 samples/sec	 accuracy=22.874410	 loss=3.594713	 lr=0.032175
Epoch[010] Batch [2699]/[3760]	Speed: 118.025238 samples/sec	 accuracy=22.861111	 loss=3.594902	 lr=0.032214
Epoch[010] Batch [2749]/[3760]	Speed: 117.480843 samples/sec	 accuracy=22.880682	 loss=3.594507	 lr=0.032253
Epoch[010] Batch [2799]/[3760]	Speed: 117.868209 samples/sec	 accuracy=22.892857	 loss=3.593692	 lr=0.032291
Epoch[010] Batch [2849]/[3760]	Speed: 117.497477 samples/sec	 accuracy=22.888158	 loss=3.594177	 lr=0.032330
Epoch[010] Batch [2899]/[3760]	Speed: 117.217683 samples/sec	 accuracy=22.903556	 loss=3.593811	 lr=0.032369
Epoch[010] Batch [2949]/[3760]	Speed: 116.790506 samples/sec	 accuracy=22.922140	 loss=3.593762	 lr=0.032408
Epoch[010] Batch [2999]/[3760]	Speed: 117.288293 samples/sec	 accuracy=22.918229	 loss=3.593598	 lr=0.032446
Epoch[010] Batch [3049]/[3760]	Speed: 117.344218 samples/sec	 accuracy=22.952869	 loss=3.592764	 lr=0.032485
Epoch[010] Batch [3099]/[3760]	Speed: 117.255345 samples/sec	 accuracy=22.955141	 loss=3.593071	 lr=0.032524
Epoch[010] Batch [3149]/[3760]	Speed: 117.704240 samples/sec	 accuracy=22.955357	 loss=3.592900	 lr=0.032563
Epoch[010] Batch [3199]/[3760]	Speed: 117.411518 samples/sec	 accuracy=22.973145	 loss=3.591451	 lr=0.032601
Epoch[010] Batch [3249]/[3760]	Speed: 117.730145 samples/sec	 accuracy=22.995673	 loss=3.590929	 lr=0.032640
Epoch[010] Batch [3299]/[3760]	Speed: 117.425933 samples/sec	 accuracy=22.996686	 loss=3.591417	 lr=0.032679
Epoch[010] Batch [3349]/[3760]	Speed: 118.055114 samples/sec	 accuracy=23.012593	 loss=3.590463	 lr=0.032717
Epoch[010] Batch [3399]/[3760]	Speed: 116.953990 samples/sec	 accuracy=23.028493	 loss=3.590341	 lr=0.032756
Epoch[010] Batch [3449]/[3760]	Speed: 117.911588 samples/sec	 accuracy=23.026268	 loss=3.589931	 lr=0.032795
Epoch[010] Batch [3499]/[3760]	Speed: 118.139017 samples/sec	 accuracy=23.012500	 loss=3.590345	 lr=0.032834
Epoch[010] Batch [3549]/[3760]	Speed: 117.737549 samples/sec	 accuracy=23.031690	 loss=3.589387	 lr=0.032872
Epoch[010] Batch [3599]/[3760]	Speed: 117.388328 samples/sec	 accuracy=23.040799	 loss=3.589392	 lr=0.032911
Epoch[010] Batch [3649]/[3760]	Speed: 117.645449 samples/sec	 accuracy=23.045377	 loss=3.589280	 lr=0.032950
Epoch[010] Batch [3699]/[3760]	Speed: 117.375849 samples/sec	 accuracy=23.029139	 loss=3.590321	 lr=0.032989
Epoch[010] Batch [3749]/[3760]	Speed: 120.278056 samples/sec	 accuracy=23.031667	 loss=3.590515	 lr=0.033027
Batch [0049]/[0303]: acc-top1=26.125000 acc-top5=51.156250
Batch [0099]/[0303]: acc-top1=24.093750 acc-top5=49.062500
Batch [0149]/[0303]: acc-top1=24.531250 acc-top5=50.781250
Batch [0199]/[0303]: acc-top1=25.960937 acc-top5=52.359375
Batch [0249]/[0303]: acc-top1=25.525000 acc-top5=52.293750
Batch [0299]/[0303]: acc-top1=25.505208 acc-top5=52.364583
[Epoch 010] training: accuracy=23.041888	 loss=3.590157
[Epoch 010] speed: 116 samples/sec	time cost: 2178.153335
[Epoch 010] validation: acc-top1=25.489893 acc-top5=52.315388 loss=3.390313
Epoch[011] Batch [0049]/[3759]	Speed: 87.460340 samples/sec	 accuracy=23.375000	 loss=3.596291	 lr=0.033074
Epoch[011] Batch [0099]/[3759]	Speed: 117.215095 samples/sec	 accuracy=24.390625	 loss=3.543345	 lr=0.033113
Epoch[011] Batch [0149]/[3759]	Speed: 118.280210 samples/sec	 accuracy=24.322917	 loss=3.533233	 lr=0.033151
Epoch[011] Batch [0199]/[3759]	Speed: 115.950086 samples/sec	 accuracy=23.984375	 loss=3.525345	 lr=0.033190
Epoch[011] Batch [0249]/[3759]	Speed: 117.964903 samples/sec	 accuracy=24.106250	 loss=3.509891	 lr=0.033229
Epoch[011] Batch [0299]/[3759]	Speed: 118.017367 samples/sec	 accuracy=24.161458	 loss=3.513058	 lr=0.033267
Epoch[011] Batch [0349]/[3759]	Speed: 118.155983 samples/sec	 accuracy=24.049107	 loss=3.518450	 lr=0.033306
Epoch[011] Batch [0399]/[3759]	Speed: 117.642730 samples/sec	 accuracy=24.078125	 loss=3.521004	 lr=0.033345
Epoch[011] Batch [0449]/[3759]	Speed: 117.347020 samples/sec	 accuracy=24.038194	 loss=3.523347	 lr=0.033384
Epoch[011] Batch [0499]/[3759]	Speed: 116.941751 samples/sec	 accuracy=24.062500	 loss=3.516692	 lr=0.033422
Epoch[011] Batch [0549]/[3759]	Speed: 116.491921 samples/sec	 accuracy=24.125000	 loss=3.516937	 lr=0.033461
Epoch[011] Batch [0599]/[3759]	Speed: 118.264694 samples/sec	 accuracy=24.111979	 loss=3.516656	 lr=0.033500
Epoch[011] Batch [0649]/[3759]	Speed: 117.093875 samples/sec	 accuracy=24.163462	 loss=3.513368	 lr=0.033539
Epoch[011] Batch [0699]/[3759]	Speed: 117.076005 samples/sec	 accuracy=24.287946	 loss=3.510611	 lr=0.033577
Epoch[011] Batch [0749]/[3759]	Speed: 117.028392 samples/sec	 accuracy=24.306250	 loss=3.514953	 lr=0.033616
Epoch[011] Batch [0799]/[3759]	Speed: 117.503218 samples/sec	 accuracy=24.273438	 loss=3.515096	 lr=0.033655
Epoch[011] Batch [0849]/[3759]	Speed: 117.694131 samples/sec	 accuracy=24.319853	 loss=3.514668	 lr=0.033694
Epoch[011] Batch [0899]/[3759]	Speed: 117.104449 samples/sec	 accuracy=24.347222	 loss=3.515366	 lr=0.033732
Epoch[011] Batch [0949]/[3759]	Speed: 117.615066 samples/sec	 accuracy=24.386513	 loss=3.513663	 lr=0.033771
Epoch[011] Batch [0999]/[3759]	Speed: 117.467697 samples/sec	 accuracy=24.309375	 loss=3.514771	 lr=0.033810
Epoch[011] Batch [1049]/[3759]	Speed: 117.873939 samples/sec	 accuracy=24.217262	 loss=3.519787	 lr=0.033848
Epoch[011] Batch [1099]/[3759]	Speed: 116.616156 samples/sec	 accuracy=24.235795	 loss=3.518248	 lr=0.033887
Epoch[011] Batch [1149]/[3759]	Speed: 117.850935 samples/sec	 accuracy=24.270380	 loss=3.516636	 lr=0.033926
Epoch[011] Batch [1199]/[3759]	Speed: 117.844265 samples/sec	 accuracy=24.270833	 loss=3.518451	 lr=0.033965
Epoch[011] Batch [1249]/[3759]	Speed: 117.463829 samples/sec	 accuracy=24.266250	 loss=3.518372	 lr=0.034003
Epoch[011] Batch [1299]/[3759]	Speed: 117.633331 samples/sec	 accuracy=24.288462	 loss=3.517294	 lr=0.034042
Epoch[011] Batch [1349]/[3759]	Speed: 117.638930 samples/sec	 accuracy=24.261574	 loss=3.519925	 lr=0.034081
Epoch[011] Batch [1399]/[3759]	Speed: 117.381466 samples/sec	 accuracy=24.255580	 loss=3.519917	 lr=0.034120
Epoch[011] Batch [1449]/[3759]	Speed: 118.082105 samples/sec	 accuracy=24.246767	 loss=3.521093	 lr=0.034158
Epoch[011] Batch [1499]/[3759]	Speed: 117.604655 samples/sec	 accuracy=24.221875	 loss=3.520926	 lr=0.034197
Epoch[011] Batch [1549]/[3759]	Speed: 117.371246 samples/sec	 accuracy=24.212702	 loss=3.521137	 lr=0.034236
Epoch[011] Batch [1599]/[3759]	Speed: 118.124275 samples/sec	 accuracy=24.212891	 loss=3.520279	 lr=0.034274
Epoch[011] Batch [1649]/[3759]	Speed: 117.656153 samples/sec	 accuracy=24.222538	 loss=3.520030	 lr=0.034313
Epoch[011] Batch [1699]/[3759]	Speed: 118.171420 samples/sec	 accuracy=24.198529	 loss=3.521051	 lr=0.034352
Epoch[011] Batch [1749]/[3759]	Speed: 117.753482 samples/sec	 accuracy=24.192857	 loss=3.521022	 lr=0.034391
Epoch[011] Batch [1799]/[3759]	Speed: 116.639066 samples/sec	 accuracy=24.171007	 loss=3.520872	 lr=0.034429
Epoch[011] Batch [1849]/[3759]	Speed: 117.311207 samples/sec	 accuracy=24.165541	 loss=3.520511	 lr=0.034468
Epoch[011] Batch [1899]/[3759]	Speed: 117.499303 samples/sec	 accuracy=24.162829	 loss=3.520058	 lr=0.034507
Epoch[011] Batch [1949]/[3759]	Speed: 117.693098 samples/sec	 accuracy=24.155449	 loss=3.520431	 lr=0.034546
Epoch[011] Batch [1999]/[3759]	Speed: 117.031688 samples/sec	 accuracy=24.153125	 loss=3.521104	 lr=0.034584
Epoch[011] Batch [2049]/[3759]	Speed: 118.010640 samples/sec	 accuracy=24.126524	 loss=3.523919	 lr=0.034623
Epoch[011] Batch [2099]/[3759]	Speed: 116.594208 samples/sec	 accuracy=24.106399	 loss=3.525388	 lr=0.034662
Epoch[011] Batch [2149]/[3759]	Speed: 117.182349 samples/sec	 accuracy=24.082849	 loss=3.526152	 lr=0.034701
Epoch[011] Batch [2199]/[3759]	Speed: 117.483043 samples/sec	 accuracy=24.084517	 loss=3.525857	 lr=0.034739
Epoch[011] Batch [2249]/[3759]	Speed: 117.809669 samples/sec	 accuracy=24.082639	 loss=3.525818	 lr=0.034778
Epoch[011] Batch [2299]/[3759]	Speed: 118.097988 samples/sec	 accuracy=24.088995	 loss=3.525883	 lr=0.034817
Epoch[011] Batch [2349]/[3759]	Speed: 117.612329 samples/sec	 accuracy=24.086436	 loss=3.525769	 lr=0.034855
Epoch[011] Batch [2399]/[3759]	Speed: 117.035224 samples/sec	 accuracy=24.079427	 loss=3.525558	 lr=0.034894
Epoch[011] Batch [2449]/[3759]	Speed: 116.742901 samples/sec	 accuracy=24.068240	 loss=3.524995	 lr=0.034933
Epoch[011] Batch [2499]/[3759]	Speed: 117.555815 samples/sec	 accuracy=24.081875	 loss=3.524369	 lr=0.034972
Epoch[011] Batch [2549]/[3759]	Speed: 117.063322 samples/sec	 accuracy=24.079044	 loss=3.524789	 lr=0.035010
Epoch[011] Batch [2599]/[3759]	Speed: 117.886124 samples/sec	 accuracy=24.064303	 loss=3.524184	 lr=0.035049
Epoch[011] Batch [2649]/[3759]	Speed: 117.629885 samples/sec	 accuracy=24.069575	 loss=3.523528	 lr=0.035088
Epoch[011] Batch [2699]/[3759]	Speed: 117.607910 samples/sec	 accuracy=24.068287	 loss=3.523764	 lr=0.035127
Epoch[011] Batch [2749]/[3759]	Speed: 117.745923 samples/sec	 accuracy=24.063636	 loss=3.523134	 lr=0.035165
Epoch[011] Batch [2799]/[3759]	Speed: 117.073787 samples/sec	 accuracy=24.057478	 loss=3.522631	 lr=0.035204
Epoch[011] Batch [2849]/[3759]	Speed: 117.213300 samples/sec	 accuracy=24.091557	 loss=3.521945	 lr=0.035243
Epoch[011] Batch [2899]/[3759]	Speed: 116.975282 samples/sec	 accuracy=24.112069	 loss=3.521377	 lr=0.035281
Epoch[011] Batch [2949]/[3759]	Speed: 117.080307 samples/sec	 accuracy=24.132415	 loss=3.520159	 lr=0.035320
Epoch[011] Batch [2999]/[3759]	Speed: 117.033244 samples/sec	 accuracy=24.126042	 loss=3.520247	 lr=0.035359
Epoch[011] Batch [3049]/[3759]	Speed: 117.600310 samples/sec	 accuracy=24.118852	 loss=3.520456	 lr=0.035398
Epoch[011] Batch [3099]/[3759]	Speed: 117.177556 samples/sec	 accuracy=24.108367	 loss=3.520854	 lr=0.035436
Epoch[011] Batch [3149]/[3759]	Speed: 117.597011 samples/sec	 accuracy=24.105655	 loss=3.521533	 lr=0.035475
Epoch[011] Batch [3199]/[3759]	Speed: 116.781930 samples/sec	 accuracy=24.125977	 loss=3.520168	 lr=0.035514
Epoch[011] Batch [3249]/[3759]	Speed: 117.561413 samples/sec	 accuracy=24.134135	 loss=3.520606	 lr=0.035553
Epoch[011] Batch [3299]/[3759]	Speed: 117.405319 samples/sec	 accuracy=24.154830	 loss=3.519633	 lr=0.035591
Epoch[011] Batch [3349]/[3759]	Speed: 117.460628 samples/sec	 accuracy=24.161381	 loss=3.518959	 lr=0.035630
Epoch[011] Batch [3399]/[3759]	Speed: 118.142791 samples/sec	 accuracy=24.167739	 loss=3.518843	 lr=0.035669
Epoch[011] Batch [3449]/[3759]	Speed: 117.684727 samples/sec	 accuracy=24.163949	 loss=3.518477	 lr=0.035708
Epoch[011] Batch [3499]/[3759]	Speed: 116.886689 samples/sec	 accuracy=24.154464	 loss=3.518716	 lr=0.035746
Epoch[011] Batch [3549]/[3759]	Speed: 117.612910 samples/sec	 accuracy=24.167694	 loss=3.518502	 lr=0.035785
Epoch[011] Batch [3599]/[3759]	Speed: 117.045256 samples/sec	 accuracy=24.179688	 loss=3.518091	 lr=0.035824
Epoch[011] Batch [3649]/[3759]	Speed: 118.254349 samples/sec	 accuracy=24.183219	 loss=3.518356	 lr=0.035862
Epoch[011] Batch [3699]/[3759]	Speed: 116.883211 samples/sec	 accuracy=24.172720	 loss=3.518812	 lr=0.035901
Epoch[011] Batch [3749]/[3759]	Speed: 120.135081 samples/sec	 accuracy=24.197083	 loss=3.517890	 lr=0.035940
Batch [0049]/[0303]: acc-top1=24.906250 acc-top5=50.156250
Batch [0099]/[0303]: acc-top1=22.281250 acc-top5=46.390625
Batch [0149]/[0303]: acc-top1=22.875000 acc-top5=47.666667
Batch [0199]/[0303]: acc-top1=25.093750 acc-top5=49.898438
Batch [0249]/[0303]: acc-top1=24.550000 acc-top5=49.456250
Batch [0299]/[0303]: acc-top1=24.083333 acc-top5=49.395833
[Epoch 011] training: accuracy=24.196512	 loss=3.517799
[Epoch 011] speed: 116 samples/sec	time cost: 2177.457937
[Epoch 011] validation: acc-top1=24.113036 acc-top5=49.386345 loss=3.568910
Epoch[012] Batch [0049]/[3760]	Speed: 88.417975 samples/sec	 accuracy=23.718750	 loss=3.533068	 lr=0.035986
Epoch[012] Batch [0099]/[3760]	Speed: 117.358489 samples/sec	 accuracy=24.265625	 loss=3.511130	 lr=0.036024
Epoch[012] Batch [0149]/[3760]	Speed: 116.754231 samples/sec	 accuracy=24.572917	 loss=3.484893	 lr=0.036063
Epoch[012] Batch [0199]/[3760]	Speed: 117.375136 samples/sec	 accuracy=24.882812	 loss=3.475378	 lr=0.036102
Epoch[012] Batch [0249]/[3760]	Speed: 117.686309 samples/sec	 accuracy=25.112500	 loss=3.466282	 lr=0.036141
Epoch[012] Batch [0299]/[3760]	Speed: 117.192188 samples/sec	 accuracy=24.895833	 loss=3.470195	 lr=0.036179
Epoch[012] Batch [0349]/[3760]	Speed: 117.646241 samples/sec	 accuracy=25.133929	 loss=3.461356	 lr=0.036218
Epoch[012] Batch [0399]/[3760]	Speed: 117.155449 samples/sec	 accuracy=25.109375	 loss=3.462676	 lr=0.036257
Epoch[012] Batch [0449]/[3760]	Speed: 117.634758 samples/sec	 accuracy=25.260417	 loss=3.458260	 lr=0.036295
Epoch[012] Batch [0499]/[3760]	Speed: 116.865280 samples/sec	 accuracy=25.187500	 loss=3.467311	 lr=0.036334
Epoch[012] Batch [0549]/[3760]	Speed: 118.641828 samples/sec	 accuracy=25.196023	 loss=3.464496	 lr=0.036373
Epoch[012] Batch [0599]/[3760]	Speed: 118.183754 samples/sec	 accuracy=25.203125	 loss=3.464635	 lr=0.036412
Epoch[012] Batch [0649]/[3760]	Speed: 117.665957 samples/sec	 accuracy=25.182692	 loss=3.463360	 lr=0.036450
Epoch[012] Batch [0699]/[3760]	Speed: 116.661845 samples/sec	 accuracy=25.095982	 loss=3.464331	 lr=0.036489
Epoch[012] Batch [0749]/[3760]	Speed: 117.407433 samples/sec	 accuracy=25.133333	 loss=3.463408	 lr=0.036528
Epoch[012] Batch [0799]/[3760]	Speed: 117.452804 samples/sec	 accuracy=25.074219	 loss=3.467201	 lr=0.036567
Epoch[012] Batch [0849]/[3760]	Speed: 117.483489 samples/sec	 accuracy=25.053309	 loss=3.466356	 lr=0.036605
Epoch[012] Batch [0899]/[3760]	Speed: 117.386964 samples/sec	 accuracy=25.114583	 loss=3.466455	 lr=0.036644
Epoch[012] Batch [0949]/[3760]	Speed: 117.298479 samples/sec	 accuracy=25.177632	 loss=3.464017	 lr=0.036683
Epoch[012] Batch [0999]/[3760]	Speed: 117.631109 samples/sec	 accuracy=25.171875	 loss=3.464928	 lr=0.036721
Epoch[012] Batch [1049]/[3760]	Speed: 117.370041 samples/sec	 accuracy=25.138393	 loss=3.464573	 lr=0.036760
Epoch[012] Batch [1099]/[3760]	Speed: 117.275631 samples/sec	 accuracy=25.139205	 loss=3.466801	 lr=0.036799
Epoch[012] Batch [1149]/[3760]	Speed: 117.839771 samples/sec	 accuracy=25.133152	 loss=3.466141	 lr=0.036838
Epoch[012] Batch [1199]/[3760]	Speed: 116.388494 samples/sec	 accuracy=25.136719	 loss=3.466205	 lr=0.036876
Epoch[012] Batch [1249]/[3760]	Speed: 117.112653 samples/sec	 accuracy=25.095000	 loss=3.468494	 lr=0.036915
Epoch[012] Batch [1299]/[3760]	Speed: 116.849932 samples/sec	 accuracy=25.132212	 loss=3.464927	 lr=0.036954
Epoch[012] Batch [1349]/[3760]	Speed: 116.837361 samples/sec	 accuracy=25.096065	 loss=3.465640	 lr=0.036993
Epoch[012] Batch [1399]/[3760]	Speed: 116.930282 samples/sec	 accuracy=25.079241	 loss=3.463784	 lr=0.037031
Epoch[012] Batch [1449]/[3760]	Speed: 116.777413 samples/sec	 accuracy=25.077586	 loss=3.464388	 lr=0.037070
Epoch[012] Batch [1499]/[3760]	Speed: 117.659801 samples/sec	 accuracy=25.057292	 loss=3.464722	 lr=0.037109
Epoch[012] Batch [1549]/[3760]	Speed: 117.721037 samples/sec	 accuracy=25.089718	 loss=3.463444	 lr=0.037148
Epoch[012] Batch [1599]/[3760]	Speed: 117.779132 samples/sec	 accuracy=25.104492	 loss=3.464218	 lr=0.037186
Epoch[012] Batch [1649]/[3760]	Speed: 117.645680 samples/sec	 accuracy=25.134470	 loss=3.463931	 lr=0.037225
Epoch[012] Batch [1699]/[3760]	Speed: 117.111933 samples/sec	 accuracy=25.145221	 loss=3.462946	 lr=0.037264
Epoch[012] Batch [1749]/[3760]	Speed: 117.622714 samples/sec	 accuracy=25.132143	 loss=3.464140	 lr=0.037302
Epoch[012] Batch [1799]/[3760]	Speed: 117.431410 samples/sec	 accuracy=25.140625	 loss=3.464294	 lr=0.037341
Epoch[012] Batch [1849]/[3760]	Speed: 117.468224 samples/sec	 accuracy=25.140203	 loss=3.465471	 lr=0.037380
Epoch[012] Batch [1899]/[3760]	Speed: 117.963330 samples/sec	 accuracy=25.144737	 loss=3.464666	 lr=0.037419
Epoch[012] Batch [1949]/[3760]	Speed: 117.370214 samples/sec	 accuracy=25.155449	 loss=3.463507	 lr=0.037457
Epoch[012] Batch [1999]/[3760]	Speed: 118.104698 samples/sec	 accuracy=25.141406	 loss=3.463368	 lr=0.037496
Epoch[012] Batch [2049]/[3760]	Speed: 117.595389 samples/sec	 accuracy=25.138720	 loss=3.463771	 lr=0.037535
Epoch[012] Batch [2099]/[3760]	Speed: 118.407997 samples/sec	 accuracy=25.128720	 loss=3.464131	 lr=0.037574
Epoch[012] Batch [2149]/[3760]	Speed: 117.402110 samples/sec	 accuracy=25.095203	 loss=3.464412	 lr=0.037612
Epoch[012] Batch [2199]/[3760]	Speed: 117.664690 samples/sec	 accuracy=25.098011	 loss=3.464098	 lr=0.037651
Epoch[012] Batch [2249]/[3760]	Speed: 117.739374 samples/sec	 accuracy=25.093750	 loss=3.464468	 lr=0.037690
Epoch[012] Batch [2299]/[3760]	Speed: 117.640533 samples/sec	 accuracy=25.122283	 loss=3.463262	 lr=0.037728
Epoch[012] Batch [2349]/[3760]	Speed: 117.455823 samples/sec	 accuracy=25.151596	 loss=3.462467	 lr=0.037767
Epoch[012] Batch [2399]/[3760]	Speed: 118.228411 samples/sec	 accuracy=25.167969	 loss=3.462604	 lr=0.037806
Epoch[012] Batch [2449]/[3760]	Speed: 117.914255 samples/sec	 accuracy=25.161990	 loss=3.462992	 lr=0.037845
Epoch[012] Batch [2499]/[3760]	Speed: 117.882000 samples/sec	 accuracy=25.143125	 loss=3.463507	 lr=0.037883
Epoch[012] Batch [2549]/[3760]	Speed: 116.993975 samples/sec	 accuracy=25.152574	 loss=3.462769	 lr=0.037922
Epoch[012] Batch [2599]/[3760]	Speed: 117.355117 samples/sec	 accuracy=25.169471	 loss=3.461943	 lr=0.037961
Epoch[012] Batch [2649]/[3760]	Speed: 117.328543 samples/sec	 accuracy=25.178066	 loss=3.462055	 lr=0.038000
Epoch[012] Batch [2699]/[3760]	Speed: 117.425392 samples/sec	 accuracy=25.181134	 loss=3.462078	 lr=0.038038
Epoch[012] Batch [2749]/[3760]	Speed: 117.485429 samples/sec	 accuracy=25.184091	 loss=3.461722	 lr=0.038077
Epoch[012] Batch [2799]/[3760]	Speed: 116.927235 samples/sec	 accuracy=25.179687	 loss=3.461768	 lr=0.038116
Epoch[012] Batch [2849]/[3760]	Speed: 118.084173 samples/sec	 accuracy=25.180373	 loss=3.461704	 lr=0.038155
Epoch[012] Batch [2899]/[3760]	Speed: 117.330497 samples/sec	 accuracy=25.183728	 loss=3.461752	 lr=0.038193
Epoch[012] Batch [2949]/[3760]	Speed: 117.699425 samples/sec	 accuracy=25.174788	 loss=3.461359	 lr=0.038232
Epoch[012] Batch [2999]/[3760]	Speed: 116.618382 samples/sec	 accuracy=25.183333	 loss=3.461222	 lr=0.038271
Epoch[012] Batch [3049]/[3760]	Speed: 117.546513 samples/sec	 accuracy=25.207992	 loss=3.460631	 lr=0.038309
Epoch[012] Batch [3099]/[3760]	Speed: 118.163798 samples/sec	 accuracy=25.229335	 loss=3.459337	 lr=0.038348
Epoch[012] Batch [3149]/[3760]	Speed: 118.183886 samples/sec	 accuracy=25.218254	 loss=3.460464	 lr=0.038387
Epoch[012] Batch [3199]/[3760]	Speed: 117.944196 samples/sec	 accuracy=25.224121	 loss=3.459853	 lr=0.038426
Epoch[012] Batch [3249]/[3760]	Speed: 117.556044 samples/sec	 accuracy=25.233173	 loss=3.459322	 lr=0.038464
Epoch[012] Batch [3299]/[3760]	Speed: 117.598784 samples/sec	 accuracy=25.242898	 loss=3.458567	 lr=0.038503
Epoch[012] Batch [3349]/[3760]	Speed: 117.409671 samples/sec	 accuracy=25.256063	 loss=3.457822	 lr=0.038542
Epoch[012] Batch [3399]/[3760]	Speed: 117.182342 samples/sec	 accuracy=25.251838	 loss=3.457342	 lr=0.038581
Epoch[012] Batch [3449]/[3760]	Speed: 117.287839 samples/sec	 accuracy=25.240036	 loss=3.458492	 lr=0.038619
Epoch[012] Batch [3499]/[3760]	Speed: 117.227578 samples/sec	 accuracy=25.243304	 loss=3.458326	 lr=0.038658
Epoch[012] Batch [3549]/[3760]	Speed: 118.100595 samples/sec	 accuracy=25.247799	 loss=3.458051	 lr=0.038697
Epoch[012] Batch [3599]/[3760]	Speed: 117.443603 samples/sec	 accuracy=25.227865	 loss=3.458838	 lr=0.038735
Epoch[012] Batch [3649]/[3760]	Speed: 116.714365 samples/sec	 accuracy=25.229880	 loss=3.458743	 lr=0.038774
Epoch[012] Batch [3699]/[3760]	Speed: 117.333437 samples/sec	 accuracy=25.237331	 loss=3.458843	 lr=0.038813
Epoch[012] Batch [3749]/[3760]	Speed: 120.373587 samples/sec	 accuracy=25.227917	 loss=3.459334	 lr=0.038852
Batch [0049]/[0303]: acc-top1=26.656250 acc-top5=53.437500
Batch [0099]/[0303]: acc-top1=24.765625 acc-top5=51.453125
Batch [0149]/[0303]: acc-top1=26.229167 acc-top5=52.885417
Batch [0199]/[0303]: acc-top1=28.984375 acc-top5=55.179688
Batch [0249]/[0303]: acc-top1=28.568750 acc-top5=55.337500
Batch [0299]/[0303]: acc-top1=27.984375 acc-top5=55.427083
[Epoch 012] training: accuracy=25.234791	 loss=3.458918
[Epoch 012] speed: 116 samples/sec	time cost: 2176.693999
[Epoch 012] validation: acc-top1=27.851691 acc-top5=55.301155 loss=3.313625
Epoch[013] Batch [0049]/[3760]	Speed: 86.145154 samples/sec	 accuracy=26.656250	 loss=3.374520	 lr=0.038898
Epoch[013] Batch [0099]/[3760]	Speed: 117.007152 samples/sec	 accuracy=25.828125	 loss=3.417770	 lr=0.038937
Epoch[013] Batch [0149]/[3760]	Speed: 116.824439 samples/sec	 accuracy=25.958333	 loss=3.390366	 lr=0.038976
Epoch[013] Batch [0199]/[3760]	Speed: 116.827180 samples/sec	 accuracy=25.906250	 loss=3.402480	 lr=0.039014
Epoch[013] Batch [0249]/[3760]	Speed: 117.563663 samples/sec	 accuracy=25.675000	 loss=3.417963	 lr=0.039053
Epoch[013] Batch [0299]/[3760]	Speed: 116.886331 samples/sec	 accuracy=25.729167	 loss=3.412087	 lr=0.039092
Epoch[013] Batch [0349]/[3760]	Speed: 117.007898 samples/sec	 accuracy=25.736607	 loss=3.414770	 lr=0.039131
Epoch[013] Batch [0399]/[3760]	Speed: 116.400168 samples/sec	 accuracy=25.949219	 loss=3.410536	 lr=0.039169
Epoch[013] Batch [0449]/[3760]	Speed: 117.879276 samples/sec	 accuracy=26.041667	 loss=3.407301	 lr=0.039208
Epoch[013] Batch [0499]/[3760]	Speed: 117.326419 samples/sec	 accuracy=25.906250	 loss=3.414359	 lr=0.039247
Epoch[013] Batch [0549]/[3760]	Speed: 117.670115 samples/sec	 accuracy=25.911932	 loss=3.412847	 lr=0.039285
Epoch[013] Batch [0599]/[3760]	Speed: 117.921433 samples/sec	 accuracy=25.893229	 loss=3.415865	 lr=0.039324
Epoch[013] Batch [0649]/[3760]	Speed: 117.490415 samples/sec	 accuracy=25.882212	 loss=3.418876	 lr=0.039363
Epoch[013] Batch [0699]/[3760]	Speed: 117.304128 samples/sec	 accuracy=25.857143	 loss=3.421525	 lr=0.039402
Epoch[013] Batch [0749]/[3760]	Speed: 118.224165 samples/sec	 accuracy=25.833333	 loss=3.420576	 lr=0.039440
Epoch[013] Batch [0799]/[3760]	Speed: 117.312532 samples/sec	 accuracy=25.878906	 loss=3.418877	 lr=0.039479
Epoch[013] Batch [0849]/[3760]	Speed: 116.596397 samples/sec	 accuracy=25.906250	 loss=3.419296	 lr=0.039518
Epoch[013] Batch [0899]/[3760]	Speed: 117.051841 samples/sec	 accuracy=25.826389	 loss=3.419445	 lr=0.039557
Epoch[013] Batch [0949]/[3760]	Speed: 116.621419 samples/sec	 accuracy=25.904605	 loss=3.414654	 lr=0.039595
Epoch[013] Batch [0999]/[3760]	Speed: 116.744680 samples/sec	 accuracy=25.798437	 loss=3.416732	 lr=0.039634
Epoch[013] Batch [1049]/[3760]	Speed: 117.530566 samples/sec	 accuracy=25.824405	 loss=3.416175	 lr=0.039673
Epoch[013] Batch [1099]/[3760]	Speed: 117.444107 samples/sec	 accuracy=25.823864	 loss=3.417116	 lr=0.039712
Epoch[013] Batch [1149]/[3760]	Speed: 117.057339 samples/sec	 accuracy=25.835598	 loss=3.417828	 lr=0.039750
Epoch[013] Batch [1199]/[3760]	Speed: 117.141327 samples/sec	 accuracy=25.912760	 loss=3.414591	 lr=0.039789
Epoch[013] Batch [1249]/[3760]	Speed: 118.167989 samples/sec	 accuracy=25.908750	 loss=3.414902	 lr=0.039828
Epoch[013] Batch [1299]/[3760]	Speed: 117.158927 samples/sec	 accuracy=25.941106	 loss=3.413276	 lr=0.039866
Epoch[013] Batch [1349]/[3760]	Speed: 116.763308 samples/sec	 accuracy=25.959491	 loss=3.412068	 lr=0.039905
Epoch[013] Batch [1399]/[3760]	Speed: 117.028248 samples/sec	 accuracy=25.976562	 loss=3.412372	 lr=0.039944
Epoch[013] Batch [1449]/[3760]	Speed: 117.239267 samples/sec	 accuracy=25.966595	 loss=3.412037	 lr=0.039983
Epoch[013] Batch [1499]/[3760]	Speed: 118.555555 samples/sec	 accuracy=25.972917	 loss=3.412083	 lr=0.040021
Epoch[013] Batch [1549]/[3760]	Speed: 118.237466 samples/sec	 accuracy=25.950605	 loss=3.413787	 lr=0.040060
Epoch[013] Batch [1599]/[3760]	Speed: 117.385326 samples/sec	 accuracy=25.956055	 loss=3.414655	 lr=0.040099
Epoch[013] Batch [1649]/[3760]	Speed: 117.327990 samples/sec	 accuracy=25.979167	 loss=3.414222	 lr=0.040138
Epoch[013] Batch [1699]/[3760]	Speed: 117.381089 samples/sec	 accuracy=25.988971	 loss=3.413146	 lr=0.040176
Epoch[013] Batch [1749]/[3760]	Speed: 117.998801 samples/sec	 accuracy=25.956250	 loss=3.413382	 lr=0.040215
Epoch[013] Batch [1799]/[3760]	Speed: 117.692595 samples/sec	 accuracy=25.960937	 loss=3.413002	 lr=0.040254
Epoch[013] Batch [1849]/[3760]	Speed: 117.583558 samples/sec	 accuracy=25.971284	 loss=3.413014	 lr=0.040292
Epoch[013] Batch [1899]/[3760]	Speed: 117.456280 samples/sec	 accuracy=25.976151	 loss=3.413581	 lr=0.040331
Epoch[013] Batch [1949]/[3760]	Speed: 116.604743 samples/sec	 accuracy=26.024038	 loss=3.413117	 lr=0.040370
Epoch[013] Batch [1999]/[3760]	Speed: 116.404161 samples/sec	 accuracy=26.027344	 loss=3.412001	 lr=0.040409
Epoch[013] Batch [2049]/[3760]	Speed: 117.728838 samples/sec	 accuracy=26.012195	 loss=3.413820	 lr=0.040447
Epoch[013] Batch [2099]/[3760]	Speed: 116.927014 samples/sec	 accuracy=26.015625	 loss=3.413162	 lr=0.040486
Epoch[013] Batch [2149]/[3760]	Speed: 117.080796 samples/sec	 accuracy=26.000727	 loss=3.413155	 lr=0.040525
Epoch[013] Batch [2199]/[3760]	Speed: 118.248353 samples/sec	 accuracy=25.968750	 loss=3.413757	 lr=0.040564
Epoch[013] Batch [2249]/[3760]	Speed: 117.518961 samples/sec	 accuracy=25.964583	 loss=3.413941	 lr=0.040602
Epoch[013] Batch [2299]/[3760]	Speed: 117.857286 samples/sec	 accuracy=25.978940	 loss=3.413346	 lr=0.040641
Epoch[013] Batch [2349]/[3760]	Speed: 117.688661 samples/sec	 accuracy=25.974069	 loss=3.413457	 lr=0.040680
Epoch[013] Batch [2399]/[3760]	Speed: 118.133378 samples/sec	 accuracy=25.975911	 loss=3.412620	 lr=0.040719
Epoch[013] Batch [2449]/[3760]	Speed: 117.991362 samples/sec	 accuracy=25.964286	 loss=3.413592	 lr=0.040757
Epoch[013] Batch [2499]/[3760]	Speed: 117.227430 samples/sec	 accuracy=25.964375	 loss=3.412991	 lr=0.040796
Epoch[013] Batch [2549]/[3760]	Speed: 117.607664 samples/sec	 accuracy=25.975490	 loss=3.412226	 lr=0.040835
Epoch[013] Batch [2599]/[3760]	Speed: 117.306727 samples/sec	 accuracy=25.980769	 loss=3.413392	 lr=0.040873
Epoch[013] Batch [2649]/[3760]	Speed: 117.954023 samples/sec	 accuracy=25.970519	 loss=3.413278	 lr=0.040912
Epoch[013] Batch [2699]/[3760]	Speed: 117.860853 samples/sec	 accuracy=26.000000	 loss=3.412470	 lr=0.040951
Epoch[013] Batch [2749]/[3760]	Speed: 117.578166 samples/sec	 accuracy=25.993182	 loss=3.413128	 lr=0.040990
Epoch[013] Batch [2799]/[3760]	Speed: 117.726751 samples/sec	 accuracy=25.991629	 loss=3.412748	 lr=0.041028
Epoch[013] Batch [2849]/[3760]	Speed: 118.202712 samples/sec	 accuracy=25.983553	 loss=3.413067	 lr=0.041067
Epoch[013] Batch [2899]/[3760]	Speed: 117.976110 samples/sec	 accuracy=25.990841	 loss=3.412790	 lr=0.041106
Epoch[013] Batch [2949]/[3760]	Speed: 117.517272 samples/sec	 accuracy=25.979873	 loss=3.413340	 lr=0.041145
Epoch[013] Batch [2999]/[3760]	Speed: 117.581517 samples/sec	 accuracy=25.955729	 loss=3.413560	 lr=0.041183
Epoch[013] Batch [3049]/[3760]	Speed: 117.871102 samples/sec	 accuracy=25.950820	 loss=3.413184	 lr=0.041222
Epoch[013] Batch [3099]/[3760]	Speed: 117.256099 samples/sec	 accuracy=25.959173	 loss=3.413092	 lr=0.041261
Epoch[013] Batch [3149]/[3760]	Speed: 117.748210 samples/sec	 accuracy=25.954365	 loss=3.413673	 lr=0.041299
Epoch[013] Batch [3199]/[3760]	Speed: 117.600956 samples/sec	 accuracy=25.945801	 loss=3.413395	 lr=0.041338
Epoch[013] Batch [3249]/[3760]	Speed: 118.121557 samples/sec	 accuracy=25.963942	 loss=3.412960	 lr=0.041377
Epoch[013] Batch [3299]/[3760]	Speed: 118.022585 samples/sec	 accuracy=25.980587	 loss=3.412601	 lr=0.041416
Epoch[013] Batch [3349]/[3760]	Speed: 117.855299 samples/sec	 accuracy=25.968284	 loss=3.413600	 lr=0.041454
Epoch[013] Batch [3399]/[3760]	Speed: 118.158810 samples/sec	 accuracy=25.986213	 loss=3.412914	 lr=0.041493
Epoch[013] Batch [3449]/[3760]	Speed: 116.989700 samples/sec	 accuracy=25.983243	 loss=3.412664	 lr=0.041532
Epoch[013] Batch [3499]/[3760]	Speed: 118.038409 samples/sec	 accuracy=25.970089	 loss=3.412392	 lr=0.041571
Epoch[013] Batch [3549]/[3760]	Speed: 117.111850 samples/sec	 accuracy=25.972711	 loss=3.412500	 lr=0.041609
Epoch[013] Batch [3599]/[3760]	Speed: 117.800063 samples/sec	 accuracy=25.990017	 loss=3.412347	 lr=0.041648
Epoch[013] Batch [3649]/[3760]	Speed: 117.163987 samples/sec	 accuracy=25.973887	 loss=3.412566	 lr=0.041687
Epoch[013] Batch [3699]/[3760]	Speed: 117.937258 samples/sec	 accuracy=25.990287	 loss=3.412109	 lr=0.041726
Epoch[013] Batch [3749]/[3760]	Speed: 121.023493 samples/sec	 accuracy=26.000000	 loss=3.412098	 lr=0.041764
Batch [0049]/[0303]: acc-top1=28.125000 acc-top5=53.875000
Batch [0099]/[0303]: acc-top1=25.718750 acc-top5=51.234375
Batch [0149]/[0303]: acc-top1=26.083333 acc-top5=52.156250
Batch [0199]/[0303]: acc-top1=28.171875 acc-top5=54.195312
Batch [0249]/[0303]: acc-top1=27.575000 acc-top5=53.981250
Batch [0299]/[0303]: acc-top1=26.984375 acc-top5=54.062500
[Epoch 013] training: accuracy=26.003989	 loss=3.412302
[Epoch 013] speed: 116 samples/sec	time cost: 2177.241477
[Epoch 013] validation: acc-top1=26.969884 acc-top5=54.037748 loss=3.324269
Epoch[014] Batch [0049]/[3759]	Speed: 88.161012 samples/sec	 accuracy=26.187500	 loss=3.353482	 lr=0.041811
Epoch[014] Batch [0099]/[3759]	Speed: 118.045259 samples/sec	 accuracy=26.218750	 loss=3.345038	 lr=0.041849
Epoch[014] Batch [0149]/[3759]	Speed: 117.471240 samples/sec	 accuracy=26.385417	 loss=3.338184	 lr=0.041888
Epoch[014] Batch [0199]/[3759]	Speed: 118.399498 samples/sec	 accuracy=26.898438	 loss=3.334413	 lr=0.041927
Epoch[014] Batch [0249]/[3759]	Speed: 118.376080 samples/sec	 accuracy=26.556250	 loss=3.363094	 lr=0.041966
Epoch[014] Batch [0299]/[3759]	Speed: 117.310314 samples/sec	 accuracy=26.463542	 loss=3.366504	 lr=0.042004
Epoch[014] Batch [0349]/[3759]	Speed: 117.482284 samples/sec	 accuracy=26.785714	 loss=3.365592	 lr=0.042043
Epoch[014] Batch [0399]/[3759]	Speed: 117.725837 samples/sec	 accuracy=26.648438	 loss=3.370055	 lr=0.042082
Epoch[014] Batch [0449]/[3759]	Speed: 117.514674 samples/sec	 accuracy=26.572917	 loss=3.373828	 lr=0.042121
Epoch[014] Batch [0499]/[3759]	Speed: 117.118252 samples/sec	 accuracy=26.612500	 loss=3.373597	 lr=0.042159
Epoch[014] Batch [0549]/[3759]	Speed: 118.119173 samples/sec	 accuracy=26.531250	 loss=3.378233	 lr=0.042198
Epoch[014] Batch [0599]/[3759]	Speed: 117.329842 samples/sec	 accuracy=26.635417	 loss=3.375495	 lr=0.042237
Epoch[014] Batch [0649]/[3759]	Speed: 117.314528 samples/sec	 accuracy=26.639423	 loss=3.376032	 lr=0.042275
Epoch[014] Batch [0699]/[3759]	Speed: 118.019092 samples/sec	 accuracy=26.667411	 loss=3.374794	 lr=0.042314
Epoch[014] Batch [0749]/[3759]	Speed: 117.659870 samples/sec	 accuracy=26.585417	 loss=3.376763	 lr=0.042353
Epoch[014] Batch [0799]/[3759]	Speed: 117.964823 samples/sec	 accuracy=26.632813	 loss=3.376217	 lr=0.042392
Epoch[014] Batch [0849]/[3759]	Speed: 117.037973 samples/sec	 accuracy=26.595588	 loss=3.376163	 lr=0.042430
Epoch[014] Batch [0899]/[3759]	Speed: 117.185548 samples/sec	 accuracy=26.583333	 loss=3.377120	 lr=0.042469
Epoch[014] Batch [0949]/[3759]	Speed: 117.209184 samples/sec	 accuracy=26.588816	 loss=3.376222	 lr=0.042508
Epoch[014] Batch [0999]/[3759]	Speed: 117.465180 samples/sec	 accuracy=26.615625	 loss=3.375858	 lr=0.042547
Epoch[014] Batch [1049]/[3759]	Speed: 117.596088 samples/sec	 accuracy=26.650298	 loss=3.376934	 lr=0.042585
Epoch[014] Batch [1099]/[3759]	Speed: 117.210376 samples/sec	 accuracy=26.639205	 loss=3.377899	 lr=0.042624
Epoch[014] Batch [1149]/[3759]	Speed: 118.338600 samples/sec	 accuracy=26.656250	 loss=3.376963	 lr=0.042663
Epoch[014] Batch [1199]/[3759]	Speed: 117.901802 samples/sec	 accuracy=26.670573	 loss=3.376826	 lr=0.042702
Epoch[014] Batch [1249]/[3759]	Speed: 118.653043 samples/sec	 accuracy=26.713750	 loss=3.375916	 lr=0.042740
Epoch[014] Batch [1299]/[3759]	Speed: 117.158069 samples/sec	 accuracy=26.694712	 loss=3.378057	 lr=0.042779
Epoch[014] Batch [1349]/[3759]	Speed: 117.124463 samples/sec	 accuracy=26.754630	 loss=3.376590	 lr=0.042818
Epoch[014] Batch [1399]/[3759]	Speed: 117.352284 samples/sec	 accuracy=26.742187	 loss=3.375518	 lr=0.042856
Epoch[014] Batch [1449]/[3759]	Speed: 117.865892 samples/sec	 accuracy=26.724138	 loss=3.373419	 lr=0.042895
Epoch[014] Batch [1499]/[3759]	Speed: 118.211354 samples/sec	 accuracy=26.726042	 loss=3.373370	 lr=0.042934
Epoch[014] Batch [1549]/[3759]	Speed: 117.585092 samples/sec	 accuracy=26.723790	 loss=3.373053	 lr=0.042973
Epoch[014] Batch [1599]/[3759]	Speed: 117.427764 samples/sec	 accuracy=26.758789	 loss=3.372378	 lr=0.043011
Epoch[014] Batch [1649]/[3759]	Speed: 118.224618 samples/sec	 accuracy=26.732955	 loss=3.374279	 lr=0.043050
Epoch[014] Batch [1699]/[3759]	Speed: 117.645740 samples/sec	 accuracy=26.730699	 loss=3.374356	 lr=0.043089
Epoch[014] Batch [1749]/[3759]	Speed: 118.020648 samples/sec	 accuracy=26.733929	 loss=3.375223	 lr=0.043128
Epoch[014] Batch [1799]/[3759]	Speed: 117.919658 samples/sec	 accuracy=26.721354	 loss=3.376695	 lr=0.043166
Epoch[014] Batch [1849]/[3759]	Speed: 117.428919 samples/sec	 accuracy=26.730574	 loss=3.376480	 lr=0.043205
Epoch[014] Batch [1899]/[3759]	Speed: 117.838762 samples/sec	 accuracy=26.796875	 loss=3.373282	 lr=0.043244
Epoch[014] Batch [1949]/[3759]	Speed: 117.731491 samples/sec	 accuracy=26.755609	 loss=3.375698	 lr=0.043283
Epoch[014] Batch [1999]/[3759]	Speed: 117.587281 samples/sec	 accuracy=26.784375	 loss=3.374369	 lr=0.043321
Epoch[014] Batch [2049]/[3759]	Speed: 118.072531 samples/sec	 accuracy=26.778201	 loss=3.373650	 lr=0.043360
Epoch[014] Batch [2099]/[3759]	Speed: 117.596923 samples/sec	 accuracy=26.779762	 loss=3.373900	 lr=0.043399
Epoch[014] Batch [2149]/[3759]	Speed: 117.084940 samples/sec	 accuracy=26.783430	 loss=3.373062	 lr=0.043437
Epoch[014] Batch [2199]/[3759]	Speed: 117.953031 samples/sec	 accuracy=26.772017	 loss=3.373943	 lr=0.043476
Epoch[014] Batch [2249]/[3759]	Speed: 117.300656 samples/sec	 accuracy=26.763889	 loss=3.374114	 lr=0.043515
Epoch[014] Batch [2299]/[3759]	Speed: 117.911023 samples/sec	 accuracy=26.769022	 loss=3.373419	 lr=0.043554
Epoch[014] Batch [2349]/[3759]	Speed: 117.498026 samples/sec	 accuracy=26.790559	 loss=3.372905	 lr=0.043592
Epoch[014] Batch [2399]/[3759]	Speed: 117.783831 samples/sec	 accuracy=26.769531	 loss=3.373970	 lr=0.043631
Epoch[014] Batch [2449]/[3759]	Speed: 117.219313 samples/sec	 accuracy=26.762117	 loss=3.373370	 lr=0.043670
Epoch[014] Batch [2499]/[3759]	Speed: 117.548880 samples/sec	 accuracy=26.765625	 loss=3.373473	 lr=0.043709
Epoch[014] Batch [2549]/[3759]	Speed: 117.074014 samples/sec	 accuracy=26.782475	 loss=3.373726	 lr=0.043747
Epoch[014] Batch [2599]/[3759]	Speed: 117.424772 samples/sec	 accuracy=26.746394	 loss=3.374433	 lr=0.043786
Epoch[014] Batch [2649]/[3759]	Speed: 117.809946 samples/sec	 accuracy=26.765330	 loss=3.374000	 lr=0.043825
Epoch[014] Batch [2699]/[3759]	Speed: 117.796836 samples/sec	 accuracy=26.791088	 loss=3.373153	 lr=0.043863
Epoch[014] Batch [2749]/[3759]	Speed: 117.462544 samples/sec	 accuracy=26.787500	 loss=3.373139	 lr=0.043902
Epoch[014] Batch [2799]/[3759]	Speed: 117.501161 samples/sec	 accuracy=26.776786	 loss=3.373280	 lr=0.043941
Epoch[014] Batch [2849]/[3759]	Speed: 117.226943 samples/sec	 accuracy=26.748355	 loss=3.374746	 lr=0.043980
Epoch[014] Batch [2899]/[3759]	Speed: 117.286692 samples/sec	 accuracy=26.771013	 loss=3.373897	 lr=0.044018
Epoch[014] Batch [2949]/[3759]	Speed: 118.152039 samples/sec	 accuracy=26.770127	 loss=3.373651	 lr=0.044057
Epoch[014] Batch [2999]/[3759]	Speed: 117.884232 samples/sec	 accuracy=26.764583	 loss=3.373738	 lr=0.044096
Epoch[014] Batch [3049]/[3759]	Speed: 117.485642 samples/sec	 accuracy=26.759734	 loss=3.373468	 lr=0.044135
Epoch[014] Batch [3099]/[3759]	Speed: 117.070277 samples/sec	 accuracy=26.772681	 loss=3.373198	 lr=0.044173
Epoch[014] Batch [3149]/[3759]	Speed: 117.585752 samples/sec	 accuracy=26.759921	 loss=3.373178	 lr=0.044212
Epoch[014] Batch [3199]/[3759]	Speed: 117.482839 samples/sec	 accuracy=26.777344	 loss=3.372987	 lr=0.044251
Epoch[014] Batch [3249]/[3759]	Speed: 116.853230 samples/sec	 accuracy=26.756250	 loss=3.373843	 lr=0.044290
Epoch[014] Batch [3299]/[3759]	Speed: 117.823542 samples/sec	 accuracy=26.744318	 loss=3.374976	 lr=0.044328
Epoch[014] Batch [3349]/[3759]	Speed: 117.632234 samples/sec	 accuracy=26.755131	 loss=3.374760	 lr=0.044367
Epoch[014] Batch [3399]/[3759]	Speed: 116.752725 samples/sec	 accuracy=26.734835	 loss=3.374937	 lr=0.044406
Epoch[014] Batch [3449]/[3759]	Speed: 117.580513 samples/sec	 accuracy=26.739130	 loss=3.374870	 lr=0.044444
Epoch[014] Batch [3499]/[3759]	Speed: 116.839447 samples/sec	 accuracy=26.740179	 loss=3.374899	 lr=0.044483
Epoch[014] Batch [3549]/[3759]	Speed: 117.725280 samples/sec	 accuracy=26.742518	 loss=3.375168	 lr=0.044522
Epoch[014] Batch [3599]/[3759]	Speed: 117.744701 samples/sec	 accuracy=26.746962	 loss=3.374911	 lr=0.044561
Epoch[014] Batch [3649]/[3759]	Speed: 117.564242 samples/sec	 accuracy=26.730736	 loss=3.375686	 lr=0.044599
Epoch[014] Batch [3699]/[3759]	Speed: 117.205725 samples/sec	 accuracy=26.727618	 loss=3.376301	 lr=0.044638
Epoch[014] Batch [3749]/[3759]	Speed: 120.742418 samples/sec	 accuracy=26.746250	 loss=3.375953	 lr=0.044677
Batch [0049]/[0303]: acc-top1=28.468750 acc-top5=52.500000
Batch [0099]/[0303]: acc-top1=24.796875 acc-top5=49.906250
Batch [0149]/[0303]: acc-top1=25.864583 acc-top5=51.614583
Batch [0199]/[0303]: acc-top1=28.226563 acc-top5=54.109375
Batch [0249]/[0303]: acc-top1=28.037500 acc-top5=54.262500
Batch [0299]/[0303]: acc-top1=27.604167 acc-top5=54.432292
[Epoch 014] training: accuracy=26.748304	 loss=3.375888
[Epoch 014] speed: 117 samples/sec	time cost: 2173.432985
[Epoch 014] validation: acc-top1=27.629950 acc-top5=54.491543 loss=3.350162
Epoch[015] Batch [0049]/[3760]	Speed: 86.943770 samples/sec	 accuracy=26.281250	 loss=3.375057	 lr=0.044723
Epoch[015] Batch [0099]/[3760]	Speed: 117.390993 samples/sec	 accuracy=26.890625	 loss=3.362496	 lr=0.044761
Epoch[015] Batch [0149]/[3760]	Speed: 117.994876 samples/sec	 accuracy=26.697917	 loss=3.368180	 lr=0.044800
Epoch[015] Batch [0199]/[3760]	Speed: 117.388444 samples/sec	 accuracy=27.140625	 loss=3.349989	 lr=0.044839
Epoch[015] Batch [0249]/[3760]	Speed: 117.730196 samples/sec	 accuracy=27.212500	 loss=3.343932	 lr=0.044877
Epoch[015] Batch [0299]/[3760]	Speed: 117.611189 samples/sec	 accuracy=27.270833	 loss=3.338472	 lr=0.044916
Epoch[015] Batch [0349]/[3760]	Speed: 117.866151 samples/sec	 accuracy=27.138393	 loss=3.346275	 lr=0.044955
Epoch[015] Batch [0399]/[3760]	Speed: 117.412075 samples/sec	 accuracy=27.152344	 loss=3.343631	 lr=0.044994
Epoch[015] Batch [0449]/[3760]	Speed: 117.500590 samples/sec	 accuracy=27.256944	 loss=3.336336	 lr=0.045032
Epoch[015] Batch [0499]/[3760]	Speed: 117.263963 samples/sec	 accuracy=27.353125	 loss=3.333451	 lr=0.045071
Epoch[015] Batch [0549]/[3760]	Speed: 117.324865 samples/sec	 accuracy=27.295455	 loss=3.334793	 lr=0.045110
Epoch[015] Batch [0599]/[3760]	Speed: 117.779565 samples/sec	 accuracy=27.432292	 loss=3.330927	 lr=0.045149
Epoch[015] Batch [0649]/[3760]	Speed: 117.468296 samples/sec	 accuracy=27.401442	 loss=3.334175	 lr=0.045187
Epoch[015] Batch [0699]/[3760]	Speed: 117.756969 samples/sec	 accuracy=27.359375	 loss=3.337297	 lr=0.045226
Epoch[015] Batch [0749]/[3760]	Speed: 117.890500 samples/sec	 accuracy=27.331250	 loss=3.338385	 lr=0.045265
Epoch[015] Batch [0799]/[3760]	Speed: 117.893046 samples/sec	 accuracy=27.275391	 loss=3.338562	 lr=0.045303
Epoch[015] Batch [0849]/[3760]	Speed: 117.002626 samples/sec	 accuracy=27.215074	 loss=3.338887	 lr=0.045342
Epoch[015] Batch [0899]/[3760]	Speed: 117.500567 samples/sec	 accuracy=27.206597	 loss=3.339503	 lr=0.045381
Epoch[015] Batch [0949]/[3760]	Speed: 117.578092 samples/sec	 accuracy=27.212171	 loss=3.340541	 lr=0.045420
Epoch[015] Batch [0999]/[3760]	Speed: 116.800294 samples/sec	 accuracy=27.273437	 loss=3.340078	 lr=0.045458
Epoch[015] Batch [1049]/[3760]	Speed: 116.757326 samples/sec	 accuracy=27.288690	 loss=3.339513	 lr=0.045497
Epoch[015] Batch [1099]/[3760]	Speed: 117.652135 samples/sec	 accuracy=27.308239	 loss=3.339367	 lr=0.045536
Epoch[015] Batch [1149]/[3760]	Speed: 117.267859 samples/sec	 accuracy=27.255435	 loss=3.340548	 lr=0.045575
Epoch[015] Batch [1199]/[3760]	Speed: 117.652258 samples/sec	 accuracy=27.273437	 loss=3.339590	 lr=0.045613
Epoch[015] Batch [1249]/[3760]	Speed: 117.481844 samples/sec	 accuracy=27.287500	 loss=3.340024	 lr=0.045652
Epoch[015] Batch [1299]/[3760]	Speed: 117.443452 samples/sec	 accuracy=27.311298	 loss=3.338920	 lr=0.045691
Epoch[015] Batch [1349]/[3760]	Speed: 117.764659 samples/sec	 accuracy=27.327546	 loss=3.337659	 lr=0.045730
Epoch[015] Batch [1399]/[3760]	Speed: 117.949243 samples/sec	 accuracy=27.354911	 loss=3.337576	 lr=0.045768
Epoch[015] Batch [1449]/[3760]	Speed: 117.754170 samples/sec	 accuracy=27.327586	 loss=3.338835	 lr=0.045807
Epoch[015] Batch [1499]/[3760]	Speed: 117.365456 samples/sec	 accuracy=27.307292	 loss=3.340826	 lr=0.045846
Epoch[015] Batch [1549]/[3760]	Speed: 116.912366 samples/sec	 accuracy=27.354839	 loss=3.338854	 lr=0.045884
Epoch[015] Batch [1599]/[3760]	Speed: 117.128375 samples/sec	 accuracy=27.341797	 loss=3.339737	 lr=0.045923
Epoch[015] Batch [1649]/[3760]	Speed: 117.255508 samples/sec	 accuracy=27.305871	 loss=3.341163	 lr=0.045962
Epoch[015] Batch [1699]/[3760]	Speed: 117.876006 samples/sec	 accuracy=27.314338	 loss=3.341059	 lr=0.046001
Epoch[015] Batch [1749]/[3760]	Speed: 117.398362 samples/sec	 accuracy=27.294643	 loss=3.341913	 lr=0.046039
Epoch[015] Batch [1799]/[3760]	Speed: 118.302728 samples/sec	 accuracy=27.310764	 loss=3.341175	 lr=0.046078
Epoch[015] Batch [1849]/[3760]	Speed: 117.412870 samples/sec	 accuracy=27.310811	 loss=3.341494	 lr=0.046117
Epoch[015] Batch [1899]/[3760]	Speed: 117.860128 samples/sec	 accuracy=27.315789	 loss=3.340884	 lr=0.046156
Epoch[015] Batch [1949]/[3760]	Speed: 117.397776 samples/sec	 accuracy=27.288462	 loss=3.342327	 lr=0.046194
Epoch[015] Batch [1999]/[3760]	Speed: 117.368854 samples/sec	 accuracy=27.308594	 loss=3.340871	 lr=0.046233
Epoch[015] Batch [2049]/[3760]	Speed: 117.706857 samples/sec	 accuracy=27.294207	 loss=3.342138	 lr=0.046272
Epoch[015] Batch [2099]/[3760]	Speed: 116.199724 samples/sec	 accuracy=27.291667	 loss=3.341834	 lr=0.046310
Epoch[015] Batch [2149]/[3760]	Speed: 117.527105 samples/sec	 accuracy=27.289971	 loss=3.342027	 lr=0.046349
Epoch[015] Batch [2199]/[3760]	Speed: 117.640035 samples/sec	 accuracy=27.282670	 loss=3.342703	 lr=0.046388
Epoch[015] Batch [2249]/[3760]	Speed: 117.512663 samples/sec	 accuracy=27.277083	 loss=3.343285	 lr=0.046427
Epoch[015] Batch [2299]/[3760]	Speed: 117.341144 samples/sec	 accuracy=27.289402	 loss=3.342661	 lr=0.046465
Epoch[015] Batch [2349]/[3760]	Speed: 117.290264 samples/sec	 accuracy=27.327128	 loss=3.342537	 lr=0.046504
Epoch[015] Batch [2399]/[3760]	Speed: 117.503104 samples/sec	 accuracy=27.320312	 loss=3.342981	 lr=0.046543
Epoch[015] Batch [2449]/[3760]	Speed: 117.669289 samples/sec	 accuracy=27.324617	 loss=3.342770	 lr=0.046582
Epoch[015] Batch [2499]/[3760]	Speed: 118.050903 samples/sec	 accuracy=27.326250	 loss=3.342711	 lr=0.046620
Epoch[015] Batch [2549]/[3760]	Speed: 117.405889 samples/sec	 accuracy=27.329657	 loss=3.343679	 lr=0.046659
Epoch[015] Batch [2599]/[3760]	Speed: 117.964548 samples/sec	 accuracy=27.334736	 loss=3.344177	 lr=0.046698
Epoch[015] Batch [2649]/[3760]	Speed: 117.004438 samples/sec	 accuracy=27.307783	 loss=3.344237	 lr=0.046737
Epoch[015] Batch [2699]/[3760]	Speed: 117.686612 samples/sec	 accuracy=27.314815	 loss=3.343612	 lr=0.046775
Epoch[015] Batch [2749]/[3760]	Speed: 117.819195 samples/sec	 accuracy=27.323864	 loss=3.343203	 lr=0.046814
Epoch[015] Batch [2799]/[3760]	Speed: 118.075885 samples/sec	 accuracy=27.325893	 loss=3.343189	 lr=0.046853
Epoch[015] Batch [2849]/[3760]	Speed: 118.075766 samples/sec	 accuracy=27.328399	 loss=3.343531	 lr=0.046891
Epoch[015] Batch [2899]/[3760]	Speed: 117.401285 samples/sec	 accuracy=27.341056	 loss=3.343384	 lr=0.046930
Epoch[015] Batch [2949]/[3760]	Speed: 117.145683 samples/sec	 accuracy=27.334746	 loss=3.343152	 lr=0.046969
Epoch[015] Batch [2999]/[3760]	Speed: 117.244077 samples/sec	 accuracy=27.334375	 loss=3.342903	 lr=0.047008
Epoch[015] Batch [3049]/[3760]	Speed: 117.377131 samples/sec	 accuracy=27.348873	 loss=3.342902	 lr=0.047046
Epoch[015] Batch [3099]/[3760]	Speed: 117.020853 samples/sec	 accuracy=27.329133	 loss=3.343899	 lr=0.047085
Epoch[015] Batch [3149]/[3760]	Speed: 116.771692 samples/sec	 accuracy=27.336310	 loss=3.343401	 lr=0.047124
Epoch[015] Batch [3199]/[3760]	Speed: 117.609365 samples/sec	 accuracy=27.334961	 loss=3.343065	 lr=0.047163
Epoch[015] Batch [3249]/[3760]	Speed: 117.466300 samples/sec	 accuracy=27.333173	 loss=3.343525	 lr=0.047201
Epoch[015] Batch [3299]/[3760]	Speed: 117.971276 samples/sec	 accuracy=27.321970	 loss=3.344105	 lr=0.047240
Epoch[015] Batch [3349]/[3760]	Speed: 117.745504 samples/sec	 accuracy=27.316698	 loss=3.343989	 lr=0.047279
Epoch[015] Batch [3399]/[3760]	Speed: 117.676797 samples/sec	 accuracy=27.313419	 loss=3.344291	 lr=0.047317
Epoch[015] Batch [3449]/[3760]	Speed: 117.410408 samples/sec	 accuracy=27.288043	 loss=3.344610	 lr=0.047356
Epoch[015] Batch [3499]/[3760]	Speed: 117.721189 samples/sec	 accuracy=27.299554	 loss=3.344020	 lr=0.047395
Epoch[015] Batch [3549]/[3760]	Speed: 117.991178 samples/sec	 accuracy=27.297535	 loss=3.343620	 lr=0.047434
Epoch[015] Batch [3599]/[3760]	Speed: 117.554715 samples/sec	 accuracy=27.296875	 loss=3.343837	 lr=0.047472
Epoch[015] Batch [3649]/[3760]	Speed: 117.986199 samples/sec	 accuracy=27.307791	 loss=3.343825	 lr=0.047511
Epoch[015] Batch [3699]/[3760]	Speed: 117.893312 samples/sec	 accuracy=27.301943	 loss=3.344391	 lr=0.047550
Epoch[015] Batch [3749]/[3760]	Speed: 120.865032 samples/sec	 accuracy=27.298333	 loss=3.344978	 lr=0.047589
Batch [0049]/[0303]: acc-top1=27.062500 acc-top5=54.750000
Batch [0099]/[0303]: acc-top1=24.937500 acc-top5=52.609375
Batch [0149]/[0303]: acc-top1=25.791667 acc-top5=53.291667
Batch [0199]/[0303]: acc-top1=28.242188 acc-top5=55.296875
Batch [0249]/[0303]: acc-top1=27.918750 acc-top5=55.125000
Batch [0299]/[0303]: acc-top1=27.718750 acc-top5=55.395833
[Epoch 015] training: accuracy=27.310505	 loss=3.344512
[Epoch 015] speed: 117 samples/sec	time cost: 2176.638997
[Epoch 015] validation: acc-top1=27.856848 acc-top5=55.548680 loss=3.278189
Epoch[016] Batch [0049]/[3760]	Speed: 87.102794 samples/sec	 accuracy=28.062500	 loss=3.282340	 lr=0.047635
Epoch[016] Batch [0099]/[3760]	Speed: 117.419967 samples/sec	 accuracy=28.078125	 loss=3.282690	 lr=0.047674
Epoch[016] Batch [0149]/[3760]	Speed: 117.525251 samples/sec	 accuracy=28.229167	 loss=3.279835	 lr=0.047713
Epoch[016] Batch [0199]/[3760]	Speed: 117.074925 samples/sec	 accuracy=27.843750	 loss=3.298086	 lr=0.047751
Epoch[016] Batch [0249]/[3760]	Speed: 118.245625 samples/sec	 accuracy=27.568750	 loss=3.305147	 lr=0.047790
Epoch[016] Batch [0299]/[3760]	Speed: 117.583835 samples/sec	 accuracy=27.588542	 loss=3.307899	 lr=0.047829
Epoch[016] Batch [0349]/[3760]	Speed: 117.073453 samples/sec	 accuracy=27.433036	 loss=3.315575	 lr=0.047867
Epoch[016] Batch [0399]/[3760]	Speed: 116.973517 samples/sec	 accuracy=27.722656	 loss=3.313123	 lr=0.047906
Epoch[016] Batch [0449]/[3760]	Speed: 117.353753 samples/sec	 accuracy=27.795139	 loss=3.315407	 lr=0.047945
Epoch[016] Batch [0499]/[3760]	Speed: 117.274747 samples/sec	 accuracy=27.746875	 loss=3.315397	 lr=0.047984
Epoch[016] Batch [0549]/[3760]	Speed: 117.704127 samples/sec	 accuracy=27.664773	 loss=3.318598	 lr=0.048022
Epoch[016] Batch [0599]/[3760]	Speed: 117.966280 samples/sec	 accuracy=27.729167	 loss=3.317462	 lr=0.048061
Epoch[016] Batch [0649]/[3760]	Speed: 117.513343 samples/sec	 accuracy=27.687500	 loss=3.319970	 lr=0.048100
Epoch[016] Batch [0699]/[3760]	Speed: 117.598224 samples/sec	 accuracy=27.660714	 loss=3.322746	 lr=0.048139
Epoch[016] Batch [0749]/[3760]	Speed: 116.560025 samples/sec	 accuracy=27.720833	 loss=3.322702	 lr=0.048177
Epoch[016] Batch [0799]/[3760]	Speed: 117.200123 samples/sec	 accuracy=27.744141	 loss=3.322027	 lr=0.048216
Epoch[016] Batch [0849]/[3760]	Speed: 117.233484 samples/sec	 accuracy=27.775735	 loss=3.318213	 lr=0.048255
Epoch[016] Batch [0899]/[3760]	Speed: 117.350444 samples/sec	 accuracy=27.824653	 loss=3.319440	 lr=0.048294
Epoch[016] Batch [0949]/[3760]	Speed: 117.386934 samples/sec	 accuracy=27.832237	 loss=3.317464	 lr=0.048332
Epoch[016] Batch [0999]/[3760]	Speed: 117.155836 samples/sec	 accuracy=27.789062	 loss=3.316989	 lr=0.048371
Epoch[016] Batch [1049]/[3760]	Speed: 117.949547 samples/sec	 accuracy=27.848214	 loss=3.316254	 lr=0.048410
Epoch[016] Batch [1099]/[3760]	Speed: 117.858464 samples/sec	 accuracy=27.825284	 loss=3.316583	 lr=0.048448
Epoch[016] Batch [1149]/[3760]	Speed: 118.250430 samples/sec	 accuracy=27.850543	 loss=3.315679	 lr=0.048487
Epoch[016] Batch [1199]/[3760]	Speed: 117.789228 samples/sec	 accuracy=27.843750	 loss=3.315044	 lr=0.048526
Epoch[016] Batch [1249]/[3760]	Speed: 118.377293 samples/sec	 accuracy=27.835000	 loss=3.315457	 lr=0.048565
Epoch[016] Batch [1299]/[3760]	Speed: 117.648768 samples/sec	 accuracy=27.865385	 loss=3.313526	 lr=0.048603
Epoch[016] Batch [1349]/[3760]	Speed: 117.684286 samples/sec	 accuracy=27.825231	 loss=3.316151	 lr=0.048642
Epoch[016] Batch [1399]/[3760]	Speed: 117.200191 samples/sec	 accuracy=27.854911	 loss=3.317145	 lr=0.048681
Epoch[016] Batch [1449]/[3760]	Speed: 117.396201 samples/sec	 accuracy=27.814655	 loss=3.319226	 lr=0.048720
Epoch[016] Batch [1499]/[3760]	Speed: 116.901686 samples/sec	 accuracy=27.806250	 loss=3.319623	 lr=0.048758
Epoch[016] Batch [1549]/[3760]	Speed: 117.126436 samples/sec	 accuracy=27.820565	 loss=3.320186	 lr=0.048797
Epoch[016] Batch [1599]/[3760]	Speed: 117.830265 samples/sec	 accuracy=27.851562	 loss=3.319042	 lr=0.048836
Epoch[016] Batch [1649]/[3760]	Speed: 117.896117 samples/sec	 accuracy=27.821023	 loss=3.319902	 lr=0.048874
Epoch[016] Batch [1699]/[3760]	Speed: 117.342646 samples/sec	 accuracy=27.812500	 loss=3.320996	 lr=0.048913
Epoch[016] Batch [1749]/[3760]	Speed: 117.362479 samples/sec	 accuracy=27.759821	 loss=3.322985	 lr=0.048952
Epoch[016] Batch [1799]/[3760]	Speed: 118.122302 samples/sec	 accuracy=27.768229	 loss=3.322630	 lr=0.048991
Epoch[016] Batch [1849]/[3760]	Speed: 117.715286 samples/sec	 accuracy=27.734797	 loss=3.323253	 lr=0.049029
Epoch[016] Batch [1899]/[3760]	Speed: 117.648235 samples/sec	 accuracy=27.755757	 loss=3.323746	 lr=0.049068
Epoch[016] Batch [1949]/[3760]	Speed: 117.070032 samples/sec	 accuracy=27.757212	 loss=3.323594	 lr=0.049107
Epoch[016] Batch [1999]/[3760]	Speed: 116.783503 samples/sec	 accuracy=27.762500	 loss=3.323854	 lr=0.049146
Epoch[016] Batch [2049]/[3760]	Speed: 116.344470 samples/sec	 accuracy=27.769817	 loss=3.324464	 lr=0.049184
Epoch[016] Batch [2099]/[3760]	Speed: 118.075790 samples/sec	 accuracy=27.769345	 loss=3.324849	 lr=0.049223
Epoch[016] Batch [2149]/[3760]	Speed: 117.786462 samples/sec	 accuracy=27.779070	 loss=3.324716	 lr=0.049262
Epoch[016] Batch [2199]/[3760]	Speed: 117.469954 samples/sec	 accuracy=27.785511	 loss=3.324465	 lr=0.049301
Epoch[016] Batch [2249]/[3760]	Speed: 117.556547 samples/sec	 accuracy=27.765278	 loss=3.325699	 lr=0.049339
Epoch[016] Batch [2299]/[3760]	Speed: 117.718774 samples/sec	 accuracy=27.766984	 loss=3.325758	 lr=0.049378
Epoch[016] Batch [2349]/[3760]	Speed: 117.525939 samples/sec	 accuracy=27.774601	 loss=3.325265	 lr=0.049417
Epoch[016] Batch [2399]/[3760]	Speed: 117.759918 samples/sec	 accuracy=27.785807	 loss=3.325504	 lr=0.049455
Epoch[016] Batch [2449]/[3760]	Speed: 117.367878 samples/sec	 accuracy=27.789541	 loss=3.325183	 lr=0.049494
Epoch[016] Batch [2499]/[3760]	Speed: 117.766096 samples/sec	 accuracy=27.793125	 loss=3.326288	 lr=0.049533
Epoch[016] Batch [2549]/[3760]	Speed: 118.128551 samples/sec	 accuracy=27.810049	 loss=3.326367	 lr=0.049572
Epoch[016] Batch [2599]/[3760]	Speed: 117.415292 samples/sec	 accuracy=27.816106	 loss=3.325580	 lr=0.049610
Epoch[016] Batch [2649]/[3760]	Speed: 117.832995 samples/sec	 accuracy=27.825472	 loss=3.324995	 lr=0.049649
Epoch[016] Batch [2699]/[3760]	Speed: 118.366200 samples/sec	 accuracy=27.830440	 loss=3.325232	 lr=0.049688
Epoch[016] Batch [2749]/[3760]	Speed: 117.597654 samples/sec	 accuracy=27.832955	 loss=3.324819	 lr=0.049727
Epoch[016] Batch [2799]/[3760]	Speed: 118.511533 samples/sec	 accuracy=27.835938	 loss=3.325049	 lr=0.049765
Epoch[016] Batch [2849]/[3760]	Speed: 117.289590 samples/sec	 accuracy=27.819627	 loss=3.326148	 lr=0.049804
Epoch[016] Batch [2899]/[3760]	Speed: 117.994292 samples/sec	 accuracy=27.810884	 loss=3.326089	 lr=0.049843
Epoch[016] Batch [2949]/[3760]	Speed: 117.694293 samples/sec	 accuracy=27.812500	 loss=3.325977	 lr=0.049881
Epoch[016] Batch [2999]/[3760]	Speed: 117.853162 samples/sec	 accuracy=27.818750	 loss=3.325998	 lr=0.049920
Epoch[016] Batch [3049]/[3760]	Speed: 117.408322 samples/sec	 accuracy=27.812500	 loss=3.326320	 lr=0.049959
Epoch[016] Batch [3099]/[3760]	Speed: 117.089525 samples/sec	 accuracy=27.823085	 loss=3.325501	 lr=0.049998
Epoch[016] Batch [3149]/[3760]	Speed: 117.613385 samples/sec	 accuracy=27.836806	 loss=3.324410	 lr=0.050036
Epoch[016] Batch [3199]/[3760]	Speed: 117.708995 samples/sec	 accuracy=27.830078	 loss=3.324145	 lr=0.050075
Epoch[016] Batch [3249]/[3760]	Speed: 117.444953 samples/sec	 accuracy=27.841346	 loss=3.323608	 lr=0.050114
Epoch[016] Batch [3299]/[3760]	Speed: 117.868929 samples/sec	 accuracy=27.857481	 loss=3.323506	 lr=0.050153
Epoch[016] Batch [3349]/[3760]	Speed: 117.863751 samples/sec	 accuracy=27.849347	 loss=3.324206	 lr=0.050191
Epoch[016] Batch [3399]/[3760]	Speed: 117.389296 samples/sec	 accuracy=27.839154	 loss=3.324500	 lr=0.050230
Epoch[016] Batch [3449]/[3760]	Speed: 117.633315 samples/sec	 accuracy=27.839221	 loss=3.324706	 lr=0.050269
Epoch[016] Batch [3499]/[3760]	Speed: 116.799517 samples/sec	 accuracy=27.826339	 loss=3.325258	 lr=0.050308
Epoch[016] Batch [3549]/[3760]	Speed: 117.935912 samples/sec	 accuracy=27.841549	 loss=3.324007	 lr=0.050346
Epoch[016] Batch [3599]/[3760]	Speed: 117.619510 samples/sec	 accuracy=27.846354	 loss=3.324452	 lr=0.050385
Epoch[016] Batch [3649]/[3760]	Speed: 117.108789 samples/sec	 accuracy=27.846318	 loss=3.324643	 lr=0.050424
Epoch[016] Batch [3699]/[3760]	Speed: 116.638913 samples/sec	 accuracy=27.843750	 loss=3.324632	 lr=0.050462
Epoch[016] Batch [3749]/[3760]	Speed: 120.399491 samples/sec	 accuracy=27.841250	 loss=3.324968	 lr=0.050501
Batch [0049]/[0303]: acc-top1=28.718750 acc-top5=56.687500
Batch [0099]/[0303]: acc-top1=26.593750 acc-top5=53.953125
Batch [0149]/[0303]: acc-top1=28.250000 acc-top5=55.260417
Batch [0199]/[0303]: acc-top1=30.625000 acc-top5=57.453125
Batch [0249]/[0303]: acc-top1=30.050000 acc-top5=57.231250
Batch [0299]/[0303]: acc-top1=29.322917 acc-top5=57.109375
[Epoch 016] training: accuracy=27.836602	 loss=3.325063
[Epoch 016] speed: 117 samples/sec	time cost: 2176.771168
[Epoch 016] validation: acc-top1=29.414191 acc-top5=57.173061 loss=3.215431
Epoch[017] Batch [0049]/[3759]	Speed: 88.384379 samples/sec	 accuracy=28.906250	 loss=3.269528	 lr=0.050548
Epoch[017] Batch [0099]/[3759]	Speed: 116.751379 samples/sec	 accuracy=27.921875	 loss=3.325232	 lr=0.050586
Epoch[017] Batch [0149]/[3759]	Speed: 117.565310 samples/sec	 accuracy=28.250000	 loss=3.320267	 lr=0.050625
Epoch[017] Batch [0199]/[3759]	Speed: 118.294189 samples/sec	 accuracy=28.281250	 loss=3.312923	 lr=0.050664
Epoch[017] Batch [0249]/[3759]	Speed: 117.199782 samples/sec	 accuracy=28.268750	 loss=3.314442	 lr=0.050703
Epoch[017] Batch [0299]/[3759]	Speed: 117.698796 samples/sec	 accuracy=28.354167	 loss=3.313958	 lr=0.050741
Epoch[017] Batch [0349]/[3759]	Speed: 117.735637 samples/sec	 accuracy=28.504464	 loss=3.310931	 lr=0.050780
Epoch[017] Batch [0399]/[3759]	Speed: 118.034208 samples/sec	 accuracy=28.449219	 loss=3.309670	 lr=0.050819
Epoch[017] Batch [0449]/[3759]	Speed: 117.375370 samples/sec	 accuracy=28.402778	 loss=3.318626	 lr=0.050857
Epoch[017] Batch [0499]/[3759]	Speed: 117.976860 samples/sec	 accuracy=28.565625	 loss=3.311516	 lr=0.050896
Epoch[017] Batch [0549]/[3759]	Speed: 118.059277 samples/sec	 accuracy=28.534091	 loss=3.306607	 lr=0.050935
Epoch[017] Batch [0599]/[3759]	Speed: 117.145206 samples/sec	 accuracy=28.419271	 loss=3.310800	 lr=0.050974
Epoch[017] Batch [0649]/[3759]	Speed: 117.219327 samples/sec	 accuracy=28.487981	 loss=3.306665	 lr=0.051012
Epoch[017] Batch [0699]/[3759]	Speed: 117.644959 samples/sec	 accuracy=28.408482	 loss=3.308008	 lr=0.051051
Epoch[017] Batch [0749]/[3759]	Speed: 117.225343 samples/sec	 accuracy=28.391667	 loss=3.308300	 lr=0.051090
Epoch[017] Batch [0799]/[3759]	Speed: 118.271137 samples/sec	 accuracy=28.398438	 loss=3.308209	 lr=0.051129
Epoch[017] Batch [0849]/[3759]	Speed: 117.540438 samples/sec	 accuracy=28.308824	 loss=3.311841	 lr=0.051167
Epoch[017] Batch [0899]/[3759]	Speed: 118.097907 samples/sec	 accuracy=28.281250	 loss=3.314919	 lr=0.051206
Epoch[017] Batch [0949]/[3759]	Speed: 118.007504 samples/sec	 accuracy=28.271382	 loss=3.313889	 lr=0.051245
Epoch[017] Batch [0999]/[3759]	Speed: 117.008680 samples/sec	 accuracy=28.292187	 loss=3.312149	 lr=0.051284
Epoch[017] Batch [1049]/[3759]	Speed: 117.386629 samples/sec	 accuracy=28.247024	 loss=3.313918	 lr=0.051322
Epoch[017] Batch [1099]/[3759]	Speed: 118.092754 samples/sec	 accuracy=28.257102	 loss=3.312994	 lr=0.051361
Epoch[017] Batch [1149]/[3759]	Speed: 118.341016 samples/sec	 accuracy=28.248641	 loss=3.311673	 lr=0.051400
Epoch[017] Batch [1199]/[3759]	Speed: 117.239191 samples/sec	 accuracy=28.236979	 loss=3.311439	 lr=0.051438
Epoch[017] Batch [1249]/[3759]	Speed: 117.436316 samples/sec	 accuracy=28.211250	 loss=3.311634	 lr=0.051477
Epoch[017] Batch [1299]/[3759]	Speed: 117.466721 samples/sec	 accuracy=28.179087	 loss=3.312044	 lr=0.051516
Epoch[017] Batch [1349]/[3759]	Speed: 116.405956 samples/sec	 accuracy=28.189815	 loss=3.313034	 lr=0.051555
Epoch[017] Batch [1399]/[3759]	Speed: 117.814994 samples/sec	 accuracy=28.189732	 loss=3.310887	 lr=0.051593
Epoch[017] Batch [1449]/[3759]	Speed: 117.363862 samples/sec	 accuracy=28.175647	 loss=3.311304	 lr=0.051632
Epoch[017] Batch [1499]/[3759]	Speed: 117.365931 samples/sec	 accuracy=28.177083	 loss=3.313170	 lr=0.051671
Epoch[017] Batch [1549]/[3759]	Speed: 117.138412 samples/sec	 accuracy=28.186492	 loss=3.313865	 lr=0.051710
Epoch[017] Batch [1599]/[3759]	Speed: 117.319611 samples/sec	 accuracy=28.183594	 loss=3.314650	 lr=0.051748
Epoch[017] Batch [1649]/[3759]	Speed: 117.467969 samples/sec	 accuracy=28.162879	 loss=3.315258	 lr=0.051787
Epoch[017] Batch [1699]/[3759]	Speed: 117.386820 samples/sec	 accuracy=28.148897	 loss=3.316454	 lr=0.051826
Epoch[017] Batch [1749]/[3759]	Speed: 117.215700 samples/sec	 accuracy=28.175000	 loss=3.315096	 lr=0.051864
Epoch[017] Batch [1799]/[3759]	Speed: 117.064058 samples/sec	 accuracy=28.202257	 loss=3.315282	 lr=0.051903
Epoch[017] Batch [1849]/[3759]	Speed: 117.378556 samples/sec	 accuracy=28.179899	 loss=3.317140	 lr=0.051942
Epoch[017] Batch [1899]/[3759]	Speed: 117.119995 samples/sec	 accuracy=28.150493	 loss=3.316611	 lr=0.051981
Epoch[017] Batch [1949]/[3759]	Speed: 117.329354 samples/sec	 accuracy=28.164263	 loss=3.316143	 lr=0.052019
Epoch[017] Batch [1999]/[3759]	Speed: 116.884758 samples/sec	 accuracy=28.168750	 loss=3.315598	 lr=0.052058
Epoch[017] Batch [2049]/[3759]	Speed: 117.373973 samples/sec	 accuracy=28.158537	 loss=3.316100	 lr=0.052097
Epoch[017] Batch [2099]/[3759]	Speed: 117.629780 samples/sec	 accuracy=28.159226	 loss=3.315432	 lr=0.052136
Epoch[017] Batch [2149]/[3759]	Speed: 117.535015 samples/sec	 accuracy=28.155523	 loss=3.315647	 lr=0.052174
Epoch[017] Batch [2199]/[3759]	Speed: 117.680119 samples/sec	 accuracy=28.159091	 loss=3.316429	 lr=0.052213
Epoch[017] Batch [2249]/[3759]	Speed: 117.346323 samples/sec	 accuracy=28.188889	 loss=3.314715	 lr=0.052252
Epoch[017] Batch [2299]/[3759]	Speed: 117.352947 samples/sec	 accuracy=28.169837	 loss=3.314075	 lr=0.052291
Epoch[017] Batch [2349]/[3759]	Speed: 118.155526 samples/sec	 accuracy=28.161569	 loss=3.314365	 lr=0.052329
Epoch[017] Batch [2399]/[3759]	Speed: 117.814683 samples/sec	 accuracy=28.167318	 loss=3.314217	 lr=0.052368
Epoch[017] Batch [2449]/[3759]	Speed: 117.526921 samples/sec	 accuracy=28.186862	 loss=3.313212	 lr=0.052407
Epoch[017] Batch [2499]/[3759]	Speed: 118.525907 samples/sec	 accuracy=28.230000	 loss=3.311604	 lr=0.052445
Epoch[017] Batch [2549]/[3759]	Speed: 117.782044 samples/sec	 accuracy=28.226716	 loss=3.310391	 lr=0.052484
Epoch[017] Batch [2599]/[3759]	Speed: 117.436885 samples/sec	 accuracy=28.215144	 loss=3.310502	 lr=0.052523
Epoch[017] Batch [2649]/[3759]	Speed: 117.552318 samples/sec	 accuracy=28.199292	 loss=3.310449	 lr=0.052562
Epoch[017] Batch [2699]/[3759]	Speed: 116.647246 samples/sec	 accuracy=28.175347	 loss=3.310923	 lr=0.052600
Epoch[017] Batch [2749]/[3759]	Speed: 117.614471 samples/sec	 accuracy=28.185795	 loss=3.309998	 lr=0.052639
Epoch[017] Batch [2799]/[3759]	Speed: 117.560184 samples/sec	 accuracy=28.175223	 loss=3.310181	 lr=0.052678
Epoch[017] Batch [2849]/[3759]	Speed: 118.553218 samples/sec	 accuracy=28.211075	 loss=3.308610	 lr=0.052717
Epoch[017] Batch [2899]/[3759]	Speed: 116.914689 samples/sec	 accuracy=28.183190	 loss=3.309710	 lr=0.052755
Epoch[017] Batch [2949]/[3759]	Speed: 117.798683 samples/sec	 accuracy=28.183792	 loss=3.309570	 lr=0.052794
Epoch[017] Batch [2999]/[3759]	Speed: 118.107128 samples/sec	 accuracy=28.163021	 loss=3.310112	 lr=0.052833
Epoch[017] Batch [3049]/[3759]	Speed: 117.689788 samples/sec	 accuracy=28.160861	 loss=3.310497	 lr=0.052871
Epoch[017] Batch [3099]/[3759]	Speed: 117.521085 samples/sec	 accuracy=28.154234	 loss=3.309947	 lr=0.052910
Epoch[017] Batch [3149]/[3759]	Speed: 117.146297 samples/sec	 accuracy=28.161210	 loss=3.310142	 lr=0.052949
Epoch[017] Batch [3199]/[3759]	Speed: 116.882810 samples/sec	 accuracy=28.154785	 loss=3.310210	 lr=0.052988
Epoch[017] Batch [3249]/[3759]	Speed: 117.362264 samples/sec	 accuracy=28.158173	 loss=3.309084	 lr=0.053026
Epoch[017] Batch [3299]/[3759]	Speed: 117.765803 samples/sec	 accuracy=28.157197	 loss=3.308967	 lr=0.053065
Epoch[017] Batch [3349]/[3759]	Speed: 117.984129 samples/sec	 accuracy=28.161847	 loss=3.308421	 lr=0.053104
Epoch[017] Batch [3399]/[3759]	Speed: 117.760313 samples/sec	 accuracy=28.159467	 loss=3.308541	 lr=0.053143
Epoch[017] Batch [3449]/[3759]	Speed: 118.268143 samples/sec	 accuracy=28.177536	 loss=3.308165	 lr=0.053181
Epoch[017] Batch [3499]/[3759]	Speed: 117.337394 samples/sec	 accuracy=28.185714	 loss=3.307367	 lr=0.053220
Epoch[017] Batch [3549]/[3759]	Speed: 118.590517 samples/sec	 accuracy=28.186180	 loss=3.306882	 lr=0.053259
Epoch[017] Batch [3599]/[3759]	Speed: 116.725036 samples/sec	 accuracy=28.191406	 loss=3.306841	 lr=0.053298
Epoch[017] Batch [3649]/[3759]	Speed: 118.204995 samples/sec	 accuracy=28.178938	 loss=3.307900	 lr=0.053336
Epoch[017] Batch [3699]/[3759]	Speed: 117.137026 samples/sec	 accuracy=28.179899	 loss=3.307503	 lr=0.053375
Epoch[017] Batch [3749]/[3759]	Speed: 120.841534 samples/sec	 accuracy=28.162917	 loss=3.308287	 lr=0.053414
Batch [0049]/[0303]: acc-top1=27.500000 acc-top5=52.906250
Batch [0099]/[0303]: acc-top1=26.312500 acc-top5=51.703125
Batch [0149]/[0303]: acc-top1=27.572917 acc-top5=53.104167
Batch [0199]/[0303]: acc-top1=29.710938 acc-top5=55.359375
Batch [0249]/[0303]: acc-top1=29.275000 acc-top5=55.268750
Batch [0299]/[0303]: acc-top1=28.729167 acc-top5=55.348958
[Epoch 017] training: accuracy=28.166151	 loss=3.308245
[Epoch 017] speed: 117 samples/sec	time cost: 2174.861529
[Epoch 017] validation: acc-top1=28.702558 acc-top5=55.280528 loss=3.343212
Epoch[018] Batch [0049]/[3760]	Speed: 88.787456 samples/sec	 accuracy=28.031250	 loss=3.296257	 lr=0.053459
Epoch[018] Batch [0099]/[3760]	Speed: 116.488649 samples/sec	 accuracy=29.093750	 loss=3.226847	 lr=0.053498
Epoch[018] Batch [0149]/[3760]	Speed: 117.071286 samples/sec	 accuracy=29.312500	 loss=3.254577	 lr=0.053537
Epoch[018] Batch [0199]/[3760]	Speed: 117.663076 samples/sec	 accuracy=28.953125	 loss=3.252585	 lr=0.053576
Epoch[018] Batch [0249]/[3760]	Speed: 117.536885 samples/sec	 accuracy=29.137500	 loss=3.246049	 lr=0.053614
Epoch[018] Batch [0299]/[3760]	Speed: 116.920397 samples/sec	 accuracy=29.015625	 loss=3.259139	 lr=0.053653
Epoch[018] Batch [0349]/[3760]	Speed: 118.491953 samples/sec	 accuracy=28.790179	 loss=3.269701	 lr=0.053692
Epoch[018] Batch [0399]/[3760]	Speed: 117.533556 samples/sec	 accuracy=28.718750	 loss=3.276497	 lr=0.053731
Epoch[018] Batch [0449]/[3760]	Speed: 117.902329 samples/sec	 accuracy=28.666667	 loss=3.279242	 lr=0.053769
Epoch[018] Batch [0499]/[3760]	Speed: 117.607093 samples/sec	 accuracy=28.584375	 loss=3.282436	 lr=0.053808
Epoch[018] Batch [0549]/[3760]	Speed: 117.552052 samples/sec	 accuracy=28.451705	 loss=3.288159	 lr=0.053847
Epoch[018] Batch [0599]/[3760]	Speed: 117.672054 samples/sec	 accuracy=28.437500	 loss=3.290808	 lr=0.053885
Epoch[018] Batch [0649]/[3760]	Speed: 117.333290 samples/sec	 accuracy=28.413462	 loss=3.290312	 lr=0.053924
Epoch[018] Batch [0699]/[3760]	Speed: 117.669464 samples/sec	 accuracy=28.462054	 loss=3.289127	 lr=0.053963
Epoch[018] Batch [0749]/[3760]	Speed: 117.016159 samples/sec	 accuracy=28.445833	 loss=3.289822	 lr=0.054002
Epoch[018] Batch [0799]/[3760]	Speed: 117.680236 samples/sec	 accuracy=28.515625	 loss=3.284553	 lr=0.054040
Epoch[018] Batch [0849]/[3760]	Speed: 117.461306 samples/sec	 accuracy=28.531250	 loss=3.286196	 lr=0.054079
Epoch[018] Batch [0899]/[3760]	Speed: 118.290586 samples/sec	 accuracy=28.539931	 loss=3.286540	 lr=0.054118
Epoch[018] Batch [0949]/[3760]	Speed: 116.967544 samples/sec	 accuracy=28.472039	 loss=3.287248	 lr=0.054157
Epoch[018] Batch [0999]/[3760]	Speed: 117.140711 samples/sec	 accuracy=28.464062	 loss=3.288245	 lr=0.054195
Epoch[018] Batch [1049]/[3760]	Speed: 117.850970 samples/sec	 accuracy=28.480655	 loss=3.290906	 lr=0.054234
Epoch[018] Batch [1099]/[3760]	Speed: 117.789377 samples/sec	 accuracy=28.458807	 loss=3.290896	 lr=0.054273
Epoch[018] Batch [1149]/[3760]	Speed: 117.781588 samples/sec	 accuracy=28.540761	 loss=3.288052	 lr=0.054312
Epoch[018] Batch [1199]/[3760]	Speed: 117.794976 samples/sec	 accuracy=28.516927	 loss=3.289265	 lr=0.054350
Epoch[018] Batch [1249]/[3760]	Speed: 117.111403 samples/sec	 accuracy=28.507500	 loss=3.289470	 lr=0.054389
Epoch[018] Batch [1299]/[3760]	Speed: 117.279342 samples/sec	 accuracy=28.479567	 loss=3.289610	 lr=0.054428
Epoch[018] Batch [1349]/[3760]	Speed: 117.037907 samples/sec	 accuracy=28.495370	 loss=3.288962	 lr=0.054466
Epoch[018] Batch [1399]/[3760]	Speed: 117.578322 samples/sec	 accuracy=28.532366	 loss=3.285918	 lr=0.054505
Epoch[018] Batch [1449]/[3760]	Speed: 116.212355 samples/sec	 accuracy=28.534483	 loss=3.285490	 lr=0.054544
Epoch[018] Batch [1499]/[3760]	Speed: 117.123336 samples/sec	 accuracy=28.537500	 loss=3.285479	 lr=0.054583
Epoch[018] Batch [1549]/[3760]	Speed: 117.767896 samples/sec	 accuracy=28.518145	 loss=3.286912	 lr=0.054621
Epoch[018] Batch [1599]/[3760]	Speed: 117.834715 samples/sec	 accuracy=28.545898	 loss=3.286124	 lr=0.054660
Epoch[018] Batch [1649]/[3760]	Speed: 117.538960 samples/sec	 accuracy=28.535038	 loss=3.287813	 lr=0.054699
Epoch[018] Batch [1699]/[3760]	Speed: 117.083267 samples/sec	 accuracy=28.557904	 loss=3.287497	 lr=0.054738
Epoch[018] Batch [1749]/[3760]	Speed: 118.084633 samples/sec	 accuracy=28.558036	 loss=3.288798	 lr=0.054776
Epoch[018] Batch [1799]/[3760]	Speed: 117.868993 samples/sec	 accuracy=28.551215	 loss=3.286961	 lr=0.054815
Epoch[018] Batch [1849]/[3760]	Speed: 117.994892 samples/sec	 accuracy=28.539696	 loss=3.285576	 lr=0.054854
Epoch[018] Batch [1899]/[3760]	Speed: 117.320376 samples/sec	 accuracy=28.511513	 loss=3.285859	 lr=0.054892
Epoch[018] Batch [1949]/[3760]	Speed: 117.699278 samples/sec	 accuracy=28.505609	 loss=3.287037	 lr=0.054931
Epoch[018] Batch [1999]/[3760]	Speed: 117.355469 samples/sec	 accuracy=28.462500	 loss=3.288578	 lr=0.054970
Epoch[018] Batch [2049]/[3760]	Speed: 118.104587 samples/sec	 accuracy=28.446646	 loss=3.287389	 lr=0.055009
Epoch[018] Batch [2099]/[3760]	Speed: 117.199084 samples/sec	 accuracy=28.450893	 loss=3.286063	 lr=0.055047
Epoch[018] Batch [2149]/[3760]	Speed: 117.962381 samples/sec	 accuracy=28.452035	 loss=3.285894	 lr=0.055086
Epoch[018] Batch [2199]/[3760]	Speed: 117.958004 samples/sec	 accuracy=28.467330	 loss=3.286322	 lr=0.055125
Epoch[018] Batch [2249]/[3760]	Speed: 116.709104 samples/sec	 accuracy=28.465278	 loss=3.287289	 lr=0.055164
Epoch[018] Batch [2299]/[3760]	Speed: 117.485492 samples/sec	 accuracy=28.451087	 loss=3.288439	 lr=0.055202
Epoch[018] Batch [2349]/[3760]	Speed: 116.842331 samples/sec	 accuracy=28.426862	 loss=3.290047	 lr=0.055241
Epoch[018] Batch [2399]/[3760]	Speed: 117.652272 samples/sec	 accuracy=28.424479	 loss=3.290464	 lr=0.055280
Epoch[018] Batch [2449]/[3760]	Speed: 116.988747 samples/sec	 accuracy=28.399872	 loss=3.292142	 lr=0.055319
Epoch[018] Batch [2499]/[3760]	Speed: 118.192830 samples/sec	 accuracy=28.420625	 loss=3.291612	 lr=0.055357
Epoch[018] Batch [2549]/[3760]	Speed: 117.821447 samples/sec	 accuracy=28.413603	 loss=3.292240	 lr=0.055396
Epoch[018] Batch [2599]/[3760]	Speed: 117.637255 samples/sec	 accuracy=28.405048	 loss=3.292569	 lr=0.055435
Epoch[018] Batch [2649]/[3760]	Speed: 117.091357 samples/sec	 accuracy=28.427476	 loss=3.292057	 lr=0.055473
Epoch[018] Batch [2699]/[3760]	Speed: 117.309482 samples/sec	 accuracy=28.410880	 loss=3.292436	 lr=0.055512
Epoch[018] Batch [2749]/[3760]	Speed: 117.351799 samples/sec	 accuracy=28.423864	 loss=3.290548	 lr=0.055551
Epoch[018] Batch [2799]/[3760]	Speed: 117.678032 samples/sec	 accuracy=28.422991	 loss=3.290399	 lr=0.055590
Epoch[018] Batch [2849]/[3760]	Speed: 117.880839 samples/sec	 accuracy=28.423794	 loss=3.289943	 lr=0.055628
Epoch[018] Batch [2899]/[3760]	Speed: 117.218379 samples/sec	 accuracy=28.421875	 loss=3.290373	 lr=0.055667
Epoch[018] Batch [2949]/[3760]	Speed: 117.232584 samples/sec	 accuracy=28.421081	 loss=3.290213	 lr=0.055706
Epoch[018] Batch [2999]/[3760]	Speed: 117.329910 samples/sec	 accuracy=28.429688	 loss=3.290538	 lr=0.055745
Epoch[018] Batch [3049]/[3760]	Speed: 117.959555 samples/sec	 accuracy=28.452357	 loss=3.289737	 lr=0.055783
Epoch[018] Batch [3099]/[3760]	Speed: 117.544763 samples/sec	 accuracy=28.426411	 loss=3.290723	 lr=0.055822
Epoch[018] Batch [3149]/[3760]	Speed: 117.929264 samples/sec	 accuracy=28.415675	 loss=3.290452	 lr=0.055861
Epoch[018] Batch [3199]/[3760]	Speed: 117.820412 samples/sec	 accuracy=28.411133	 loss=3.290099	 lr=0.055899
Epoch[018] Batch [3249]/[3760]	Speed: 117.425164 samples/sec	 accuracy=28.413942	 loss=3.289874	 lr=0.055938
Epoch[018] Batch [3299]/[3760]	Speed: 117.658426 samples/sec	 accuracy=28.394413	 loss=3.290423	 lr=0.055977
Epoch[018] Batch [3349]/[3760]	Speed: 117.490640 samples/sec	 accuracy=28.389459	 loss=3.291369	 lr=0.056016
Epoch[018] Batch [3399]/[3760]	Speed: 117.704798 samples/sec	 accuracy=28.404412	 loss=3.290683	 lr=0.056054
Epoch[018] Batch [3449]/[3760]	Speed: 117.917827 samples/sec	 accuracy=28.422554	 loss=3.290483	 lr=0.056093
Epoch[018] Batch [3499]/[3760]	Speed: 117.630980 samples/sec	 accuracy=28.436607	 loss=3.289875	 lr=0.056132
Epoch[018] Batch [3549]/[3760]	Speed: 118.137284 samples/sec	 accuracy=28.435739	 loss=3.290064	 lr=0.056171
Epoch[018] Batch [3599]/[3760]	Speed: 116.926471 samples/sec	 accuracy=28.440104	 loss=3.290704	 lr=0.056209
Epoch[018] Batch [3649]/[3760]	Speed: 117.015722 samples/sec	 accuracy=28.431507	 loss=3.290511	 lr=0.056248
Epoch[018] Batch [3699]/[3760]	Speed: 117.083800 samples/sec	 accuracy=28.413429	 loss=3.291010	 lr=0.056287
Epoch[018] Batch [3749]/[3760]	Speed: 120.489464 samples/sec	 accuracy=28.438333	 loss=3.290330	 lr=0.056326
Batch [0049]/[0303]: acc-top1=27.625000 acc-top5=54.625000
Batch [0099]/[0303]: acc-top1=25.765625 acc-top5=52.750000
Batch [0149]/[0303]: acc-top1=27.781250 acc-top5=53.916667
Batch [0199]/[0303]: acc-top1=30.093750 acc-top5=56.007812
Batch [0249]/[0303]: acc-top1=29.468750 acc-top5=55.875000
Batch [0299]/[0303]: acc-top1=29.000000 acc-top5=56.239583
[Epoch 018] training: accuracy=28.430436	 loss=3.290447
[Epoch 018] speed: 117 samples/sec	time cost: 2175.731277
[Epoch 018] validation: acc-top1=29.063531 acc-top5=56.301568 loss=3.315158
Epoch[019] Batch [0049]/[3760]	Speed: 88.400361 samples/sec	 accuracy=29.437500	 loss=3.264239	 lr=0.056372
Epoch[019] Batch [0099]/[3760]	Speed: 117.326573 samples/sec	 accuracy=28.828125	 loss=3.288594	 lr=0.056411
Epoch[019] Batch [0149]/[3760]	Speed: 117.065198 samples/sec	 accuracy=29.083333	 loss=3.275201	 lr=0.056449
Epoch[019] Batch [0199]/[3760]	Speed: 117.178538 samples/sec	 accuracy=29.390625	 loss=3.267518	 lr=0.056488
Epoch[019] Batch [0249]/[3760]	Speed: 117.884674 samples/sec	 accuracy=29.287500	 loss=3.269736	 lr=0.056527
Epoch[019] Batch [0299]/[3760]	Speed: 117.525514 samples/sec	 accuracy=29.265625	 loss=3.266623	 lr=0.056566
Epoch[019] Batch [0349]/[3760]	Speed: 117.686638 samples/sec	 accuracy=29.205357	 loss=3.262271	 lr=0.056604
Epoch[019] Batch [0399]/[3760]	Speed: 117.693909 samples/sec	 accuracy=29.140625	 loss=3.260353	 lr=0.056643
Epoch[019] Batch [0449]/[3760]	Speed: 117.137253 samples/sec	 accuracy=29.111111	 loss=3.264434	 lr=0.056682
Epoch[019] Batch [0499]/[3760]	Speed: 117.663224 samples/sec	 accuracy=29.175000	 loss=3.259531	 lr=0.056721
Epoch[019] Batch [0549]/[3760]	Speed: 116.898933 samples/sec	 accuracy=29.196023	 loss=3.258681	 lr=0.056759
Epoch[019] Batch [0599]/[3760]	Speed: 117.450011 samples/sec	 accuracy=29.117188	 loss=3.260274	 lr=0.056798
Epoch[019] Batch [0649]/[3760]	Speed: 117.470717 samples/sec	 accuracy=29.122596	 loss=3.261112	 lr=0.056837
Epoch[019] Batch [0699]/[3760]	Speed: 117.812821 samples/sec	 accuracy=28.995536	 loss=3.264840	 lr=0.056875
Epoch[019] Batch [0749]/[3760]	Speed: 116.856244 samples/sec	 accuracy=28.904167	 loss=3.263396	 lr=0.056914
Epoch[019] Batch [0799]/[3760]	Speed: 116.785533 samples/sec	 accuracy=28.888672	 loss=3.264660	 lr=0.056953
Epoch[019] Batch [0849]/[3760]	Speed: 117.471948 samples/sec	 accuracy=28.900735	 loss=3.265385	 lr=0.056992
Epoch[019] Batch [0899]/[3760]	Speed: 117.080034 samples/sec	 accuracy=28.925347	 loss=3.261621	 lr=0.057030
Epoch[019] Batch [0949]/[3760]	Speed: 116.965637 samples/sec	 accuracy=28.847039	 loss=3.266166	 lr=0.057069
Epoch[019] Batch [0999]/[3760]	Speed: 117.923604 samples/sec	 accuracy=28.762500	 loss=3.270913	 lr=0.057108
Epoch[019] Batch [1049]/[3760]	Speed: 117.918784 samples/sec	 accuracy=28.889881	 loss=3.265692	 lr=0.057147
Epoch[019] Batch [1099]/[3760]	Speed: 117.637656 samples/sec	 accuracy=28.940341	 loss=3.262710	 lr=0.057185
Epoch[019] Batch [1149]/[3760]	Speed: 117.894828 samples/sec	 accuracy=28.899457	 loss=3.265334	 lr=0.057224
Epoch[019] Batch [1199]/[3760]	Speed: 117.268019 samples/sec	 accuracy=28.932292	 loss=3.264795	 lr=0.057263
Epoch[019] Batch [1249]/[3760]	Speed: 117.507282 samples/sec	 accuracy=28.870000	 loss=3.264875	 lr=0.057302
Epoch[019] Batch [1299]/[3760]	Speed: 117.165011 samples/sec	 accuracy=28.802885	 loss=3.266596	 lr=0.057340
Epoch[019] Batch [1349]/[3760]	Speed: 117.225586 samples/sec	 accuracy=28.840278	 loss=3.267254	 lr=0.057379
Epoch[019] Batch [1399]/[3760]	Speed: 117.856500 samples/sec	 accuracy=28.868304	 loss=3.266078	 lr=0.057418
Epoch[019] Batch [1449]/[3760]	Speed: 116.846558 samples/sec	 accuracy=28.860991	 loss=3.267632	 lr=0.057456
Epoch[019] Batch [1499]/[3760]	Speed: 117.354463 samples/sec	 accuracy=28.866667	 loss=3.268131	 lr=0.057495
Epoch[019] Batch [1549]/[3760]	Speed: 117.562535 samples/sec	 accuracy=28.864919	 loss=3.268762	 lr=0.057534
Epoch[019] Batch [1599]/[3760]	Speed: 117.182373 samples/sec	 accuracy=28.848633	 loss=3.269131	 lr=0.057573
Epoch[019] Batch [1649]/[3760]	Speed: 117.210474 samples/sec	 accuracy=28.850379	 loss=3.267730	 lr=0.057611
Epoch[019] Batch [1699]/[3760]	Speed: 117.748191 samples/sec	 accuracy=28.858456	 loss=3.266222	 lr=0.057650
Epoch[019] Batch [1749]/[3760]	Speed: 118.299896 samples/sec	 accuracy=28.875893	 loss=3.265578	 lr=0.057689
Epoch[019] Batch [1799]/[3760]	Speed: 117.782084 samples/sec	 accuracy=28.860243	 loss=3.267680	 lr=0.057728
Epoch[019] Batch [1849]/[3760]	Speed: 117.413421 samples/sec	 accuracy=28.907939	 loss=3.266624	 lr=0.057766
Epoch[019] Batch [1899]/[3760]	Speed: 117.233426 samples/sec	 accuracy=28.859375	 loss=3.268104	 lr=0.057805
Epoch[019] Batch [1949]/[3760]	Speed: 117.621333 samples/sec	 accuracy=28.883013	 loss=3.267504	 lr=0.057844
Epoch[019] Batch [1999]/[3760]	Speed: 118.210817 samples/sec	 accuracy=28.870313	 loss=3.267643	 lr=0.057882
Epoch[019] Batch [2049]/[3760]	Speed: 116.965026 samples/sec	 accuracy=28.879573	 loss=3.267008	 lr=0.057921
Epoch[019] Batch [2099]/[3760]	Speed: 118.213362 samples/sec	 accuracy=28.878720	 loss=3.266991	 lr=0.057960
Epoch[019] Batch [2149]/[3760]	Speed: 117.675465 samples/sec	 accuracy=28.863372	 loss=3.267539	 lr=0.057999
Epoch[019] Batch [2199]/[3760]	Speed: 117.928122 samples/sec	 accuracy=28.828125	 loss=3.267888	 lr=0.058037
Epoch[019] Batch [2249]/[3760]	Speed: 117.258534 samples/sec	 accuracy=28.842361	 loss=3.268618	 lr=0.058076
Epoch[019] Batch [2299]/[3760]	Speed: 117.285149 samples/sec	 accuracy=28.845109	 loss=3.268526	 lr=0.058115
Epoch[019] Batch [2349]/[3760]	Speed: 117.676352 samples/sec	 accuracy=28.845745	 loss=3.268999	 lr=0.058154
Epoch[019] Batch [2399]/[3760]	Speed: 117.297101 samples/sec	 accuracy=28.856771	 loss=3.268850	 lr=0.058192
Epoch[019] Batch [2449]/[3760]	Speed: 117.452753 samples/sec	 accuracy=28.853954	 loss=3.269777	 lr=0.058231
Epoch[019] Batch [2499]/[3760]	Speed: 118.541324 samples/sec	 accuracy=28.856250	 loss=3.270841	 lr=0.058270
Epoch[019] Batch [2549]/[3760]	Speed: 117.753162 samples/sec	 accuracy=28.863971	 loss=3.269283	 lr=0.058309
Epoch[019] Batch [2599]/[3760]	Speed: 117.250012 samples/sec	 accuracy=28.879207	 loss=3.268745	 lr=0.058347
Epoch[019] Batch [2649]/[3760]	Speed: 117.969313 samples/sec	 accuracy=28.864976	 loss=3.269186	 lr=0.058386
Epoch[019] Batch [2699]/[3760]	Speed: 117.789857 samples/sec	 accuracy=28.843750	 loss=3.269952	 lr=0.058425
Epoch[019] Batch [2749]/[3760]	Speed: 117.852387 samples/sec	 accuracy=28.841477	 loss=3.269788	 lr=0.058463
Epoch[019] Batch [2799]/[3760]	Speed: 117.672496 samples/sec	 accuracy=28.821429	 loss=3.270194	 lr=0.058502
Epoch[019] Batch [2849]/[3760]	Speed: 117.308312 samples/sec	 accuracy=28.826206	 loss=3.270067	 lr=0.058541
Epoch[019] Batch [2899]/[3760]	Speed: 117.636349 samples/sec	 accuracy=28.806034	 loss=3.270624	 lr=0.058580
Epoch[019] Batch [2949]/[3760]	Speed: 117.130255 samples/sec	 accuracy=28.797140	 loss=3.271845	 lr=0.058618
Epoch[019] Batch [2999]/[3760]	Speed: 117.834461 samples/sec	 accuracy=28.784375	 loss=3.272986	 lr=0.058657
Epoch[019] Batch [3049]/[3760]	Speed: 117.716877 samples/sec	 accuracy=28.803279	 loss=3.272234	 lr=0.058696
Epoch[019] Batch [3099]/[3760]	Speed: 116.806894 samples/sec	 accuracy=28.788306	 loss=3.272090	 lr=0.058735
Epoch[019] Batch [3149]/[3760]	Speed: 117.087215 samples/sec	 accuracy=28.780258	 loss=3.272429	 lr=0.058773
Epoch[019] Batch [3199]/[3760]	Speed: 117.309183 samples/sec	 accuracy=28.792969	 loss=3.272074	 lr=0.058812
Epoch[019] Batch [3249]/[3760]	Speed: 117.595190 samples/sec	 accuracy=28.803846	 loss=3.271904	 lr=0.058851
Epoch[019] Batch [3299]/[3760]	Speed: 117.160508 samples/sec	 accuracy=28.785038	 loss=3.272938	 lr=0.058889
Epoch[019] Batch [3349]/[3760]	Speed: 117.821111 samples/sec	 accuracy=28.787313	 loss=3.273057	 lr=0.058928
Epoch[019] Batch [3399]/[3760]	Speed: 117.772872 samples/sec	 accuracy=28.797794	 loss=3.272860	 lr=0.058967
Epoch[019] Batch [3449]/[3760]	Speed: 117.850372 samples/sec	 accuracy=28.798913	 loss=3.272584	 lr=0.059006
Epoch[019] Batch [3499]/[3760]	Speed: 117.583351 samples/sec	 accuracy=28.803571	 loss=3.272230	 lr=0.059044
Epoch[019] Batch [3549]/[3760]	Speed: 117.655103 samples/sec	 accuracy=28.810299	 loss=3.272828	 lr=0.059083
Epoch[019] Batch [3599]/[3760]	Speed: 117.633734 samples/sec	 accuracy=28.813802	 loss=3.272378	 lr=0.059122
Epoch[019] Batch [3649]/[3760]	Speed: 117.639617 samples/sec	 accuracy=28.815068	 loss=3.272512	 lr=0.059161
Epoch[019] Batch [3699]/[3760]	Speed: 117.991877 samples/sec	 accuracy=28.804054	 loss=3.273349	 lr=0.059199
Epoch[019] Batch [3749]/[3760]	Speed: 120.878451 samples/sec	 accuracy=28.805833	 loss=3.273990	 lr=0.059238
Batch [0049]/[0303]: acc-top1=28.000000 acc-top5=54.656250
Batch [0099]/[0303]: acc-top1=27.093750 acc-top5=52.859375
Batch [0149]/[0303]: acc-top1=27.937500 acc-top5=54.750000
Batch [0199]/[0303]: acc-top1=30.882812 acc-top5=57.132812
Batch [0249]/[0303]: acc-top1=30.450000 acc-top5=57.231250
Batch [0299]/[0303]: acc-top1=29.906250 acc-top5=57.239583
[Epoch 019] training: accuracy=28.799867	 loss=3.274255
[Epoch 019] speed: 117 samples/sec	time cost: 2176.280480
[Epoch 019] validation: acc-top1=29.935025 acc-top5=57.245256 loss=3.227044
Epoch[020] Batch [0049]/[3759]	Speed: 86.728801 samples/sec	 accuracy=29.625000	 loss=3.168458	 lr=0.059285
Epoch[020] Batch [0099]/[3759]	Speed: 117.666564 samples/sec	 accuracy=29.390625	 loss=3.198060	 lr=0.059323
Epoch[020] Batch [0149]/[3759]	Speed: 117.870307 samples/sec	 accuracy=29.010417	 loss=3.198979	 lr=0.059362
Epoch[020] Batch [0199]/[3759]	Speed: 118.024923 samples/sec	 accuracy=28.929688	 loss=3.212035	 lr=0.059401
Epoch[020] Batch [0249]/[3759]	Speed: 117.869592 samples/sec	 accuracy=29.143750	 loss=3.213904	 lr=0.059439
Epoch[020] Batch [0299]/[3759]	Speed: 118.010136 samples/sec	 accuracy=28.916667	 loss=3.226053	 lr=0.059478
Epoch[020] Batch [0349]/[3759]	Speed: 117.781650 samples/sec	 accuracy=28.776786	 loss=3.237343	 lr=0.059517
Epoch[020] Batch [0399]/[3759]	Speed: 117.009375 samples/sec	 accuracy=28.679688	 loss=3.243799	 lr=0.059556
Epoch[020] Batch [0449]/[3759]	Speed: 117.296284 samples/sec	 accuracy=28.809028	 loss=3.239820	 lr=0.059594
Epoch[020] Batch [0499]/[3759]	Speed: 118.233817 samples/sec	 accuracy=28.765625	 loss=3.246742	 lr=0.059633
Epoch[020] Batch [0549]/[3759]	Speed: 118.074503 samples/sec	 accuracy=28.721591	 loss=3.249752	 lr=0.059672
Epoch[020] Batch [0599]/[3759]	Speed: 117.953044 samples/sec	 accuracy=28.734375	 loss=3.255390	 lr=0.059711
Epoch[020] Batch [0649]/[3759]	Speed: 117.761730 samples/sec	 accuracy=28.740385	 loss=3.255628	 lr=0.059749
Epoch[020] Batch [0699]/[3759]	Speed: 117.797563 samples/sec	 accuracy=28.716518	 loss=3.259267	 lr=0.059788
Epoch[020] Batch [0749]/[3759]	Speed: 118.081303 samples/sec	 accuracy=28.785417	 loss=3.257124	 lr=0.059827
Epoch[020] Batch [0799]/[3759]	Speed: 117.868976 samples/sec	 accuracy=28.839844	 loss=3.254043	 lr=0.059866
Epoch[020] Batch [0849]/[3759]	Speed: 116.740948 samples/sec	 accuracy=28.858456	 loss=3.250822	 lr=0.059904
Epoch[020] Batch [0899]/[3759]	Speed: 117.086436 samples/sec	 accuracy=28.940972	 loss=3.248582	 lr=0.059943
Epoch[020] Batch [0949]/[3759]	Speed: 117.134625 samples/sec	 accuracy=28.932566	 loss=3.248823	 lr=0.059982
Epoch[020] Batch [0999]/[3759]	Speed: 117.190764 samples/sec	 accuracy=28.951563	 loss=3.250056	 lr=0.060020
Epoch[020] Batch [1049]/[3759]	Speed: 117.302851 samples/sec	 accuracy=28.906250	 loss=3.251640	 lr=0.060059
Epoch[020] Batch [1099]/[3759]	Speed: 117.223012 samples/sec	 accuracy=28.886364	 loss=3.254910	 lr=0.060098
Epoch[020] Batch [1149]/[3759]	Speed: 117.554148 samples/sec	 accuracy=28.933424	 loss=3.253408	 lr=0.060137
Epoch[020] Batch [1199]/[3759]	Speed: 117.027397 samples/sec	 accuracy=28.938802	 loss=3.254069	 lr=0.060175
Epoch[020] Batch [1249]/[3759]	Speed: 117.343540 samples/sec	 accuracy=28.927500	 loss=3.254608	 lr=0.060214
Epoch[020] Batch [1299]/[3759]	Speed: 117.720363 samples/sec	 accuracy=28.901442	 loss=3.256048	 lr=0.060253
Epoch[020] Batch [1349]/[3759]	Speed: 117.784125 samples/sec	 accuracy=28.875000	 loss=3.258056	 lr=0.060292
Epoch[020] Batch [1399]/[3759]	Speed: 118.066214 samples/sec	 accuracy=28.897321	 loss=3.258491	 lr=0.060330
Epoch[020] Batch [1449]/[3759]	Speed: 118.029864 samples/sec	 accuracy=28.906250	 loss=3.259262	 lr=0.060369
Epoch[020] Batch [1499]/[3759]	Speed: 117.328207 samples/sec	 accuracy=28.932292	 loss=3.260008	 lr=0.060408
Epoch[020] Batch [1549]/[3759]	Speed: 117.457311 samples/sec	 accuracy=28.941532	 loss=3.260418	 lr=0.060446
Epoch[020] Batch [1599]/[3759]	Speed: 117.033735 samples/sec	 accuracy=28.879883	 loss=3.262564	 lr=0.060485
Epoch[020] Batch [1649]/[3759]	Speed: 116.646286 samples/sec	 accuracy=28.837121	 loss=3.263178	 lr=0.060524
Epoch[020] Batch [1699]/[3759]	Speed: 117.518092 samples/sec	 accuracy=28.839154	 loss=3.264327	 lr=0.060563
Epoch[020] Batch [1749]/[3759]	Speed: 116.908185 samples/sec	 accuracy=28.837500	 loss=3.264492	 lr=0.060601
Epoch[020] Batch [1799]/[3759]	Speed: 117.494336 samples/sec	 accuracy=28.810764	 loss=3.265475	 lr=0.060640
Epoch[020] Batch [1849]/[3759]	Speed: 116.550702 samples/sec	 accuracy=28.778716	 loss=3.266426	 lr=0.060679
Epoch[020] Batch [1899]/[3759]	Speed: 117.734166 samples/sec	 accuracy=28.800987	 loss=3.266013	 lr=0.060718
Epoch[020] Batch [1949]/[3759]	Speed: 117.435758 samples/sec	 accuracy=28.831731	 loss=3.266288	 lr=0.060756
Epoch[020] Batch [1999]/[3759]	Speed: 116.624959 samples/sec	 accuracy=28.846094	 loss=3.265829	 lr=0.060795
Epoch[020] Batch [2049]/[3759]	Speed: 117.606245 samples/sec	 accuracy=28.870427	 loss=3.266177	 lr=0.060834
Epoch[020] Batch [2099]/[3759]	Speed: 117.509753 samples/sec	 accuracy=28.867560	 loss=3.265971	 lr=0.060873
Epoch[020] Batch [2149]/[3759]	Speed: 118.197768 samples/sec	 accuracy=28.856105	 loss=3.266244	 lr=0.060911
Epoch[020] Batch [2199]/[3759]	Speed: 117.838725 samples/sec	 accuracy=28.857244	 loss=3.266527	 lr=0.060950
Epoch[020] Batch [2249]/[3759]	Speed: 117.741586 samples/sec	 accuracy=28.893056	 loss=3.265812	 lr=0.060989
Epoch[020] Batch [2299]/[3759]	Speed: 116.864686 samples/sec	 accuracy=28.896739	 loss=3.266052	 lr=0.061027
Epoch[020] Batch [2349]/[3759]	Speed: 117.365694 samples/sec	 accuracy=28.908245	 loss=3.266031	 lr=0.061066
Epoch[020] Batch [2399]/[3759]	Speed: 117.524990 samples/sec	 accuracy=28.903646	 loss=3.266745	 lr=0.061105
Epoch[020] Batch [2449]/[3759]	Speed: 118.261477 samples/sec	 accuracy=28.917092	 loss=3.266207	 lr=0.061144
Epoch[020] Batch [2499]/[3759]	Speed: 117.325272 samples/sec	 accuracy=28.928125	 loss=3.266023	 lr=0.061182
Epoch[020] Batch [2549]/[3759]	Speed: 117.326055 samples/sec	 accuracy=28.958946	 loss=3.265544	 lr=0.061221
Epoch[020] Batch [2599]/[3759]	Speed: 117.099233 samples/sec	 accuracy=28.927885	 loss=3.266211	 lr=0.061260
Epoch[020] Batch [2649]/[3759]	Speed: 118.049031 samples/sec	 accuracy=28.943396	 loss=3.265229	 lr=0.061299
Epoch[020] Batch [2699]/[3759]	Speed: 116.843077 samples/sec	 accuracy=28.929977	 loss=3.266082	 lr=0.061337
Epoch[020] Batch [2749]/[3759]	Speed: 117.332816 samples/sec	 accuracy=28.950568	 loss=3.265562	 lr=0.061376
Epoch[020] Batch [2799]/[3759]	Speed: 117.448942 samples/sec	 accuracy=28.933036	 loss=3.266256	 lr=0.061415
Epoch[020] Batch [2849]/[3759]	Speed: 117.014612 samples/sec	 accuracy=28.945724	 loss=3.265853	 lr=0.061453
Epoch[020] Batch [2899]/[3759]	Speed: 117.565584 samples/sec	 accuracy=28.943966	 loss=3.265597	 lr=0.061492
Epoch[020] Batch [2949]/[3759]	Speed: 117.282453 samples/sec	 accuracy=28.941208	 loss=3.265753	 lr=0.061531
Epoch[020] Batch [2999]/[3759]	Speed: 117.430395 samples/sec	 accuracy=28.943229	 loss=3.265068	 lr=0.061570
Epoch[020] Batch [3049]/[3759]	Speed: 117.106558 samples/sec	 accuracy=28.936475	 loss=3.265686	 lr=0.061608
Epoch[020] Batch [3099]/[3759]	Speed: 117.686794 samples/sec	 accuracy=28.935484	 loss=3.265395	 lr=0.061647
Epoch[020] Batch [3149]/[3759]	Speed: 117.674586 samples/sec	 accuracy=28.934524	 loss=3.265741	 lr=0.061686
Epoch[020] Batch [3199]/[3759]	Speed: 117.999301 samples/sec	 accuracy=28.916992	 loss=3.265729	 lr=0.061725
Epoch[020] Batch [3249]/[3759]	Speed: 117.112072 samples/sec	 accuracy=28.932692	 loss=3.265889	 lr=0.061763
Epoch[020] Batch [3299]/[3759]	Speed: 116.936062 samples/sec	 accuracy=28.947917	 loss=3.265718	 lr=0.061802
Epoch[020] Batch [3349]/[3759]	Speed: 117.735020 samples/sec	 accuracy=28.951959	 loss=3.265618	 lr=0.061841
Epoch[020] Batch [3399]/[3759]	Speed: 117.497206 samples/sec	 accuracy=28.946232	 loss=3.265913	 lr=0.061880
Epoch[020] Batch [3449]/[3759]	Speed: 118.182850 samples/sec	 accuracy=28.942935	 loss=3.265994	 lr=0.061918
Epoch[020] Batch [3499]/[3759]	Speed: 117.440490 samples/sec	 accuracy=28.927679	 loss=3.265739	 lr=0.061957
Epoch[020] Batch [3549]/[3759]	Speed: 117.554010 samples/sec	 accuracy=28.944102	 loss=3.265379	 lr=0.061996
Epoch[020] Batch [3599]/[3759]	Speed: 116.942440 samples/sec	 accuracy=28.939670	 loss=3.265532	 lr=0.062034
Epoch[020] Batch [3649]/[3759]	Speed: 116.947648 samples/sec	 accuracy=28.937500	 loss=3.265550	 lr=0.062073
Epoch[020] Batch [3699]/[3759]	Speed: 117.812204 samples/sec	 accuracy=28.945101	 loss=3.265229	 lr=0.062112
Epoch[020] Batch [3749]/[3759]	Speed: 120.875264 samples/sec	 accuracy=28.940000	 loss=3.265834	 lr=0.062151
Batch [0049]/[0303]: acc-top1=26.781250 acc-top5=54.687500
Batch [0099]/[0303]: acc-top1=26.187500 acc-top5=52.750000
Batch [0149]/[0303]: acc-top1=27.708333 acc-top5=54.156250
Batch [0199]/[0303]: acc-top1=29.812500 acc-top5=56.195313
Batch [0249]/[0303]: acc-top1=29.056250 acc-top5=56.275000
Batch [0299]/[0303]: acc-top1=28.645833 acc-top5=56.000000
[Epoch 020] training: accuracy=28.933476	 loss=3.265928
[Epoch 020] speed: 116 samples/sec	time cost: 2176.806476
[Epoch 020] validation: acc-top1=28.620050 acc-top5=55.981848 loss=3.299429
Epoch[021] Batch [0049]/[3760]	Speed: 88.928245 samples/sec	 accuracy=29.437500	 loss=3.207445	 lr=0.062196
Epoch[021] Batch [0099]/[3760]	Speed: 117.028628 samples/sec	 accuracy=29.578125	 loss=3.227711	 lr=0.062235
Epoch[021] Batch [0149]/[3760]	Speed: 116.790292 samples/sec	 accuracy=29.031250	 loss=3.243746	 lr=0.062274
Epoch[021] Batch [0199]/[3760]	Speed: 117.189407 samples/sec	 accuracy=29.343750	 loss=3.237076	 lr=0.062313
Epoch[021] Batch [0249]/[3760]	Speed: 117.637381 samples/sec	 accuracy=29.468750	 loss=3.237012	 lr=0.062351
Epoch[021] Batch [0299]/[3760]	Speed: 117.805236 samples/sec	 accuracy=29.401042	 loss=3.245127	 lr=0.062390
Epoch[021] Batch [0349]/[3760]	Speed: 116.665938 samples/sec	 accuracy=29.285714	 loss=3.247315	 lr=0.062429
Epoch[021] Batch [0399]/[3760]	Speed: 117.832873 samples/sec	 accuracy=29.367188	 loss=3.241144	 lr=0.062467
Epoch[021] Batch [0449]/[3760]	Speed: 117.894491 samples/sec	 accuracy=29.381944	 loss=3.243318	 lr=0.062506
Epoch[021] Batch [0499]/[3760]	Speed: 117.437942 samples/sec	 accuracy=29.428125	 loss=3.241331	 lr=0.062545
Epoch[021] Batch [0549]/[3760]	Speed: 117.611991 samples/sec	 accuracy=29.343750	 loss=3.245957	 lr=0.062584
Epoch[021] Batch [0599]/[3760]	Speed: 117.663902 samples/sec	 accuracy=29.398438	 loss=3.241656	 lr=0.062622
Epoch[021] Batch [0649]/[3760]	Speed: 117.547295 samples/sec	 accuracy=29.406250	 loss=3.243231	 lr=0.062661
Epoch[021] Batch [0699]/[3760]	Speed: 117.512508 samples/sec	 accuracy=29.381696	 loss=3.244570	 lr=0.062700
Epoch[021] Batch [0749]/[3760]	Speed: 116.937468 samples/sec	 accuracy=29.356250	 loss=3.244237	 lr=0.062739
Epoch[021] Batch [0799]/[3760]	Speed: 116.607839 samples/sec	 accuracy=29.353516	 loss=3.243458	 lr=0.062777
Epoch[021] Batch [0849]/[3760]	Speed: 117.326994 samples/sec	 accuracy=29.329044	 loss=3.245389	 lr=0.062816
Epoch[021] Batch [0899]/[3760]	Speed: 117.626861 samples/sec	 accuracy=29.317708	 loss=3.247956	 lr=0.062855
Epoch[021] Batch [0949]/[3760]	Speed: 116.608164 samples/sec	 accuracy=29.366776	 loss=3.247307	 lr=0.062893
Epoch[021] Batch [0999]/[3760]	Speed: 117.988114 samples/sec	 accuracy=29.301562	 loss=3.250063	 lr=0.062932
Epoch[021] Batch [1049]/[3760]	Speed: 116.728206 samples/sec	 accuracy=29.282738	 loss=3.250231	 lr=0.062971
Epoch[021] Batch [1099]/[3760]	Speed: 117.994349 samples/sec	 accuracy=29.269886	 loss=3.250622	 lr=0.063010
Epoch[021] Batch [1149]/[3760]	Speed: 117.414823 samples/sec	 accuracy=29.221467	 loss=3.252079	 lr=0.063048
Epoch[021] Batch [1199]/[3760]	Speed: 118.237542 samples/sec	 accuracy=29.248698	 loss=3.250374	 lr=0.063087
Epoch[021] Batch [1249]/[3760]	Speed: 117.030816 samples/sec	 accuracy=29.302500	 loss=3.247845	 lr=0.063126
Epoch[021] Batch [1299]/[3760]	Speed: 117.041193 samples/sec	 accuracy=29.280048	 loss=3.248572	 lr=0.063165
Epoch[021] Batch [1349]/[3760]	Speed: 117.219010 samples/sec	 accuracy=29.238426	 loss=3.249327	 lr=0.063203
Epoch[021] Batch [1399]/[3760]	Speed: 117.407104 samples/sec	 accuracy=29.226562	 loss=3.249460	 lr=0.063242
Epoch[021] Batch [1449]/[3760]	Speed: 117.213521 samples/sec	 accuracy=29.221983	 loss=3.249463	 lr=0.063281
Epoch[021] Batch [1499]/[3760]	Speed: 117.798686 samples/sec	 accuracy=29.203125	 loss=3.250331	 lr=0.063320
Epoch[021] Batch [1549]/[3760]	Speed: 117.110509 samples/sec	 accuracy=29.208669	 loss=3.250172	 lr=0.063358
Epoch[021] Batch [1599]/[3760]	Speed: 116.923793 samples/sec	 accuracy=29.196289	 loss=3.250348	 lr=0.063397
Epoch[021] Batch [1649]/[3760]	Speed: 117.434523 samples/sec	 accuracy=29.144886	 loss=3.251178	 lr=0.063436
Epoch[021] Batch [1699]/[3760]	Speed: 117.815087 samples/sec	 accuracy=29.126838	 loss=3.252870	 lr=0.063474
Epoch[021] Batch [1749]/[3760]	Speed: 117.592591 samples/sec	 accuracy=29.117857	 loss=3.254252	 lr=0.063513
Epoch[021] Batch [1799]/[3760]	Speed: 118.495521 samples/sec	 accuracy=29.105903	 loss=3.254905	 lr=0.063552
Epoch[021] Batch [1849]/[3760]	Speed: 117.007923 samples/sec	 accuracy=29.108108	 loss=3.255541	 lr=0.063591
Epoch[021] Batch [1899]/[3760]	Speed: 118.427398 samples/sec	 accuracy=29.115132	 loss=3.255993	 lr=0.063629
Epoch[021] Batch [1949]/[3760]	Speed: 116.714526 samples/sec	 accuracy=29.095353	 loss=3.257891	 lr=0.063668
Epoch[021] Batch [1999]/[3760]	Speed: 117.313790 samples/sec	 accuracy=29.066406	 loss=3.258933	 lr=0.063707
Epoch[021] Batch [2049]/[3760]	Speed: 117.614780 samples/sec	 accuracy=29.105183	 loss=3.256765	 lr=0.063746
Epoch[021] Batch [2099]/[3760]	Speed: 117.445503 samples/sec	 accuracy=29.116071	 loss=3.255837	 lr=0.063784
Epoch[021] Batch [2149]/[3760]	Speed: 116.747872 samples/sec	 accuracy=29.116279	 loss=3.255840	 lr=0.063823
Epoch[021] Batch [2199]/[3760]	Speed: 116.997394 samples/sec	 accuracy=29.142045	 loss=3.254754	 lr=0.063862
Epoch[021] Batch [2249]/[3760]	Speed: 118.151603 samples/sec	 accuracy=29.115278	 loss=3.255856	 lr=0.063900
Epoch[021] Batch [2299]/[3760]	Speed: 117.330323 samples/sec	 accuracy=29.135190	 loss=3.255891	 lr=0.063939
Epoch[021] Batch [2349]/[3760]	Speed: 116.716314 samples/sec	 accuracy=29.145612	 loss=3.254858	 lr=0.063978
Epoch[021] Batch [2399]/[3760]	Speed: 117.562391 samples/sec	 accuracy=29.119141	 loss=3.255087	 lr=0.064017
Epoch[021] Batch [2449]/[3760]	Speed: 117.519474 samples/sec	 accuracy=29.102679	 loss=3.256771	 lr=0.064055
Epoch[021] Batch [2499]/[3760]	Speed: 117.499167 samples/sec	 accuracy=29.100000	 loss=3.257743	 lr=0.064094
Epoch[021] Batch [2549]/[3760]	Speed: 118.934129 samples/sec	 accuracy=29.095588	 loss=3.257922	 lr=0.064133
Epoch[021] Batch [2599]/[3760]	Speed: 117.170020 samples/sec	 accuracy=29.100361	 loss=3.257665	 lr=0.064172
Epoch[021] Batch [2649]/[3760]	Speed: 117.149398 samples/sec	 accuracy=29.098467	 loss=3.257438	 lr=0.064210
Epoch[021] Batch [2699]/[3760]	Speed: 117.860566 samples/sec	 accuracy=29.093750	 loss=3.257265	 lr=0.064249
Epoch[021] Batch [2749]/[3760]	Speed: 117.828332 samples/sec	 accuracy=29.090909	 loss=3.257630	 lr=0.064288
Epoch[021] Batch [2799]/[3760]	Speed: 118.164646 samples/sec	 accuracy=29.089844	 loss=3.258383	 lr=0.064327
Epoch[021] Batch [2849]/[3760]	Speed: 117.201992 samples/sec	 accuracy=29.087171	 loss=3.258878	 lr=0.064365
Epoch[021] Batch [2899]/[3760]	Speed: 117.296617 samples/sec	 accuracy=29.077047	 loss=3.260612	 lr=0.064404
Epoch[021] Batch [2949]/[3760]	Speed: 117.368486 samples/sec	 accuracy=29.070975	 loss=3.260408	 lr=0.064443
Epoch[021] Batch [2999]/[3760]	Speed: 117.920598 samples/sec	 accuracy=29.068229	 loss=3.260263	 lr=0.064481
Epoch[021] Batch [3049]/[3760]	Speed: 117.295867 samples/sec	 accuracy=29.074795	 loss=3.260046	 lr=0.064520
Epoch[021] Batch [3099]/[3760]	Speed: 117.804537 samples/sec	 accuracy=29.092742	 loss=3.259632	 lr=0.064559
Epoch[021] Batch [3149]/[3760]	Speed: 117.875573 samples/sec	 accuracy=29.104663	 loss=3.259735	 lr=0.064598
Epoch[021] Batch [3199]/[3760]	Speed: 117.417251 samples/sec	 accuracy=29.086914	 loss=3.260078	 lr=0.064636
Epoch[021] Batch [3249]/[3760]	Speed: 117.213550 samples/sec	 accuracy=29.067788	 loss=3.260668	 lr=0.064675
Epoch[021] Batch [3299]/[3760]	Speed: 117.517166 samples/sec	 accuracy=29.065814	 loss=3.260387	 lr=0.064714
Epoch[021] Batch [3349]/[3760]	Speed: 117.721231 samples/sec	 accuracy=29.062500	 loss=3.259826	 lr=0.064753
Epoch[021] Batch [3399]/[3760]	Speed: 117.415276 samples/sec	 accuracy=29.049173	 loss=3.259672	 lr=0.064791
Epoch[021] Batch [3449]/[3760]	Speed: 117.167997 samples/sec	 accuracy=29.032156	 loss=3.260416	 lr=0.064830
Epoch[021] Batch [3499]/[3760]	Speed: 116.835171 samples/sec	 accuracy=29.028571	 loss=3.259931	 lr=0.064869
Epoch[021] Batch [3549]/[3760]	Speed: 117.614615 samples/sec	 accuracy=29.021567	 loss=3.260139	 lr=0.064907
Epoch[021] Batch [3599]/[3760]	Speed: 117.207838 samples/sec	 accuracy=29.050781	 loss=3.259140	 lr=0.064946
Epoch[021] Batch [3649]/[3760]	Speed: 117.277498 samples/sec	 accuracy=29.046233	 loss=3.259507	 lr=0.064985
Epoch[021] Batch [3699]/[3760]	Speed: 118.155043 samples/sec	 accuracy=29.046875	 loss=3.259889	 lr=0.065024
Epoch[021] Batch [3749]/[3760]	Speed: 120.543606 samples/sec	 accuracy=29.035417	 loss=3.260189	 lr=0.065062
Batch [0049]/[0303]: acc-top1=27.125000 acc-top5=55.406250
Batch [0099]/[0303]: acc-top1=24.609375 acc-top5=51.484375
Batch [0149]/[0303]: acc-top1=26.177083 acc-top5=52.468750
Batch [0199]/[0303]: acc-top1=28.968750 acc-top5=54.945312
Batch [0249]/[0303]: acc-top1=28.825000 acc-top5=55.318750
Batch [0299]/[0303]: acc-top1=28.098958 acc-top5=54.989583
[Epoch 021] training: accuracy=29.041307	 loss=3.260012
[Epoch 021] speed: 116 samples/sec	time cost: 2177.306501
[Epoch 021] validation: acc-top1=28.243606 acc-top5=55.110355 loss=3.416849
Epoch[022] Batch [0049]/[3760]	Speed: 86.547879 samples/sec	 accuracy=30.406250	 loss=3.249706	 lr=0.065109
Epoch[022] Batch [0099]/[3760]	Speed: 117.303328 samples/sec	 accuracy=30.328125	 loss=3.225940	 lr=0.065148
Epoch[022] Batch [0149]/[3760]	Speed: 116.880136 samples/sec	 accuracy=29.989583	 loss=3.236050	 lr=0.065186
Epoch[022] Batch [0199]/[3760]	Speed: 117.171999 samples/sec	 accuracy=29.867187	 loss=3.228094	 lr=0.065225
Epoch[022] Batch [0249]/[3760]	Speed: 118.211578 samples/sec	 accuracy=29.687500	 loss=3.235393	 lr=0.065264
Epoch[022] Batch [0299]/[3760]	Speed: 116.927302 samples/sec	 accuracy=29.677083	 loss=3.232831	 lr=0.065303
Epoch[022] Batch [0349]/[3760]	Speed: 117.300870 samples/sec	 accuracy=29.562500	 loss=3.231063	 lr=0.065341
Epoch[022] Batch [0399]/[3760]	Speed: 117.433377 samples/sec	 accuracy=29.527344	 loss=3.239288	 lr=0.065380
Epoch[022] Batch [0449]/[3760]	Speed: 116.364836 samples/sec	 accuracy=29.628472	 loss=3.239554	 lr=0.065419
Epoch[022] Batch [0499]/[3760]	Speed: 116.811667 samples/sec	 accuracy=29.534375	 loss=3.241043	 lr=0.065457
Epoch[022] Batch [0549]/[3760]	Speed: 117.583022 samples/sec	 accuracy=29.627841	 loss=3.239926	 lr=0.065496
Epoch[022] Batch [0599]/[3760]	Speed: 116.898922 samples/sec	 accuracy=29.609375	 loss=3.238756	 lr=0.065535
Epoch[022] Batch [0649]/[3760]	Speed: 117.597763 samples/sec	 accuracy=29.557692	 loss=3.242492	 lr=0.065574
Epoch[022] Batch [0699]/[3760]	Speed: 118.190000 samples/sec	 accuracy=29.694196	 loss=3.238134	 lr=0.065612
Epoch[022] Batch [0749]/[3760]	Speed: 117.518510 samples/sec	 accuracy=29.681250	 loss=3.238879	 lr=0.065651
Epoch[022] Batch [0799]/[3760]	Speed: 117.345280 samples/sec	 accuracy=29.611328	 loss=3.240686	 lr=0.065690
Epoch[022] Batch [0849]/[3760]	Speed: 116.918710 samples/sec	 accuracy=29.553309	 loss=3.241860	 lr=0.065729
Epoch[022] Batch [0899]/[3760]	Speed: 117.727927 samples/sec	 accuracy=29.449653	 loss=3.243512	 lr=0.065767
Epoch[022] Batch [0949]/[3760]	Speed: 117.639779 samples/sec	 accuracy=29.455592	 loss=3.244588	 lr=0.065806
Epoch[022] Batch [0999]/[3760]	Speed: 117.579097 samples/sec	 accuracy=29.459375	 loss=3.245465	 lr=0.065845
Epoch[022] Batch [1049]/[3760]	Speed: 117.699959 samples/sec	 accuracy=29.494048	 loss=3.245591	 lr=0.065884
Epoch[022] Batch [1099]/[3760]	Speed: 116.880556 samples/sec	 accuracy=29.501420	 loss=3.245205	 lr=0.065922
Epoch[022] Batch [1149]/[3760]	Speed: 117.774395 samples/sec	 accuracy=29.471467	 loss=3.244797	 lr=0.065961
Epoch[022] Batch [1199]/[3760]	Speed: 117.093320 samples/sec	 accuracy=29.438802	 loss=3.246641	 lr=0.066000
Epoch[022] Batch [1249]/[3760]	Speed: 117.495675 samples/sec	 accuracy=29.360000	 loss=3.249276	 lr=0.066038
Epoch[022] Batch [1299]/[3760]	Speed: 117.457454 samples/sec	 accuracy=29.407452	 loss=3.248349	 lr=0.066077
Epoch[022] Batch [1349]/[3760]	Speed: 116.840858 samples/sec	 accuracy=29.427083	 loss=3.248769	 lr=0.066116
Epoch[022] Batch [1399]/[3760]	Speed: 117.572323 samples/sec	 accuracy=29.393973	 loss=3.249486	 lr=0.066155
Epoch[022] Batch [1449]/[3760]	Speed: 117.899948 samples/sec	 accuracy=29.431034	 loss=3.247996	 lr=0.066193
Epoch[022] Batch [1499]/[3760]	Speed: 117.639538 samples/sec	 accuracy=29.440625	 loss=3.247474	 lr=0.066232
Epoch[022] Batch [1549]/[3760]	Speed: 117.614587 samples/sec	 accuracy=29.420363	 loss=3.247935	 lr=0.066271
Epoch[022] Batch [1599]/[3760]	Speed: 117.439942 samples/sec	 accuracy=29.418945	 loss=3.249303	 lr=0.066310
Epoch[022] Batch [1649]/[3760]	Speed: 117.040239 samples/sec	 accuracy=29.384470	 loss=3.250951	 lr=0.066348
Epoch[022] Batch [1699]/[3760]	Speed: 117.682970 samples/sec	 accuracy=29.423713	 loss=3.249492	 lr=0.066387
Epoch[022] Batch [1749]/[3760]	Speed: 116.468178 samples/sec	 accuracy=29.408036	 loss=3.250884	 lr=0.066426
Epoch[022] Batch [1799]/[3760]	Speed: 117.492062 samples/sec	 accuracy=29.389757	 loss=3.251805	 lr=0.066464
Epoch[022] Batch [1849]/[3760]	Speed: 117.386388 samples/sec	 accuracy=29.397804	 loss=3.250699	 lr=0.066503
Epoch[022] Batch [1899]/[3760]	Speed: 116.683469 samples/sec	 accuracy=29.390625	 loss=3.251398	 lr=0.066542
Epoch[022] Batch [1949]/[3760]	Speed: 117.107695 samples/sec	 accuracy=29.383814	 loss=3.251804	 lr=0.066581
Epoch[022] Batch [1999]/[3760]	Speed: 117.504446 samples/sec	 accuracy=29.399219	 loss=3.250760	 lr=0.066619
Epoch[022] Batch [2049]/[3760]	Speed: 117.952471 samples/sec	 accuracy=29.380335	 loss=3.251979	 lr=0.066658
Epoch[022] Batch [2099]/[3760]	Speed: 116.509463 samples/sec	 accuracy=29.351190	 loss=3.252954	 lr=0.066697
Epoch[022] Batch [2149]/[3760]	Speed: 117.437756 samples/sec	 accuracy=29.385901	 loss=3.252434	 lr=0.066736
Epoch[022] Batch [2199]/[3760]	Speed: 118.047905 samples/sec	 accuracy=29.391335	 loss=3.251536	 lr=0.066774
Epoch[022] Batch [2249]/[3760]	Speed: 117.123095 samples/sec	 accuracy=29.367361	 loss=3.251618	 lr=0.066813
Epoch[022] Batch [2299]/[3760]	Speed: 117.574129 samples/sec	 accuracy=29.363451	 loss=3.251962	 lr=0.066852
Epoch[022] Batch [2349]/[3760]	Speed: 117.176062 samples/sec	 accuracy=29.375000	 loss=3.251721	 lr=0.066891
Epoch[022] Batch [2399]/[3760]	Speed: 117.492505 samples/sec	 accuracy=29.362630	 loss=3.252745	 lr=0.066929
Epoch[022] Batch [2449]/[3760]	Speed: 117.421814 samples/sec	 accuracy=29.373087	 loss=3.251194	 lr=0.066968
Epoch[022] Batch [2499]/[3760]	Speed: 117.945228 samples/sec	 accuracy=29.370625	 loss=3.251261	 lr=0.067007
Epoch[022] Batch [2549]/[3760]	Speed: 117.434201 samples/sec	 accuracy=29.367034	 loss=3.251869	 lr=0.067045
Epoch[022] Batch [2599]/[3760]	Speed: 117.207442 samples/sec	 accuracy=29.353966	 loss=3.253974	 lr=0.067084
Epoch[022] Batch [2649]/[3760]	Speed: 117.757727 samples/sec	 accuracy=29.346108	 loss=3.253789	 lr=0.067123
Epoch[022] Batch [2699]/[3760]	Speed: 117.105896 samples/sec	 accuracy=29.336227	 loss=3.254520	 lr=0.067162
Epoch[022] Batch [2749]/[3760]	Speed: 117.302560 samples/sec	 accuracy=29.349432	 loss=3.253444	 lr=0.067200
Epoch[022] Batch [2799]/[3760]	Speed: 117.655788 samples/sec	 accuracy=29.333705	 loss=3.254133	 lr=0.067239
Epoch[022] Batch [2849]/[3760]	Speed: 116.585470 samples/sec	 accuracy=29.317434	 loss=3.254335	 lr=0.067278
Epoch[022] Batch [2899]/[3760]	Speed: 117.465718 samples/sec	 accuracy=29.286099	 loss=3.255754	 lr=0.067317
Epoch[022] Batch [2949]/[3760]	Speed: 117.890505 samples/sec	 accuracy=29.279661	 loss=3.255837	 lr=0.067355
Epoch[022] Batch [2999]/[3760]	Speed: 117.206649 samples/sec	 accuracy=29.271354	 loss=3.255778	 lr=0.067394
Epoch[022] Batch [3049]/[3760]	Speed: 117.580230 samples/sec	 accuracy=29.267930	 loss=3.255993	 lr=0.067433
Epoch[022] Batch [3099]/[3760]	Speed: 117.061555 samples/sec	 accuracy=29.277722	 loss=3.255961	 lr=0.067471
Epoch[022] Batch [3149]/[3760]	Speed: 117.222404 samples/sec	 accuracy=29.269841	 loss=3.256697	 lr=0.067510
Epoch[022] Batch [3199]/[3760]	Speed: 117.395920 samples/sec	 accuracy=29.244629	 loss=3.257170	 lr=0.067549
Epoch[022] Batch [3249]/[3760]	Speed: 117.539807 samples/sec	 accuracy=29.247115	 loss=3.256858	 lr=0.067588
Epoch[022] Batch [3299]/[3760]	Speed: 117.298962 samples/sec	 accuracy=29.261364	 loss=3.256222	 lr=0.067626
Epoch[022] Batch [3349]/[3760]	Speed: 117.653044 samples/sec	 accuracy=29.270056	 loss=3.255485	 lr=0.067665
Epoch[022] Batch [3399]/[3760]	Speed: 117.075979 samples/sec	 accuracy=29.247243	 loss=3.255574	 lr=0.067704
Epoch[022] Batch [3449]/[3760]	Speed: 116.920962 samples/sec	 accuracy=29.244112	 loss=3.255600	 lr=0.067743
Epoch[022] Batch [3499]/[3760]	Speed: 117.433329 samples/sec	 accuracy=29.261161	 loss=3.254693	 lr=0.067781
Epoch[022] Batch [3549]/[3760]	Speed: 117.851316 samples/sec	 accuracy=29.264965	 loss=3.254257	 lr=0.067820
Epoch[022] Batch [3599]/[3760]	Speed: 117.986626 samples/sec	 accuracy=29.260851	 loss=3.254177	 lr=0.067859
Epoch[022] Batch [3649]/[3760]	Speed: 117.936923 samples/sec	 accuracy=29.249144	 loss=3.254790	 lr=0.067898
Epoch[022] Batch [3699]/[3760]	Speed: 117.377800 samples/sec	 accuracy=29.262669	 loss=3.254766	 lr=0.067936
Epoch[022] Batch [3749]/[3760]	Speed: 120.596178 samples/sec	 accuracy=29.279583	 loss=3.254928	 lr=0.067975
Batch [0049]/[0303]: acc-top1=28.125000 acc-top5=55.781250
Batch [0099]/[0303]: acc-top1=27.640625 acc-top5=54.578125
Batch [0149]/[0303]: acc-top1=28.604167 acc-top5=55.166667
Batch [0199]/[0303]: acc-top1=30.164062 acc-top5=55.765625
Batch [0249]/[0303]: acc-top1=29.768750 acc-top5=55.593750
Batch [0299]/[0303]: acc-top1=28.890625 acc-top5=55.406250
[Epoch 022] training: accuracy=29.280668	 loss=3.255220
[Epoch 022] speed: 116 samples/sec	time cost: 2178.031094
[Epoch 022] validation: acc-top1=29.011964 acc-top5=55.517739 loss=3.476618
Epoch[023] Batch [0049]/[3759]	Speed: 86.952307 samples/sec	 accuracy=30.875000	 loss=3.185671	 lr=0.068021
Epoch[023] Batch [0099]/[3759]	Speed: 117.774203 samples/sec	 accuracy=30.281250	 loss=3.198768	 lr=0.068060
Epoch[023] Batch [0149]/[3759]	Speed: 116.965569 samples/sec	 accuracy=29.697917	 loss=3.207026	 lr=0.068099
Epoch[023] Batch [0199]/[3759]	Speed: 117.974863 samples/sec	 accuracy=29.710938	 loss=3.218320	 lr=0.068138
Epoch[023] Batch [0249]/[3759]	Speed: 117.662461 samples/sec	 accuracy=29.693750	 loss=3.213209	 lr=0.068176
Epoch[023] Batch [0299]/[3759]	Speed: 117.492837 samples/sec	 accuracy=29.505208	 loss=3.215667	 lr=0.068215
Epoch[023] Batch [0349]/[3759]	Speed: 117.313027 samples/sec	 accuracy=29.321429	 loss=3.228233	 lr=0.068254
Epoch[023] Batch [0399]/[3759]	Speed: 117.145766 samples/sec	 accuracy=29.464844	 loss=3.226799	 lr=0.068293
Epoch[023] Batch [0449]/[3759]	Speed: 118.206608 samples/sec	 accuracy=29.482639	 loss=3.223977	 lr=0.068331
Epoch[023] Batch [0499]/[3759]	Speed: 117.183247 samples/sec	 accuracy=29.540625	 loss=3.219781	 lr=0.068370
Epoch[023] Batch [0549]/[3759]	Speed: 117.612688 samples/sec	 accuracy=29.491477	 loss=3.224535	 lr=0.068409
Epoch[023] Batch [0599]/[3759]	Speed: 116.983431 samples/sec	 accuracy=29.427083	 loss=3.228746	 lr=0.068448
Epoch[023] Batch [0649]/[3759]	Speed: 117.346866 samples/sec	 accuracy=29.387019	 loss=3.229099	 lr=0.068486
Epoch[023] Batch [0699]/[3759]	Speed: 117.045692 samples/sec	 accuracy=29.419643	 loss=3.230373	 lr=0.068525
Epoch[023] Batch [0749]/[3759]	Speed: 116.885043 samples/sec	 accuracy=29.441667	 loss=3.228389	 lr=0.068564
Epoch[023] Batch [0799]/[3759]	Speed: 117.435514 samples/sec	 accuracy=29.378906	 loss=3.229952	 lr=0.068602
Epoch[023] Batch [0849]/[3759]	Speed: 117.591367 samples/sec	 accuracy=29.435662	 loss=3.228858	 lr=0.068641
Epoch[023] Batch [0899]/[3759]	Speed: 117.122801 samples/sec	 accuracy=29.421875	 loss=3.231704	 lr=0.068680
Epoch[023] Batch [0949]/[3759]	Speed: 117.723531 samples/sec	 accuracy=29.419408	 loss=3.233837	 lr=0.068719
Epoch[023] Batch [0999]/[3759]	Speed: 117.369334 samples/sec	 accuracy=29.409375	 loss=3.235112	 lr=0.068757
Epoch[023] Batch [1049]/[3759]	Speed: 117.152560 samples/sec	 accuracy=29.441964	 loss=3.234148	 lr=0.068796
Epoch[023] Batch [1099]/[3759]	Speed: 117.481385 samples/sec	 accuracy=29.443182	 loss=3.234838	 lr=0.068835
Epoch[023] Batch [1149]/[3759]	Speed: 117.521925 samples/sec	 accuracy=29.418478	 loss=3.237184	 lr=0.068874
Epoch[023] Batch [1199]/[3759]	Speed: 117.286512 samples/sec	 accuracy=29.442708	 loss=3.235690	 lr=0.068912
Epoch[023] Batch [1249]/[3759]	Speed: 116.543865 samples/sec	 accuracy=29.483750	 loss=3.233864	 lr=0.068951
Epoch[023] Batch [1299]/[3759]	Speed: 116.753247 samples/sec	 accuracy=29.427885	 loss=3.236726	 lr=0.068990
Epoch[023] Batch [1349]/[3759]	Speed: 117.847313 samples/sec	 accuracy=29.415509	 loss=3.238443	 lr=0.069028
Epoch[023] Batch [1399]/[3759]	Speed: 116.995300 samples/sec	 accuracy=29.456473	 loss=3.235652	 lr=0.069067
Epoch[023] Batch [1449]/[3759]	Speed: 117.237351 samples/sec	 accuracy=29.495690	 loss=3.234150	 lr=0.069106
Epoch[023] Batch [1499]/[3759]	Speed: 117.499746 samples/sec	 accuracy=29.447917	 loss=3.235736	 lr=0.069145
Epoch[023] Batch [1549]/[3759]	Speed: 117.617372 samples/sec	 accuracy=29.417339	 loss=3.236927	 lr=0.069183
Epoch[023] Batch [1599]/[3759]	Speed: 116.634879 samples/sec	 accuracy=29.388672	 loss=3.237897	 lr=0.069222
Epoch[023] Batch [1649]/[3759]	Speed: 117.028733 samples/sec	 accuracy=29.397727	 loss=3.238121	 lr=0.069261
Epoch[023] Batch [1699]/[3759]	Speed: 116.228203 samples/sec	 accuracy=29.402574	 loss=3.237125	 lr=0.069300
Epoch[023] Batch [1749]/[3759]	Speed: 117.863424 samples/sec	 accuracy=29.400893	 loss=3.236446	 lr=0.069338
Epoch[023] Batch [1799]/[3759]	Speed: 117.410417 samples/sec	 accuracy=29.400174	 loss=3.236661	 lr=0.069377
Epoch[023] Batch [1849]/[3759]	Speed: 117.217310 samples/sec	 accuracy=29.392736	 loss=3.236013	 lr=0.069416
Epoch[023] Batch [1899]/[3759]	Speed: 117.007454 samples/sec	 accuracy=29.383224	 loss=3.236710	 lr=0.069455
Epoch[023] Batch [1949]/[3759]	Speed: 117.093758 samples/sec	 accuracy=29.408654	 loss=3.236309	 lr=0.069493
Epoch[023] Batch [1999]/[3759]	Speed: 117.697285 samples/sec	 accuracy=29.420313	 loss=3.235982	 lr=0.069532
Epoch[023] Batch [2049]/[3759]	Speed: 117.023716 samples/sec	 accuracy=29.423018	 loss=3.236317	 lr=0.069571
Epoch[023] Batch [2099]/[3759]	Speed: 117.451497 samples/sec	 accuracy=29.443452	 loss=3.235519	 lr=0.069609
Epoch[023] Batch [2149]/[3759]	Speed: 117.295049 samples/sec	 accuracy=29.409884	 loss=3.237958	 lr=0.069648
Epoch[023] Batch [2199]/[3759]	Speed: 116.431766 samples/sec	 accuracy=29.402699	 loss=3.238338	 lr=0.069687
Epoch[023] Batch [2249]/[3759]	Speed: 117.687180 samples/sec	 accuracy=29.405556	 loss=3.238136	 lr=0.069726
Epoch[023] Batch [2299]/[3759]	Speed: 116.735165 samples/sec	 accuracy=29.438179	 loss=3.237386	 lr=0.069764
Epoch[023] Batch [2349]/[3759]	Speed: 117.655856 samples/sec	 accuracy=29.437500	 loss=3.238079	 lr=0.069803
Epoch[023] Batch [2399]/[3759]	Speed: 117.668176 samples/sec	 accuracy=29.444010	 loss=3.238072	 lr=0.069842
Epoch[023] Batch [2449]/[3759]	Speed: 117.839816 samples/sec	 accuracy=29.430485	 loss=3.238399	 lr=0.069881
Epoch[023] Batch [2499]/[3759]	Speed: 117.214657 samples/sec	 accuracy=29.405625	 loss=3.239306	 lr=0.069919
Epoch[023] Batch [2549]/[3759]	Speed: 117.361140 samples/sec	 accuracy=29.367034	 loss=3.241303	 lr=0.069958
Epoch[023] Batch [2599]/[3759]	Speed: 117.183257 samples/sec	 accuracy=29.385817	 loss=3.241067	 lr=0.069997
Epoch[023] Batch [2649]/[3759]	Speed: 116.849975 samples/sec	 accuracy=29.373821	 loss=3.241737	 lr=0.070035
Epoch[023] Batch [2699]/[3759]	Speed: 117.908482 samples/sec	 accuracy=29.372106	 loss=3.242277	 lr=0.070074
Epoch[023] Batch [2749]/[3759]	Speed: 116.581068 samples/sec	 accuracy=29.384091	 loss=3.242629	 lr=0.070113
Epoch[023] Batch [2799]/[3759]	Speed: 117.521327 samples/sec	 accuracy=29.395647	 loss=3.242150	 lr=0.070152
Epoch[023] Batch [2849]/[3759]	Speed: 117.045225 samples/sec	 accuracy=29.396382	 loss=3.242005	 lr=0.070190
Epoch[023] Batch [2899]/[3759]	Speed: 117.146117 samples/sec	 accuracy=29.403017	 loss=3.242222	 lr=0.070229
Epoch[023] Batch [2949]/[3759]	Speed: 116.671618 samples/sec	 accuracy=29.400953	 loss=3.242117	 lr=0.070268
Epoch[023] Batch [2999]/[3759]	Speed: 117.666129 samples/sec	 accuracy=29.393750	 loss=3.242572	 lr=0.070307
Epoch[023] Batch [3049]/[3759]	Speed: 117.257661 samples/sec	 accuracy=29.391906	 loss=3.243098	 lr=0.070345
Epoch[023] Batch [3099]/[3759]	Speed: 117.468524 samples/sec	 accuracy=29.366431	 loss=3.243272	 lr=0.070384
Epoch[023] Batch [3149]/[3759]	Speed: 117.205772 samples/sec	 accuracy=29.366567	 loss=3.243760	 lr=0.070423
Epoch[023] Batch [3199]/[3759]	Speed: 116.799893 samples/sec	 accuracy=29.361816	 loss=3.244228	 lr=0.070462
Epoch[023] Batch [3249]/[3759]	Speed: 116.935347 samples/sec	 accuracy=29.358173	 loss=3.244447	 lr=0.070500
Epoch[023] Batch [3299]/[3759]	Speed: 117.339417 samples/sec	 accuracy=29.340436	 loss=3.245598	 lr=0.070539
Epoch[023] Batch [3349]/[3759]	Speed: 117.879501 samples/sec	 accuracy=29.349813	 loss=3.245864	 lr=0.070578
Epoch[023] Batch [3399]/[3759]	Speed: 118.296524 samples/sec	 accuracy=29.354779	 loss=3.246218	 lr=0.070616
Epoch[023] Batch [3449]/[3759]	Speed: 117.439100 samples/sec	 accuracy=29.357790	 loss=3.246062	 lr=0.070655
Epoch[023] Batch [3499]/[3759]	Speed: 117.408094 samples/sec	 accuracy=29.338839	 loss=3.246934	 lr=0.070694
Epoch[023] Batch [3549]/[3759]	Speed: 117.711440 samples/sec	 accuracy=29.351232	 loss=3.247030	 lr=0.070733
Epoch[023] Batch [3599]/[3759]	Speed: 117.232005 samples/sec	 accuracy=29.343316	 loss=3.247109	 lr=0.070771
Epoch[023] Batch [3649]/[3759]	Speed: 117.131220 samples/sec	 accuracy=29.348459	 loss=3.247136	 lr=0.070810
Epoch[023] Batch [3699]/[3759]	Speed: 117.133994 samples/sec	 accuracy=29.335304	 loss=3.247753	 lr=0.070849
Epoch[023] Batch [3749]/[3759]	Speed: 120.295480 samples/sec	 accuracy=29.342500	 loss=3.247883	 lr=0.070888
Batch [0049]/[0303]: acc-top1=31.468750 acc-top5=55.968750
Batch [0099]/[0303]: acc-top1=28.578125 acc-top5=53.953125
Batch [0149]/[0303]: acc-top1=29.895833 acc-top5=54.927083
Batch [0199]/[0303]: acc-top1=31.796875 acc-top5=57.148438
Batch [0249]/[0303]: acc-top1=30.712500 acc-top5=56.681250
Batch [0299]/[0303]: acc-top1=30.479167 acc-top5=57.010417
[Epoch 023] training: accuracy=29.344157	 loss=3.247924
[Epoch 023] speed: 116 samples/sec	time cost: 2179.905723
[Epoch 023] validation: acc-top1=30.404290 acc-top5=56.961634 loss=3.274981
Epoch[024] Batch [0049]/[3760]	Speed: 86.995256 samples/sec	 accuracy=30.343750	 loss=3.193202	 lr=0.070933
Epoch[024] Batch [0099]/[3760]	Speed: 117.323395 samples/sec	 accuracy=30.046875	 loss=3.192946	 lr=0.070972
Epoch[024] Batch [0149]/[3760]	Speed: 117.643449 samples/sec	 accuracy=29.812500	 loss=3.206385	 lr=0.071011
Epoch[024] Batch [0199]/[3760]	Speed: 117.154043 samples/sec	 accuracy=29.875000	 loss=3.212578	 lr=0.071049
Epoch[024] Batch [0249]/[3760]	Speed: 117.822116 samples/sec	 accuracy=30.106250	 loss=3.206889	 lr=0.071088
Epoch[024] Batch [0299]/[3760]	Speed: 117.239589 samples/sec	 accuracy=29.927083	 loss=3.219145	 lr=0.071127
Epoch[024] Batch [0349]/[3760]	Speed: 117.745550 samples/sec	 accuracy=29.906250	 loss=3.223808	 lr=0.071166
Epoch[024] Batch [0399]/[3760]	Speed: 118.160323 samples/sec	 accuracy=30.031250	 loss=3.219773	 lr=0.071204
Epoch[024] Batch [0449]/[3760]	Speed: 117.774651 samples/sec	 accuracy=29.982639	 loss=3.220469	 lr=0.071243
Epoch[024] Batch [0499]/[3760]	Speed: 117.383246 samples/sec	 accuracy=29.790625	 loss=3.227099	 lr=0.071282
Epoch[024] Batch [0549]/[3760]	Speed: 116.944380 samples/sec	 accuracy=29.732955	 loss=3.233038	 lr=0.071321
Epoch[024] Batch [0599]/[3760]	Speed: 118.289963 samples/sec	 accuracy=29.682292	 loss=3.233452	 lr=0.071359
Epoch[024] Batch [0649]/[3760]	Speed: 117.472621 samples/sec	 accuracy=29.627404	 loss=3.235241	 lr=0.071398
Epoch[024] Batch [0699]/[3760]	Speed: 117.773732 samples/sec	 accuracy=29.582589	 loss=3.240264	 lr=0.071437
Epoch[024] Batch [0749]/[3760]	Speed: 116.252152 samples/sec	 accuracy=29.518750	 loss=3.243369	 lr=0.071475
Epoch[024] Batch [0799]/[3760]	Speed: 117.296806 samples/sec	 accuracy=29.488281	 loss=3.245316	 lr=0.071514
Epoch[024] Batch [0849]/[3760]	Speed: 117.913023 samples/sec	 accuracy=29.498162	 loss=3.245396	 lr=0.071553
Epoch[024] Batch [0899]/[3760]	Speed: 117.553574 samples/sec	 accuracy=29.517361	 loss=3.245676	 lr=0.071592
Epoch[024] Batch [0949]/[3760]	Speed: 117.677953 samples/sec	 accuracy=29.544408	 loss=3.246830	 lr=0.071630
Epoch[024] Batch [0999]/[3760]	Speed: 117.319307 samples/sec	 accuracy=29.542188	 loss=3.247419	 lr=0.071669
Epoch[024] Batch [1049]/[3760]	Speed: 117.250594 samples/sec	 accuracy=29.547619	 loss=3.247060	 lr=0.071708
Epoch[024] Batch [1099]/[3760]	Speed: 117.400274 samples/sec	 accuracy=29.511364	 loss=3.248671	 lr=0.071747
Epoch[024] Batch [1149]/[3760]	Speed: 117.725756 samples/sec	 accuracy=29.468750	 loss=3.249362	 lr=0.071785
Epoch[024] Batch [1199]/[3760]	Speed: 116.753888 samples/sec	 accuracy=29.412760	 loss=3.251919	 lr=0.071824
Epoch[024] Batch [1249]/[3760]	Speed: 117.512030 samples/sec	 accuracy=29.367500	 loss=3.255538	 lr=0.071863
Epoch[024] Batch [1299]/[3760]	Speed: 117.511169 samples/sec	 accuracy=29.358173	 loss=3.256884	 lr=0.071902
Epoch[024] Batch [1349]/[3760]	Speed: 118.177073 samples/sec	 accuracy=29.410880	 loss=3.254190	 lr=0.071940
Epoch[024] Batch [1399]/[3760]	Speed: 117.490519 samples/sec	 accuracy=29.414062	 loss=3.252677	 lr=0.071979
Epoch[024] Batch [1449]/[3760]	Speed: 117.519291 samples/sec	 accuracy=29.400862	 loss=3.253317	 lr=0.072018
Epoch[024] Batch [1499]/[3760]	Speed: 117.859221 samples/sec	 accuracy=29.373958	 loss=3.255001	 lr=0.072056
Epoch[024] Batch [1549]/[3760]	Speed: 116.816471 samples/sec	 accuracy=29.394153	 loss=3.253264	 lr=0.072095
Epoch[024] Batch [1599]/[3760]	Speed: 117.396022 samples/sec	 accuracy=29.379883	 loss=3.254959	 lr=0.072134
Epoch[024] Batch [1649]/[3760]	Speed: 117.439718 samples/sec	 accuracy=29.381629	 loss=3.255073	 lr=0.072173
Epoch[024] Batch [1699]/[3760]	Speed: 117.026636 samples/sec	 accuracy=29.349265	 loss=3.256031	 lr=0.072211
Epoch[024] Batch [1749]/[3760]	Speed: 118.179560 samples/sec	 accuracy=29.373214	 loss=3.255018	 lr=0.072250
Epoch[024] Batch [1799]/[3760]	Speed: 117.179829 samples/sec	 accuracy=29.383681	 loss=3.254153	 lr=0.072289
Epoch[024] Batch [1849]/[3760]	Speed: 117.325660 samples/sec	 accuracy=29.433277	 loss=3.252040	 lr=0.072328
Epoch[024] Batch [1899]/[3760]	Speed: 117.834782 samples/sec	 accuracy=29.443257	 loss=3.251883	 lr=0.072366
Epoch[024] Batch [1949]/[3760]	Speed: 117.620220 samples/sec	 accuracy=29.447917	 loss=3.252434	 lr=0.072405
Epoch[024] Batch [1999]/[3760]	Speed: 117.000804 samples/sec	 accuracy=29.439844	 loss=3.250911	 lr=0.072444
Epoch[024] Batch [2049]/[3760]	Speed: 117.230981 samples/sec	 accuracy=29.439787	 loss=3.251109	 lr=0.072482
Epoch[024] Batch [2099]/[3760]	Speed: 117.556554 samples/sec	 accuracy=29.403274	 loss=3.252945	 lr=0.072521
Epoch[024] Batch [2149]/[3760]	Speed: 117.609551 samples/sec	 accuracy=29.419331	 loss=3.252464	 lr=0.072560
Epoch[024] Batch [2199]/[3760]	Speed: 117.708239 samples/sec	 accuracy=29.419744	 loss=3.252040	 lr=0.072599
Epoch[024] Batch [2249]/[3760]	Speed: 117.877767 samples/sec	 accuracy=29.417361	 loss=3.251561	 lr=0.072637
Epoch[024] Batch [2299]/[3760]	Speed: 116.937463 samples/sec	 accuracy=29.422554	 loss=3.251488	 lr=0.072676
Epoch[024] Batch [2349]/[3760]	Speed: 117.067219 samples/sec	 accuracy=29.407580	 loss=3.251303	 lr=0.072715
Epoch[024] Batch [2399]/[3760]	Speed: 117.647771 samples/sec	 accuracy=29.440104	 loss=3.250085	 lr=0.072754
Epoch[024] Batch [2449]/[3760]	Speed: 117.096180 samples/sec	 accuracy=29.419643	 loss=3.250768	 lr=0.072792
Epoch[024] Batch [2499]/[3760]	Speed: 116.903095 samples/sec	 accuracy=29.400625	 loss=3.251571	 lr=0.072831
Epoch[024] Batch [2549]/[3760]	Speed: 117.409889 samples/sec	 accuracy=29.402574	 loss=3.251114	 lr=0.072870
Epoch[024] Batch [2599]/[3760]	Speed: 116.994978 samples/sec	 accuracy=29.443510	 loss=3.250318	 lr=0.072909
Epoch[024] Batch [2649]/[3760]	Speed: 117.008774 samples/sec	 accuracy=29.439269	 loss=3.251023	 lr=0.072947
Epoch[024] Batch [2699]/[3760]	Speed: 117.426548 samples/sec	 accuracy=29.442708	 loss=3.250526	 lr=0.072986
Epoch[024] Batch [2749]/[3760]	Speed: 117.669499 samples/sec	 accuracy=29.433523	 loss=3.250370	 lr=0.073025
Epoch[024] Batch [2799]/[3760]	Speed: 117.894310 samples/sec	 accuracy=29.409598	 loss=3.251748	 lr=0.073063
Epoch[024] Batch [2849]/[3760]	Speed: 117.269273 samples/sec	 accuracy=29.445175	 loss=3.250397	 lr=0.073102
Epoch[024] Batch [2899]/[3760]	Speed: 116.769602 samples/sec	 accuracy=29.449892	 loss=3.250916	 lr=0.073141
Epoch[024] Batch [2949]/[3760]	Speed: 117.697214 samples/sec	 accuracy=29.472987	 loss=3.249335	 lr=0.073180
Epoch[024] Batch [2999]/[3760]	Speed: 117.829799 samples/sec	 accuracy=29.468750	 loss=3.249271	 lr=0.073218
Epoch[024] Batch [3049]/[3760]	Speed: 117.706102 samples/sec	 accuracy=29.448258	 loss=3.250171	 lr=0.073257
Epoch[024] Batch [3099]/[3760]	Speed: 116.919007 samples/sec	 accuracy=29.458669	 loss=3.249813	 lr=0.073296
Epoch[024] Batch [3149]/[3760]	Speed: 118.337010 samples/sec	 accuracy=29.422123	 loss=3.251205	 lr=0.073335
Epoch[024] Batch [3199]/[3760]	Speed: 117.810748 samples/sec	 accuracy=29.441406	 loss=3.250467	 lr=0.073373
Epoch[024] Batch [3249]/[3760]	Speed: 117.518625 samples/sec	 accuracy=29.443269	 loss=3.250190	 lr=0.073412
Epoch[024] Batch [3299]/[3760]	Speed: 117.137907 samples/sec	 accuracy=29.459754	 loss=3.249299	 lr=0.073451
Epoch[024] Batch [3349]/[3760]	Speed: 118.393011 samples/sec	 accuracy=29.447761	 loss=3.249548	 lr=0.073489
Epoch[024] Batch [3399]/[3760]	Speed: 117.669265 samples/sec	 accuracy=29.452206	 loss=3.249236	 lr=0.073528
Epoch[024] Batch [3449]/[3760]	Speed: 117.624517 samples/sec	 accuracy=29.435688	 loss=3.250621	 lr=0.073567
Epoch[024] Batch [3499]/[3760]	Speed: 117.719290 samples/sec	 accuracy=29.437946	 loss=3.250286	 lr=0.073606
Epoch[024] Batch [3549]/[3760]	Speed: 118.207974 samples/sec	 accuracy=29.436620	 loss=3.250745	 lr=0.073644
Epoch[024] Batch [3599]/[3760]	Speed: 117.540251 samples/sec	 accuracy=29.438802	 loss=3.250936	 lr=0.073683
Epoch[024] Batch [3649]/[3760]	Speed: 117.000475 samples/sec	 accuracy=29.433647	 loss=3.251216	 lr=0.073722
Epoch[024] Batch [3699]/[3760]	Speed: 117.112934 samples/sec	 accuracy=29.432432	 loss=3.252093	 lr=0.073761
Epoch[024] Batch [3749]/[3760]	Speed: 120.475075 samples/sec	 accuracy=29.433333	 loss=3.252515	 lr=0.073799
Batch [0049]/[0303]: acc-top1=27.687500 acc-top5=53.718750
Batch [0099]/[0303]: acc-top1=26.953125 acc-top5=53.625000
Batch [0149]/[0303]: acc-top1=27.760417 acc-top5=54.729167
Batch [0199]/[0303]: acc-top1=30.335938 acc-top5=56.515625
Batch [0249]/[0303]: acc-top1=29.962500 acc-top5=56.281250
Batch [0299]/[0303]: acc-top1=29.354167 acc-top5=56.401042
[Epoch 024] training: accuracy=29.441489	 loss=3.252359
[Epoch 024] speed: 116 samples/sec	time cost: 2177.888368
[Epoch 024] validation: acc-top1=29.398721 acc-top5=56.430487 loss=3.349370
Epoch[025] Batch [0049]/[3760]	Speed: 87.652667 samples/sec	 accuracy=31.062500	 loss=3.193035	 lr=0.073846
Epoch[025] Batch [0099]/[3760]	Speed: 117.027972 samples/sec	 accuracy=30.093750	 loss=3.215108	 lr=0.073885
Epoch[025] Batch [0149]/[3760]	Speed: 117.436417 samples/sec	 accuracy=29.843750	 loss=3.222456	 lr=0.073923
Epoch[025] Batch [0199]/[3760]	Speed: 116.619713 samples/sec	 accuracy=29.835938	 loss=3.212817	 lr=0.073962
Epoch[025] Batch [0249]/[3760]	Speed: 117.045176 samples/sec	 accuracy=29.931250	 loss=3.220397	 lr=0.074001
Epoch[025] Batch [0299]/[3760]	Speed: 117.161867 samples/sec	 accuracy=30.046875	 loss=3.208316	 lr=0.074039
Epoch[025] Batch [0349]/[3760]	Speed: 117.999404 samples/sec	 accuracy=30.227679	 loss=3.203290	 lr=0.074078
Epoch[025] Batch [0399]/[3760]	Speed: 116.612150 samples/sec	 accuracy=30.195312	 loss=3.201556	 lr=0.074117
Epoch[025] Batch [0449]/[3760]	Speed: 118.057354 samples/sec	 accuracy=30.072917	 loss=3.203142	 lr=0.074156
Epoch[025] Batch [0499]/[3760]	Speed: 117.115141 samples/sec	 accuracy=30.065625	 loss=3.205813	 lr=0.074194
Epoch[025] Batch [0549]/[3760]	Speed: 117.535522 samples/sec	 accuracy=29.889205	 loss=3.216591	 lr=0.074233
Epoch[025] Batch [0599]/[3760]	Speed: 117.224707 samples/sec	 accuracy=29.880208	 loss=3.219733	 lr=0.074272
Epoch[025] Batch [0649]/[3760]	Speed: 116.440705 samples/sec	 accuracy=29.838942	 loss=3.221330	 lr=0.074311
Epoch[025] Batch [0699]/[3760]	Speed: 117.721441 samples/sec	 accuracy=29.792411	 loss=3.226367	 lr=0.074349
Epoch[025] Batch [0749]/[3760]	Speed: 116.967205 samples/sec	 accuracy=29.675000	 loss=3.234127	 lr=0.074388
Epoch[025] Batch [0799]/[3760]	Speed: 117.936061 samples/sec	 accuracy=29.599609	 loss=3.237621	 lr=0.074427
Epoch[025] Batch [0849]/[3760]	Speed: 117.194039 samples/sec	 accuracy=29.586397	 loss=3.235889	 lr=0.074466
Epoch[025] Batch [0899]/[3760]	Speed: 117.618275 samples/sec	 accuracy=29.585069	 loss=3.237928	 lr=0.074504
Epoch[025] Batch [0949]/[3760]	Speed: 117.314419 samples/sec	 accuracy=29.557566	 loss=3.240013	 lr=0.074543
Epoch[025] Batch [0999]/[3760]	Speed: 117.380467 samples/sec	 accuracy=29.528125	 loss=3.238495	 lr=0.074582
Epoch[025] Batch [1049]/[3760]	Speed: 117.704086 samples/sec	 accuracy=29.546131	 loss=3.237320	 lr=0.074620
Epoch[025] Batch [1099]/[3760]	Speed: 117.259475 samples/sec	 accuracy=29.492898	 loss=3.237680	 lr=0.074659
Epoch[025] Batch [1149]/[3760]	Speed: 117.416122 samples/sec	 accuracy=29.487772	 loss=3.238741	 lr=0.074698
Epoch[025] Batch [1199]/[3760]	Speed: 116.768775 samples/sec	 accuracy=29.473958	 loss=3.238914	 lr=0.074737
Epoch[025] Batch [1249]/[3760]	Speed: 118.084334 samples/sec	 accuracy=29.422500	 loss=3.240338	 lr=0.074775
Epoch[025] Batch [1299]/[3760]	Speed: 117.024735 samples/sec	 accuracy=29.424279	 loss=3.238253	 lr=0.074814
Epoch[025] Batch [1349]/[3760]	Speed: 116.956988 samples/sec	 accuracy=29.405093	 loss=3.239099	 lr=0.074853
Epoch[025] Batch [1399]/[3760]	Speed: 116.445600 samples/sec	 accuracy=29.414062	 loss=3.237665	 lr=0.074892
Epoch[025] Batch [1449]/[3760]	Speed: 118.200958 samples/sec	 accuracy=29.419181	 loss=3.238732	 lr=0.074930
Epoch[025] Batch [1499]/[3760]	Speed: 117.344212 samples/sec	 accuracy=29.396875	 loss=3.238676	 lr=0.074969
Epoch[025] Batch [1549]/[3760]	Speed: 116.967124 samples/sec	 accuracy=29.407258	 loss=3.238909	 lr=0.075008
Epoch[025] Batch [1599]/[3760]	Speed: 117.485419 samples/sec	 accuracy=29.392578	 loss=3.238751	 lr=0.075046
Epoch[025] Batch [1649]/[3760]	Speed: 117.044196 samples/sec	 accuracy=29.368371	 loss=3.241013	 lr=0.075085
Epoch[025] Batch [1699]/[3760]	Speed: 116.902758 samples/sec	 accuracy=29.370404	 loss=3.242138	 lr=0.075124
Epoch[025] Batch [1749]/[3760]	Speed: 117.469869 samples/sec	 accuracy=29.384821	 loss=3.241283	 lr=0.075163
Epoch[025] Batch [1799]/[3760]	Speed: 116.969035 samples/sec	 accuracy=29.389757	 loss=3.240956	 lr=0.075201
Epoch[025] Batch [1849]/[3760]	Speed: 117.551450 samples/sec	 accuracy=29.440034	 loss=3.239996	 lr=0.075240
Epoch[025] Batch [1899]/[3760]	Speed: 116.659332 samples/sec	 accuracy=29.458059	 loss=3.239839	 lr=0.075279
Epoch[025] Batch [1949]/[3760]	Speed: 117.104328 samples/sec	 accuracy=29.479968	 loss=3.240537	 lr=0.075318
Epoch[025] Batch [1999]/[3760]	Speed: 117.349697 samples/sec	 accuracy=29.453125	 loss=3.242236	 lr=0.075356
Epoch[025] Batch [2049]/[3760]	Speed: 117.889626 samples/sec	 accuracy=29.490854	 loss=3.240922	 lr=0.075395
Epoch[025] Batch [2099]/[3760]	Speed: 117.900666 samples/sec	 accuracy=29.493304	 loss=3.240926	 lr=0.075434
Epoch[025] Batch [2149]/[3760]	Speed: 117.244833 samples/sec	 accuracy=29.514535	 loss=3.239738	 lr=0.075473
Epoch[025] Batch [2199]/[3760]	Speed: 117.713359 samples/sec	 accuracy=29.464489	 loss=3.240852	 lr=0.075511
Epoch[025] Batch [2249]/[3760]	Speed: 117.242432 samples/sec	 accuracy=29.430556	 loss=3.241522	 lr=0.075550
Epoch[025] Batch [2299]/[3760]	Speed: 116.985398 samples/sec	 accuracy=29.427989	 loss=3.242286	 lr=0.075589
Epoch[025] Batch [2349]/[3760]	Speed: 117.319810 samples/sec	 accuracy=29.443484	 loss=3.242220	 lr=0.075627
Epoch[025] Batch [2399]/[3760]	Speed: 118.064387 samples/sec	 accuracy=29.421875	 loss=3.243341	 lr=0.075666
Epoch[025] Batch [2449]/[3760]	Speed: 117.812276 samples/sec	 accuracy=29.427296	 loss=3.242718	 lr=0.075705
Epoch[025] Batch [2499]/[3760]	Speed: 118.147677 samples/sec	 accuracy=29.423125	 loss=3.243319	 lr=0.075744
Epoch[025] Batch [2549]/[3760]	Speed: 118.419018 samples/sec	 accuracy=29.423407	 loss=3.243562	 lr=0.075782
Epoch[025] Batch [2599]/[3760]	Speed: 116.785616 samples/sec	 accuracy=29.441707	 loss=3.242683	 lr=0.075821
Epoch[025] Batch [2649]/[3760]	Speed: 116.746205 samples/sec	 accuracy=29.441038	 loss=3.241923	 lr=0.075860
Epoch[025] Batch [2699]/[3760]	Speed: 118.057340 samples/sec	 accuracy=29.447338	 loss=3.242050	 lr=0.075899
Epoch[025] Batch [2749]/[3760]	Speed: 117.720132 samples/sec	 accuracy=29.438068	 loss=3.242923	 lr=0.075937
Epoch[025] Batch [2799]/[3760]	Speed: 117.805614 samples/sec	 accuracy=29.407366	 loss=3.245066	 lr=0.075976
Epoch[025] Batch [2849]/[3760]	Speed: 117.586056 samples/sec	 accuracy=29.391996	 loss=3.245382	 lr=0.076015
Epoch[025] Batch [2899]/[3760]	Speed: 117.141273 samples/sec	 accuracy=29.392241	 loss=3.245447	 lr=0.076053
Epoch[025] Batch [2949]/[3760]	Speed: 117.514663 samples/sec	 accuracy=29.387712	 loss=3.245704	 lr=0.076092
Epoch[025] Batch [2999]/[3760]	Speed: 116.810879 samples/sec	 accuracy=29.394792	 loss=3.245813	 lr=0.076131
Epoch[025] Batch [3049]/[3760]	Speed: 117.062410 samples/sec	 accuracy=29.380123	 loss=3.246911	 lr=0.076170
Epoch[025] Batch [3099]/[3760]	Speed: 116.822737 samples/sec	 accuracy=29.351310	 loss=3.248554	 lr=0.076208
Epoch[025] Batch [3149]/[3760]	Speed: 116.871456 samples/sec	 accuracy=29.339782	 loss=3.248790	 lr=0.076247
Epoch[025] Batch [3199]/[3760]	Speed: 117.342474 samples/sec	 accuracy=29.335449	 loss=3.248835	 lr=0.076286
Epoch[025] Batch [3249]/[3760]	Speed: 117.671937 samples/sec	 accuracy=29.340865	 loss=3.247733	 lr=0.076325
Epoch[025] Batch [3299]/[3760]	Speed: 117.596536 samples/sec	 accuracy=29.319602	 loss=3.247692	 lr=0.076363
Epoch[025] Batch [3349]/[3760]	Speed: 117.207696 samples/sec	 accuracy=29.320429	 loss=3.247914	 lr=0.076402
Epoch[025] Batch [3399]/[3760]	Speed: 117.635007 samples/sec	 accuracy=29.319393	 loss=3.248840	 lr=0.076441
Epoch[025] Batch [3449]/[3760]	Speed: 118.020605 samples/sec	 accuracy=29.283967	 loss=3.250394	 lr=0.076480
Epoch[025] Batch [3499]/[3760]	Speed: 117.348004 samples/sec	 accuracy=29.276339	 loss=3.250736	 lr=0.076518
Epoch[025] Batch [3549]/[3760]	Speed: 117.811729 samples/sec	 accuracy=29.257482	 loss=3.251472	 lr=0.076557
Epoch[025] Batch [3599]/[3760]	Speed: 118.453682 samples/sec	 accuracy=29.251302	 loss=3.251557	 lr=0.076596
Epoch[025] Batch [3649]/[3760]	Speed: 117.383222 samples/sec	 accuracy=29.252140	 loss=3.251966	 lr=0.076634
Epoch[025] Batch [3699]/[3760]	Speed: 117.273044 samples/sec	 accuracy=29.237753	 loss=3.251935	 lr=0.076673
Epoch[025] Batch [3749]/[3760]	Speed: 120.072127 samples/sec	 accuracy=29.244583	 loss=3.252210	 lr=0.076712
Batch [0049]/[0303]: acc-top1=30.281250 acc-top5=57.906250
Batch [0099]/[0303]: acc-top1=28.609375 acc-top5=55.390625
Batch [0149]/[0303]: acc-top1=29.343750 acc-top5=55.395833
Batch [0199]/[0303]: acc-top1=31.265625 acc-top5=56.968750
Batch [0249]/[0303]: acc-top1=30.537500 acc-top5=56.631250
Batch [0299]/[0303]: acc-top1=29.765625 acc-top5=56.421875
[Epoch 025] training: accuracy=29.238281	 loss=3.252568
[Epoch 025] speed: 116 samples/sec	time cost: 2179.541485
[Epoch 025] validation: acc-top1=29.666873 acc-top5=56.327351 loss=3.274901
Epoch[026] Batch [0049]/[3759]	Speed: 86.718207 samples/sec	 accuracy=29.531250	 loss=3.211357	 lr=0.076758
Epoch[026] Batch [0099]/[3759]	Speed: 118.215786 samples/sec	 accuracy=29.921875	 loss=3.188262	 lr=0.076797
Epoch[026] Batch [0149]/[3759]	Speed: 116.711700 samples/sec	 accuracy=29.281250	 loss=3.203169	 lr=0.076836
Epoch[026] Batch [0199]/[3759]	Speed: 117.266432 samples/sec	 accuracy=28.976563	 loss=3.222393	 lr=0.076875
Epoch[026] Batch [0249]/[3759]	Speed: 117.381229 samples/sec	 accuracy=29.262500	 loss=3.221224	 lr=0.076913
Epoch[026] Batch [0299]/[3759]	Speed: 118.441147 samples/sec	 accuracy=29.317708	 loss=3.226183	 lr=0.076952
Epoch[026] Batch [0349]/[3759]	Speed: 117.475650 samples/sec	 accuracy=29.245536	 loss=3.241012	 lr=0.076991
Epoch[026] Batch [0399]/[3759]	Speed: 117.634545 samples/sec	 accuracy=29.160156	 loss=3.242853	 lr=0.077029
Epoch[026] Batch [0449]/[3759]	Speed: 117.278100 samples/sec	 accuracy=29.385417	 loss=3.232996	 lr=0.077068
Epoch[026] Batch [0499]/[3759]	Speed: 117.231552 samples/sec	 accuracy=29.400000	 loss=3.239686	 lr=0.077107
Epoch[026] Batch [0549]/[3759]	Speed: 117.865048 samples/sec	 accuracy=29.392045	 loss=3.238556	 lr=0.077146
Epoch[026] Batch [0599]/[3759]	Speed: 117.855802 samples/sec	 accuracy=29.424479	 loss=3.237088	 lr=0.077184
Epoch[026] Batch [0649]/[3759]	Speed: 117.376522 samples/sec	 accuracy=29.413462	 loss=3.237389	 lr=0.077223
Epoch[026] Batch [0699]/[3759]	Speed: 117.387049 samples/sec	 accuracy=29.345982	 loss=3.243112	 lr=0.077262
Epoch[026] Batch [0749]/[3759]	Speed: 117.100681 samples/sec	 accuracy=29.287500	 loss=3.246487	 lr=0.077301
Epoch[026] Batch [0799]/[3759]	Speed: 117.095239 samples/sec	 accuracy=29.302734	 loss=3.248432	 lr=0.077339
Epoch[026] Batch [0849]/[3759]	Speed: 117.295061 samples/sec	 accuracy=29.244485	 loss=3.249157	 lr=0.077378
Epoch[026] Batch [0899]/[3759]	Speed: 117.364031 samples/sec	 accuracy=29.255208	 loss=3.249205	 lr=0.077417
Epoch[026] Batch [0949]/[3759]	Speed: 116.954479 samples/sec	 accuracy=29.250000	 loss=3.249621	 lr=0.077456
Epoch[026] Batch [0999]/[3759]	Speed: 117.525788 samples/sec	 accuracy=29.226562	 loss=3.249200	 lr=0.077494
Epoch[026] Batch [1049]/[3759]	Speed: 117.314219 samples/sec	 accuracy=29.313988	 loss=3.248010	 lr=0.077533
Epoch[026] Batch [1099]/[3759]	Speed: 116.539794 samples/sec	 accuracy=29.353693	 loss=3.244672	 lr=0.077572
Epoch[026] Batch [1149]/[3759]	Speed: 117.786878 samples/sec	 accuracy=29.339674	 loss=3.245147	 lr=0.077610
Epoch[026] Batch [1199]/[3759]	Speed: 117.341382 samples/sec	 accuracy=29.321615	 loss=3.244653	 lr=0.077649
Epoch[026] Batch [1249]/[3759]	Speed: 117.428933 samples/sec	 accuracy=29.316250	 loss=3.244487	 lr=0.077688
Epoch[026] Batch [1299]/[3759]	Speed: 117.845498 samples/sec	 accuracy=29.334135	 loss=3.242811	 lr=0.077727
Epoch[026] Batch [1349]/[3759]	Speed: 117.940520 samples/sec	 accuracy=29.342593	 loss=3.243611	 lr=0.077765
Epoch[026] Batch [1399]/[3759]	Speed: 117.325445 samples/sec	 accuracy=29.361607	 loss=3.242727	 lr=0.077804
Epoch[026] Batch [1449]/[3759]	Speed: 117.706923 samples/sec	 accuracy=29.368534	 loss=3.243683	 lr=0.077843
Epoch[026] Batch [1499]/[3759]	Speed: 117.875075 samples/sec	 accuracy=29.342708	 loss=3.245396	 lr=0.077882
Epoch[026] Batch [1549]/[3759]	Speed: 117.390181 samples/sec	 accuracy=29.355847	 loss=3.245036	 lr=0.077920
Epoch[026] Batch [1599]/[3759]	Speed: 115.746859 samples/sec	 accuracy=29.360352	 loss=3.243979	 lr=0.077959
Epoch[026] Batch [1649]/[3759]	Speed: 116.967212 samples/sec	 accuracy=29.358902	 loss=3.244278	 lr=0.077998
Epoch[026] Batch [1699]/[3759]	Speed: 116.825453 samples/sec	 accuracy=29.362132	 loss=3.244165	 lr=0.078036
Epoch[026] Batch [1749]/[3759]	Speed: 117.482393 samples/sec	 accuracy=29.366071	 loss=3.244214	 lr=0.078075
Epoch[026] Batch [1799]/[3759]	Speed: 117.640682 samples/sec	 accuracy=29.354167	 loss=3.245087	 lr=0.078114
Epoch[026] Batch [1849]/[3759]	Speed: 116.724970 samples/sec	 accuracy=29.332770	 loss=3.245898	 lr=0.078153
Epoch[026] Batch [1899]/[3759]	Speed: 118.148126 samples/sec	 accuracy=29.315789	 loss=3.246399	 lr=0.078191
Epoch[026] Batch [1949]/[3759]	Speed: 117.200361 samples/sec	 accuracy=29.323718	 loss=3.246755	 lr=0.078230
Epoch[026] Batch [1999]/[3759]	Speed: 117.183911 samples/sec	 accuracy=29.321875	 loss=3.246999	 lr=0.078269
Epoch[026] Batch [2049]/[3759]	Speed: 117.895925 samples/sec	 accuracy=29.289634	 loss=3.247457	 lr=0.078308
Epoch[026] Batch [2099]/[3759]	Speed: 117.642262 samples/sec	 accuracy=29.298363	 loss=3.248249	 lr=0.078346
Epoch[026] Batch [2149]/[3759]	Speed: 117.398221 samples/sec	 accuracy=29.258721	 loss=3.249530	 lr=0.078385
Epoch[026] Batch [2199]/[3759]	Speed: 118.259995 samples/sec	 accuracy=29.269886	 loss=3.250222	 lr=0.078424
Epoch[026] Batch [2249]/[3759]	Speed: 117.285247 samples/sec	 accuracy=29.263889	 loss=3.250953	 lr=0.078463
Epoch[026] Batch [2299]/[3759]	Speed: 117.339969 samples/sec	 accuracy=29.262908	 loss=3.250805	 lr=0.078501
Epoch[026] Batch [2349]/[3759]	Speed: 117.491227 samples/sec	 accuracy=29.287899	 loss=3.250791	 lr=0.078540
Epoch[026] Batch [2399]/[3759]	Speed: 117.439396 samples/sec	 accuracy=29.283203	 loss=3.251792	 lr=0.078579
Epoch[026] Batch [2449]/[3759]	Speed: 117.473396 samples/sec	 accuracy=29.278699	 loss=3.252111	 lr=0.078617
Epoch[026] Batch [2499]/[3759]	Speed: 116.866389 samples/sec	 accuracy=29.292500	 loss=3.252345	 lr=0.078656
Epoch[026] Batch [2549]/[3759]	Speed: 117.429221 samples/sec	 accuracy=29.294730	 loss=3.251742	 lr=0.078695
Epoch[026] Batch [2599]/[3759]	Speed: 116.933759 samples/sec	 accuracy=29.317308	 loss=3.251180	 lr=0.078734
Epoch[026] Batch [2649]/[3759]	Speed: 117.145512 samples/sec	 accuracy=29.330778	 loss=3.249917	 lr=0.078772
Epoch[026] Batch [2699]/[3759]	Speed: 118.085967 samples/sec	 accuracy=29.312500	 loss=3.251311	 lr=0.078811
Epoch[026] Batch [2749]/[3759]	Speed: 117.382995 samples/sec	 accuracy=29.309659	 loss=3.251459	 lr=0.078850
Epoch[026] Batch [2799]/[3759]	Speed: 118.160838 samples/sec	 accuracy=29.312500	 loss=3.251944	 lr=0.078889
Epoch[026] Batch [2849]/[3759]	Speed: 117.204071 samples/sec	 accuracy=29.328947	 loss=3.251755	 lr=0.078927
Epoch[026] Batch [2899]/[3759]	Speed: 118.118717 samples/sec	 accuracy=29.335129	 loss=3.251698	 lr=0.078966
Epoch[026] Batch [2949]/[3759]	Speed: 116.947009 samples/sec	 accuracy=29.336335	 loss=3.251685	 lr=0.079005
Epoch[026] Batch [2999]/[3759]	Speed: 117.501096 samples/sec	 accuracy=29.348958	 loss=3.251260	 lr=0.079043
Epoch[026] Batch [3049]/[3759]	Speed: 117.921146 samples/sec	 accuracy=29.345287	 loss=3.251975	 lr=0.079082
Epoch[026] Batch [3099]/[3759]	Speed: 117.424255 samples/sec	 accuracy=29.348286	 loss=3.252177	 lr=0.079121
Epoch[026] Batch [3149]/[3759]	Speed: 117.379832 samples/sec	 accuracy=29.349206	 loss=3.252091	 lr=0.079160
Epoch[026] Batch [3199]/[3759]	Speed: 117.175433 samples/sec	 accuracy=29.351562	 loss=3.252111	 lr=0.079198
Epoch[026] Batch [3249]/[3759]	Speed: 117.508827 samples/sec	 accuracy=29.362981	 loss=3.251783	 lr=0.079237
Epoch[026] Batch [3299]/[3759]	Speed: 117.358390 samples/sec	 accuracy=29.350852	 loss=3.252029	 lr=0.079276
Epoch[026] Batch [3349]/[3759]	Speed: 117.848863 samples/sec	 accuracy=29.354011	 loss=3.251876	 lr=0.079315
Epoch[026] Batch [3399]/[3759]	Speed: 117.660644 samples/sec	 accuracy=29.373621	 loss=3.251752	 lr=0.079353
Epoch[026] Batch [3449]/[3759]	Speed: 117.555745 samples/sec	 accuracy=29.374547	 loss=3.252775	 lr=0.079392
Epoch[026] Batch [3499]/[3759]	Speed: 117.892826 samples/sec	 accuracy=29.368304	 loss=3.252833	 lr=0.079431
Epoch[026] Batch [3549]/[3759]	Speed: 117.540280 samples/sec	 accuracy=29.383363	 loss=3.253235	 lr=0.079470
Epoch[026] Batch [3599]/[3759]	Speed: 117.086101 samples/sec	 accuracy=29.362847	 loss=3.253793	 lr=0.079508
Epoch[026] Batch [3649]/[3759]	Speed: 117.375671 samples/sec	 accuracy=29.354024	 loss=3.254410	 lr=0.079547
Epoch[026] Batch [3699]/[3759]	Speed: 117.214845 samples/sec	 accuracy=29.327280	 loss=3.255255	 lr=0.079586
Epoch[026] Batch [3749]/[3759]	Speed: 119.448104 samples/sec	 accuracy=29.332083	 loss=3.255129	 lr=0.079624
Batch [0049]/[0303]: acc-top1=27.531250 acc-top5=54.375000
Batch [0099]/[0303]: acc-top1=26.718750 acc-top5=53.515625
Batch [0149]/[0303]: acc-top1=28.010417 acc-top5=54.895833
Batch [0199]/[0303]: acc-top1=30.851562 acc-top5=57.085937
Batch [0249]/[0303]: acc-top1=30.393750 acc-top5=56.875000
Batch [0299]/[0303]: acc-top1=29.942708 acc-top5=56.984375
[Epoch 026] training: accuracy=29.333766	 loss=3.255221
[Epoch 026] speed: 116 samples/sec	time cost: 2178.394973
[Epoch 026] validation: acc-top1=29.883457 acc-top5=56.915223 loss=3.291931
Epoch[027] Batch [0049]/[3760]	Speed: 86.846662 samples/sec	 accuracy=29.312500	 loss=3.249164	 lr=0.079670
Epoch[027] Batch [0099]/[3760]	Speed: 118.796303 samples/sec	 accuracy=29.500000	 loss=3.240174	 lr=0.079709
Epoch[027] Batch [0149]/[3760]	Speed: 118.010375 samples/sec	 accuracy=29.937500	 loss=3.226227	 lr=0.079748
Epoch[027] Batch [0199]/[3760]	Speed: 117.536817 samples/sec	 accuracy=30.117187	 loss=3.210691	 lr=0.079786
Epoch[027] Batch [0249]/[3760]	Speed: 117.492509 samples/sec	 accuracy=29.575000	 loss=3.227838	 lr=0.079825
Epoch[027] Batch [0299]/[3760]	Speed: 118.354029 samples/sec	 accuracy=29.468750	 loss=3.232829	 lr=0.079864
Epoch[027] Batch [0349]/[3760]	Speed: 118.227549 samples/sec	 accuracy=29.584821	 loss=3.227882	 lr=0.079903
Epoch[027] Batch [0399]/[3760]	Speed: 117.320438 samples/sec	 accuracy=29.652344	 loss=3.225056	 lr=0.079941
Epoch[027] Batch [0449]/[3760]	Speed: 117.466389 samples/sec	 accuracy=29.534722	 loss=3.233702	 lr=0.079980
Epoch[027] Batch [0499]/[3760]	Speed: 117.179515 samples/sec	 accuracy=29.559375	 loss=3.232481	 lr=0.080019
Epoch[027] Batch [0549]/[3760]	Speed: 116.827016 samples/sec	 accuracy=29.534091	 loss=3.236712	 lr=0.080057
Epoch[027] Batch [0599]/[3760]	Speed: 117.071346 samples/sec	 accuracy=29.554688	 loss=3.234549	 lr=0.080096
Epoch[027] Batch [0649]/[3760]	Speed: 117.944590 samples/sec	 accuracy=29.569712	 loss=3.235131	 lr=0.080135
Epoch[027] Batch [0699]/[3760]	Speed: 117.810196 samples/sec	 accuracy=29.533482	 loss=3.239325	 lr=0.080174
Epoch[027] Batch [0749]/[3760]	Speed: 116.915627 samples/sec	 accuracy=29.458333	 loss=3.242456	 lr=0.080212
Epoch[027] Batch [0799]/[3760]	Speed: 118.796276 samples/sec	 accuracy=29.380859	 loss=3.244166	 lr=0.080251
Epoch[027] Batch [0849]/[3760]	Speed: 117.887570 samples/sec	 accuracy=29.461397	 loss=3.239869	 lr=0.080290
Epoch[027] Batch [0899]/[3760]	Speed: 117.558623 samples/sec	 accuracy=29.371528	 loss=3.241919	 lr=0.080329
Epoch[027] Batch [0949]/[3760]	Speed: 118.087168 samples/sec	 accuracy=29.422697	 loss=3.240018	 lr=0.080367
Epoch[027] Batch [0999]/[3760]	Speed: 117.861006 samples/sec	 accuracy=29.423437	 loss=3.242789	 lr=0.080406
Epoch[027] Batch [1049]/[3760]	Speed: 117.960156 samples/sec	 accuracy=29.413690	 loss=3.245906	 lr=0.080445
Epoch[027] Batch [1099]/[3760]	Speed: 117.743299 samples/sec	 accuracy=29.411932	 loss=3.248053	 lr=0.080484
Epoch[027] Batch [1149]/[3760]	Speed: 117.563343 samples/sec	 accuracy=29.398098	 loss=3.249849	 lr=0.080522
Epoch[027] Batch [1199]/[3760]	Speed: 117.413388 samples/sec	 accuracy=29.391927	 loss=3.250485	 lr=0.080561
Epoch[027] Batch [1249]/[3760]	Speed: 117.524064 samples/sec	 accuracy=29.410000	 loss=3.251229	 lr=0.080600
Epoch[027] Batch [1299]/[3760]	Speed: 117.671358 samples/sec	 accuracy=29.417067	 loss=3.250543	 lr=0.080638
Epoch[027] Batch [1349]/[3760]	Speed: 118.138657 samples/sec	 accuracy=29.415509	 loss=3.250813	 lr=0.080677
Epoch[027] Batch [1399]/[3760]	Speed: 117.735060 samples/sec	 accuracy=29.421875	 loss=3.249124	 lr=0.080716
Epoch[027] Batch [1449]/[3760]	Speed: 117.683950 samples/sec	 accuracy=29.385776	 loss=3.249899	 lr=0.080755
Epoch[027] Batch [1499]/[3760]	Speed: 117.275619 samples/sec	 accuracy=29.363542	 loss=3.249881	 lr=0.080793
Epoch[027] Batch [1549]/[3760]	Speed: 118.094315 samples/sec	 accuracy=29.352823	 loss=3.250536	 lr=0.080832
Epoch[027] Batch [1599]/[3760]	Speed: 117.319405 samples/sec	 accuracy=29.367188	 loss=3.250861	 lr=0.080871
Epoch[027] Batch [1649]/[3760]	Speed: 117.400569 samples/sec	 accuracy=29.367424	 loss=3.250723	 lr=0.080910
Epoch[027] Batch [1699]/[3760]	Speed: 117.491936 samples/sec	 accuracy=29.395221	 loss=3.251058	 lr=0.080948
Epoch[027] Batch [1749]/[3760]	Speed: 117.334587 samples/sec	 accuracy=29.384821	 loss=3.252141	 lr=0.080987
Epoch[027] Batch [1799]/[3760]	Speed: 117.644568 samples/sec	 accuracy=29.390625	 loss=3.252451	 lr=0.081026
Epoch[027] Batch [1849]/[3760]	Speed: 117.379521 samples/sec	 accuracy=29.348818	 loss=3.253856	 lr=0.081064
Epoch[027] Batch [1899]/[3760]	Speed: 117.537296 samples/sec	 accuracy=29.339638	 loss=3.254279	 lr=0.081103
Epoch[027] Batch [1949]/[3760]	Speed: 117.638755 samples/sec	 accuracy=29.370192	 loss=3.253696	 lr=0.081142
Epoch[027] Batch [1999]/[3760]	Speed: 117.312464 samples/sec	 accuracy=29.375000	 loss=3.254085	 lr=0.081181
Epoch[027] Batch [2049]/[3760]	Speed: 118.367863 samples/sec	 accuracy=29.393293	 loss=3.252865	 lr=0.081219
Epoch[027] Batch [2099]/[3760]	Speed: 117.583729 samples/sec	 accuracy=29.384673	 loss=3.252540	 lr=0.081258
Epoch[027] Batch [2149]/[3760]	Speed: 117.781693 samples/sec	 accuracy=29.372093	 loss=3.252375	 lr=0.081297
Epoch[027] Batch [2199]/[3760]	Speed: 117.285965 samples/sec	 accuracy=29.372869	 loss=3.252849	 lr=0.081336
Epoch[027] Batch [2249]/[3760]	Speed: 116.990361 samples/sec	 accuracy=29.333333	 loss=3.254579	 lr=0.081374
Epoch[027] Batch [2299]/[3760]	Speed: 117.621373 samples/sec	 accuracy=29.341033	 loss=3.253792	 lr=0.081413
Epoch[027] Batch [2349]/[3760]	Speed: 117.578655 samples/sec	 accuracy=29.329787	 loss=3.254466	 lr=0.081452
Epoch[027] Batch [2399]/[3760]	Speed: 117.437584 samples/sec	 accuracy=29.345052	 loss=3.254189	 lr=0.081491
Epoch[027] Batch [2449]/[3760]	Speed: 117.134829 samples/sec	 accuracy=29.358418	 loss=3.253951	 lr=0.081529
Epoch[027] Batch [2499]/[3760]	Speed: 117.973618 samples/sec	 accuracy=29.350000	 loss=3.254369	 lr=0.081568
Epoch[027] Batch [2549]/[3760]	Speed: 117.320502 samples/sec	 accuracy=29.367647	 loss=3.254030	 lr=0.081607
Epoch[027] Batch [2599]/[3760]	Speed: 117.964742 samples/sec	 accuracy=29.343750	 loss=3.254409	 lr=0.081645
Epoch[027] Batch [2649]/[3760]	Speed: 118.100957 samples/sec	 accuracy=29.325472	 loss=3.254835	 lr=0.081684
Epoch[027] Batch [2699]/[3760]	Speed: 118.127511 samples/sec	 accuracy=29.322917	 loss=3.255196	 lr=0.081723
Epoch[027] Batch [2749]/[3760]	Speed: 118.117884 samples/sec	 accuracy=29.325568	 loss=3.254699	 lr=0.081762
Epoch[027] Batch [2799]/[3760]	Speed: 117.936002 samples/sec	 accuracy=29.334263	 loss=3.254088	 lr=0.081800
Epoch[027] Batch [2849]/[3760]	Speed: 117.952834 samples/sec	 accuracy=29.326206	 loss=3.255057	 lr=0.081839
Epoch[027] Batch [2899]/[3760]	Speed: 117.052514 samples/sec	 accuracy=29.340517	 loss=3.254906	 lr=0.081878
Epoch[027] Batch [2949]/[3760]	Speed: 118.048502 samples/sec	 accuracy=29.319386	 loss=3.255240	 lr=0.081917
Epoch[027] Batch [2999]/[3760]	Speed: 117.906537 samples/sec	 accuracy=29.320833	 loss=3.255742	 lr=0.081955
Epoch[027] Batch [3049]/[3760]	Speed: 118.031559 samples/sec	 accuracy=29.329406	 loss=3.255460	 lr=0.081994
Epoch[027] Batch [3099]/[3760]	Speed: 117.584852 samples/sec	 accuracy=29.311996	 loss=3.256099	 lr=0.082033
Epoch[027] Batch [3149]/[3760]	Speed: 117.463750 samples/sec	 accuracy=29.316468	 loss=3.255933	 lr=0.082071
Epoch[027] Batch [3199]/[3760]	Speed: 118.003424 samples/sec	 accuracy=29.331055	 loss=3.255798	 lr=0.082110
Epoch[027] Batch [3249]/[3760]	Speed: 118.063600 samples/sec	 accuracy=29.316827	 loss=3.255780	 lr=0.082149
Epoch[027] Batch [3299]/[3760]	Speed: 117.548550 samples/sec	 accuracy=29.297348	 loss=3.256577	 lr=0.082188
Epoch[027] Batch [3349]/[3760]	Speed: 117.849696 samples/sec	 accuracy=29.290578	 loss=3.257088	 lr=0.082226
Epoch[027] Batch [3399]/[3760]	Speed: 117.812506 samples/sec	 accuracy=29.280331	 loss=3.257355	 lr=0.082265
Epoch[027] Batch [3449]/[3760]	Speed: 117.586774 samples/sec	 accuracy=29.284420	 loss=3.257129	 lr=0.082304
Epoch[027] Batch [3499]/[3760]	Speed: 117.559497 samples/sec	 accuracy=29.289286	 loss=3.256960	 lr=0.082343
Epoch[027] Batch [3549]/[3760]	Speed: 117.982085 samples/sec	 accuracy=29.299736	 loss=3.256729	 lr=0.082381
Epoch[027] Batch [3599]/[3760]	Speed: 117.240226 samples/sec	 accuracy=29.291233	 loss=3.257009	 lr=0.082420
Epoch[027] Batch [3649]/[3760]	Speed: 118.361060 samples/sec	 accuracy=29.278682	 loss=3.258284	 lr=0.082459
Epoch[027] Batch [3699]/[3760]	Speed: 117.748622 samples/sec	 accuracy=29.287584	 loss=3.257999	 lr=0.082498
Epoch[027] Batch [3749]/[3760]	Speed: 120.615891 samples/sec	 accuracy=29.290833	 loss=3.257731	 lr=0.082536
Batch [0049]/[0303]: acc-top1=28.250000 acc-top5=54.968750
Batch [0099]/[0303]: acc-top1=27.968750 acc-top5=53.625000
Batch [0149]/[0303]: acc-top1=29.093750 acc-top5=54.625000
Batch [0199]/[0303]: acc-top1=32.085938 acc-top5=57.656250
Batch [0249]/[0303]: acc-top1=30.981250 acc-top5=56.868750
Batch [0299]/[0303]: acc-top1=30.416667 acc-top5=57.161458
[Epoch 027] training: accuracy=29.279837	 loss=3.258193
[Epoch 027] speed: 117 samples/sec	time cost: 2173.500536
[Epoch 027] validation: acc-top1=30.409447 acc-top5=57.152434 loss=3.260846
Epoch[028] Batch [0049]/[3760]	Speed: 87.045277 samples/sec	 accuracy=28.906250	 loss=3.228402	 lr=0.082583
Epoch[028] Batch [0099]/[3760]	Speed: 117.656571 samples/sec	 accuracy=28.875000	 loss=3.233313	 lr=0.082621
Epoch[028] Batch [0149]/[3760]	Speed: 117.305489 samples/sec	 accuracy=28.895833	 loss=3.259703	 lr=0.082660
Epoch[028] Batch [0199]/[3760]	Speed: 117.812294 samples/sec	 accuracy=28.710938	 loss=3.261844	 lr=0.082699
Epoch[028] Batch [0249]/[3760]	Speed: 117.492065 samples/sec	 accuracy=28.825000	 loss=3.261028	 lr=0.082738
Epoch[028] Batch [0299]/[3760]	Speed: 116.630112 samples/sec	 accuracy=28.906250	 loss=3.264533	 lr=0.082776
Epoch[028] Batch [0349]/[3760]	Speed: 117.980719 samples/sec	 accuracy=29.058036	 loss=3.265553	 lr=0.082815
Epoch[028] Batch [0399]/[3760]	Speed: 117.376308 samples/sec	 accuracy=29.285156	 loss=3.254312	 lr=0.082854
Epoch[028] Batch [0449]/[3760]	Speed: 117.002591 samples/sec	 accuracy=29.177083	 loss=3.259643	 lr=0.082893
Epoch[028] Batch [0499]/[3760]	Speed: 116.343138 samples/sec	 accuracy=29.171875	 loss=3.256250	 lr=0.082931
Epoch[028] Batch [0549]/[3760]	Speed: 117.247510 samples/sec	 accuracy=29.250000	 loss=3.255156	 lr=0.082970
Epoch[028] Batch [0599]/[3760]	Speed: 117.405869 samples/sec	 accuracy=29.299479	 loss=3.248277	 lr=0.083009
Epoch[028] Batch [0649]/[3760]	Speed: 116.265914 samples/sec	 accuracy=29.310096	 loss=3.248179	 lr=0.083047
Epoch[028] Batch [0699]/[3760]	Speed: 117.010613 samples/sec	 accuracy=29.281250	 loss=3.251186	 lr=0.083086
Epoch[028] Batch [0749]/[3760]	Speed: 117.155785 samples/sec	 accuracy=29.297917	 loss=3.251918	 lr=0.083125
Epoch[028] Batch [0799]/[3760]	Speed: 116.826687 samples/sec	 accuracy=29.298828	 loss=3.251873	 lr=0.083164
Epoch[028] Batch [0849]/[3760]	Speed: 117.178329 samples/sec	 accuracy=29.340074	 loss=3.251338	 lr=0.083202
Epoch[028] Batch [0899]/[3760]	Speed: 117.270916 samples/sec	 accuracy=29.348958	 loss=3.250311	 lr=0.083241
Epoch[028] Batch [0949]/[3760]	Speed: 117.432263 samples/sec	 accuracy=29.347039	 loss=3.249405	 lr=0.083280
Epoch[028] Batch [0999]/[3760]	Speed: 117.346695 samples/sec	 accuracy=29.345312	 loss=3.251603	 lr=0.083319
Epoch[028] Batch [1049]/[3760]	Speed: 117.661835 samples/sec	 accuracy=29.367560	 loss=3.250203	 lr=0.083357
Epoch[028] Batch [1099]/[3760]	Speed: 117.315063 samples/sec	 accuracy=29.419034	 loss=3.248344	 lr=0.083396
Epoch[028] Batch [1149]/[3760]	Speed: 116.786818 samples/sec	 accuracy=29.413043	 loss=3.247533	 lr=0.083435
Epoch[028] Batch [1199]/[3760]	Speed: 117.407516 samples/sec	 accuracy=29.419271	 loss=3.247792	 lr=0.083474
Epoch[028] Batch [1249]/[3760]	Speed: 117.178110 samples/sec	 accuracy=29.436250	 loss=3.247712	 lr=0.083512
Epoch[028] Batch [1299]/[3760]	Speed: 117.095178 samples/sec	 accuracy=29.453125	 loss=3.248394	 lr=0.083551
Epoch[028] Batch [1349]/[3760]	Speed: 117.560976 samples/sec	 accuracy=29.464120	 loss=3.249459	 lr=0.083590
Epoch[028] Batch [1399]/[3760]	Speed: 117.350457 samples/sec	 accuracy=29.450893	 loss=3.249414	 lr=0.083628
Epoch[028] Batch [1449]/[3760]	Speed: 117.569059 samples/sec	 accuracy=29.477371	 loss=3.250211	 lr=0.083667
Epoch[028] Batch [1499]/[3760]	Speed: 117.395793 samples/sec	 accuracy=29.468750	 loss=3.250585	 lr=0.083706
Epoch[028] Batch [1549]/[3760]	Speed: 117.772191 samples/sec	 accuracy=29.495968	 loss=3.247863	 lr=0.083745
Epoch[028] Batch [1599]/[3760]	Speed: 116.701007 samples/sec	 accuracy=29.484375	 loss=3.249144	 lr=0.083783
Epoch[028] Batch [1649]/[3760]	Speed: 117.330281 samples/sec	 accuracy=29.455492	 loss=3.250978	 lr=0.083822
Epoch[028] Batch [1699]/[3760]	Speed: 117.955751 samples/sec	 accuracy=29.422794	 loss=3.251486	 lr=0.083861
Epoch[028] Batch [1749]/[3760]	Speed: 118.035528 samples/sec	 accuracy=29.411607	 loss=3.251099	 lr=0.083900
Epoch[028] Batch [1799]/[3760]	Speed: 116.873368 samples/sec	 accuracy=29.411458	 loss=3.250515	 lr=0.083938
Epoch[028] Batch [1849]/[3760]	Speed: 117.967881 samples/sec	 accuracy=29.420608	 loss=3.249770	 lr=0.083977
Epoch[028] Batch [1899]/[3760]	Speed: 117.894041 samples/sec	 accuracy=29.421053	 loss=3.248802	 lr=0.084016
Epoch[028] Batch [1949]/[3760]	Speed: 117.382095 samples/sec	 accuracy=29.381410	 loss=3.251228	 lr=0.084054
Epoch[028] Batch [1999]/[3760]	Speed: 117.290955 samples/sec	 accuracy=29.362500	 loss=3.251392	 lr=0.084093
Epoch[028] Batch [2049]/[3760]	Speed: 117.748940 samples/sec	 accuracy=29.333079	 loss=3.252465	 lr=0.084132
Epoch[028] Batch [2099]/[3760]	Speed: 117.486280 samples/sec	 accuracy=29.331101	 loss=3.253493	 lr=0.084171
Epoch[028] Batch [2149]/[3760]	Speed: 117.477480 samples/sec	 accuracy=29.366279	 loss=3.252985	 lr=0.084209
Epoch[028] Batch [2199]/[3760]	Speed: 118.304610 samples/sec	 accuracy=29.369318	 loss=3.253722	 lr=0.084248
Epoch[028] Batch [2249]/[3760]	Speed: 116.578805 samples/sec	 accuracy=29.377778	 loss=3.252825	 lr=0.084287
Epoch[028] Batch [2299]/[3760]	Speed: 117.496300 samples/sec	 accuracy=29.376359	 loss=3.253626	 lr=0.084326
Epoch[028] Batch [2349]/[3760]	Speed: 117.306761 samples/sec	 accuracy=29.373005	 loss=3.253619	 lr=0.084364
Epoch[028] Batch [2399]/[3760]	Speed: 118.410331 samples/sec	 accuracy=29.354167	 loss=3.255092	 lr=0.084403
Epoch[028] Batch [2449]/[3760]	Speed: 117.050136 samples/sec	 accuracy=29.350765	 loss=3.255358	 lr=0.084442
Epoch[028] Batch [2499]/[3760]	Speed: 117.239883 samples/sec	 accuracy=29.333125	 loss=3.256353	 lr=0.084481
Epoch[028] Batch [2549]/[3760]	Speed: 117.356476 samples/sec	 accuracy=29.344363	 loss=3.256463	 lr=0.084519
Epoch[028] Batch [2599]/[3760]	Speed: 117.309509 samples/sec	 accuracy=29.333534	 loss=3.257002	 lr=0.084558
Epoch[028] Batch [2649]/[3760]	Speed: 118.365859 samples/sec	 accuracy=29.339033	 loss=3.256219	 lr=0.084597
Epoch[028] Batch [2699]/[3760]	Speed: 118.176425 samples/sec	 accuracy=29.330440	 loss=3.256349	 lr=0.084635
Epoch[028] Batch [2749]/[3760]	Speed: 117.338475 samples/sec	 accuracy=29.305114	 loss=3.257080	 lr=0.084674
Epoch[028] Batch [2799]/[3760]	Speed: 117.565734 samples/sec	 accuracy=29.319196	 loss=3.257151	 lr=0.084713
Epoch[028] Batch [2849]/[3760]	Speed: 118.467129 samples/sec	 accuracy=29.324013	 loss=3.257216	 lr=0.084752
Epoch[028] Batch [2899]/[3760]	Speed: 117.083183 samples/sec	 accuracy=29.304957	 loss=3.257197	 lr=0.084790
Epoch[028] Batch [2949]/[3760]	Speed: 117.856800 samples/sec	 accuracy=29.306674	 loss=3.257692	 lr=0.084829
Epoch[028] Batch [2999]/[3760]	Speed: 117.272184 samples/sec	 accuracy=29.284375	 loss=3.258316	 lr=0.084868
Epoch[028] Batch [3049]/[3760]	Speed: 117.160405 samples/sec	 accuracy=29.265369	 loss=3.259435	 lr=0.084907
Epoch[028] Batch [3099]/[3760]	Speed: 117.315662 samples/sec	 accuracy=29.251512	 loss=3.259976	 lr=0.084945
Epoch[028] Batch [3149]/[3760]	Speed: 117.495765 samples/sec	 accuracy=29.239583	 loss=3.260886	 lr=0.084984
Epoch[028] Batch [3199]/[3760]	Speed: 116.364157 samples/sec	 accuracy=29.260742	 loss=3.259947	 lr=0.085023
Epoch[028] Batch [3249]/[3760]	Speed: 116.360589 samples/sec	 accuracy=29.277885	 loss=3.259551	 lr=0.085061
Epoch[028] Batch [3299]/[3760]	Speed: 117.556896 samples/sec	 accuracy=29.275568	 loss=3.260841	 lr=0.085100
Epoch[028] Batch [3349]/[3760]	Speed: 117.927929 samples/sec	 accuracy=29.271455	 loss=3.260854	 lr=0.085139
Epoch[028] Batch [3399]/[3760]	Speed: 116.556873 samples/sec	 accuracy=29.263327	 loss=3.261525	 lr=0.085178
Epoch[028] Batch [3449]/[3760]	Speed: 117.624775 samples/sec	 accuracy=29.270833	 loss=3.261364	 lr=0.085216
Epoch[028] Batch [3499]/[3760]	Speed: 118.192044 samples/sec	 accuracy=29.280804	 loss=3.260589	 lr=0.085255
Epoch[028] Batch [3549]/[3760]	Speed: 117.537844 samples/sec	 accuracy=29.287412	 loss=3.260936	 lr=0.085294
Epoch[028] Batch [3599]/[3760]	Speed: 117.890758 samples/sec	 accuracy=29.266927	 loss=3.261826	 lr=0.085333
Epoch[028] Batch [3649]/[3760]	Speed: 117.685984 samples/sec	 accuracy=29.277397	 loss=3.261822	 lr=0.085371
Epoch[028] Batch [3699]/[3760]	Speed: 117.089399 samples/sec	 accuracy=29.255912	 loss=3.262887	 lr=0.085410
Epoch[028] Batch [3749]/[3760]	Speed: 121.016878 samples/sec	 accuracy=29.263333	 loss=3.263013	 lr=0.085449
Batch [0049]/[0303]: acc-top1=28.906250 acc-top5=53.843750
Batch [0099]/[0303]: acc-top1=26.828125 acc-top5=52.734375
Batch [0149]/[0303]: acc-top1=28.875000 acc-top5=54.500000
Batch [0199]/[0303]: acc-top1=30.562500 acc-top5=55.945312
Batch [0249]/[0303]: acc-top1=29.687500 acc-top5=55.568750
Batch [0299]/[0303]: acc-top1=29.010417 acc-top5=55.291667
[Epoch 028] training: accuracy=29.266539	 loss=3.262945
[Epoch 028] speed: 116 samples/sec	time cost: 2178.753781
[Epoch 028] validation: acc-top1=28.924299 acc-top5=55.244431 loss=3.427814
Epoch[029] Batch [0049]/[3759]	Speed: 87.638454 samples/sec	 accuracy=26.937500	 loss=3.369577	 lr=0.085495
Epoch[029] Batch [0099]/[3759]	Speed: 117.312584 samples/sec	 accuracy=28.421875	 loss=3.303307	 lr=0.085534
Epoch[029] Batch [0149]/[3759]	Speed: 117.106437 samples/sec	 accuracy=28.354167	 loss=3.279266	 lr=0.085573
Epoch[029] Batch [0199]/[3759]	Speed: 117.639003 samples/sec	 accuracy=29.015625	 loss=3.265750	 lr=0.085611
Epoch[029] Batch [0249]/[3759]	Speed: 118.489219 samples/sec	 accuracy=29.337500	 loss=3.247750	 lr=0.085650
Epoch[029] Batch [0299]/[3759]	Speed: 117.050716 samples/sec	 accuracy=29.203125	 loss=3.251616	 lr=0.085689
Epoch[029] Batch [0349]/[3759]	Speed: 117.591299 samples/sec	 accuracy=29.174107	 loss=3.257757	 lr=0.085728
Epoch[029] Batch [0399]/[3759]	Speed: 117.569714 samples/sec	 accuracy=29.277344	 loss=3.253686	 lr=0.085766
Epoch[029] Batch [0449]/[3759]	Speed: 117.612280 samples/sec	 accuracy=29.281250	 loss=3.255292	 lr=0.085805
Epoch[029] Batch [0499]/[3759]	Speed: 117.558423 samples/sec	 accuracy=29.384375	 loss=3.254588	 lr=0.085844
Epoch[029] Batch [0549]/[3759]	Speed: 118.124110 samples/sec	 accuracy=29.269886	 loss=3.254169	 lr=0.085883
Epoch[029] Batch [0599]/[3759]	Speed: 116.938339 samples/sec	 accuracy=29.177083	 loss=3.257165	 lr=0.085921
Epoch[029] Batch [0649]/[3759]	Speed: 117.254778 samples/sec	 accuracy=29.189904	 loss=3.255862	 lr=0.085960
Epoch[029] Batch [0699]/[3759]	Speed: 117.528101 samples/sec	 accuracy=29.223214	 loss=3.256407	 lr=0.085999
Epoch[029] Batch [0749]/[3759]	Speed: 117.813444 samples/sec	 accuracy=29.081250	 loss=3.263829	 lr=0.086038
Epoch[029] Batch [0799]/[3759]	Speed: 117.477413 samples/sec	 accuracy=28.980469	 loss=3.265188	 lr=0.086076
Epoch[029] Batch [0849]/[3759]	Speed: 117.201718 samples/sec	 accuracy=28.979779	 loss=3.265605	 lr=0.086115
Epoch[029] Batch [0899]/[3759]	Speed: 117.213068 samples/sec	 accuracy=29.015625	 loss=3.265107	 lr=0.086154
Epoch[029] Batch [0949]/[3759]	Speed: 117.244810 samples/sec	 accuracy=29.014803	 loss=3.266026	 lr=0.086192
Epoch[029] Batch [0999]/[3759]	Speed: 117.478668 samples/sec	 accuracy=29.054688	 loss=3.265450	 lr=0.086231
Epoch[029] Batch [1049]/[3759]	Speed: 117.837157 samples/sec	 accuracy=29.125000	 loss=3.262979	 lr=0.086270
Epoch[029] Batch [1099]/[3759]	Speed: 116.691078 samples/sec	 accuracy=29.137784	 loss=3.263518	 lr=0.086309
Epoch[029] Batch [1149]/[3759]	Speed: 117.389488 samples/sec	 accuracy=29.131793	 loss=3.262640	 lr=0.086347
Epoch[029] Batch [1199]/[3759]	Speed: 117.152952 samples/sec	 accuracy=29.138021	 loss=3.261613	 lr=0.086386
Epoch[029] Batch [1249]/[3759]	Speed: 117.139270 samples/sec	 accuracy=29.101250	 loss=3.263113	 lr=0.086425
Epoch[029] Batch [1299]/[3759]	Speed: 117.121243 samples/sec	 accuracy=29.078125	 loss=3.263677	 lr=0.086464
Epoch[029] Batch [1349]/[3759]	Speed: 117.376553 samples/sec	 accuracy=29.024306	 loss=3.265461	 lr=0.086502
Epoch[029] Batch [1399]/[3759]	Speed: 117.641015 samples/sec	 accuracy=29.065848	 loss=3.265841	 lr=0.086541
Epoch[029] Batch [1449]/[3759]	Speed: 117.349183 samples/sec	 accuracy=29.071121	 loss=3.266757	 lr=0.086580
Epoch[029] Batch [1499]/[3759]	Speed: 117.226884 samples/sec	 accuracy=29.020833	 loss=3.269183	 lr=0.086618
Epoch[029] Batch [1549]/[3759]	Speed: 116.748895 samples/sec	 accuracy=29.018145	 loss=3.269087	 lr=0.086657
Epoch[029] Batch [1599]/[3759]	Speed: 117.475391 samples/sec	 accuracy=29.056641	 loss=3.266817	 lr=0.086696
Epoch[029] Batch [1649]/[3759]	Speed: 117.050567 samples/sec	 accuracy=29.124053	 loss=3.264801	 lr=0.086735
Epoch[029] Batch [1699]/[3759]	Speed: 116.592870 samples/sec	 accuracy=29.117647	 loss=3.265048	 lr=0.086773
Epoch[029] Batch [1749]/[3759]	Speed: 117.501569 samples/sec	 accuracy=29.160714	 loss=3.263389	 lr=0.086812
Epoch[029] Batch [1799]/[3759]	Speed: 118.042478 samples/sec	 accuracy=29.123264	 loss=3.265223	 lr=0.086851
Epoch[029] Batch [1849]/[3759]	Speed: 117.792754 samples/sec	 accuracy=29.110642	 loss=3.266092	 lr=0.086890
Epoch[029] Batch [1899]/[3759]	Speed: 118.057061 samples/sec	 accuracy=29.109375	 loss=3.267364	 lr=0.086928
Epoch[029] Batch [1949]/[3759]	Speed: 117.010036 samples/sec	 accuracy=29.092147	 loss=3.268321	 lr=0.086967
Epoch[029] Batch [1999]/[3759]	Speed: 116.895832 samples/sec	 accuracy=29.082812	 loss=3.269375	 lr=0.087006
Epoch[029] Batch [2049]/[3759]	Speed: 117.562744 samples/sec	 accuracy=29.088415	 loss=3.269498	 lr=0.087045
Epoch[029] Batch [2099]/[3759]	Speed: 117.283201 samples/sec	 accuracy=29.118304	 loss=3.267763	 lr=0.087083
Epoch[029] Batch [2149]/[3759]	Speed: 118.079058 samples/sec	 accuracy=29.100291	 loss=3.268000	 lr=0.087122
Epoch[029] Batch [2199]/[3759]	Speed: 116.868334 samples/sec	 accuracy=29.086648	 loss=3.268105	 lr=0.087161
Epoch[029] Batch [2249]/[3759]	Speed: 117.365590 samples/sec	 accuracy=29.088194	 loss=3.267997	 lr=0.087199
Epoch[029] Batch [2299]/[3759]	Speed: 117.962436 samples/sec	 accuracy=29.096467	 loss=3.267949	 lr=0.087238
Epoch[029] Batch [2349]/[3759]	Speed: 117.940891 samples/sec	 accuracy=29.059176	 loss=3.269229	 lr=0.087277
Epoch[029] Batch [2399]/[3759]	Speed: 116.879799 samples/sec	 accuracy=29.078125	 loss=3.268148	 lr=0.087316
Epoch[029] Batch [2449]/[3759]	Speed: 117.564012 samples/sec	 accuracy=29.099490	 loss=3.267035	 lr=0.087354
Epoch[029] Batch [2499]/[3759]	Speed: 118.134417 samples/sec	 accuracy=29.116875	 loss=3.267180	 lr=0.087393
Epoch[029] Batch [2549]/[3759]	Speed: 117.374164 samples/sec	 accuracy=29.124387	 loss=3.267588	 lr=0.087432
Epoch[029] Batch [2599]/[3759]	Speed: 117.448610 samples/sec	 accuracy=29.106370	 loss=3.268976	 lr=0.087471
Epoch[029] Batch [2649]/[3759]	Speed: 117.265549 samples/sec	 accuracy=29.061321	 loss=3.270805	 lr=0.087509
Epoch[029] Batch [2699]/[3759]	Speed: 117.182665 samples/sec	 accuracy=29.031829	 loss=3.272365	 lr=0.087548
Epoch[029] Batch [2749]/[3759]	Speed: 117.384889 samples/sec	 accuracy=29.042614	 loss=3.271726	 lr=0.087587
Epoch[029] Batch [2799]/[3759]	Speed: 117.147903 samples/sec	 accuracy=29.027902	 loss=3.272399	 lr=0.087625
Epoch[029] Batch [2849]/[3759]	Speed: 116.863069 samples/sec	 accuracy=29.010417	 loss=3.272090	 lr=0.087664
Epoch[029] Batch [2899]/[3759]	Speed: 117.872510 samples/sec	 accuracy=29.014009	 loss=3.272041	 lr=0.087703
Epoch[029] Batch [2949]/[3759]	Speed: 117.311955 samples/sec	 accuracy=28.997881	 loss=3.272495	 lr=0.087742
Epoch[029] Batch [2999]/[3759]	Speed: 117.122977 samples/sec	 accuracy=28.988542	 loss=3.273222	 lr=0.087780
Epoch[029] Batch [3049]/[3759]	Speed: 117.429714 samples/sec	 accuracy=28.981557	 loss=3.274304	 lr=0.087819
Epoch[029] Batch [3099]/[3759]	Speed: 116.407033 samples/sec	 accuracy=28.982359	 loss=3.273914	 lr=0.087858
Epoch[029] Batch [3149]/[3759]	Speed: 116.855333 samples/sec	 accuracy=29.005456	 loss=3.273264	 lr=0.087897
Epoch[029] Batch [3199]/[3759]	Speed: 117.176334 samples/sec	 accuracy=28.988770	 loss=3.274306	 lr=0.087935
Epoch[029] Batch [3249]/[3759]	Speed: 117.744438 samples/sec	 accuracy=28.993750	 loss=3.273452	 lr=0.087974
Epoch[029] Batch [3299]/[3759]	Speed: 116.899677 samples/sec	 accuracy=28.983902	 loss=3.273797	 lr=0.088013
Epoch[029] Batch [3349]/[3759]	Speed: 117.166274 samples/sec	 accuracy=29.004198	 loss=3.273608	 lr=0.088052
Epoch[029] Batch [3399]/[3759]	Speed: 117.486176 samples/sec	 accuracy=29.009191	 loss=3.273719	 lr=0.088090
Epoch[029] Batch [3449]/[3759]	Speed: 116.932665 samples/sec	 accuracy=29.038496	 loss=3.272288	 lr=0.088129
Epoch[029] Batch [3499]/[3759]	Speed: 117.604762 samples/sec	 accuracy=29.019643	 loss=3.272944	 lr=0.088168
Epoch[029] Batch [3549]/[3759]	Speed: 117.891457 samples/sec	 accuracy=29.008363	 loss=3.274200	 lr=0.088206
Epoch[029] Batch [3599]/[3759]	Speed: 117.679729 samples/sec	 accuracy=28.988281	 loss=3.275231	 lr=0.088245
Epoch[029] Batch [3649]/[3759]	Speed: 117.721655 samples/sec	 accuracy=29.028682	 loss=3.273775	 lr=0.088284
Epoch[029] Batch [3699]/[3759]	Speed: 117.287151 samples/sec	 accuracy=29.010135	 loss=3.274252	 lr=0.088323
Epoch[029] Batch [3749]/[3759]	Speed: 120.642507 samples/sec	 accuracy=28.986667	 loss=3.274720	 lr=0.088361
Batch [0049]/[0303]: acc-top1=29.375000 acc-top5=53.812500
Batch [0099]/[0303]: acc-top1=25.906250 acc-top5=50.078125
Batch [0149]/[0303]: acc-top1=26.875000 acc-top5=51.604167
Batch [0199]/[0303]: acc-top1=29.742188 acc-top5=54.406250
Batch [0249]/[0303]: acc-top1=28.731250 acc-top5=54.412500
Batch [0299]/[0303]: acc-top1=28.468750 acc-top5=54.692708
[Epoch 029] training: accuracy=28.982941	 loss=3.274763
[Epoch 029] speed: 116 samples/sec	time cost: 2178.515051
[Epoch 029] validation: acc-top1=28.460190 acc-top5=54.739068 loss=3.458164
Epoch[030] Batch [0049]/[3760]	Speed: 87.153144 samples/sec	 accuracy=29.718750	 loss=3.199839	 lr=0.088407
Epoch[030] Batch [0099]/[3760]	Speed: 117.829349 samples/sec	 accuracy=28.781250	 loss=3.249110	 lr=0.088446
Epoch[030] Batch [0149]/[3760]	Speed: 117.729848 samples/sec	 accuracy=28.822917	 loss=3.264166	 lr=0.088485
Epoch[030] Batch [0199]/[3760]	Speed: 117.009176 samples/sec	 accuracy=29.429688	 loss=3.252455	 lr=0.088523
Epoch[030] Batch [0249]/[3760]	Speed: 116.961379 samples/sec	 accuracy=29.231250	 loss=3.248280	 lr=0.088562
Epoch[030] Batch [0299]/[3760]	Speed: 118.353144 samples/sec	 accuracy=29.380208	 loss=3.240166	 lr=0.088601
Epoch[030] Batch [0349]/[3760]	Speed: 117.754191 samples/sec	 accuracy=29.120536	 loss=3.251941	 lr=0.088639
Epoch[030] Batch [0399]/[3760]	Speed: 117.231232 samples/sec	 accuracy=29.117188	 loss=3.254675	 lr=0.088678
Epoch[030] Batch [0449]/[3760]	Speed: 117.837547 samples/sec	 accuracy=29.260417	 loss=3.254546	 lr=0.088717
Epoch[030] Batch [0499]/[3760]	Speed: 116.586944 samples/sec	 accuracy=29.262500	 loss=3.256906	 lr=0.088756
Epoch[030] Batch [0549]/[3760]	Speed: 117.210849 samples/sec	 accuracy=29.292614	 loss=3.260899	 lr=0.088794
Epoch[030] Batch [0599]/[3760]	Speed: 117.109818 samples/sec	 accuracy=29.401042	 loss=3.254414	 lr=0.088833
Epoch[030] Batch [0649]/[3760]	Speed: 117.980294 samples/sec	 accuracy=29.365385	 loss=3.258449	 lr=0.088872
Epoch[030] Batch [0699]/[3760]	Speed: 116.959327 samples/sec	 accuracy=29.303571	 loss=3.261072	 lr=0.088911
Epoch[030] Batch [0749]/[3760]	Speed: 117.931295 samples/sec	 accuracy=29.381250	 loss=3.260591	 lr=0.088949
Epoch[030] Batch [0799]/[3760]	Speed: 117.689395 samples/sec	 accuracy=29.349609	 loss=3.265107	 lr=0.088988
Epoch[030] Batch [0849]/[3760]	Speed: 116.748912 samples/sec	 accuracy=29.308824	 loss=3.264342	 lr=0.089027
Epoch[030] Batch [0899]/[3760]	Speed: 117.325448 samples/sec	 accuracy=29.225694	 loss=3.271786	 lr=0.089065
Epoch[030] Batch [0949]/[3760]	Speed: 116.731677 samples/sec	 accuracy=29.190789	 loss=3.272508	 lr=0.089104
Epoch[030] Batch [0999]/[3760]	Speed: 117.154723 samples/sec	 accuracy=29.131250	 loss=3.275943	 lr=0.089143
Epoch[030] Batch [1049]/[3760]	Speed: 117.589799 samples/sec	 accuracy=29.159226	 loss=3.276524	 lr=0.089182
Epoch[030] Batch [1099]/[3760]	Speed: 117.811635 samples/sec	 accuracy=29.132102	 loss=3.278252	 lr=0.089220
Epoch[030] Batch [1149]/[3760]	Speed: 117.510579 samples/sec	 accuracy=29.082880	 loss=3.279924	 lr=0.089259
Epoch[030] Batch [1199]/[3760]	Speed: 117.977800 samples/sec	 accuracy=29.111979	 loss=3.278466	 lr=0.089298
Epoch[030] Batch [1249]/[3760]	Speed: 117.364603 samples/sec	 accuracy=29.142500	 loss=3.277635	 lr=0.089337
Epoch[030] Batch [1299]/[3760]	Speed: 117.488206 samples/sec	 accuracy=29.129808	 loss=3.278161	 lr=0.089375
Epoch[030] Batch [1349]/[3760]	Speed: 117.461211 samples/sec	 accuracy=29.104167	 loss=3.278762	 lr=0.089414
Epoch[030] Batch [1399]/[3760]	Speed: 117.407581 samples/sec	 accuracy=29.110491	 loss=3.279791	 lr=0.089453
Epoch[030] Batch [1449]/[3760]	Speed: 117.846361 samples/sec	 accuracy=29.066810	 loss=3.282172	 lr=0.089492
Epoch[030] Batch [1499]/[3760]	Speed: 117.532334 samples/sec	 accuracy=29.110417	 loss=3.279889	 lr=0.089530
Epoch[030] Batch [1549]/[3760]	Speed: 117.768026 samples/sec	 accuracy=29.100806	 loss=3.280213	 lr=0.089569
Epoch[030] Batch [1599]/[3760]	Speed: 117.240350 samples/sec	 accuracy=29.125977	 loss=3.279187	 lr=0.089608
Epoch[030] Batch [1649]/[3760]	Speed: 117.461730 samples/sec	 accuracy=29.076705	 loss=3.279068	 lr=0.089646
Epoch[030] Batch [1699]/[3760]	Speed: 116.897917 samples/sec	 accuracy=29.086397	 loss=3.277743	 lr=0.089685
Epoch[030] Batch [1749]/[3760]	Speed: 117.125940 samples/sec	 accuracy=29.096429	 loss=3.278406	 lr=0.089724
Epoch[030] Batch [1799]/[3760]	Speed: 117.176812 samples/sec	 accuracy=29.071181	 loss=3.277821	 lr=0.089763
Epoch[030] Batch [1849]/[3760]	Speed: 116.447460 samples/sec	 accuracy=29.083615	 loss=3.277322	 lr=0.089801
Epoch[030] Batch [1899]/[3760]	Speed: 117.274502 samples/sec	 accuracy=29.081414	 loss=3.277660	 lr=0.089840
Epoch[030] Batch [1949]/[3760]	Speed: 116.778488 samples/sec	 accuracy=29.068109	 loss=3.277334	 lr=0.089879
Epoch[030] Batch [1999]/[3760]	Speed: 117.481807 samples/sec	 accuracy=29.036719	 loss=3.279617	 lr=0.089918
Epoch[030] Batch [2049]/[3760]	Speed: 117.664303 samples/sec	 accuracy=29.040396	 loss=3.278816	 lr=0.089956
Epoch[030] Batch [2099]/[3760]	Speed: 117.870328 samples/sec	 accuracy=28.991815	 loss=3.281283	 lr=0.089995
Epoch[030] Batch [2149]/[3760]	Speed: 117.709428 samples/sec	 accuracy=28.991279	 loss=3.280392	 lr=0.090034
Epoch[030] Batch [2199]/[3760]	Speed: 117.206093 samples/sec	 accuracy=29.018466	 loss=3.278798	 lr=0.090073
Epoch[030] Batch [2249]/[3760]	Speed: 116.372659 samples/sec	 accuracy=28.997222	 loss=3.279071	 lr=0.090111
Epoch[030] Batch [2299]/[3760]	Speed: 116.835926 samples/sec	 accuracy=28.988451	 loss=3.279357	 lr=0.090150
Epoch[030] Batch [2349]/[3760]	Speed: 116.783485 samples/sec	 accuracy=28.953457	 loss=3.281022	 lr=0.090189
Epoch[030] Batch [2399]/[3760]	Speed: 117.045719 samples/sec	 accuracy=28.904297	 loss=3.282425	 lr=0.090227
Epoch[030] Batch [2449]/[3760]	Speed: 117.390655 samples/sec	 accuracy=28.922832	 loss=3.281840	 lr=0.090266
Epoch[030] Batch [2499]/[3760]	Speed: 117.064231 samples/sec	 accuracy=28.904375	 loss=3.282208	 lr=0.090305
Epoch[030] Batch [2549]/[3760]	Speed: 117.653250 samples/sec	 accuracy=28.915441	 loss=3.282366	 lr=0.090344
Epoch[030] Batch [2599]/[3760]	Speed: 117.126680 samples/sec	 accuracy=28.937500	 loss=3.281560	 lr=0.090382
Epoch[030] Batch [2649]/[3760]	Speed: 117.724790 samples/sec	 accuracy=28.952241	 loss=3.280790	 lr=0.090421
Epoch[030] Batch [2699]/[3760]	Speed: 117.276058 samples/sec	 accuracy=28.944444	 loss=3.281344	 lr=0.090460
Epoch[030] Batch [2749]/[3760]	Speed: 117.365705 samples/sec	 accuracy=28.943182	 loss=3.281421	 lr=0.090499
Epoch[030] Batch [2799]/[3760]	Speed: 117.380771 samples/sec	 accuracy=28.927455	 loss=3.281536	 lr=0.090537
Epoch[030] Batch [2849]/[3760]	Speed: 117.015470 samples/sec	 accuracy=28.916118	 loss=3.282241	 lr=0.090576
Epoch[030] Batch [2899]/[3760]	Speed: 117.487673 samples/sec	 accuracy=28.903556	 loss=3.282940	 lr=0.090615
Epoch[030] Batch [2949]/[3760]	Speed: 117.332556 samples/sec	 accuracy=28.906780	 loss=3.282753	 lr=0.090653
Epoch[030] Batch [2999]/[3760]	Speed: 117.760297 samples/sec	 accuracy=28.870313	 loss=3.283641	 lr=0.090692
Epoch[030] Batch [3049]/[3760]	Speed: 117.712517 samples/sec	 accuracy=28.900102	 loss=3.282287	 lr=0.090731
Epoch[030] Batch [3099]/[3760]	Speed: 117.218462 samples/sec	 accuracy=28.898690	 loss=3.282421	 lr=0.090770
Epoch[030] Batch [3149]/[3760]	Speed: 117.589916 samples/sec	 accuracy=28.893849	 loss=3.282511	 lr=0.090808
Epoch[030] Batch [3199]/[3760]	Speed: 117.177966 samples/sec	 accuracy=28.903320	 loss=3.282328	 lr=0.090847
Epoch[030] Batch [3249]/[3760]	Speed: 118.169979 samples/sec	 accuracy=28.923558	 loss=3.281621	 lr=0.090886
Epoch[030] Batch [3299]/[3760]	Speed: 116.058466 samples/sec	 accuracy=28.919981	 loss=3.281637	 lr=0.090925
Epoch[030] Batch [3349]/[3760]	Speed: 117.005243 samples/sec	 accuracy=28.904851	 loss=3.283194	 lr=0.090963
Epoch[030] Batch [3399]/[3760]	Speed: 117.182059 samples/sec	 accuracy=28.904871	 loss=3.283300	 lr=0.091002
Epoch[030] Batch [3449]/[3760]	Speed: 116.858223 samples/sec	 accuracy=28.900362	 loss=3.283035	 lr=0.091041
Epoch[030] Batch [3499]/[3760]	Speed: 117.598607 samples/sec	 accuracy=28.894196	 loss=3.283280	 lr=0.091080
Epoch[030] Batch [3549]/[3760]	Speed: 117.752062 samples/sec	 accuracy=28.903609	 loss=3.283011	 lr=0.091118
Epoch[030] Batch [3599]/[3760]	Speed: 117.827638 samples/sec	 accuracy=28.887153	 loss=3.283354	 lr=0.091157
Epoch[030] Batch [3649]/[3760]	Speed: 117.950750 samples/sec	 accuracy=28.872860	 loss=3.284099	 lr=0.091196
Epoch[030] Batch [3699]/[3760]	Speed: 117.626919 samples/sec	 accuracy=28.892314	 loss=3.283548	 lr=0.091234
Epoch[030] Batch [3749]/[3760]	Speed: 121.144042 samples/sec	 accuracy=28.890000	 loss=3.283570	 lr=0.091273
Batch [0049]/[0303]: acc-top1=29.281250 acc-top5=56.187500
Batch [0099]/[0303]: acc-top1=25.890625 acc-top5=51.875000
Batch [0149]/[0303]: acc-top1=27.531250 acc-top5=52.833333
Batch [0199]/[0303]: acc-top1=30.234375 acc-top5=55.656250
Batch [0249]/[0303]: acc-top1=29.350000 acc-top5=55.475000
Batch [0299]/[0303]: acc-top1=29.000000 acc-top5=55.567708
[Epoch 030] training: accuracy=28.892952	 loss=3.283366
[Epoch 030] speed: 116 samples/sec	time cost: 2179.399014
[Epoch 030] validation: acc-top1=28.981023 acc-top5=55.558993 loss=3.417623
Epoch[031] Batch [0049]/[3760]	Speed: 88.198947 samples/sec	 accuracy=29.937500	 loss=3.228246	 lr=0.091320
Epoch[031] Batch [0099]/[3760]	Speed: 117.390249 samples/sec	 accuracy=29.046875	 loss=3.260470	 lr=0.091358
Epoch[031] Batch [0149]/[3760]	Speed: 117.263373 samples/sec	 accuracy=28.937500	 loss=3.271877	 lr=0.091397
Epoch[031] Batch [0199]/[3760]	Speed: 116.437788 samples/sec	 accuracy=29.023438	 loss=3.257923	 lr=0.091436
Epoch[031] Batch [0249]/[3760]	Speed: 117.141087 samples/sec	 accuracy=29.137500	 loss=3.260931	 lr=0.091475
Epoch[031] Batch [0299]/[3760]	Speed: 117.111224 samples/sec	 accuracy=29.078125	 loss=3.257338	 lr=0.091513
Epoch[031] Batch [0349]/[3760]	Speed: 117.720743 samples/sec	 accuracy=29.071429	 loss=3.250137	 lr=0.091552
Epoch[031] Batch [0399]/[3760]	Speed: 116.926738 samples/sec	 accuracy=29.105469	 loss=3.248574	 lr=0.091591
Epoch[031] Batch [0449]/[3760]	Speed: 116.632432 samples/sec	 accuracy=29.055556	 loss=3.258630	 lr=0.091629
Epoch[031] Batch [0499]/[3760]	Speed: 117.161094 samples/sec	 accuracy=29.084375	 loss=3.256986	 lr=0.091668
Epoch[031] Batch [0549]/[3760]	Speed: 118.228381 samples/sec	 accuracy=29.076705	 loss=3.253291	 lr=0.091707
Epoch[031] Batch [0599]/[3760]	Speed: 117.302192 samples/sec	 accuracy=29.075521	 loss=3.256303	 lr=0.091746
Epoch[031] Batch [0649]/[3760]	Speed: 117.181232 samples/sec	 accuracy=29.187500	 loss=3.256617	 lr=0.091784
Epoch[031] Batch [0699]/[3760]	Speed: 117.924362 samples/sec	 accuracy=29.029018	 loss=3.263134	 lr=0.091823
Epoch[031] Batch [0749]/[3760]	Speed: 117.258700 samples/sec	 accuracy=28.989583	 loss=3.268134	 lr=0.091862
Epoch[031] Batch [0799]/[3760]	Speed: 116.928147 samples/sec	 accuracy=29.027344	 loss=3.268300	 lr=0.091901
Epoch[031] Batch [0849]/[3760]	Speed: 116.407070 samples/sec	 accuracy=29.117647	 loss=3.265261	 lr=0.091939
Epoch[031] Batch [0899]/[3760]	Speed: 117.080938 samples/sec	 accuracy=29.147569	 loss=3.266160	 lr=0.091978
Epoch[031] Batch [0949]/[3760]	Speed: 117.791870 samples/sec	 accuracy=29.101974	 loss=3.268931	 lr=0.092017
Epoch[031] Batch [0999]/[3760]	Speed: 117.209158 samples/sec	 accuracy=29.060938	 loss=3.272253	 lr=0.092056
Epoch[031] Batch [1049]/[3760]	Speed: 117.487892 samples/sec	 accuracy=29.129464	 loss=3.269752	 lr=0.092094
Epoch[031] Batch [1099]/[3760]	Speed: 117.349223 samples/sec	 accuracy=29.160511	 loss=3.269166	 lr=0.092133
Epoch[031] Batch [1149]/[3760]	Speed: 117.536084 samples/sec	 accuracy=29.172554	 loss=3.269289	 lr=0.092172
Epoch[031] Batch [1199]/[3760]	Speed: 117.863583 samples/sec	 accuracy=29.113281	 loss=3.272827	 lr=0.092210
Epoch[031] Batch [1249]/[3760]	Speed: 116.930202 samples/sec	 accuracy=29.102500	 loss=3.275048	 lr=0.092249
Epoch[031] Batch [1299]/[3760]	Speed: 118.018699 samples/sec	 accuracy=29.108173	 loss=3.275319	 lr=0.092288
Epoch[031] Batch [1349]/[3760]	Speed: 116.587127 samples/sec	 accuracy=29.060185	 loss=3.278035	 lr=0.092327
Epoch[031] Batch [1399]/[3760]	Speed: 118.136651 samples/sec	 accuracy=29.016741	 loss=3.279652	 lr=0.092365
Epoch[031] Batch [1449]/[3760]	Speed: 117.533705 samples/sec	 accuracy=29.042026	 loss=3.279183	 lr=0.092404
Epoch[031] Batch [1499]/[3760]	Speed: 117.159561 samples/sec	 accuracy=29.018750	 loss=3.280725	 lr=0.092443
Epoch[031] Batch [1549]/[3760]	Speed: 117.935245 samples/sec	 accuracy=29.033266	 loss=3.279018	 lr=0.092482
Epoch[031] Batch [1599]/[3760]	Speed: 117.618153 samples/sec	 accuracy=28.997070	 loss=3.280493	 lr=0.092520
Epoch[031] Batch [1649]/[3760]	Speed: 117.031405 samples/sec	 accuracy=28.995265	 loss=3.281323	 lr=0.092559
Epoch[031] Batch [1699]/[3760]	Speed: 117.132227 samples/sec	 accuracy=28.983456	 loss=3.281993	 lr=0.092598
Epoch[031] Batch [1749]/[3760]	Speed: 117.233713 samples/sec	 accuracy=29.030357	 loss=3.282283	 lr=0.092636
Epoch[031] Batch [1799]/[3760]	Speed: 117.833029 samples/sec	 accuracy=29.020833	 loss=3.280968	 lr=0.092675
Epoch[031] Batch [1849]/[3760]	Speed: 117.388467 samples/sec	 accuracy=28.994932	 loss=3.281821	 lr=0.092714
Epoch[031] Batch [1899]/[3760]	Speed: 117.664417 samples/sec	 accuracy=29.034539	 loss=3.280914	 lr=0.092753
Epoch[031] Batch [1949]/[3760]	Speed: 117.180261 samples/sec	 accuracy=29.036859	 loss=3.280725	 lr=0.092791
Epoch[031] Batch [1999]/[3760]	Speed: 117.464765 samples/sec	 accuracy=29.019531	 loss=3.281232	 lr=0.092830
Epoch[031] Batch [2049]/[3760]	Speed: 117.043402 samples/sec	 accuracy=28.993902	 loss=3.283032	 lr=0.092869
Epoch[031] Batch [2099]/[3760]	Speed: 117.283913 samples/sec	 accuracy=28.956845	 loss=3.284904	 lr=0.092908
Epoch[031] Batch [2149]/[3760]	Speed: 116.669372 samples/sec	 accuracy=28.937500	 loss=3.285444	 lr=0.092946
Epoch[031] Batch [2199]/[3760]	Speed: 117.979983 samples/sec	 accuracy=28.957386	 loss=3.284651	 lr=0.092985
Epoch[031] Batch [2249]/[3760]	Speed: 118.202515 samples/sec	 accuracy=28.980556	 loss=3.284076	 lr=0.093024
Epoch[031] Batch [2299]/[3760]	Speed: 117.550325 samples/sec	 accuracy=28.982337	 loss=3.284798	 lr=0.093063
Epoch[031] Batch [2349]/[3760]	Speed: 117.377009 samples/sec	 accuracy=28.976064	 loss=3.284643	 lr=0.093101
Epoch[031] Batch [2399]/[3760]	Speed: 117.249877 samples/sec	 accuracy=28.954427	 loss=3.284741	 lr=0.093140
Epoch[031] Batch [2449]/[3760]	Speed: 116.801305 samples/sec	 accuracy=28.932398	 loss=3.284992	 lr=0.093179
Epoch[031] Batch [2499]/[3760]	Speed: 117.889440 samples/sec	 accuracy=28.920625	 loss=3.285475	 lr=0.093217
Epoch[031] Batch [2549]/[3760]	Speed: 118.146105 samples/sec	 accuracy=28.905637	 loss=3.285767	 lr=0.093256
Epoch[031] Batch [2599]/[3760]	Speed: 117.274842 samples/sec	 accuracy=28.918269	 loss=3.286011	 lr=0.093295
Epoch[031] Batch [2649]/[3760]	Speed: 116.927921 samples/sec	 accuracy=28.926887	 loss=3.286096	 lr=0.093334
Epoch[031] Batch [2699]/[3760]	Speed: 117.494055 samples/sec	 accuracy=28.910880	 loss=3.286325	 lr=0.093372
Epoch[031] Batch [2749]/[3760]	Speed: 117.149784 samples/sec	 accuracy=28.886364	 loss=3.286925	 lr=0.093411
Epoch[031] Batch [2799]/[3760]	Speed: 117.160362 samples/sec	 accuracy=28.872210	 loss=3.287874	 lr=0.093450
Epoch[031] Batch [2849]/[3760]	Speed: 117.828907 samples/sec	 accuracy=28.856360	 loss=3.288575	 lr=0.093489
Epoch[031] Batch [2899]/[3760]	Speed: 117.572456 samples/sec	 accuracy=28.869073	 loss=3.288638	 lr=0.093527
Epoch[031] Batch [2949]/[3760]	Speed: 117.207596 samples/sec	 accuracy=28.859640	 loss=3.289010	 lr=0.093566
Epoch[031] Batch [2999]/[3760]	Speed: 118.125391 samples/sec	 accuracy=28.848438	 loss=3.288632	 lr=0.093605
Epoch[031] Batch [3049]/[3760]	Speed: 117.987180 samples/sec	 accuracy=28.840676	 loss=3.289907	 lr=0.093643
Epoch[031] Batch [3099]/[3760]	Speed: 117.393714 samples/sec	 accuracy=28.839718	 loss=3.291150	 lr=0.093682
Epoch[031] Batch [3149]/[3760]	Speed: 116.588744 samples/sec	 accuracy=28.845238	 loss=3.291042	 lr=0.093721
Epoch[031] Batch [3199]/[3760]	Speed: 117.417463 samples/sec	 accuracy=28.841309	 loss=3.290483	 lr=0.093760
Epoch[031] Batch [3249]/[3760]	Speed: 117.046745 samples/sec	 accuracy=28.837981	 loss=3.290264	 lr=0.093798
Epoch[031] Batch [3299]/[3760]	Speed: 117.299364 samples/sec	 accuracy=28.834754	 loss=3.290152	 lr=0.093837
Epoch[031] Batch [3349]/[3760]	Speed: 117.318621 samples/sec	 accuracy=28.831623	 loss=3.290532	 lr=0.093876
Epoch[031] Batch [3399]/[3760]	Speed: 117.232753 samples/sec	 accuracy=28.846967	 loss=3.290266	 lr=0.093915
Epoch[031] Batch [3449]/[3760]	Speed: 116.476288 samples/sec	 accuracy=28.840127	 loss=3.290569	 lr=0.093953
Epoch[031] Batch [3499]/[3760]	Speed: 118.020707 samples/sec	 accuracy=28.838839	 loss=3.290668	 lr=0.093992
Epoch[031] Batch [3549]/[3760]	Speed: 117.375159 samples/sec	 accuracy=28.824824	 loss=3.291276	 lr=0.094031
Epoch[031] Batch [3599]/[3760]	Speed: 117.891272 samples/sec	 accuracy=28.831597	 loss=3.291036	 lr=0.094070
Epoch[031] Batch [3649]/[3760]	Speed: 117.662544 samples/sec	 accuracy=28.829623	 loss=3.290273	 lr=0.094108
Epoch[031] Batch [3699]/[3760]	Speed: 117.543423 samples/sec	 accuracy=28.827703	 loss=3.290530	 lr=0.094147
Epoch[031] Batch [3749]/[3760]	Speed: 120.396260 samples/sec	 accuracy=28.835000	 loss=3.291150	 lr=0.094186
Batch [0049]/[0303]: acc-top1=26.718750 acc-top5=53.968750
Batch [0099]/[0303]: acc-top1=26.187500 acc-top5=52.906250
Batch [0149]/[0303]: acc-top1=27.427083 acc-top5=53.322917
Batch [0199]/[0303]: acc-top1=29.867187 acc-top5=56.031250
Batch [0249]/[0303]: acc-top1=29.262500 acc-top5=56.056250
Batch [0299]/[0303]: acc-top1=28.932292 acc-top5=56.187500
[Epoch 031] training: accuracy=28.826463	 loss=3.291470
[Epoch 031] speed: 116 samples/sec	time cost: 2179.462208
[Epoch 031] validation: acc-top1=28.852104 acc-top5=56.074670 loss=3.398228
Epoch[032] Batch [0049]/[3759]	Speed: 86.353268 samples/sec	 accuracy=28.968750	 loss=3.260516	 lr=0.094232
Epoch[032] Batch [0099]/[3759]	Speed: 116.690499 samples/sec	 accuracy=29.640625	 loss=3.240853	 lr=0.094271
Epoch[032] Batch [0149]/[3759]	Speed: 117.663396 samples/sec	 accuracy=28.802083	 loss=3.263959	 lr=0.094310
Epoch[032] Batch [0199]/[3759]	Speed: 116.549252 samples/sec	 accuracy=28.976563	 loss=3.267118	 lr=0.094348
Epoch[032] Batch [0249]/[3759]	Speed: 117.745234 samples/sec	 accuracy=28.937500	 loss=3.264065	 lr=0.094387
Epoch[032] Batch [0299]/[3759]	Speed: 117.059402 samples/sec	 accuracy=28.651042	 loss=3.271700	 lr=0.094426
Epoch[032] Batch [0349]/[3759]	Speed: 117.595949 samples/sec	 accuracy=28.861607	 loss=3.261027	 lr=0.094465
Epoch[032] Batch [0399]/[3759]	Speed: 117.147494 samples/sec	 accuracy=28.699219	 loss=3.269722	 lr=0.094503
Epoch[032] Batch [0449]/[3759]	Speed: 117.989957 samples/sec	 accuracy=28.711806	 loss=3.273697	 lr=0.094542
Epoch[032] Batch [0499]/[3759]	Speed: 118.240471 samples/sec	 accuracy=28.712500	 loss=3.278662	 lr=0.094581
Epoch[032] Batch [0549]/[3759]	Speed: 116.632496 samples/sec	 accuracy=28.786932	 loss=3.275899	 lr=0.094620
Epoch[032] Batch [0599]/[3759]	Speed: 117.727484 samples/sec	 accuracy=28.815104	 loss=3.276294	 lr=0.094658
Epoch[032] Batch [0649]/[3759]	Speed: 117.471570 samples/sec	 accuracy=28.790865	 loss=3.279019	 lr=0.094697
Epoch[032] Batch [0699]/[3759]	Speed: 117.297737 samples/sec	 accuracy=28.783482	 loss=3.279146	 lr=0.094736
Epoch[032] Batch [0749]/[3759]	Speed: 117.781967 samples/sec	 accuracy=28.820833	 loss=3.277450	 lr=0.094774
Epoch[032] Batch [0799]/[3759]	Speed: 117.286315 samples/sec	 accuracy=28.820312	 loss=3.278232	 lr=0.094813
Epoch[032] Batch [0849]/[3759]	Speed: 117.505251 samples/sec	 accuracy=28.895221	 loss=3.277449	 lr=0.094852
Epoch[032] Batch [0899]/[3759]	Speed: 116.942200 samples/sec	 accuracy=28.977431	 loss=3.274464	 lr=0.094891
Epoch[032] Batch [0949]/[3759]	Speed: 117.527727 samples/sec	 accuracy=28.888158	 loss=3.277310	 lr=0.094929
Epoch[032] Batch [0999]/[3759]	Speed: 116.818411 samples/sec	 accuracy=28.915625	 loss=3.277253	 lr=0.094968
Epoch[032] Batch [1049]/[3759]	Speed: 117.320339 samples/sec	 accuracy=28.909226	 loss=3.276713	 lr=0.095007
Epoch[032] Batch [1099]/[3759]	Speed: 117.094854 samples/sec	 accuracy=28.906250	 loss=3.279409	 lr=0.095046
Epoch[032] Batch [1149]/[3759]	Speed: 117.653635 samples/sec	 accuracy=28.908967	 loss=3.279388	 lr=0.095084
Epoch[032] Batch [1199]/[3759]	Speed: 116.989294 samples/sec	 accuracy=28.873698	 loss=3.278070	 lr=0.095123
Epoch[032] Batch [1249]/[3759]	Speed: 117.431041 samples/sec	 accuracy=28.901250	 loss=3.277936	 lr=0.095162
Epoch[032] Batch [1299]/[3759]	Speed: 118.534743 samples/sec	 accuracy=28.912260	 loss=3.277621	 lr=0.095200
Epoch[032] Batch [1349]/[3759]	Speed: 117.537862 samples/sec	 accuracy=28.945602	 loss=3.276398	 lr=0.095239
Epoch[032] Batch [1399]/[3759]	Speed: 117.567669 samples/sec	 accuracy=28.915179	 loss=3.277487	 lr=0.095278
Epoch[032] Batch [1449]/[3759]	Speed: 117.325070 samples/sec	 accuracy=28.870690	 loss=3.277287	 lr=0.095317
Epoch[032] Batch [1499]/[3759]	Speed: 117.975305 samples/sec	 accuracy=28.886458	 loss=3.278756	 lr=0.095355
Epoch[032] Batch [1549]/[3759]	Speed: 118.200266 samples/sec	 accuracy=28.882056	 loss=3.278834	 lr=0.095394
Epoch[032] Batch [1599]/[3759]	Speed: 117.613441 samples/sec	 accuracy=28.812500	 loss=3.280381	 lr=0.095433
Epoch[032] Batch [1649]/[3759]	Speed: 117.815664 samples/sec	 accuracy=28.818182	 loss=3.282031	 lr=0.095472
Epoch[032] Batch [1699]/[3759]	Speed: 117.516003 samples/sec	 accuracy=28.821691	 loss=3.281964	 lr=0.095510
Epoch[032] Batch [1749]/[3759]	Speed: 117.959705 samples/sec	 accuracy=28.806250	 loss=3.281985	 lr=0.095549
Epoch[032] Batch [1799]/[3759]	Speed: 117.664871 samples/sec	 accuracy=28.822049	 loss=3.281566	 lr=0.095588
Epoch[032] Batch [1849]/[3759]	Speed: 117.485103 samples/sec	 accuracy=28.826014	 loss=3.281910	 lr=0.095627
Epoch[032] Batch [1899]/[3759]	Speed: 116.679917 samples/sec	 accuracy=28.837993	 loss=3.281861	 lr=0.095665
Epoch[032] Batch [1949]/[3759]	Speed: 117.787737 samples/sec	 accuracy=28.834135	 loss=3.280852	 lr=0.095704
Epoch[032] Batch [1999]/[3759]	Speed: 117.355268 samples/sec	 accuracy=28.827344	 loss=3.281738	 lr=0.095743
Epoch[032] Batch [2049]/[3759]	Speed: 117.472073 samples/sec	 accuracy=28.807927	 loss=3.283550	 lr=0.095781
Epoch[032] Batch [2099]/[3759]	Speed: 117.178222 samples/sec	 accuracy=28.783482	 loss=3.284066	 lr=0.095820
Epoch[032] Batch [2149]/[3759]	Speed: 117.069720 samples/sec	 accuracy=28.767442	 loss=3.284965	 lr=0.095859
Epoch[032] Batch [2199]/[3759]	Speed: 116.687046 samples/sec	 accuracy=28.769886	 loss=3.285063	 lr=0.095898
Epoch[032] Batch [2249]/[3759]	Speed: 117.357953 samples/sec	 accuracy=28.773611	 loss=3.283848	 lr=0.095936
Epoch[032] Batch [2299]/[3759]	Speed: 117.107308 samples/sec	 accuracy=28.752717	 loss=3.284583	 lr=0.095975
Epoch[032] Batch [2349]/[3759]	Speed: 116.926643 samples/sec	 accuracy=28.726064	 loss=3.286127	 lr=0.096014
Epoch[032] Batch [2399]/[3759]	Speed: 117.140250 samples/sec	 accuracy=28.712240	 loss=3.287120	 lr=0.096053
Epoch[032] Batch [2449]/[3759]	Speed: 117.177656 samples/sec	 accuracy=28.724490	 loss=3.286881	 lr=0.096091
Epoch[032] Batch [2499]/[3759]	Speed: 117.280863 samples/sec	 accuracy=28.703750	 loss=3.288040	 lr=0.096130
Epoch[032] Batch [2549]/[3759]	Speed: 118.379260 samples/sec	 accuracy=28.699755	 loss=3.288261	 lr=0.096169
Epoch[032] Batch [2599]/[3759]	Speed: 116.929973 samples/sec	 accuracy=28.710337	 loss=3.288871	 lr=0.096207
Epoch[032] Batch [2649]/[3759]	Speed: 118.246048 samples/sec	 accuracy=28.732311	 loss=3.288956	 lr=0.096246
Epoch[032] Batch [2699]/[3759]	Speed: 117.645907 samples/sec	 accuracy=28.747106	 loss=3.288290	 lr=0.096285
Epoch[032] Batch [2749]/[3759]	Speed: 117.548305 samples/sec	 accuracy=28.751136	 loss=3.288670	 lr=0.096324
Epoch[032] Batch [2799]/[3759]	Speed: 117.468151 samples/sec	 accuracy=28.751674	 loss=3.288463	 lr=0.096362
Epoch[032] Batch [2849]/[3759]	Speed: 117.603428 samples/sec	 accuracy=28.780702	 loss=3.288430	 lr=0.096401
Epoch[032] Batch [2899]/[3759]	Speed: 117.258218 samples/sec	 accuracy=28.792565	 loss=3.288969	 lr=0.096440
Epoch[032] Batch [2949]/[3759]	Speed: 116.689458 samples/sec	 accuracy=28.774364	 loss=3.289641	 lr=0.096479
Epoch[032] Batch [2999]/[3759]	Speed: 118.078631 samples/sec	 accuracy=28.774479	 loss=3.289479	 lr=0.096517
Epoch[032] Batch [3049]/[3759]	Speed: 117.204241 samples/sec	 accuracy=28.752561	 loss=3.289930	 lr=0.096556
Epoch[032] Batch [3099]/[3759]	Speed: 117.917934 samples/sec	 accuracy=28.741935	 loss=3.290781	 lr=0.096595
Epoch[032] Batch [3149]/[3759]	Speed: 118.145491 samples/sec	 accuracy=28.740575	 loss=3.289978	 lr=0.096634
Epoch[032] Batch [3199]/[3759]	Speed: 117.437540 samples/sec	 accuracy=28.731445	 loss=3.290248	 lr=0.096672
Epoch[032] Batch [3249]/[3759]	Speed: 117.461860 samples/sec	 accuracy=28.718750	 loss=3.290667	 lr=0.096711
Epoch[032] Batch [3299]/[3759]	Speed: 117.441328 samples/sec	 accuracy=28.711174	 loss=3.290434	 lr=0.096750
Epoch[032] Batch [3349]/[3759]	Speed: 117.607491 samples/sec	 accuracy=28.710354	 loss=3.290679	 lr=0.096788
Epoch[032] Batch [3399]/[3759]	Speed: 117.725024 samples/sec	 accuracy=28.711397	 loss=3.291241	 lr=0.096827
Epoch[032] Batch [3449]/[3759]	Speed: 116.648668 samples/sec	 accuracy=28.711504	 loss=3.291179	 lr=0.096866
Epoch[032] Batch [3499]/[3759]	Speed: 117.499055 samples/sec	 accuracy=28.713393	 loss=3.291318	 lr=0.096905
Epoch[032] Batch [3549]/[3759]	Speed: 117.786291 samples/sec	 accuracy=28.705106	 loss=3.291931	 lr=0.096943
Epoch[032] Batch [3599]/[3759]	Speed: 116.465728 samples/sec	 accuracy=28.710938	 loss=3.291817	 lr=0.096982
Epoch[032] Batch [3649]/[3759]	Speed: 117.324337 samples/sec	 accuracy=28.709760	 loss=3.291805	 lr=0.097021
Epoch[032] Batch [3699]/[3759]	Speed: 116.747884 samples/sec	 accuracy=28.701858	 loss=3.291837	 lr=0.097060
Epoch[032] Batch [3749]/[3759]	Speed: 120.885544 samples/sec	 accuracy=28.692083	 loss=3.293108	 lr=0.097098
Batch [0049]/[0303]: acc-top1=27.906250 acc-top5=53.187500
Batch [0099]/[0303]: acc-top1=26.109375 acc-top5=52.906250
Batch [0149]/[0303]: acc-top1=27.520833 acc-top5=53.885417
Batch [0199]/[0303]: acc-top1=30.093750 acc-top5=56.148438
Batch [0249]/[0303]: acc-top1=29.068750 acc-top5=55.875000
Batch [0299]/[0303]: acc-top1=28.734375 acc-top5=56.333333
[Epoch 032] training: accuracy=28.689063	 loss=3.293456
[Epoch 032] speed: 116 samples/sec	time cost: 2178.298553
[Epoch 032] validation: acc-top1=28.723185 acc-top5=56.250000 loss=3.403583
Epoch[033] Batch [0049]/[3760]	Speed: 87.015792 samples/sec	 accuracy=28.843750	 loss=3.318358	 lr=0.097144
Epoch[033] Batch [0099]/[3760]	Speed: 117.584760 samples/sec	 accuracy=28.562500	 loss=3.312726	 lr=0.097183
Epoch[033] Batch [0149]/[3760]	Speed: 117.503717 samples/sec	 accuracy=28.687500	 loss=3.303225	 lr=0.097221
Epoch[033] Batch [0199]/[3760]	Speed: 117.652881 samples/sec	 accuracy=28.812500	 loss=3.294435	 lr=0.097260
Epoch[033] Batch [0249]/[3760]	Speed: 117.713663 samples/sec	 accuracy=28.500000	 loss=3.292895	 lr=0.097299
Epoch[033] Batch [0299]/[3760]	Speed: 117.650734 samples/sec	 accuracy=28.302083	 loss=3.295629	 lr=0.097338
Epoch[033] Batch [0349]/[3760]	Speed: 117.060530 samples/sec	 accuracy=28.473214	 loss=3.284877	 lr=0.097376
Epoch[033] Batch [0399]/[3760]	Speed: 117.102895 samples/sec	 accuracy=28.410156	 loss=3.294096	 lr=0.097415
Epoch[033] Batch [0449]/[3760]	Speed: 117.912778 samples/sec	 accuracy=28.663194	 loss=3.286707	 lr=0.097454
Epoch[033] Batch [0499]/[3760]	Speed: 117.292116 samples/sec	 accuracy=28.662500	 loss=3.293584	 lr=0.097493
Epoch[033] Batch [0549]/[3760]	Speed: 116.742751 samples/sec	 accuracy=28.815341	 loss=3.290848	 lr=0.097531
Epoch[033] Batch [0599]/[3760]	Speed: 118.040843 samples/sec	 accuracy=28.713542	 loss=3.293975	 lr=0.097570
Epoch[033] Batch [0649]/[3760]	Speed: 117.014218 samples/sec	 accuracy=28.608173	 loss=3.301725	 lr=0.097609
Epoch[033] Batch [0699]/[3760]	Speed: 116.942985 samples/sec	 accuracy=28.595982	 loss=3.302350	 lr=0.097647
Epoch[033] Batch [0749]/[3760]	Speed: 116.657230 samples/sec	 accuracy=28.643750	 loss=3.300767	 lr=0.097686
Epoch[033] Batch [0799]/[3760]	Speed: 118.104030 samples/sec	 accuracy=28.687500	 loss=3.296691	 lr=0.097725
Epoch[033] Batch [0849]/[3760]	Speed: 116.702273 samples/sec	 accuracy=28.709559	 loss=3.297563	 lr=0.097764
Epoch[033] Batch [0899]/[3760]	Speed: 117.316760 samples/sec	 accuracy=28.777778	 loss=3.295197	 lr=0.097802
Epoch[033] Batch [0949]/[3760]	Speed: 117.690828 samples/sec	 accuracy=28.777961	 loss=3.297116	 lr=0.097841
Epoch[033] Batch [0999]/[3760]	Speed: 116.658798 samples/sec	 accuracy=28.820312	 loss=3.297789	 lr=0.097880
Epoch[033] Batch [1049]/[3760]	Speed: 117.266585 samples/sec	 accuracy=28.845238	 loss=3.295827	 lr=0.097919
Epoch[033] Batch [1099]/[3760]	Speed: 117.156852 samples/sec	 accuracy=28.815341	 loss=3.295929	 lr=0.097957
Epoch[033] Batch [1149]/[3760]	Speed: 117.895002 samples/sec	 accuracy=28.792120	 loss=3.295639	 lr=0.097996
Epoch[033] Batch [1199]/[3760]	Speed: 117.505589 samples/sec	 accuracy=28.777344	 loss=3.296260	 lr=0.098035
Epoch[033] Batch [1249]/[3760]	Speed: 117.784363 samples/sec	 accuracy=28.751250	 loss=3.297863	 lr=0.098074
Epoch[033] Batch [1299]/[3760]	Speed: 117.278721 samples/sec	 accuracy=28.700721	 loss=3.299681	 lr=0.098112
Epoch[033] Batch [1349]/[3760]	Speed: 116.916573 samples/sec	 accuracy=28.692130	 loss=3.300530	 lr=0.098151
Epoch[033] Batch [1399]/[3760]	Speed: 117.323260 samples/sec	 accuracy=28.661830	 loss=3.300052	 lr=0.098190
Epoch[033] Batch [1449]/[3760]	Speed: 117.125018 samples/sec	 accuracy=28.726293	 loss=3.298120	 lr=0.098228
Epoch[033] Batch [1499]/[3760]	Speed: 117.707015 samples/sec	 accuracy=28.729167	 loss=3.297704	 lr=0.098267
Epoch[033] Batch [1549]/[3760]	Speed: 117.544115 samples/sec	 accuracy=28.729839	 loss=3.301033	 lr=0.098306
Epoch[033] Batch [1599]/[3760]	Speed: 117.175774 samples/sec	 accuracy=28.717773	 loss=3.300882	 lr=0.098345
Epoch[033] Batch [1649]/[3760]	Speed: 117.515205 samples/sec	 accuracy=28.696970	 loss=3.301180	 lr=0.098383
Epoch[033] Batch [1699]/[3760]	Speed: 117.294299 samples/sec	 accuracy=28.696691	 loss=3.302302	 lr=0.098422
Epoch[033] Batch [1749]/[3760]	Speed: 117.138441 samples/sec	 accuracy=28.692857	 loss=3.302124	 lr=0.098461
Epoch[033] Batch [1799]/[3760]	Speed: 117.315872 samples/sec	 accuracy=28.691840	 loss=3.302429	 lr=0.098500
Epoch[033] Batch [1849]/[3760]	Speed: 116.975870 samples/sec	 accuracy=28.696791	 loss=3.301385	 lr=0.098538
Epoch[033] Batch [1899]/[3760]	Speed: 117.421324 samples/sec	 accuracy=28.704770	 loss=3.301664	 lr=0.098577
Epoch[033] Batch [1949]/[3760]	Speed: 117.864271 samples/sec	 accuracy=28.717949	 loss=3.301950	 lr=0.098616
Epoch[033] Batch [1999]/[3760]	Speed: 116.968434 samples/sec	 accuracy=28.725000	 loss=3.301752	 lr=0.098654
Epoch[033] Batch [2049]/[3760]	Speed: 117.567984 samples/sec	 accuracy=28.713415	 loss=3.303130	 lr=0.098693
Epoch[033] Batch [2099]/[3760]	Speed: 117.640864 samples/sec	 accuracy=28.702381	 loss=3.303890	 lr=0.098732
Epoch[033] Batch [2149]/[3760]	Speed: 117.830827 samples/sec	 accuracy=28.731831	 loss=3.303125	 lr=0.098771
Epoch[033] Batch [2199]/[3760]	Speed: 117.652024 samples/sec	 accuracy=28.735795	 loss=3.302768	 lr=0.098809
Epoch[033] Batch [2249]/[3760]	Speed: 117.605436 samples/sec	 accuracy=28.726389	 loss=3.303041	 lr=0.098848
Epoch[033] Batch [2299]/[3760]	Speed: 117.796217 samples/sec	 accuracy=28.717391	 loss=3.302964	 lr=0.098887
Epoch[033] Batch [2349]/[3760]	Speed: 117.707453 samples/sec	 accuracy=28.704787	 loss=3.303942	 lr=0.098926
Epoch[033] Batch [2399]/[3760]	Speed: 117.617223 samples/sec	 accuracy=28.694661	 loss=3.304076	 lr=0.098964
Epoch[033] Batch [2449]/[3760]	Speed: 117.225859 samples/sec	 accuracy=28.692602	 loss=3.304460	 lr=0.099003
Epoch[033] Batch [2499]/[3760]	Speed: 117.349339 samples/sec	 accuracy=28.663125	 loss=3.305732	 lr=0.099042
Epoch[033] Batch [2549]/[3760]	Speed: 118.313614 samples/sec	 accuracy=28.677083	 loss=3.305435	 lr=0.099081
Epoch[033] Batch [2599]/[3760]	Speed: 117.392737 samples/sec	 accuracy=28.688101	 loss=3.305174	 lr=0.099119
Epoch[033] Batch [2649]/[3760]	Speed: 117.093030 samples/sec	 accuracy=28.699882	 loss=3.304060	 lr=0.099158
Epoch[033] Batch [2699]/[3760]	Speed: 117.558573 samples/sec	 accuracy=28.715278	 loss=3.303670	 lr=0.099197
Epoch[033] Batch [2749]/[3760]	Speed: 117.783251 samples/sec	 accuracy=28.709659	 loss=3.303877	 lr=0.099235
Epoch[033] Batch [2799]/[3760]	Speed: 117.646739 samples/sec	 accuracy=28.662388	 loss=3.305001	 lr=0.099274
Epoch[033] Batch [2849]/[3760]	Speed: 117.311744 samples/sec	 accuracy=28.669956	 loss=3.304777	 lr=0.099313
Epoch[033] Batch [2899]/[3760]	Speed: 117.587171 samples/sec	 accuracy=28.692349	 loss=3.304061	 lr=0.099352
Epoch[033] Batch [2949]/[3760]	Speed: 117.431610 samples/sec	 accuracy=28.690678	 loss=3.305153	 lr=0.099390
Epoch[033] Batch [2999]/[3760]	Speed: 116.547971 samples/sec	 accuracy=28.684375	 loss=3.306300	 lr=0.099429
Epoch[033] Batch [3049]/[3760]	Speed: 117.273024 samples/sec	 accuracy=28.668545	 loss=3.306907	 lr=0.099468
Epoch[033] Batch [3099]/[3760]	Speed: 117.856769 samples/sec	 accuracy=28.658770	 loss=3.307655	 lr=0.099507
Epoch[033] Batch [3149]/[3760]	Speed: 117.826415 samples/sec	 accuracy=28.635913	 loss=3.308509	 lr=0.099545
Epoch[033] Batch [3199]/[3760]	Speed: 117.464670 samples/sec	 accuracy=28.638672	 loss=3.308470	 lr=0.099584
Epoch[033] Batch [3249]/[3760]	Speed: 117.006746 samples/sec	 accuracy=28.625000	 loss=3.308531	 lr=0.099623
Epoch[033] Batch [3299]/[3760]	Speed: 118.271206 samples/sec	 accuracy=28.624527	 loss=3.308543	 lr=0.099661
Epoch[033] Batch [3349]/[3760]	Speed: 117.304075 samples/sec	 accuracy=28.634795	 loss=3.308033	 lr=0.099700
Epoch[033] Batch [3399]/[3760]	Speed: 117.504399 samples/sec	 accuracy=28.639246	 loss=3.307974	 lr=0.099739
Epoch[033] Batch [3449]/[3760]	Speed: 117.175696 samples/sec	 accuracy=28.624547	 loss=3.308486	 lr=0.099778
Epoch[033] Batch [3499]/[3760]	Speed: 117.045660 samples/sec	 accuracy=28.629464	 loss=3.308768	 lr=0.099816
Epoch[033] Batch [3549]/[3760]	Speed: 117.020706 samples/sec	 accuracy=28.639965	 loss=3.308557	 lr=0.099855
Epoch[033] Batch [3599]/[3760]	Speed: 117.499094 samples/sec	 accuracy=28.647569	 loss=3.308439	 lr=0.099894
Epoch[033] Batch [3649]/[3760]	Speed: 117.981146 samples/sec	 accuracy=28.633990	 loss=3.308901	 lr=0.099933
Epoch[033] Batch [3699]/[3760]	Speed: 117.000949 samples/sec	 accuracy=28.624578	 loss=3.309075	 lr=0.099971
Epoch[033] Batch [3749]/[3760]	Speed: 120.148023 samples/sec	 accuracy=28.633333	 loss=3.308991	 lr=0.100000
Batch [0049]/[0303]: acc-top1=24.062500 acc-top5=47.625000
Batch [0099]/[0303]: acc-top1=24.312500 acc-top5=49.250000
Batch [0149]/[0303]: acc-top1=25.166667 acc-top5=50.500000
Batch [0199]/[0303]: acc-top1=27.343750 acc-top5=52.476562
Batch [0249]/[0303]: acc-top1=27.068750 acc-top5=52.668750
Batch [0299]/[0303]: acc-top1=26.421875 acc-top5=52.104167
[Epoch 033] training: accuracy=28.632397	 loss=3.308936
[Epoch 033] speed: 116 samples/sec	time cost: 2178.888978
[Epoch 033] validation: acc-top1=26.474835 acc-top5=52.129744 loss=3.727392
Epoch[034] Batch [0049]/[3759]	Speed: 86.747178 samples/sec	 accuracy=28.375000	 loss=3.330405	 lr=0.100000
Epoch[034] Batch [0099]/[3759]	Speed: 118.166821 samples/sec	 accuracy=29.218750	 loss=3.287044	 lr=0.100000
Epoch[034] Batch [0149]/[3759]	Speed: 117.228121 samples/sec	 accuracy=29.250000	 loss=3.271065	 lr=0.100000
Epoch[034] Batch [0199]/[3759]	Speed: 117.755122 samples/sec	 accuracy=29.171875	 loss=3.274963	 lr=0.100000
Epoch[034] Batch [0249]/[3759]	Speed: 117.259292 samples/sec	 accuracy=29.131250	 loss=3.273931	 lr=0.100000
Epoch[034] Batch [0299]/[3759]	Speed: 117.264734 samples/sec	 accuracy=28.859375	 loss=3.285290	 lr=0.100000
Epoch[034] Batch [0349]/[3759]	Speed: 117.018874 samples/sec	 accuracy=28.767857	 loss=3.284599	 lr=0.100000
Epoch[034] Batch [0399]/[3759]	Speed: 116.979456 samples/sec	 accuracy=28.722656	 loss=3.284443	 lr=0.100000
Epoch[034] Batch [0449]/[3759]	Speed: 117.707493 samples/sec	 accuracy=28.795139	 loss=3.281997	 lr=0.100000
Epoch[034] Batch [0499]/[3759]	Speed: 117.661083 samples/sec	 accuracy=28.781250	 loss=3.286134	 lr=0.100000
Epoch[034] Batch [0549]/[3759]	Speed: 116.747115 samples/sec	 accuracy=28.801136	 loss=3.284454	 lr=0.100000
Epoch[034] Batch [0599]/[3759]	Speed: 117.436781 samples/sec	 accuracy=28.861979	 loss=3.281905	 lr=0.100000
Epoch[034] Batch [0649]/[3759]	Speed: 116.650420 samples/sec	 accuracy=28.903846	 loss=3.279962	 lr=0.100000
Epoch[034] Batch [0699]/[3759]	Speed: 117.210262 samples/sec	 accuracy=28.848214	 loss=3.283499	 lr=0.100000
Epoch[034] Batch [0749]/[3759]	Speed: 116.886834 samples/sec	 accuracy=28.791667	 loss=3.287216	 lr=0.100000
Epoch[034] Batch [0799]/[3759]	Speed: 117.710413 samples/sec	 accuracy=28.830078	 loss=3.287527	 lr=0.100000
Epoch[034] Batch [0849]/[3759]	Speed: 117.071963 samples/sec	 accuracy=28.836397	 loss=3.287112	 lr=0.099999
Epoch[034] Batch [0899]/[3759]	Speed: 117.269501 samples/sec	 accuracy=28.795139	 loss=3.285161	 lr=0.099999
Epoch[034] Batch [0949]/[3759]	Speed: 116.902990 samples/sec	 accuracy=28.796053	 loss=3.284729	 lr=0.099999
Epoch[034] Batch [0999]/[3759]	Speed: 117.755729 samples/sec	 accuracy=28.832812	 loss=3.284720	 lr=0.099999
Epoch[034] Batch [1049]/[3759]	Speed: 116.566610 samples/sec	 accuracy=28.843750	 loss=3.281193	 lr=0.099999
Epoch[034] Batch [1099]/[3759]	Speed: 116.926942 samples/sec	 accuracy=28.887784	 loss=3.278227	 lr=0.099999
Epoch[034] Batch [1149]/[3759]	Speed: 117.393061 samples/sec	 accuracy=28.870924	 loss=3.281042	 lr=0.099999
Epoch[034] Batch [1199]/[3759]	Speed: 116.721769 samples/sec	 accuracy=28.854167	 loss=3.283006	 lr=0.099999
Epoch[034] Batch [1249]/[3759]	Speed: 117.537263 samples/sec	 accuracy=28.745000	 loss=3.287457	 lr=0.099999
Epoch[034] Batch [1299]/[3759]	Speed: 117.161537 samples/sec	 accuracy=28.681490	 loss=3.289136	 lr=0.099999
Epoch[034] Batch [1349]/[3759]	Speed: 117.299169 samples/sec	 accuracy=28.630787	 loss=3.293447	 lr=0.099999
Epoch[034] Batch [1399]/[3759]	Speed: 118.060453 samples/sec	 accuracy=28.601562	 loss=3.295671	 lr=0.099999
Epoch[034] Batch [1449]/[3759]	Speed: 117.859616 samples/sec	 accuracy=28.660560	 loss=3.293722	 lr=0.099999
Epoch[034] Batch [1499]/[3759]	Speed: 117.491580 samples/sec	 accuracy=28.655208	 loss=3.293887	 lr=0.099998
Epoch[034] Batch [1549]/[3759]	Speed: 117.141255 samples/sec	 accuracy=28.661290	 loss=3.293469	 lr=0.099998
Epoch[034] Batch [1599]/[3759]	Speed: 118.089230 samples/sec	 accuracy=28.669922	 loss=3.295197	 lr=0.099998
Epoch[034] Batch [1649]/[3759]	Speed: 117.214133 samples/sec	 accuracy=28.670455	 loss=3.295991	 lr=0.099998
Epoch[034] Batch [1699]/[3759]	Speed: 117.671893 samples/sec	 accuracy=28.681985	 loss=3.293919	 lr=0.099998
Epoch[034] Batch [1749]/[3759]	Speed: 117.302246 samples/sec	 accuracy=28.648214	 loss=3.294918	 lr=0.099998
Epoch[034] Batch [1799]/[3759]	Speed: 117.400112 samples/sec	 accuracy=28.603299	 loss=3.296872	 lr=0.099998
Epoch[034] Batch [1849]/[3759]	Speed: 117.107898 samples/sec	 accuracy=28.577703	 loss=3.296986	 lr=0.099998
Epoch[034] Batch [1899]/[3759]	Speed: 116.121007 samples/sec	 accuracy=28.588816	 loss=3.296750	 lr=0.099998
Epoch[034] Batch [1949]/[3759]	Speed: 117.585170 samples/sec	 accuracy=28.570513	 loss=3.297741	 lr=0.099997
Epoch[034] Batch [1999]/[3759]	Speed: 117.483085 samples/sec	 accuracy=28.592187	 loss=3.296826	 lr=0.099997
Epoch[034] Batch [2049]/[3759]	Speed: 117.368287 samples/sec	 accuracy=28.541159	 loss=3.299526	 lr=0.099997
Epoch[034] Batch [2099]/[3759]	Speed: 117.514363 samples/sec	 accuracy=28.531994	 loss=3.300913	 lr=0.099997
Epoch[034] Batch [2149]/[3759]	Speed: 117.851819 samples/sec	 accuracy=28.524709	 loss=3.300814	 lr=0.099997
Epoch[034] Batch [2199]/[3759]	Speed: 117.391063 samples/sec	 accuracy=28.521307	 loss=3.300939	 lr=0.099997
Epoch[034] Batch [2249]/[3759]	Speed: 117.710980 samples/sec	 accuracy=28.528472	 loss=3.301259	 lr=0.099997
Epoch[034] Batch [2299]/[3759]	Speed: 117.149060 samples/sec	 accuracy=28.511549	 loss=3.302513	 lr=0.099996
Epoch[034] Batch [2349]/[3759]	Speed: 117.928864 samples/sec	 accuracy=28.509309	 loss=3.303005	 lr=0.099996
Epoch[034] Batch [2399]/[3759]	Speed: 117.003197 samples/sec	 accuracy=28.484375	 loss=3.304518	 lr=0.099996
Epoch[034] Batch [2449]/[3759]	Speed: 117.750588 samples/sec	 accuracy=28.508291	 loss=3.302845	 lr=0.099996
Epoch[034] Batch [2499]/[3759]	Speed: 117.712901 samples/sec	 accuracy=28.516250	 loss=3.302437	 lr=0.099996
Epoch[034] Batch [2549]/[3759]	Speed: 116.760477 samples/sec	 accuracy=28.527574	 loss=3.302680	 lr=0.099996
Epoch[034] Batch [2599]/[3759]	Speed: 117.525369 samples/sec	 accuracy=28.524639	 loss=3.303373	 lr=0.099995
Epoch[034] Batch [2649]/[3759]	Speed: 117.877776 samples/sec	 accuracy=28.489976	 loss=3.305423	 lr=0.099995
Epoch[034] Batch [2699]/[3759]	Speed: 117.701166 samples/sec	 accuracy=28.473958	 loss=3.305901	 lr=0.099995
Epoch[034] Batch [2749]/[3759]	Speed: 116.771384 samples/sec	 accuracy=28.461364	 loss=3.305990	 lr=0.099995
Epoch[034] Batch [2799]/[3759]	Speed: 117.412092 samples/sec	 accuracy=28.436384	 loss=3.306156	 lr=0.099995
Epoch[034] Batch [2849]/[3759]	Speed: 117.542210 samples/sec	 accuracy=28.434211	 loss=3.306251	 lr=0.099995
Epoch[034] Batch [2899]/[3759]	Speed: 117.136801 samples/sec	 accuracy=28.455819	 loss=3.305936	 lr=0.099994
Epoch[034] Batch [2949]/[3759]	Speed: 117.259061 samples/sec	 accuracy=28.431144	 loss=3.306674	 lr=0.099994
Epoch[034] Batch [2999]/[3759]	Speed: 117.435581 samples/sec	 accuracy=28.437500	 loss=3.305962	 lr=0.099994
Epoch[034] Batch [3049]/[3759]	Speed: 117.527900 samples/sec	 accuracy=28.434939	 loss=3.306235	 lr=0.099994
Epoch[034] Batch [3099]/[3759]	Speed: 117.818127 samples/sec	 accuracy=28.424395	 loss=3.306969	 lr=0.099994
Epoch[034] Batch [3149]/[3759]	Speed: 117.223683 samples/sec	 accuracy=28.417659	 loss=3.307310	 lr=0.099993
Epoch[034] Batch [3199]/[3759]	Speed: 117.054676 samples/sec	 accuracy=28.403320	 loss=3.308358	 lr=0.099993
Epoch[034] Batch [3249]/[3759]	Speed: 116.953291 samples/sec	 accuracy=28.386058	 loss=3.309073	 lr=0.099993
Epoch[034] Batch [3299]/[3759]	Speed: 117.979383 samples/sec	 accuracy=28.388258	 loss=3.308937	 lr=0.099993
Epoch[034] Batch [3349]/[3759]	Speed: 117.603626 samples/sec	 accuracy=28.388060	 loss=3.308755	 lr=0.099992
Epoch[034] Batch [3399]/[3759]	Speed: 117.038620 samples/sec	 accuracy=28.377757	 loss=3.308738	 lr=0.099992
Epoch[034] Batch [3449]/[3759]	Speed: 117.219960 samples/sec	 accuracy=28.360054	 loss=3.309735	 lr=0.099992
Epoch[034] Batch [3499]/[3759]	Speed: 117.830282 samples/sec	 accuracy=28.368304	 loss=3.309979	 lr=0.099992
Epoch[034] Batch [3549]/[3759]	Speed: 117.701317 samples/sec	 accuracy=28.348592	 loss=3.310640	 lr=0.099992
Epoch[034] Batch [3599]/[3759]	Speed: 117.516884 samples/sec	 accuracy=28.347656	 loss=3.310993	 lr=0.099991
Epoch[034] Batch [3649]/[3759]	Speed: 117.300577 samples/sec	 accuracy=28.336473	 loss=3.311023	 lr=0.099991
Epoch[034] Batch [3699]/[3759]	Speed: 117.775599 samples/sec	 accuracy=28.346706	 loss=3.310857	 lr=0.099991
Epoch[034] Batch [3749]/[3759]	Speed: 120.417196 samples/sec	 accuracy=28.341667	 loss=3.311015	 lr=0.099991
Batch [0049]/[0303]: acc-top1=25.437500 acc-top5=49.500000
Batch [0099]/[0303]: acc-top1=23.968750 acc-top5=48.046875
Batch [0149]/[0303]: acc-top1=25.145833 acc-top5=49.479167
Batch [0199]/[0303]: acc-top1=27.367188 acc-top5=51.023438
Batch [0249]/[0303]: acc-top1=26.643750 acc-top5=50.450000
Batch [0299]/[0303]: acc-top1=26.098958 acc-top5=50.687500
[Epoch 034] training: accuracy=28.337823	 loss=3.311336
[Epoch 034] speed: 116 samples/sec	time cost: 2179.295706
[Epoch 034] validation: acc-top1=26.088078 acc-top5=50.732261 loss=3.723547
Epoch[035] Batch [0049]/[3760]	Speed: 88.031995 samples/sec	 accuracy=27.187500	 loss=3.333143	 lr=0.099990
Epoch[035] Batch [0099]/[3760]	Speed: 117.604604 samples/sec	 accuracy=28.078125	 loss=3.297981	 lr=0.099990
Epoch[035] Batch [0149]/[3760]	Speed: 117.619449 samples/sec	 accuracy=28.718750	 loss=3.280316	 lr=0.099990
Epoch[035] Batch [0199]/[3760]	Speed: 116.713601 samples/sec	 accuracy=28.796875	 loss=3.289316	 lr=0.099989
Epoch[035] Batch [0249]/[3760]	Speed: 116.986292 samples/sec	 accuracy=28.762500	 loss=3.288583	 lr=0.099989
Epoch[035] Batch [0299]/[3760]	Speed: 116.891117 samples/sec	 accuracy=28.770833	 loss=3.293052	 lr=0.099989
Epoch[035] Batch [0349]/[3760]	Speed: 117.709172 samples/sec	 accuracy=28.790179	 loss=3.294013	 lr=0.099989
Epoch[035] Batch [0399]/[3760]	Speed: 117.064300 samples/sec	 accuracy=28.757812	 loss=3.297664	 lr=0.099988
Epoch[035] Batch [0449]/[3760]	Speed: 116.356936 samples/sec	 accuracy=28.784722	 loss=3.300388	 lr=0.099988
Epoch[035] Batch [0499]/[3760]	Speed: 117.270139 samples/sec	 accuracy=28.878125	 loss=3.297545	 lr=0.099988
Epoch[035] Batch [0549]/[3760]	Speed: 116.836624 samples/sec	 accuracy=28.914773	 loss=3.292474	 lr=0.099988
Epoch[035] Batch [0599]/[3760]	Speed: 117.609383 samples/sec	 accuracy=28.802083	 loss=3.298411	 lr=0.099987
Epoch[035] Batch [0649]/[3760]	Speed: 117.082600 samples/sec	 accuracy=28.853365	 loss=3.294558	 lr=0.099987
Epoch[035] Batch [0699]/[3760]	Speed: 117.184448 samples/sec	 accuracy=28.859375	 loss=3.290369	 lr=0.099987
Epoch[035] Batch [0749]/[3760]	Speed: 116.850669 samples/sec	 accuracy=28.816667	 loss=3.288933	 lr=0.099986
Epoch[035] Batch [0799]/[3760]	Speed: 117.864570 samples/sec	 accuracy=28.779297	 loss=3.290496	 lr=0.099986
Epoch[035] Batch [0849]/[3760]	Speed: 116.777701 samples/sec	 accuracy=28.772059	 loss=3.290610	 lr=0.099986
Epoch[035] Batch [0899]/[3760]	Speed: 117.528392 samples/sec	 accuracy=28.750000	 loss=3.291199	 lr=0.099985
Epoch[035] Batch [0949]/[3760]	Speed: 117.764736 samples/sec	 accuracy=28.776316	 loss=3.290853	 lr=0.099985
Epoch[035] Batch [0999]/[3760]	Speed: 116.631391 samples/sec	 accuracy=28.701563	 loss=3.293468	 lr=0.099985
Epoch[035] Batch [1049]/[3760]	Speed: 117.286698 samples/sec	 accuracy=28.708333	 loss=3.291693	 lr=0.099984
Epoch[035] Batch [1099]/[3760]	Speed: 116.364712 samples/sec	 accuracy=28.715909	 loss=3.289675	 lr=0.099984
Epoch[035] Batch [1149]/[3760]	Speed: 117.775144 samples/sec	 accuracy=28.691576	 loss=3.291915	 lr=0.099984
Epoch[035] Batch [1199]/[3760]	Speed: 116.851533 samples/sec	 accuracy=28.755208	 loss=3.288204	 lr=0.099983
Epoch[035] Batch [1249]/[3760]	Speed: 117.114968 samples/sec	 accuracy=28.726250	 loss=3.289182	 lr=0.099983
Epoch[035] Batch [1299]/[3760]	Speed: 117.372167 samples/sec	 accuracy=28.746394	 loss=3.290290	 lr=0.099983
Epoch[035] Batch [1349]/[3760]	Speed: 117.245262 samples/sec	 accuracy=28.800926	 loss=3.289078	 lr=0.099982
Epoch[035] Batch [1399]/[3760]	Speed: 117.377481 samples/sec	 accuracy=28.782366	 loss=3.289546	 lr=0.099982
Epoch[035] Batch [1449]/[3760]	Speed: 116.826951 samples/sec	 accuracy=28.781250	 loss=3.289393	 lr=0.099982
Epoch[035] Batch [1499]/[3760]	Speed: 117.745703 samples/sec	 accuracy=28.841667	 loss=3.287637	 lr=0.099981
Epoch[035] Batch [1549]/[3760]	Speed: 117.263940 samples/sec	 accuracy=28.820565	 loss=3.289427	 lr=0.099981
Epoch[035] Batch [1599]/[3760]	Speed: 117.544527 samples/sec	 accuracy=28.791016	 loss=3.290125	 lr=0.099981
Epoch[035] Batch [1649]/[3760]	Speed: 117.224347 samples/sec	 accuracy=28.796402	 loss=3.291331	 lr=0.099980
Epoch[035] Batch [1699]/[3760]	Speed: 117.513463 samples/sec	 accuracy=28.784926	 loss=3.292148	 lr=0.099980
Epoch[035] Batch [1749]/[3760]	Speed: 116.723563 samples/sec	 accuracy=28.798214	 loss=3.293337	 lr=0.099980
Epoch[035] Batch [1799]/[3760]	Speed: 116.949160 samples/sec	 accuracy=28.791667	 loss=3.294096	 lr=0.099979
Epoch[035] Batch [1849]/[3760]	Speed: 117.878009 samples/sec	 accuracy=28.815034	 loss=3.292918	 lr=0.099979
Epoch[035] Batch [1899]/[3760]	Speed: 117.539884 samples/sec	 accuracy=28.814967	 loss=3.292432	 lr=0.099979
Epoch[035] Batch [1949]/[3760]	Speed: 117.611849 samples/sec	 accuracy=28.811699	 loss=3.293532	 lr=0.099978
Epoch[035] Batch [1999]/[3760]	Speed: 117.723521 samples/sec	 accuracy=28.791406	 loss=3.294163	 lr=0.099978
Epoch[035] Batch [2049]/[3760]	Speed: 117.361649 samples/sec	 accuracy=28.799543	 loss=3.293769	 lr=0.099977
Epoch[035] Batch [2099]/[3760]	Speed: 117.527799 samples/sec	 accuracy=28.741071	 loss=3.296624	 lr=0.099977
Epoch[035] Batch [2149]/[3760]	Speed: 117.605042 samples/sec	 accuracy=28.726744	 loss=3.296798	 lr=0.099977
Epoch[035] Batch [2199]/[3760]	Speed: 116.976643 samples/sec	 accuracy=28.725852	 loss=3.295980	 lr=0.099976
Epoch[035] Batch [2249]/[3760]	Speed: 117.768383 samples/sec	 accuracy=28.745833	 loss=3.295499	 lr=0.099976
Epoch[035] Batch [2299]/[3760]	Speed: 117.672520 samples/sec	 accuracy=28.726223	 loss=3.296584	 lr=0.099975
Epoch[035] Batch [2349]/[3760]	Speed: 117.370298 samples/sec	 accuracy=28.707447	 loss=3.297693	 lr=0.099975
Epoch[035] Batch [2399]/[3760]	Speed: 117.717961 samples/sec	 accuracy=28.706380	 loss=3.298078	 lr=0.099975
Epoch[035] Batch [2449]/[3760]	Speed: 117.539011 samples/sec	 accuracy=28.700255	 loss=3.298323	 lr=0.099974
Epoch[035] Batch [2499]/[3760]	Speed: 116.718442 samples/sec	 accuracy=28.700000	 loss=3.298376	 lr=0.099974
Epoch[035] Batch [2549]/[3760]	Speed: 118.034233 samples/sec	 accuracy=28.691789	 loss=3.299377	 lr=0.099973
Epoch[035] Batch [2599]/[3760]	Speed: 117.123865 samples/sec	 accuracy=28.701322	 loss=3.299377	 lr=0.099973
Epoch[035] Batch [2649]/[3760]	Speed: 117.159155 samples/sec	 accuracy=28.698703	 loss=3.300114	 lr=0.099972
Epoch[035] Batch [2699]/[3760]	Speed: 116.943797 samples/sec	 accuracy=28.704282	 loss=3.300289	 lr=0.099972
Epoch[035] Batch [2749]/[3760]	Speed: 117.890427 samples/sec	 accuracy=28.714773	 loss=3.299128	 lr=0.099972
Epoch[035] Batch [2799]/[3760]	Speed: 117.032565 samples/sec	 accuracy=28.691406	 loss=3.299907	 lr=0.099971
Epoch[035] Batch [2849]/[3760]	Speed: 117.344556 samples/sec	 accuracy=28.709430	 loss=3.299654	 lr=0.099971
Epoch[035] Batch [2899]/[3760]	Speed: 118.203265 samples/sec	 accuracy=28.705280	 loss=3.299704	 lr=0.099970
Epoch[035] Batch [2949]/[3760]	Speed: 117.478424 samples/sec	 accuracy=28.696504	 loss=3.300025	 lr=0.099970
Epoch[035] Batch [2999]/[3760]	Speed: 118.152656 samples/sec	 accuracy=28.684896	 loss=3.300117	 lr=0.099969
Epoch[035] Batch [3049]/[3760]	Speed: 118.006300 samples/sec	 accuracy=28.688012	 loss=3.300131	 lr=0.099969
Epoch[035] Batch [3099]/[3760]	Speed: 117.335180 samples/sec	 accuracy=28.672379	 loss=3.300350	 lr=0.099968
Epoch[035] Batch [3149]/[3760]	Speed: 117.880591 samples/sec	 accuracy=28.660714	 loss=3.300379	 lr=0.099968
Epoch[035] Batch [3199]/[3760]	Speed: 117.526671 samples/sec	 accuracy=28.643066	 loss=3.301016	 lr=0.099968
Epoch[035] Batch [3249]/[3760]	Speed: 117.971112 samples/sec	 accuracy=28.644712	 loss=3.301495	 lr=0.099967
Epoch[035] Batch [3299]/[3760]	Speed: 117.773451 samples/sec	 accuracy=28.649148	 loss=3.301736	 lr=0.099967
Epoch[035] Batch [3349]/[3760]	Speed: 117.610376 samples/sec	 accuracy=28.638060	 loss=3.301849	 lr=0.099966
Epoch[035] Batch [3399]/[3760]	Speed: 117.464984 samples/sec	 accuracy=28.626379	 loss=3.302432	 lr=0.099966
Epoch[035] Batch [3449]/[3760]	Speed: 117.255526 samples/sec	 accuracy=28.639493	 loss=3.302120	 lr=0.099965
Epoch[035] Batch [3499]/[3760]	Speed: 117.251292 samples/sec	 accuracy=28.634821	 loss=3.301935	 lr=0.099965
Epoch[035] Batch [3549]/[3760]	Speed: 117.472252 samples/sec	 accuracy=28.623680	 loss=3.302584	 lr=0.099964
Epoch[035] Batch [3599]/[3760]	Speed: 117.195159 samples/sec	 accuracy=28.611545	 loss=3.302793	 lr=0.099964
Epoch[035] Batch [3649]/[3760]	Speed: 117.690729 samples/sec	 accuracy=28.621147	 loss=3.302641	 lr=0.099963
Epoch[035] Batch [3699]/[3760]	Speed: 117.205986 samples/sec	 accuracy=28.635557	 loss=3.302378	 lr=0.099963
Epoch[035] Batch [3749]/[3760]	Speed: 120.513385 samples/sec	 accuracy=28.652083	 loss=3.302115	 lr=0.099962
Batch [0049]/[0303]: acc-top1=27.531250 acc-top5=53.718750
Batch [0099]/[0303]: acc-top1=26.578125 acc-top5=51.406250
Batch [0149]/[0303]: acc-top1=28.072917 acc-top5=53.166667
Batch [0199]/[0303]: acc-top1=30.828125 acc-top5=55.804688
Batch [0249]/[0303]: acc-top1=30.087500 acc-top5=56.156250
Batch [0299]/[0303]: acc-top1=29.953125 acc-top5=56.494792
[Epoch 035] training: accuracy=28.658993	 loss=3.301917
[Epoch 035] speed: 116 samples/sec	time cost: 2178.725304
[Epoch 035] validation: acc-top1=29.909241 acc-top5=56.487211 loss=3.312148
Epoch[036] Batch [0049]/[3760]	Speed: 86.645670 samples/sec	 accuracy=28.437500	 loss=3.271186	 lr=0.099962
Epoch[036] Batch [0099]/[3760]	Speed: 116.914694 samples/sec	 accuracy=28.734375	 loss=3.270162	 lr=0.099961
Epoch[036] Batch [0149]/[3760]	Speed: 117.683512 samples/sec	 accuracy=29.020833	 loss=3.262809	 lr=0.099961
Epoch[036] Batch [0199]/[3760]	Speed: 117.023027 samples/sec	 accuracy=28.695312	 loss=3.282252	 lr=0.099960
Epoch[036] Batch [0249]/[3760]	Speed: 117.258616 samples/sec	 accuracy=28.787500	 loss=3.269111	 lr=0.099960
Epoch[036] Batch [0299]/[3760]	Speed: 117.397444 samples/sec	 accuracy=28.817708	 loss=3.267150	 lr=0.099959
Epoch[036] Batch [0349]/[3760]	Speed: 117.306903 samples/sec	 accuracy=28.821429	 loss=3.262550	 lr=0.099959
Epoch[036] Batch [0399]/[3760]	Speed: 115.159473 samples/sec	 accuracy=28.960938	 loss=3.259001	 lr=0.099958
Epoch[036] Batch [0449]/[3760]	Speed: 117.671379 samples/sec	 accuracy=29.121528	 loss=3.263608	 lr=0.099958
Epoch[036] Batch [0499]/[3760]	Speed: 117.869018 samples/sec	 accuracy=29.043750	 loss=3.266829	 lr=0.099957
Epoch[036] Batch [0549]/[3760]	Speed: 116.958662 samples/sec	 accuracy=29.011364	 loss=3.270679	 lr=0.099956
Epoch[036] Batch [0599]/[3760]	Speed: 117.800067 samples/sec	 accuracy=28.966146	 loss=3.276675	 lr=0.099956
Epoch[036] Batch [0649]/[3760]	Speed: 116.855100 samples/sec	 accuracy=28.889423	 loss=3.279952	 lr=0.099955
Epoch[036] Batch [0699]/[3760]	Speed: 118.121101 samples/sec	 accuracy=28.875000	 loss=3.279069	 lr=0.099955
Epoch[036] Batch [0749]/[3760]	Speed: 117.378993 samples/sec	 accuracy=28.941667	 loss=3.277361	 lr=0.099954
Epoch[036] Batch [0799]/[3760]	Speed: 116.733826 samples/sec	 accuracy=28.923828	 loss=3.274928	 lr=0.099954
Epoch[036] Batch [0849]/[3760]	Speed: 117.260926 samples/sec	 accuracy=29.007353	 loss=3.275250	 lr=0.099953
Epoch[036] Batch [0899]/[3760]	Speed: 117.998312 samples/sec	 accuracy=28.963542	 loss=3.277030	 lr=0.099953
Epoch[036] Batch [0949]/[3760]	Speed: 117.651766 samples/sec	 accuracy=28.935855	 loss=3.280634	 lr=0.099952
Epoch[036] Batch [0999]/[3760]	Speed: 116.873154 samples/sec	 accuracy=28.992188	 loss=3.277645	 lr=0.099951
Epoch[036] Batch [1049]/[3760]	Speed: 116.461933 samples/sec	 accuracy=28.979167	 loss=3.280150	 lr=0.099951
Epoch[036] Batch [1099]/[3760]	Speed: 116.879084 samples/sec	 accuracy=28.982955	 loss=3.280421	 lr=0.099950
Epoch[036] Batch [1149]/[3760]	Speed: 117.852863 samples/sec	 accuracy=29.006793	 loss=3.279868	 lr=0.099950
Epoch[036] Batch [1199]/[3760]	Speed: 117.049121 samples/sec	 accuracy=29.029948	 loss=3.278857	 lr=0.099949
Epoch[036] Batch [1249]/[3760]	Speed: 117.055312 samples/sec	 accuracy=29.001250	 loss=3.280606	 lr=0.099949
Epoch[036] Batch [1299]/[3760]	Speed: 117.467848 samples/sec	 accuracy=29.027644	 loss=3.280079	 lr=0.099948
Epoch[036] Batch [1349]/[3760]	Speed: 117.215139 samples/sec	 accuracy=28.995370	 loss=3.280660	 lr=0.099947
Epoch[036] Batch [1399]/[3760]	Speed: 116.929313 samples/sec	 accuracy=28.972098	 loss=3.280536	 lr=0.099947
Epoch[036] Batch [1449]/[3760]	Speed: 117.291553 samples/sec	 accuracy=28.941810	 loss=3.280700	 lr=0.099946
Epoch[036] Batch [1499]/[3760]	Speed: 117.513370 samples/sec	 accuracy=28.968750	 loss=3.279677	 lr=0.099946
Epoch[036] Batch [1549]/[3760]	Speed: 116.419662 samples/sec	 accuracy=28.924395	 loss=3.284000	 lr=0.099945
Epoch[036] Batch [1599]/[3760]	Speed: 117.816037 samples/sec	 accuracy=28.909180	 loss=3.284371	 lr=0.099944
Epoch[036] Batch [1649]/[3760]	Speed: 116.753880 samples/sec	 accuracy=28.892992	 loss=3.285667	 lr=0.099944
Epoch[036] Batch [1699]/[3760]	Speed: 117.079430 samples/sec	 accuracy=28.921875	 loss=3.285053	 lr=0.099943
Epoch[036] Batch [1749]/[3760]	Speed: 116.516790 samples/sec	 accuracy=28.937500	 loss=3.284871	 lr=0.099943
Epoch[036] Batch [1799]/[3760]	Speed: 117.278125 samples/sec	 accuracy=28.915799	 loss=3.284269	 lr=0.099942
Epoch[036] Batch [1849]/[3760]	Speed: 118.351545 samples/sec	 accuracy=28.898649	 loss=3.284407	 lr=0.099941
Epoch[036] Batch [1899]/[3760]	Speed: 117.473333 samples/sec	 accuracy=28.836349	 loss=3.287060	 lr=0.099941
Epoch[036] Batch [1949]/[3760]	Speed: 117.236139 samples/sec	 accuracy=28.836538	 loss=3.287729	 lr=0.099940
Epoch[036] Batch [1999]/[3760]	Speed: 116.394724 samples/sec	 accuracy=28.805469	 loss=3.289012	 lr=0.099939
Epoch[036] Batch [2049]/[3760]	Speed: 118.060301 samples/sec	 accuracy=28.796494	 loss=3.289133	 lr=0.099939
Epoch[036] Batch [2099]/[3760]	Speed: 115.907174 samples/sec	 accuracy=28.773065	 loss=3.289090	 lr=0.099938
Epoch[036] Batch [2149]/[3760]	Speed: 116.638598 samples/sec	 accuracy=28.779797	 loss=3.288911	 lr=0.099938
Epoch[036] Batch [2199]/[3760]	Speed: 117.598162 samples/sec	 accuracy=28.777699	 loss=3.288025	 lr=0.099937
Epoch[036] Batch [2249]/[3760]	Speed: 117.058258 samples/sec	 accuracy=28.770833	 loss=3.288971	 lr=0.099936
Epoch[036] Batch [2299]/[3760]	Speed: 117.586116 samples/sec	 accuracy=28.754755	 loss=3.289532	 lr=0.099936
Epoch[036] Batch [2349]/[3760]	Speed: 117.143530 samples/sec	 accuracy=28.767952	 loss=3.288220	 lr=0.099935
Epoch[036] Batch [2399]/[3760]	Speed: 116.432561 samples/sec	 accuracy=28.786458	 loss=3.288129	 lr=0.099934
Epoch[036] Batch [2449]/[3760]	Speed: 117.004712 samples/sec	 accuracy=28.764668	 loss=3.289366	 lr=0.099934
Epoch[036] Batch [2499]/[3760]	Speed: 117.423834 samples/sec	 accuracy=28.783125	 loss=3.288593	 lr=0.099933
Epoch[036] Batch [2549]/[3760]	Speed: 116.807944 samples/sec	 accuracy=28.778186	 loss=3.289354	 lr=0.099932
Epoch[036] Batch [2599]/[3760]	Speed: 116.814858 samples/sec	 accuracy=28.758413	 loss=3.289800	 lr=0.099932
Epoch[036] Batch [2649]/[3760]	Speed: 117.285255 samples/sec	 accuracy=28.742335	 loss=3.289613	 lr=0.099931
Epoch[036] Batch [2699]/[3760]	Speed: 117.449599 samples/sec	 accuracy=28.730324	 loss=3.290283	 lr=0.099930
Epoch[036] Batch [2749]/[3760]	Speed: 117.925721 samples/sec	 accuracy=28.740341	 loss=3.289718	 lr=0.099930
Epoch[036] Batch [2799]/[3760]	Speed: 117.034103 samples/sec	 accuracy=28.762835	 loss=3.289014	 lr=0.099929
Epoch[036] Batch [2849]/[3760]	Speed: 118.453123 samples/sec	 accuracy=28.762061	 loss=3.289415	 lr=0.099928
Epoch[036] Batch [2899]/[3760]	Speed: 117.317212 samples/sec	 accuracy=28.766703	 loss=3.289383	 lr=0.099927
Epoch[036] Batch [2949]/[3760]	Speed: 117.397547 samples/sec	 accuracy=28.762712	 loss=3.290263	 lr=0.099927
Epoch[036] Batch [2999]/[3760]	Speed: 116.539821 samples/sec	 accuracy=28.764583	 loss=3.290109	 lr=0.099926
Epoch[036] Batch [3049]/[3760]	Speed: 118.022913 samples/sec	 accuracy=28.779201	 loss=3.290579	 lr=0.099925
Epoch[036] Batch [3099]/[3760]	Speed: 117.469768 samples/sec	 accuracy=28.775202	 loss=3.290840	 lr=0.099925
Epoch[036] Batch [3149]/[3760]	Speed: 117.000098 samples/sec	 accuracy=28.751984	 loss=3.292069	 lr=0.099924
Epoch[036] Batch [3199]/[3760]	Speed: 116.714112 samples/sec	 accuracy=28.750488	 loss=3.292113	 lr=0.099923
Epoch[036] Batch [3249]/[3760]	Speed: 116.891625 samples/sec	 accuracy=28.765385	 loss=3.291151	 lr=0.099923
Epoch[036] Batch [3299]/[3760]	Speed: 117.658384 samples/sec	 accuracy=28.760890	 loss=3.291605	 lr=0.099922
Epoch[036] Batch [3349]/[3760]	Speed: 116.868052 samples/sec	 accuracy=28.756063	 loss=3.292636	 lr=0.099921
Epoch[036] Batch [3399]/[3760]	Speed: 117.393654 samples/sec	 accuracy=28.768382	 loss=3.291954	 lr=0.099920
Epoch[036] Batch [3449]/[3760]	Speed: 117.107907 samples/sec	 accuracy=28.754982	 loss=3.292043	 lr=0.099920
Epoch[036] Batch [3499]/[3760]	Speed: 117.818932 samples/sec	 accuracy=28.749107	 loss=3.292408	 lr=0.099919
Epoch[036] Batch [3549]/[3760]	Speed: 117.532689 samples/sec	 accuracy=28.772887	 loss=3.291578	 lr=0.099918
Epoch[036] Batch [3599]/[3760]	Speed: 116.986593 samples/sec	 accuracy=28.770833	 loss=3.291737	 lr=0.099917
Epoch[036] Batch [3649]/[3760]	Speed: 117.541750 samples/sec	 accuracy=28.770548	 loss=3.291975	 lr=0.099917
Epoch[036] Batch [3699]/[3760]	Speed: 117.467845 samples/sec	 accuracy=28.752111	 loss=3.291981	 lr=0.099916
Epoch[036] Batch [3749]/[3760]	Speed: 120.522585 samples/sec	 accuracy=28.779583	 loss=3.291776	 lr=0.099915
Batch [0049]/[0303]: acc-top1=27.156250 acc-top5=55.187500
Batch [0099]/[0303]: acc-top1=26.187500 acc-top5=53.000000
Batch [0149]/[0303]: acc-top1=27.468750 acc-top5=54.354167
Batch [0199]/[0303]: acc-top1=30.570313 acc-top5=56.976563
Batch [0249]/[0303]: acc-top1=29.625000 acc-top5=56.712500
Batch [0299]/[0303]: acc-top1=29.192708 acc-top5=56.489583
[Epoch 036] training: accuracy=28.782829	 loss=3.291787
[Epoch 036] speed: 116 samples/sec	time cost: 2181.849386
[Epoch 036] validation: acc-top1=29.202764 acc-top5=56.435644 loss=3.308287
Epoch[037] Batch [0049]/[3759]	Speed: 88.037554 samples/sec	 accuracy=27.031250	 loss=3.387527	 lr=0.099914
Epoch[037] Batch [0099]/[3759]	Speed: 116.954141 samples/sec	 accuracy=28.421875	 loss=3.327434	 lr=0.099914
Epoch[037] Batch [0149]/[3759]	Speed: 116.925624 samples/sec	 accuracy=28.468750	 loss=3.316905	 lr=0.099913
Epoch[037] Batch [0199]/[3759]	Speed: 116.641482 samples/sec	 accuracy=28.742188	 loss=3.304829	 lr=0.099912
Epoch[037] Batch [0249]/[3759]	Speed: 116.762062 samples/sec	 accuracy=28.606250	 loss=3.300091	 lr=0.099911
Epoch[037] Batch [0299]/[3759]	Speed: 115.624732 samples/sec	 accuracy=28.390625	 loss=3.311062	 lr=0.099910
Epoch[037] Batch [0349]/[3759]	Speed: 117.546509 samples/sec	 accuracy=28.468750	 loss=3.309567	 lr=0.099910
Epoch[037] Batch [0399]/[3759]	Speed: 117.026339 samples/sec	 accuracy=28.500000	 loss=3.310678	 lr=0.099909
Epoch[037] Batch [0449]/[3759]	Speed: 116.992801 samples/sec	 accuracy=28.559028	 loss=3.305438	 lr=0.099908
Epoch[037] Batch [0499]/[3759]	Speed: 117.198421 samples/sec	 accuracy=28.537500	 loss=3.306657	 lr=0.099907
Epoch[037] Batch [0549]/[3759]	Speed: 116.471978 samples/sec	 accuracy=28.656250	 loss=3.300642	 lr=0.099907
Epoch[037] Batch [0599]/[3759]	Speed: 116.833946 samples/sec	 accuracy=28.658854	 loss=3.297797	 lr=0.099906
Epoch[037] Batch [0649]/[3759]	Speed: 116.093714 samples/sec	 accuracy=28.649038	 loss=3.295900	 lr=0.099905
Epoch[037] Batch [0699]/[3759]	Speed: 116.114159 samples/sec	 accuracy=28.767857	 loss=3.293871	 lr=0.099904
Epoch[037] Batch [0749]/[3759]	Speed: 116.669986 samples/sec	 accuracy=28.785417	 loss=3.293736	 lr=0.099903
Epoch[037] Batch [0799]/[3759]	Speed: 117.757493 samples/sec	 accuracy=28.830078	 loss=3.294564	 lr=0.099903
Epoch[037] Batch [0849]/[3759]	Speed: 116.915256 samples/sec	 accuracy=28.862132	 loss=3.293878	 lr=0.099902
Epoch[037] Batch [0899]/[3759]	Speed: 117.432282 samples/sec	 accuracy=28.958333	 loss=3.290969	 lr=0.099901
Epoch[037] Batch [0949]/[3759]	Speed: 117.387872 samples/sec	 accuracy=28.927632	 loss=3.290307	 lr=0.099900
Epoch[037] Batch [0999]/[3759]	Speed: 117.223998 samples/sec	 accuracy=28.928125	 loss=3.288872	 lr=0.099899
Epoch[037] Batch [1049]/[3759]	Speed: 117.905255 samples/sec	 accuracy=28.919643	 loss=3.287192	 lr=0.099899
Epoch[037] Batch [1099]/[3759]	Speed: 116.549235 samples/sec	 accuracy=28.948864	 loss=3.286291	 lr=0.099898
Epoch[037] Batch [1149]/[3759]	Speed: 116.889721 samples/sec	 accuracy=28.948370	 loss=3.285508	 lr=0.099897
Epoch[037] Batch [1199]/[3759]	Speed: 116.850608 samples/sec	 accuracy=28.921875	 loss=3.288100	 lr=0.099896
Epoch[037] Batch [1249]/[3759]	Speed: 116.834920 samples/sec	 accuracy=28.878750	 loss=3.291129	 lr=0.099895
Epoch[037] Batch [1299]/[3759]	Speed: 117.311891 samples/sec	 accuracy=28.913462	 loss=3.289995	 lr=0.099894
Epoch[037] Batch [1349]/[3759]	Speed: 117.164636 samples/sec	 accuracy=28.958333	 loss=3.288859	 lr=0.099894
Epoch[037] Batch [1399]/[3759]	Speed: 117.273861 samples/sec	 accuracy=28.919643	 loss=3.291129	 lr=0.099893
Epoch[037] Batch [1449]/[3759]	Speed: 117.205160 samples/sec	 accuracy=28.935345	 loss=3.290256	 lr=0.099892
Epoch[037] Batch [1499]/[3759]	Speed: 116.507655 samples/sec	 accuracy=28.993750	 loss=3.286934	 lr=0.099891
Epoch[037] Batch [1549]/[3759]	Speed: 117.308508 samples/sec	 accuracy=29.011089	 loss=3.285703	 lr=0.099890
Epoch[037] Batch [1599]/[3759]	Speed: 117.358842 samples/sec	 accuracy=28.985352	 loss=3.286476	 lr=0.099889
Epoch[037] Batch [1649]/[3759]	Speed: 116.802304 samples/sec	 accuracy=28.984848	 loss=3.287144	 lr=0.099888
Epoch[037] Batch [1699]/[3759]	Speed: 117.270662 samples/sec	 accuracy=28.988051	 loss=3.287945	 lr=0.099888
Epoch[037] Batch [1749]/[3759]	Speed: 117.740717 samples/sec	 accuracy=28.981250	 loss=3.289349	 lr=0.099887
Epoch[037] Batch [1799]/[3759]	Speed: 117.660596 samples/sec	 accuracy=29.004340	 loss=3.288064	 lr=0.099886
Epoch[037] Batch [1849]/[3759]	Speed: 117.596246 samples/sec	 accuracy=28.998311	 loss=3.289127	 lr=0.099885
Epoch[037] Batch [1899]/[3759]	Speed: 116.873151 samples/sec	 accuracy=28.965461	 loss=3.289783	 lr=0.099884
Epoch[037] Batch [1949]/[3759]	Speed: 117.247615 samples/sec	 accuracy=28.959936	 loss=3.289942	 lr=0.099883
Epoch[037] Batch [1999]/[3759]	Speed: 117.473039 samples/sec	 accuracy=28.945312	 loss=3.290227	 lr=0.099882
Epoch[037] Batch [2049]/[3759]	Speed: 117.423275 samples/sec	 accuracy=28.968750	 loss=3.289005	 lr=0.099881
Epoch[037] Batch [2099]/[3759]	Speed: 117.144436 samples/sec	 accuracy=28.962054	 loss=3.288958	 lr=0.099881
Epoch[037] Batch [2149]/[3759]	Speed: 116.946764 samples/sec	 accuracy=28.922238	 loss=3.290507	 lr=0.099880
Epoch[037] Batch [2199]/[3759]	Speed: 117.295072 samples/sec	 accuracy=28.883523	 loss=3.291953	 lr=0.099879
Epoch[037] Batch [2249]/[3759]	Speed: 117.518131 samples/sec	 accuracy=28.861806	 loss=3.292158	 lr=0.099878
Epoch[037] Batch [2299]/[3759]	Speed: 118.226742 samples/sec	 accuracy=28.867527	 loss=3.291216	 lr=0.099877
Epoch[037] Batch [2349]/[3759]	Speed: 117.149839 samples/sec	 accuracy=28.875000	 loss=3.291400	 lr=0.099876
Epoch[037] Batch [2399]/[3759]	Speed: 117.825229 samples/sec	 accuracy=28.910156	 loss=3.289922	 lr=0.099875
Epoch[037] Batch [2449]/[3759]	Speed: 117.563684 samples/sec	 accuracy=28.910077	 loss=3.289907	 lr=0.099874
Epoch[037] Batch [2499]/[3759]	Speed: 117.354094 samples/sec	 accuracy=28.928125	 loss=3.289675	 lr=0.099873
Epoch[037] Batch [2549]/[3759]	Speed: 116.927278 samples/sec	 accuracy=28.901961	 loss=3.289721	 lr=0.099872
Epoch[037] Batch [2599]/[3759]	Speed: 118.048910 samples/sec	 accuracy=28.929688	 loss=3.289009	 lr=0.099871
Epoch[037] Batch [2649]/[3759]	Speed: 117.291953 samples/sec	 accuracy=28.919811	 loss=3.289683	 lr=0.099871
Epoch[037] Batch [2699]/[3759]	Speed: 117.455528 samples/sec	 accuracy=28.929398	 loss=3.289222	 lr=0.099870
Epoch[037] Batch [2749]/[3759]	Speed: 118.462796 samples/sec	 accuracy=28.936364	 loss=3.289347	 lr=0.099869
Epoch[037] Batch [2799]/[3759]	Speed: 117.021336 samples/sec	 accuracy=28.926339	 loss=3.289460	 lr=0.099868
Epoch[037] Batch [2849]/[3759]	Speed: 116.859088 samples/sec	 accuracy=28.921601	 loss=3.288891	 lr=0.099867
Epoch[037] Batch [2899]/[3759]	Speed: 116.810286 samples/sec	 accuracy=28.913793	 loss=3.287511	 lr=0.099866
Epoch[037] Batch [2949]/[3759]	Speed: 116.796380 samples/sec	 accuracy=28.885064	 loss=3.288139	 lr=0.099865
Epoch[037] Batch [2999]/[3759]	Speed: 117.546207 samples/sec	 accuracy=28.882812	 loss=3.288180	 lr=0.099864
Epoch[037] Batch [3049]/[3759]	Speed: 117.111973 samples/sec	 accuracy=28.881148	 loss=3.288673	 lr=0.099863
Epoch[037] Batch [3099]/[3759]	Speed: 117.289714 samples/sec	 accuracy=28.889113	 loss=3.288073	 lr=0.099862
Epoch[037] Batch [3149]/[3759]	Speed: 116.740869 samples/sec	 accuracy=28.901786	 loss=3.288517	 lr=0.099861
Epoch[037] Batch [3199]/[3759]	Speed: 118.072941 samples/sec	 accuracy=28.871582	 loss=3.289501	 lr=0.099860
Epoch[037] Batch [3249]/[3759]	Speed: 117.180331 samples/sec	 accuracy=28.868269	 loss=3.290116	 lr=0.099859
Epoch[037] Batch [3299]/[3759]	Speed: 117.226187 samples/sec	 accuracy=28.856534	 loss=3.289950	 lr=0.099858
Epoch[037] Batch [3349]/[3759]	Speed: 117.045590 samples/sec	 accuracy=28.854478	 loss=3.290557	 lr=0.099857
Epoch[037] Batch [3399]/[3759]	Speed: 117.188685 samples/sec	 accuracy=28.862592	 loss=3.290015	 lr=0.099856
Epoch[037] Batch [3449]/[3759]	Speed: 117.308194 samples/sec	 accuracy=28.865036	 loss=3.290539	 lr=0.099855
Epoch[037] Batch [3499]/[3759]	Speed: 117.660163 samples/sec	 accuracy=28.883036	 loss=3.290349	 lr=0.099854
Epoch[037] Batch [3549]/[3759]	Speed: 117.222409 samples/sec	 accuracy=28.898327	 loss=3.289989	 lr=0.099853
Epoch[037] Batch [3599]/[3759]	Speed: 117.929725 samples/sec	 accuracy=28.893229	 loss=3.290108	 lr=0.099852
Epoch[037] Batch [3649]/[3759]	Speed: 117.372819 samples/sec	 accuracy=28.896832	 loss=3.289160	 lr=0.099851
Epoch[037] Batch [3699]/[3759]	Speed: 117.639269 samples/sec	 accuracy=28.898226	 loss=3.289796	 lr=0.099850
Epoch[037] Batch [3749]/[3759]	Speed: 120.685823 samples/sec	 accuracy=28.884583	 loss=3.290758	 lr=0.099849
Batch [0049]/[0303]: acc-top1=29.093750 acc-top5=55.968750
Batch [0099]/[0303]: acc-top1=26.609375 acc-top5=53.093750
Batch [0149]/[0303]: acc-top1=28.166667 acc-top5=54.395833
Batch [0199]/[0303]: acc-top1=31.375000 acc-top5=56.820312
Batch [0249]/[0303]: acc-top1=30.875000 acc-top5=56.925000
Batch [0299]/[0303]: acc-top1=30.250000 acc-top5=56.963542
[Epoch 037] training: accuracy=28.889831	 loss=3.290673
[Epoch 037] speed: 116 samples/sec	time cost: 2181.729977
[Epoch 037] validation: acc-top1=30.337252 acc-top5=57.023515 loss=3.295321
Epoch[038] Batch [0049]/[3760]	Speed: 86.606035 samples/sec	 accuracy=28.812500	 loss=3.282275	 lr=0.099848
Epoch[038] Batch [0099]/[3760]	Speed: 117.771641 samples/sec	 accuracy=29.609375	 loss=3.245500	 lr=0.099847
Epoch[038] Batch [0149]/[3760]	Speed: 117.162376 samples/sec	 accuracy=29.447917	 loss=3.252926	 lr=0.099846
Epoch[038] Batch [0199]/[3760]	Speed: 117.131882 samples/sec	 accuracy=29.218750	 loss=3.257660	 lr=0.099845
Epoch[038] Batch [0249]/[3760]	Speed: 116.724871 samples/sec	 accuracy=29.325000	 loss=3.254263	 lr=0.099844
Epoch[038] Batch [0299]/[3760]	Speed: 117.208339 samples/sec	 accuracy=29.322917	 loss=3.249357	 lr=0.099843
Epoch[038] Batch [0349]/[3760]	Speed: 117.545485 samples/sec	 accuracy=29.227679	 loss=3.256825	 lr=0.099842
Epoch[038] Batch [0399]/[3760]	Speed: 117.304560 samples/sec	 accuracy=29.261719	 loss=3.257036	 lr=0.099841
Epoch[038] Batch [0449]/[3760]	Speed: 117.406888 samples/sec	 accuracy=29.218750	 loss=3.258177	 lr=0.099840
Epoch[038] Batch [0499]/[3760]	Speed: 117.279604 samples/sec	 accuracy=29.281250	 loss=3.257581	 lr=0.099839
Epoch[038] Batch [0549]/[3760]	Speed: 116.601287 samples/sec	 accuracy=29.241477	 loss=3.261702	 lr=0.099838
Epoch[038] Batch [0599]/[3760]	Speed: 116.509115 samples/sec	 accuracy=29.367188	 loss=3.257481	 lr=0.099837
Epoch[038] Batch [0649]/[3760]	Speed: 116.342332 samples/sec	 accuracy=29.314904	 loss=3.260404	 lr=0.099836
Epoch[038] Batch [0699]/[3760]	Speed: 117.184606 samples/sec	 accuracy=29.227679	 loss=3.263884	 lr=0.099835
Epoch[038] Batch [0749]/[3760]	Speed: 117.060415 samples/sec	 accuracy=29.220833	 loss=3.262753	 lr=0.099834
Epoch[038] Batch [0799]/[3760]	Speed: 116.634060 samples/sec	 accuracy=29.191406	 loss=3.261881	 lr=0.099833
Epoch[038] Batch [0849]/[3760]	Speed: 117.060378 samples/sec	 accuracy=29.187500	 loss=3.263177	 lr=0.099832
Epoch[038] Batch [0899]/[3760]	Speed: 117.472183 samples/sec	 accuracy=29.156250	 loss=3.267973	 lr=0.099831
Epoch[038] Batch [0949]/[3760]	Speed: 116.608984 samples/sec	 accuracy=29.134868	 loss=3.267574	 lr=0.099830
Epoch[038] Batch [0999]/[3760]	Speed: 117.189569 samples/sec	 accuracy=29.145312	 loss=3.268102	 lr=0.099828
Epoch[038] Batch [1049]/[3760]	Speed: 116.989120 samples/sec	 accuracy=29.107143	 loss=3.269408	 lr=0.099827
Epoch[038] Batch [1099]/[3760]	Speed: 117.850332 samples/sec	 accuracy=29.150568	 loss=3.268611	 lr=0.099826
Epoch[038] Batch [1149]/[3760]	Speed: 117.002274 samples/sec	 accuracy=29.191576	 loss=3.270022	 lr=0.099825
Epoch[038] Batch [1199]/[3760]	Speed: 117.620835 samples/sec	 accuracy=29.131510	 loss=3.273010	 lr=0.099824
Epoch[038] Batch [1249]/[3760]	Speed: 117.222608 samples/sec	 accuracy=29.098750	 loss=3.274006	 lr=0.099823
Epoch[038] Batch [1299]/[3760]	Speed: 117.815533 samples/sec	 accuracy=29.100962	 loss=3.275735	 lr=0.099822
Epoch[038] Batch [1349]/[3760]	Speed: 116.543925 samples/sec	 accuracy=29.106481	 loss=3.275621	 lr=0.099821
Epoch[038] Batch [1399]/[3760]	Speed: 117.642361 samples/sec	 accuracy=29.097098	 loss=3.274348	 lr=0.099820
Epoch[038] Batch [1449]/[3760]	Speed: 117.397965 samples/sec	 accuracy=29.059267	 loss=3.275937	 lr=0.099819
Epoch[038] Batch [1499]/[3760]	Speed: 117.298796 samples/sec	 accuracy=29.011458	 loss=3.277894	 lr=0.099818
Epoch[038] Batch [1549]/[3760]	Speed: 117.687154 samples/sec	 accuracy=29.003024	 loss=3.276846	 lr=0.099817
Epoch[038] Batch [1599]/[3760]	Speed: 116.992679 samples/sec	 accuracy=29.002930	 loss=3.276374	 lr=0.099815
Epoch[038] Batch [1649]/[3760]	Speed: 117.421692 samples/sec	 accuracy=29.000947	 loss=3.275279	 lr=0.099814
Epoch[038] Batch [1699]/[3760]	Speed: 117.842955 samples/sec	 accuracy=29.010110	 loss=3.276478	 lr=0.099813
Epoch[038] Batch [1749]/[3760]	Speed: 117.329632 samples/sec	 accuracy=29.000000	 loss=3.276285	 lr=0.099812
Epoch[038] Batch [1799]/[3760]	Speed: 117.216081 samples/sec	 accuracy=28.973090	 loss=3.277650	 lr=0.099811
Epoch[038] Batch [1849]/[3760]	Speed: 117.280582 samples/sec	 accuracy=28.934966	 loss=3.278720	 lr=0.099810
Epoch[038] Batch [1899]/[3760]	Speed: 117.335920 samples/sec	 accuracy=28.938322	 loss=3.279199	 lr=0.099809
Epoch[038] Batch [1949]/[3760]	Speed: 116.938874 samples/sec	 accuracy=28.935897	 loss=3.279564	 lr=0.099808
Epoch[038] Batch [1999]/[3760]	Speed: 117.222477 samples/sec	 accuracy=28.928125	 loss=3.278862	 lr=0.099806
Epoch[038] Batch [2049]/[3760]	Speed: 116.490903 samples/sec	 accuracy=28.903963	 loss=3.280322	 lr=0.099805
Epoch[038] Batch [2099]/[3760]	Speed: 117.174200 samples/sec	 accuracy=28.874256	 loss=3.280919	 lr=0.099804
Epoch[038] Batch [2149]/[3760]	Speed: 117.381880 samples/sec	 accuracy=28.832849	 loss=3.282909	 lr=0.099803
Epoch[038] Batch [2199]/[3760]	Speed: 117.355291 samples/sec	 accuracy=28.826705	 loss=3.284955	 lr=0.099802
Epoch[038] Batch [2249]/[3760]	Speed: 117.726417 samples/sec	 accuracy=28.829167	 loss=3.285804	 lr=0.099801
Epoch[038] Batch [2299]/[3760]	Speed: 116.723604 samples/sec	 accuracy=28.851902	 loss=3.285435	 lr=0.099800
Epoch[038] Batch [2349]/[3760]	Speed: 117.505929 samples/sec	 accuracy=28.839096	 loss=3.286884	 lr=0.099798
Epoch[038] Batch [2399]/[3760]	Speed: 117.819857 samples/sec	 accuracy=28.840495	 loss=3.285993	 lr=0.099797
Epoch[038] Batch [2449]/[3760]	Speed: 117.337499 samples/sec	 accuracy=28.843750	 loss=3.286374	 lr=0.099796
Epoch[038] Batch [2499]/[3760]	Speed: 117.877263 samples/sec	 accuracy=28.845625	 loss=3.286365	 lr=0.099795
Epoch[038] Batch [2549]/[3760]	Speed: 116.841742 samples/sec	 accuracy=28.832108	 loss=3.286285	 lr=0.099794
Epoch[038] Batch [2599]/[3760]	Speed: 116.349508 samples/sec	 accuracy=28.826322	 loss=3.285593	 lr=0.099793
Epoch[038] Batch [2649]/[3760]	Speed: 118.216870 samples/sec	 accuracy=28.836085	 loss=3.284808	 lr=0.099791
Epoch[038] Batch [2699]/[3760]	Speed: 117.911625 samples/sec	 accuracy=28.828704	 loss=3.285603	 lr=0.099790
Epoch[038] Batch [2749]/[3760]	Speed: 116.982708 samples/sec	 accuracy=28.850568	 loss=3.286348	 lr=0.099789
Epoch[038] Batch [2799]/[3760]	Speed: 117.608103 samples/sec	 accuracy=28.832589	 loss=3.287353	 lr=0.099788
Epoch[038] Batch [2849]/[3760]	Speed: 116.961120 samples/sec	 accuracy=28.832237	 loss=3.286606	 lr=0.099787
Epoch[038] Batch [2899]/[3760]	Speed: 117.440184 samples/sec	 accuracy=28.857220	 loss=3.286120	 lr=0.099786
Epoch[038] Batch [2949]/[3760]	Speed: 117.138863 samples/sec	 accuracy=28.853284	 loss=3.286965	 lr=0.099784
Epoch[038] Batch [2999]/[3760]	Speed: 117.415432 samples/sec	 accuracy=28.877083	 loss=3.286158	 lr=0.099783
Epoch[038] Batch [3049]/[3760]	Speed: 117.441432 samples/sec	 accuracy=28.877049	 loss=3.286046	 lr=0.099782
Epoch[038] Batch [3099]/[3760]	Speed: 117.551018 samples/sec	 accuracy=28.901210	 loss=3.285605	 lr=0.099781
Epoch[038] Batch [3149]/[3760]	Speed: 117.394543 samples/sec	 accuracy=28.904266	 loss=3.284821	 lr=0.099780
Epoch[038] Batch [3199]/[3760]	Speed: 117.134762 samples/sec	 accuracy=28.895508	 loss=3.286327	 lr=0.099778
Epoch[038] Batch [3249]/[3760]	Speed: 117.147106 samples/sec	 accuracy=28.891827	 loss=3.286849	 lr=0.099777
Epoch[038] Batch [3299]/[3760]	Speed: 117.478448 samples/sec	 accuracy=28.864110	 loss=3.287819	 lr=0.099776
Epoch[038] Batch [3349]/[3760]	Speed: 116.683158 samples/sec	 accuracy=28.856810	 loss=3.287727	 lr=0.099775
Epoch[038] Batch [3399]/[3760]	Speed: 117.732171 samples/sec	 accuracy=28.829044	 loss=3.288498	 lr=0.099773
Epoch[038] Batch [3449]/[3760]	Speed: 117.551197 samples/sec	 accuracy=28.834692	 loss=3.288545	 lr=0.099772
Epoch[038] Batch [3499]/[3760]	Speed: 116.697882 samples/sec	 accuracy=28.843750	 loss=3.287867	 lr=0.099771
Epoch[038] Batch [3549]/[3760]	Speed: 116.837937 samples/sec	 accuracy=28.850352	 loss=3.288145	 lr=0.099770
Epoch[038] Batch [3599]/[3760]	Speed: 117.828783 samples/sec	 accuracy=28.846354	 loss=3.288016	 lr=0.099768
Epoch[038] Batch [3649]/[3760]	Speed: 116.967695 samples/sec	 accuracy=28.854024	 loss=3.287734	 lr=0.099767
Epoch[038] Batch [3699]/[3760]	Speed: 117.710845 samples/sec	 accuracy=28.842483	 loss=3.287897	 lr=0.099766
Epoch[038] Batch [3749]/[3760]	Speed: 120.203809 samples/sec	 accuracy=28.836250	 loss=3.288290	 lr=0.099765
Batch [0049]/[0303]: acc-top1=30.218750 acc-top5=57.000000
Batch [0099]/[0303]: acc-top1=28.750000 acc-top5=55.218750
Batch [0149]/[0303]: acc-top1=29.604167 acc-top5=55.791667
Batch [0199]/[0303]: acc-top1=32.250000 acc-top5=57.734375
Batch [0249]/[0303]: acc-top1=30.956250 acc-top5=57.787500
Batch [0299]/[0303]: acc-top1=30.489583 acc-top5=57.734375
[Epoch 038] training: accuracy=28.843085	 loss=3.288165
[Epoch 038] speed: 116 samples/sec	time cost: 2182.094711
[Epoch 038] validation: acc-top1=30.486799 acc-top5=57.735149 loss=3.218359
Epoch[039] Batch [0049]/[3760]	Speed: 88.968714 samples/sec	 accuracy=27.718750	 loss=3.276131	 lr=0.099763
Epoch[039] Batch [0099]/[3760]	Speed: 117.827290 samples/sec	 accuracy=28.703125	 loss=3.255109	 lr=0.099762
Epoch[039] Batch [0149]/[3760]	Speed: 117.671103 samples/sec	 accuracy=28.562500	 loss=3.265388	 lr=0.099761
Epoch[039] Batch [0199]/[3760]	Speed: 117.380311 samples/sec	 accuracy=28.585938	 loss=3.257746	 lr=0.099759
Epoch[039] Batch [0249]/[3760]	Speed: 117.984899 samples/sec	 accuracy=28.606250	 loss=3.269641	 lr=0.099758
Epoch[039] Batch [0299]/[3760]	Speed: 117.166608 samples/sec	 accuracy=28.515625	 loss=3.277160	 lr=0.099757
Epoch[039] Batch [0349]/[3760]	Speed: 118.036607 samples/sec	 accuracy=28.589286	 loss=3.277812	 lr=0.099756
Epoch[039] Batch [0399]/[3760]	Speed: 116.862713 samples/sec	 accuracy=28.695312	 loss=3.269609	 lr=0.099754
Epoch[039] Batch [0449]/[3760]	Speed: 116.928496 samples/sec	 accuracy=28.763889	 loss=3.271164	 lr=0.099753
Epoch[039] Batch [0499]/[3760]	Speed: 117.529563 samples/sec	 accuracy=28.818750	 loss=3.263874	 lr=0.099752
Epoch[039] Batch [0549]/[3760]	Speed: 116.775340 samples/sec	 accuracy=28.846591	 loss=3.264232	 lr=0.099751
Epoch[039] Batch [0599]/[3760]	Speed: 117.191761 samples/sec	 accuracy=28.895833	 loss=3.265439	 lr=0.099749
Epoch[039] Batch [0649]/[3760]	Speed: 117.164357 samples/sec	 accuracy=28.778846	 loss=3.272296	 lr=0.099748
Epoch[039] Batch [0699]/[3760]	Speed: 117.105876 samples/sec	 accuracy=28.839286	 loss=3.271299	 lr=0.099747
Epoch[039] Batch [0749]/[3760]	Speed: 117.361837 samples/sec	 accuracy=28.820833	 loss=3.272026	 lr=0.099745
Epoch[039] Batch [0799]/[3760]	Speed: 118.261584 samples/sec	 accuracy=28.833984	 loss=3.269935	 lr=0.099744
Epoch[039] Batch [0849]/[3760]	Speed: 116.613856 samples/sec	 accuracy=28.878676	 loss=3.269364	 lr=0.099743
Epoch[039] Batch [0899]/[3760]	Speed: 117.398400 samples/sec	 accuracy=28.871528	 loss=3.271278	 lr=0.099741
Epoch[039] Batch [0949]/[3760]	Speed: 117.159416 samples/sec	 accuracy=28.893092	 loss=3.268844	 lr=0.099740
Epoch[039] Batch [0999]/[3760]	Speed: 117.106405 samples/sec	 accuracy=28.935937	 loss=3.267675	 lr=0.099739
Epoch[039] Batch [1049]/[3760]	Speed: 117.435436 samples/sec	 accuracy=28.913690	 loss=3.267590	 lr=0.099738
Epoch[039] Batch [1099]/[3760]	Speed: 117.479359 samples/sec	 accuracy=28.913352	 loss=3.269141	 lr=0.099736
Epoch[039] Batch [1149]/[3760]	Speed: 117.177044 samples/sec	 accuracy=28.922554	 loss=3.271935	 lr=0.099735
Epoch[039] Batch [1199]/[3760]	Speed: 116.988433 samples/sec	 accuracy=28.912760	 loss=3.272009	 lr=0.099734
Epoch[039] Batch [1249]/[3760]	Speed: 117.301033 samples/sec	 accuracy=28.908750	 loss=3.273066	 lr=0.099732
Epoch[039] Batch [1299]/[3760]	Speed: 117.492278 samples/sec	 accuracy=28.847356	 loss=3.275925	 lr=0.099731
Epoch[039] Batch [1349]/[3760]	Speed: 117.640441 samples/sec	 accuracy=28.855324	 loss=3.278148	 lr=0.099730
Epoch[039] Batch [1399]/[3760]	Speed: 116.807926 samples/sec	 accuracy=28.850446	 loss=3.276707	 lr=0.099728
Epoch[039] Batch [1449]/[3760]	Speed: 117.458097 samples/sec	 accuracy=28.849138	 loss=3.277122	 lr=0.099727
Epoch[039] Batch [1499]/[3760]	Speed: 117.649923 samples/sec	 accuracy=28.866667	 loss=3.276305	 lr=0.099726
Epoch[039] Batch [1549]/[3760]	Speed: 117.356251 samples/sec	 accuracy=28.886089	 loss=3.274028	 lr=0.099724
Epoch[039] Batch [1599]/[3760]	Speed: 117.411200 samples/sec	 accuracy=28.851562	 loss=3.275605	 lr=0.099723
Epoch[039] Batch [1649]/[3760]	Speed: 117.549523 samples/sec	 accuracy=28.863636	 loss=3.277342	 lr=0.099721
Epoch[039] Batch [1699]/[3760]	Speed: 116.852346 samples/sec	 accuracy=28.875919	 loss=3.276297	 lr=0.099720
Epoch[039] Batch [1749]/[3760]	Speed: 117.576240 samples/sec	 accuracy=28.880357	 loss=3.276291	 lr=0.099719
Epoch[039] Batch [1799]/[3760]	Speed: 117.545871 samples/sec	 accuracy=28.895833	 loss=3.275736	 lr=0.099717
Epoch[039] Batch [1849]/[3760]	Speed: 117.647338 samples/sec	 accuracy=28.910473	 loss=3.275645	 lr=0.099716
Epoch[039] Batch [1899]/[3760]	Speed: 117.557702 samples/sec	 accuracy=28.890625	 loss=3.277147	 lr=0.099715
Epoch[039] Batch [1949]/[3760]	Speed: 117.827956 samples/sec	 accuracy=28.927083	 loss=3.277479	 lr=0.099713
Epoch[039] Batch [1999]/[3760]	Speed: 117.677028 samples/sec	 accuracy=28.881250	 loss=3.277871	 lr=0.099712
Epoch[039] Batch [2049]/[3760]	Speed: 117.331140 samples/sec	 accuracy=28.864329	 loss=3.279978	 lr=0.099710
Epoch[039] Batch [2099]/[3760]	Speed: 117.637340 samples/sec	 accuracy=28.848958	 loss=3.281177	 lr=0.099709
Epoch[039] Batch [2149]/[3760]	Speed: 117.711890 samples/sec	 accuracy=28.865552	 loss=3.280813	 lr=0.099708
Epoch[039] Batch [2199]/[3760]	Speed: 117.857558 samples/sec	 accuracy=28.846591	 loss=3.282195	 lr=0.099706
Epoch[039] Batch [2249]/[3760]	Speed: 117.392675 samples/sec	 accuracy=28.829167	 loss=3.283060	 lr=0.099705
Epoch[039] Batch [2299]/[3760]	Speed: 118.234458 samples/sec	 accuracy=28.840353	 loss=3.282565	 lr=0.099704
Epoch[039] Batch [2349]/[3760]	Speed: 118.011771 samples/sec	 accuracy=28.851064	 loss=3.281322	 lr=0.099702
Epoch[039] Batch [2399]/[3760]	Speed: 117.842112 samples/sec	 accuracy=28.876953	 loss=3.280902	 lr=0.099701
Epoch[039] Batch [2449]/[3760]	Speed: 117.444413 samples/sec	 accuracy=28.882015	 loss=3.281864	 lr=0.099699
Epoch[039] Batch [2499]/[3760]	Speed: 117.392796 samples/sec	 accuracy=28.903750	 loss=3.281282	 lr=0.099698
Epoch[039] Batch [2549]/[3760]	Speed: 117.049951 samples/sec	 accuracy=28.915441	 loss=3.281766	 lr=0.099696
Epoch[039] Batch [2599]/[3760]	Speed: 117.626456 samples/sec	 accuracy=28.900841	 loss=3.281933	 lr=0.099695
Epoch[039] Batch [2649]/[3760]	Speed: 117.102420 samples/sec	 accuracy=28.893278	 loss=3.281612	 lr=0.099694
Epoch[039] Batch [2699]/[3760]	Speed: 117.180621 samples/sec	 accuracy=28.907986	 loss=3.281154	 lr=0.099692
Epoch[039] Batch [2749]/[3760]	Speed: 117.077957 samples/sec	 accuracy=28.920455	 loss=3.281111	 lr=0.099691
Epoch[039] Batch [2799]/[3760]	Speed: 117.397914 samples/sec	 accuracy=28.936942	 loss=3.280180	 lr=0.099689
Epoch[039] Batch [2849]/[3760]	Speed: 117.192889 samples/sec	 accuracy=28.939145	 loss=3.279943	 lr=0.099688
Epoch[039] Batch [2899]/[3760]	Speed: 117.097616 samples/sec	 accuracy=28.973060	 loss=3.278943	 lr=0.099686
Epoch[039] Batch [2949]/[3760]	Speed: 117.562621 samples/sec	 accuracy=28.973517	 loss=3.278670	 lr=0.099685
Epoch[039] Batch [2999]/[3760]	Speed: 118.056830 samples/sec	 accuracy=28.966667	 loss=3.280051	 lr=0.099684
Epoch[039] Batch [3049]/[3760]	Speed: 117.368921 samples/sec	 accuracy=28.955943	 loss=3.280764	 lr=0.099682
Epoch[039] Batch [3099]/[3760]	Speed: 117.709668 samples/sec	 accuracy=28.960181	 loss=3.280457	 lr=0.099681
Epoch[039] Batch [3149]/[3760]	Speed: 117.744913 samples/sec	 accuracy=28.958333	 loss=3.281053	 lr=0.099679
Epoch[039] Batch [3199]/[3760]	Speed: 117.627432 samples/sec	 accuracy=28.950684	 loss=3.281320	 lr=0.099678
Epoch[039] Batch [3249]/[3760]	Speed: 116.985288 samples/sec	 accuracy=28.943750	 loss=3.281938	 lr=0.099676
Epoch[039] Batch [3299]/[3760]	Speed: 117.651523 samples/sec	 accuracy=28.943182	 loss=3.282249	 lr=0.099675
Epoch[039] Batch [3349]/[3760]	Speed: 117.808839 samples/sec	 accuracy=28.952892	 loss=3.282015	 lr=0.099673
Epoch[039] Batch [3399]/[3760]	Speed: 117.372992 samples/sec	 accuracy=28.957721	 loss=3.281656	 lr=0.099672
Epoch[039] Batch [3449]/[3760]	Speed: 117.452466 samples/sec	 accuracy=28.940217	 loss=3.282310	 lr=0.099670
Epoch[039] Batch [3499]/[3760]	Speed: 117.951347 samples/sec	 accuracy=28.936607	 loss=3.282434	 lr=0.099669
Epoch[039] Batch [3549]/[3760]	Speed: 117.379863 samples/sec	 accuracy=28.935299	 loss=3.282876	 lr=0.099667
Epoch[039] Batch [3599]/[3760]	Speed: 118.026905 samples/sec	 accuracy=28.931424	 loss=3.282893	 lr=0.099666
Epoch[039] Batch [3649]/[3760]	Speed: 117.417026 samples/sec	 accuracy=28.926370	 loss=3.283188	 lr=0.099664
Epoch[039] Batch [3699]/[3760]	Speed: 117.315481 samples/sec	 accuracy=28.924831	 loss=3.283667	 lr=0.099663
Epoch[039] Batch [3749]/[3760]	Speed: 120.334297 samples/sec	 accuracy=28.920417	 loss=3.283561	 lr=0.099661
Batch [0049]/[0303]: acc-top1=25.625000 acc-top5=52.531250
Batch [0099]/[0303]: acc-top1=24.203125 acc-top5=50.125000
Batch [0149]/[0303]: acc-top1=25.770833 acc-top5=52.322917
Batch [0199]/[0303]: acc-top1=28.093750 acc-top5=54.085938
Batch [0249]/[0303]: acc-top1=27.731250 acc-top5=54.168750
Batch [0299]/[0303]: acc-top1=27.364583 acc-top5=53.770833
[Epoch 039] training: accuracy=28.918717	 loss=3.283395
[Epoch 039] speed: 116 samples/sec	time cost: 2176.714819
[Epoch 039] validation: acc-top1=27.310231 acc-top5=53.759282 loss=3.487343
Epoch[040] Batch [0049]/[3759]	Speed: 88.643856 samples/sec	 accuracy=30.218750	 loss=3.232169	 lr=0.099660
Epoch[040] Batch [0099]/[3759]	Speed: 117.843205 samples/sec	 accuracy=29.484375	 loss=3.261567	 lr=0.099658
Epoch[040] Batch [0149]/[3759]	Speed: 117.834588 samples/sec	 accuracy=29.375000	 loss=3.258743	 lr=0.099657
Epoch[040] Batch [0199]/[3759]	Speed: 117.435204 samples/sec	 accuracy=29.367188	 loss=3.268436	 lr=0.099655
Epoch[040] Batch [0249]/[3759]	Speed: 118.554289 samples/sec	 accuracy=29.212500	 loss=3.267192	 lr=0.099654
Epoch[040] Batch [0299]/[3759]	Speed: 116.806924 samples/sec	 accuracy=29.125000	 loss=3.271523	 lr=0.099652
Epoch[040] Batch [0349]/[3759]	Speed: 118.105147 samples/sec	 accuracy=29.236607	 loss=3.274109	 lr=0.099651
Epoch[040] Batch [0399]/[3759]	Speed: 117.927254 samples/sec	 accuracy=29.230469	 loss=3.274284	 lr=0.099649
Epoch[040] Batch [0449]/[3759]	Speed: 117.247492 samples/sec	 accuracy=29.142361	 loss=3.276324	 lr=0.099648
Epoch[040] Batch [0499]/[3759]	Speed: 117.231584 samples/sec	 accuracy=29.346875	 loss=3.271751	 lr=0.099646
Epoch[040] Batch [0549]/[3759]	Speed: 117.150093 samples/sec	 accuracy=29.312500	 loss=3.271490	 lr=0.099644
Epoch[040] Batch [0599]/[3759]	Speed: 117.310615 samples/sec	 accuracy=29.294271	 loss=3.271190	 lr=0.099643
Epoch[040] Batch [0649]/[3759]	Speed: 116.863368 samples/sec	 accuracy=29.295673	 loss=3.269632	 lr=0.099641
Epoch[040] Batch [0699]/[3759]	Speed: 117.202889 samples/sec	 accuracy=29.339286	 loss=3.265921	 lr=0.099640
Epoch[040] Batch [0749]/[3759]	Speed: 117.246216 samples/sec	 accuracy=29.408333	 loss=3.260959	 lr=0.099638
Epoch[040] Batch [0799]/[3759]	Speed: 117.639624 samples/sec	 accuracy=29.361328	 loss=3.262517	 lr=0.099637
Epoch[040] Batch [0849]/[3759]	Speed: 118.041126 samples/sec	 accuracy=29.319853	 loss=3.263034	 lr=0.099635
Epoch[040] Batch [0899]/[3759]	Speed: 117.320070 samples/sec	 accuracy=29.255208	 loss=3.265991	 lr=0.099634
Epoch[040] Batch [0949]/[3759]	Speed: 117.946159 samples/sec	 accuracy=29.245066	 loss=3.267233	 lr=0.099632
Epoch[040] Batch [0999]/[3759]	Speed: 118.174491 samples/sec	 accuracy=29.271875	 loss=3.268086	 lr=0.099630
Epoch[040] Batch [1049]/[3759]	Speed: 117.567119 samples/sec	 accuracy=29.206845	 loss=3.272443	 lr=0.099629
Epoch[040] Batch [1099]/[3759]	Speed: 117.447914 samples/sec	 accuracy=29.201705	 loss=3.275841	 lr=0.099627
Epoch[040] Batch [1149]/[3759]	Speed: 118.104130 samples/sec	 accuracy=29.202446	 loss=3.276144	 lr=0.099626
Epoch[040] Batch [1199]/[3759]	Speed: 117.548952 samples/sec	 accuracy=29.201823	 loss=3.275907	 lr=0.099624
Epoch[040] Batch [1249]/[3759]	Speed: 117.138783 samples/sec	 accuracy=29.180000	 loss=3.276920	 lr=0.099623
Epoch[040] Batch [1299]/[3759]	Speed: 116.840300 samples/sec	 accuracy=29.157452	 loss=3.277482	 lr=0.099621
Epoch[040] Batch [1349]/[3759]	Speed: 117.934367 samples/sec	 accuracy=29.142361	 loss=3.276231	 lr=0.099619
Epoch[040] Batch [1399]/[3759]	Speed: 116.235849 samples/sec	 accuracy=29.084821	 loss=3.277309	 lr=0.099618
Epoch[040] Batch [1449]/[3759]	Speed: 118.041515 samples/sec	 accuracy=29.145474	 loss=3.274844	 lr=0.099616
Epoch[040] Batch [1499]/[3759]	Speed: 117.237255 samples/sec	 accuracy=29.129167	 loss=3.275234	 lr=0.099615
Epoch[040] Batch [1549]/[3759]	Speed: 117.620745 samples/sec	 accuracy=29.111895	 loss=3.276989	 lr=0.099613
Epoch[040] Batch [1599]/[3759]	Speed: 117.728715 samples/sec	 accuracy=29.133789	 loss=3.276755	 lr=0.099611
Epoch[040] Batch [1649]/[3759]	Speed: 117.631008 samples/sec	 accuracy=29.130682	 loss=3.276718	 lr=0.099610
Epoch[040] Batch [1699]/[3759]	Speed: 117.161716 samples/sec	 accuracy=29.115809	 loss=3.278402	 lr=0.099608
Epoch[040] Batch [1749]/[3759]	Speed: 117.269550 samples/sec	 accuracy=29.086607	 loss=3.279923	 lr=0.099607
Epoch[040] Batch [1799]/[3759]	Speed: 116.622792 samples/sec	 accuracy=29.118056	 loss=3.278473	 lr=0.099605
Epoch[040] Batch [1849]/[3759]	Speed: 117.934266 samples/sec	 accuracy=29.093750	 loss=3.279651	 lr=0.099603
Epoch[040] Batch [1899]/[3759]	Speed: 117.913395 samples/sec	 accuracy=29.121711	 loss=3.277791	 lr=0.099602
Epoch[040] Batch [1949]/[3759]	Speed: 117.321892 samples/sec	 accuracy=29.104968	 loss=3.278576	 lr=0.099600
Epoch[040] Batch [1999]/[3759]	Speed: 117.859383 samples/sec	 accuracy=29.100781	 loss=3.279240	 lr=0.099599
Epoch[040] Batch [2049]/[3759]	Speed: 116.845410 samples/sec	 accuracy=29.108994	 loss=3.279119	 lr=0.099597
Epoch[040] Batch [2099]/[3759]	Speed: 113.981684 samples/sec	 accuracy=29.078869	 loss=3.280176	 lr=0.099595
Epoch[040] Batch [2149]/[3759]	Speed: 117.174507 samples/sec	 accuracy=29.066860	 loss=3.279939	 lr=0.099594
Epoch[040] Batch [2199]/[3759]	Speed: 117.420833 samples/sec	 accuracy=29.091619	 loss=3.279118	 lr=0.099592
Epoch[040] Batch [2249]/[3759]	Speed: 117.305038 samples/sec	 accuracy=29.064583	 loss=3.279902	 lr=0.099590
Epoch[040] Batch [2299]/[3759]	Speed: 117.186677 samples/sec	 accuracy=29.054348	 loss=3.280810	 lr=0.099589
Epoch[040] Batch [2349]/[3759]	Speed: 116.984615 samples/sec	 accuracy=29.036569	 loss=3.281599	 lr=0.099587
Epoch[040] Batch [2399]/[3759]	Speed: 117.250517 samples/sec	 accuracy=29.074219	 loss=3.279574	 lr=0.099585
Epoch[040] Batch [2449]/[3759]	Speed: 117.476293 samples/sec	 accuracy=29.071429	 loss=3.279384	 lr=0.099584
Epoch[040] Batch [2499]/[3759]	Speed: 117.471717 samples/sec	 accuracy=29.066250	 loss=3.279077	 lr=0.099582
Epoch[040] Batch [2549]/[3759]	Speed: 116.971289 samples/sec	 accuracy=29.064338	 loss=3.279658	 lr=0.099580
Epoch[040] Batch [2599]/[3759]	Speed: 117.872105 samples/sec	 accuracy=29.082933	 loss=3.279388	 lr=0.099579
Epoch[040] Batch [2649]/[3759]	Speed: 117.036558 samples/sec	 accuracy=29.103774	 loss=3.278203	 lr=0.099577
Epoch[040] Batch [2699]/[3759]	Speed: 117.804695 samples/sec	 accuracy=29.104167	 loss=3.277947	 lr=0.099575
Epoch[040] Batch [2749]/[3759]	Speed: 116.752632 samples/sec	 accuracy=29.083523	 loss=3.278876	 lr=0.099574
Epoch[040] Batch [2799]/[3759]	Speed: 117.690475 samples/sec	 accuracy=29.097656	 loss=3.278513	 lr=0.099572
Epoch[040] Batch [2849]/[3759]	Speed: 117.957917 samples/sec	 accuracy=29.110746	 loss=3.277650	 lr=0.099570
Epoch[040] Batch [2899]/[3759]	Speed: 116.945936 samples/sec	 accuracy=29.123384	 loss=3.276426	 lr=0.099569
Epoch[040] Batch [2949]/[3759]	Speed: 117.906898 samples/sec	 accuracy=29.118114	 loss=3.276190	 lr=0.099567
Epoch[040] Batch [2999]/[3759]	Speed: 116.778891 samples/sec	 accuracy=29.096875	 loss=3.277165	 lr=0.099565
Epoch[040] Batch [3049]/[3759]	Speed: 117.352320 samples/sec	 accuracy=29.073770	 loss=3.277844	 lr=0.099564
Epoch[040] Batch [3099]/[3759]	Speed: 116.712978 samples/sec	 accuracy=29.072581	 loss=3.277763	 lr=0.099562
Epoch[040] Batch [3149]/[3759]	Speed: 117.250649 samples/sec	 accuracy=29.053571	 loss=3.278254	 lr=0.099560
Epoch[040] Batch [3199]/[3759]	Speed: 117.992188 samples/sec	 accuracy=29.057617	 loss=3.277480	 lr=0.099558
Epoch[040] Batch [3249]/[3759]	Speed: 117.818905 samples/sec	 accuracy=29.046154	 loss=3.277968	 lr=0.099557
Epoch[040] Batch [3299]/[3759]	Speed: 117.381737 samples/sec	 accuracy=29.021780	 loss=3.278405	 lr=0.099555
Epoch[040] Batch [3349]/[3759]	Speed: 117.470499 samples/sec	 accuracy=29.027052	 loss=3.278171	 lr=0.099553
Epoch[040] Batch [3399]/[3759]	Speed: 117.260511 samples/sec	 accuracy=29.005974	 loss=3.278910	 lr=0.099552
Epoch[040] Batch [3449]/[3759]	Speed: 117.555491 samples/sec	 accuracy=29.015851	 loss=3.278714	 lr=0.099550
Epoch[040] Batch [3499]/[3759]	Speed: 117.351001 samples/sec	 accuracy=28.994643	 loss=3.279545	 lr=0.099548
Epoch[040] Batch [3549]/[3759]	Speed: 117.592802 samples/sec	 accuracy=28.989437	 loss=3.279596	 lr=0.099546
Epoch[040] Batch [3599]/[3759]	Speed: 117.410918 samples/sec	 accuracy=29.004340	 loss=3.279359	 lr=0.099545
Epoch[040] Batch [3649]/[3759]	Speed: 117.013322 samples/sec	 accuracy=28.992723	 loss=3.279634	 lr=0.099543
Epoch[040] Batch [3699]/[3759]	Speed: 117.484594 samples/sec	 accuracy=28.997044	 loss=3.279804	 lr=0.099541
Epoch[040] Batch [3749]/[3759]	Speed: 121.008723 samples/sec	 accuracy=28.999167	 loss=3.279812	 lr=0.099539
Batch [0049]/[0303]: acc-top1=27.000000 acc-top5=54.000000
Batch [0099]/[0303]: acc-top1=26.281250 acc-top5=52.078125
Batch [0149]/[0303]: acc-top1=27.281250 acc-top5=53.187500
Batch [0199]/[0303]: acc-top1=29.851562 acc-top5=54.992188
Batch [0249]/[0303]: acc-top1=28.875000 acc-top5=54.825000
Batch [0299]/[0303]: acc-top1=28.786458 acc-top5=55.171875
[Epoch 040] training: accuracy=29.000815	 loss=3.279935
[Epoch 040] speed: 116 samples/sec	time cost: 2177.636565
[Epoch 040] validation: acc-top1=28.846947 acc-top5=55.192863 loss=3.351056
Epoch[041] Batch [0049]/[3760]	Speed: 88.393528 samples/sec	 accuracy=30.406250	 loss=3.230937	 lr=0.099537
Epoch[041] Batch [0099]/[3760]	Speed: 117.312888 samples/sec	 accuracy=30.015625	 loss=3.248406	 lr=0.099536
Epoch[041] Batch [0149]/[3760]	Speed: 116.953877 samples/sec	 accuracy=29.854167	 loss=3.243721	 lr=0.099534
Epoch[041] Batch [0199]/[3760]	Speed: 117.274674 samples/sec	 accuracy=29.609375	 loss=3.257981	 lr=0.099532
Epoch[041] Batch [0249]/[3760]	Speed: 117.425316 samples/sec	 accuracy=29.450000	 loss=3.256056	 lr=0.099530
Epoch[041] Batch [0299]/[3760]	Speed: 116.568208 samples/sec	 accuracy=29.515625	 loss=3.252452	 lr=0.099529
Epoch[041] Batch [0349]/[3760]	Speed: 116.578614 samples/sec	 accuracy=29.526786	 loss=3.255035	 lr=0.099527
Epoch[041] Batch [0399]/[3760]	Speed: 116.674636 samples/sec	 accuracy=29.378906	 loss=3.250889	 lr=0.099525
Epoch[041] Batch [0449]/[3760]	Speed: 117.927946 samples/sec	 accuracy=29.319444	 loss=3.258062	 lr=0.099523
Epoch[041] Batch [0499]/[3760]	Speed: 117.692197 samples/sec	 accuracy=29.428125	 loss=3.251736	 lr=0.099521
Epoch[041] Batch [0549]/[3760]	Speed: 117.640451 samples/sec	 accuracy=29.457386	 loss=3.249497	 lr=0.099520
Epoch[041] Batch [0599]/[3760]	Speed: 118.048166 samples/sec	 accuracy=29.369792	 loss=3.257195	 lr=0.099518
Epoch[041] Batch [0649]/[3760]	Speed: 117.235290 samples/sec	 accuracy=29.382212	 loss=3.256059	 lr=0.099516
Epoch[041] Batch [0699]/[3760]	Speed: 117.236441 samples/sec	 accuracy=29.348214	 loss=3.258725	 lr=0.099514
Epoch[041] Batch [0749]/[3760]	Speed: 116.320654 samples/sec	 accuracy=29.333333	 loss=3.261333	 lr=0.099513
Epoch[041] Batch [0799]/[3760]	Speed: 117.271861 samples/sec	 accuracy=29.298828	 loss=3.262249	 lr=0.099511
Epoch[041] Batch [0849]/[3760]	Speed: 116.791671 samples/sec	 accuracy=29.316176	 loss=3.257722	 lr=0.099509
Epoch[041] Batch [0899]/[3760]	Speed: 117.666703 samples/sec	 accuracy=29.253472	 loss=3.262812	 lr=0.099507
Epoch[041] Batch [0949]/[3760]	Speed: 117.234285 samples/sec	 accuracy=29.273026	 loss=3.262033	 lr=0.099505
Epoch[041] Batch [0999]/[3760]	Speed: 117.431908 samples/sec	 accuracy=29.289062	 loss=3.262651	 lr=0.099504
Epoch[041] Batch [1049]/[3760]	Speed: 117.323601 samples/sec	 accuracy=29.342262	 loss=3.259289	 lr=0.099502
Epoch[041] Batch [1099]/[3760]	Speed: 117.325896 samples/sec	 accuracy=29.306818	 loss=3.263248	 lr=0.099500
Epoch[041] Batch [1149]/[3760]	Speed: 117.584994 samples/sec	 accuracy=29.345109	 loss=3.262435	 lr=0.099498
Epoch[041] Batch [1199]/[3760]	Speed: 117.685077 samples/sec	 accuracy=29.278646	 loss=3.265823	 lr=0.099496
Epoch[041] Batch [1249]/[3760]	Speed: 117.649846 samples/sec	 accuracy=29.315000	 loss=3.263034	 lr=0.099494
Epoch[041] Batch [1299]/[3760]	Speed: 117.529022 samples/sec	 accuracy=29.293269	 loss=3.262812	 lr=0.099493
Epoch[041] Batch [1349]/[3760]	Speed: 116.903419 samples/sec	 accuracy=29.252315	 loss=3.265584	 lr=0.099491
Epoch[041] Batch [1399]/[3760]	Speed: 117.143554 samples/sec	 accuracy=29.216518	 loss=3.266774	 lr=0.099489
Epoch[041] Batch [1449]/[3760]	Speed: 116.604103 samples/sec	 accuracy=29.198276	 loss=3.268816	 lr=0.099487
Epoch[041] Batch [1499]/[3760]	Speed: 116.729841 samples/sec	 accuracy=29.230208	 loss=3.268529	 lr=0.099485
Epoch[041] Batch [1549]/[3760]	Speed: 117.933749 samples/sec	 accuracy=29.209677	 loss=3.269100	 lr=0.099483
Epoch[041] Batch [1599]/[3760]	Speed: 117.900707 samples/sec	 accuracy=29.227539	 loss=3.268473	 lr=0.099482
Epoch[041] Batch [1649]/[3760]	Speed: 117.146777 samples/sec	 accuracy=29.219697	 loss=3.268345	 lr=0.099480
Epoch[041] Batch [1699]/[3760]	Speed: 117.533364 samples/sec	 accuracy=29.204044	 loss=3.269689	 lr=0.099478
Epoch[041] Batch [1749]/[3760]	Speed: 116.433097 samples/sec	 accuracy=29.191964	 loss=3.270062	 lr=0.099476
Epoch[041] Batch [1799]/[3760]	Speed: 117.105510 samples/sec	 accuracy=29.208333	 loss=3.270120	 lr=0.099474
Epoch[041] Batch [1849]/[3760]	Speed: 116.952455 samples/sec	 accuracy=29.179054	 loss=3.270589	 lr=0.099472
Epoch[041] Batch [1899]/[3760]	Speed: 116.948932 samples/sec	 accuracy=29.196546	 loss=3.270477	 lr=0.099470
Epoch[041] Batch [1949]/[3760]	Speed: 117.217326 samples/sec	 accuracy=29.192308	 loss=3.269977	 lr=0.099468
Epoch[041] Batch [1999]/[3760]	Speed: 116.761803 samples/sec	 accuracy=29.161719	 loss=3.271887	 lr=0.099467
Epoch[041] Batch [2049]/[3760]	Speed: 116.949266 samples/sec	 accuracy=29.163872	 loss=3.271443	 lr=0.099465
Epoch[041] Batch [2099]/[3760]	Speed: 117.012952 samples/sec	 accuracy=29.182292	 loss=3.270328	 lr=0.099463
Epoch[041] Batch [2149]/[3760]	Speed: 118.203009 samples/sec	 accuracy=29.169331	 loss=3.271050	 lr=0.099461
Epoch[041] Batch [2199]/[3760]	Speed: 117.669854 samples/sec	 accuracy=29.142045	 loss=3.272450	 lr=0.099459
Epoch[041] Batch [2249]/[3760]	Speed: 117.019779 samples/sec	 accuracy=29.150694	 loss=3.271507	 lr=0.099457
Epoch[041] Batch [2299]/[3760]	Speed: 118.414790 samples/sec	 accuracy=29.133832	 loss=3.272272	 lr=0.099455
Epoch[041] Batch [2349]/[3760]	Speed: 117.406743 samples/sec	 accuracy=29.114362	 loss=3.273235	 lr=0.099453
Epoch[041] Batch [2399]/[3760]	Speed: 118.197402 samples/sec	 accuracy=29.092448	 loss=3.273871	 lr=0.099451
Epoch[041] Batch [2449]/[3760]	Speed: 116.167140 samples/sec	 accuracy=29.064413	 loss=3.273773	 lr=0.099450
Epoch[041] Batch [2499]/[3760]	Speed: 117.358425 samples/sec	 accuracy=29.045625	 loss=3.274284	 lr=0.099448
Epoch[041] Batch [2549]/[3760]	Speed: 117.205279 samples/sec	 accuracy=29.057598	 loss=3.273545	 lr=0.099446
Epoch[041] Batch [2599]/[3760]	Speed: 116.659336 samples/sec	 accuracy=29.075120	 loss=3.273091	 lr=0.099444
Epoch[041] Batch [2649]/[3760]	Speed: 116.522171 samples/sec	 accuracy=29.047170	 loss=3.274426	 lr=0.099442
Epoch[041] Batch [2699]/[3760]	Speed: 117.386331 samples/sec	 accuracy=29.068866	 loss=3.273852	 lr=0.099440
Epoch[041] Batch [2749]/[3760]	Speed: 119.017000 samples/sec	 accuracy=29.069318	 loss=3.273309	 lr=0.099438
Epoch[041] Batch [2799]/[3760]	Speed: 117.297526 samples/sec	 accuracy=29.069754	 loss=3.273408	 lr=0.099436
Epoch[041] Batch [2849]/[3760]	Speed: 116.540693 samples/sec	 accuracy=29.072917	 loss=3.272750	 lr=0.099434
Epoch[041] Batch [2899]/[3760]	Speed: 116.581057 samples/sec	 accuracy=29.068966	 loss=3.273458	 lr=0.099432
Epoch[041] Batch [2949]/[3760]	Speed: 117.469061 samples/sec	 accuracy=29.064089	 loss=3.273001	 lr=0.099430
Epoch[041] Batch [2999]/[3760]	Speed: 117.460262 samples/sec	 accuracy=29.083333	 loss=3.271955	 lr=0.099428
Epoch[041] Batch [3049]/[3760]	Speed: 117.851579 samples/sec	 accuracy=29.094262	 loss=3.272064	 lr=0.099426
Epoch[041] Batch [3099]/[3760]	Speed: 117.263428 samples/sec	 accuracy=29.123488	 loss=3.271314	 lr=0.099424
Epoch[041] Batch [3149]/[3760]	Speed: 116.618374 samples/sec	 accuracy=29.111607	 loss=3.271837	 lr=0.099423
Epoch[041] Batch [3199]/[3760]	Speed: 117.020204 samples/sec	 accuracy=29.100098	 loss=3.272512	 lr=0.099421
Epoch[041] Batch [3249]/[3760]	Speed: 116.405554 samples/sec	 accuracy=29.105288	 loss=3.272119	 lr=0.099419
Epoch[041] Batch [3299]/[3760]	Speed: 117.419447 samples/sec	 accuracy=29.116477	 loss=3.270714	 lr=0.099417
Epoch[041] Batch [3349]/[3760]	Speed: 117.596508 samples/sec	 accuracy=29.131996	 loss=3.270202	 lr=0.099415
Epoch[041] Batch [3399]/[3760]	Speed: 117.713300 samples/sec	 accuracy=29.134651	 loss=3.270425	 lr=0.099413
Epoch[041] Batch [3449]/[3760]	Speed: 117.462027 samples/sec	 accuracy=29.152174	 loss=3.269299	 lr=0.099411
Epoch[041] Batch [3499]/[3760]	Speed: 117.845933 samples/sec	 accuracy=29.140625	 loss=3.269605	 lr=0.099409
Epoch[041] Batch [3549]/[3760]	Speed: 116.850909 samples/sec	 accuracy=29.151408	 loss=3.269645	 lr=0.099407
Epoch[041] Batch [3599]/[3760]	Speed: 117.850785 samples/sec	 accuracy=29.134115	 loss=3.270305	 lr=0.099405
Epoch[041] Batch [3649]/[3760]	Speed: 117.710432 samples/sec	 accuracy=29.119007	 loss=3.271232	 lr=0.099403
Epoch[041] Batch [3699]/[3760]	Speed: 116.767283 samples/sec	 accuracy=29.123311	 loss=3.271443	 lr=0.099401
Epoch[041] Batch [3749]/[3760]	Speed: 120.751823 samples/sec	 accuracy=29.120000	 loss=3.272321	 lr=0.099399
Batch [0049]/[0303]: acc-top1=31.156250 acc-top5=56.718750
Batch [0099]/[0303]: acc-top1=28.156250 acc-top5=53.140625
Batch [0149]/[0303]: acc-top1=30.135417 acc-top5=55.708333
Batch [0199]/[0303]: acc-top1=32.695312 acc-top5=58.031250
Batch [0249]/[0303]: acc-top1=31.731250 acc-top5=57.631250
Batch [0299]/[0303]: acc-top1=31.588542 acc-top5=58.182292
[Epoch 041] training: accuracy=29.114445	 loss=3.272503
[Epoch 041] speed: 116 samples/sec	time cost: 2180.641979
[Epoch 041] validation: acc-top1=31.616130 acc-top5=58.281766 loss=3.111665
Epoch[042] Batch [0049]/[3760]	Speed: 87.683585 samples/sec	 accuracy=29.093750	 loss=3.264186	 lr=0.099396
Epoch[042] Batch [0099]/[3760]	Speed: 118.147815 samples/sec	 accuracy=28.703125	 loss=3.291468	 lr=0.099394
Epoch[042] Batch [0149]/[3760]	Speed: 117.088427 samples/sec	 accuracy=28.593750	 loss=3.276522	 lr=0.099392
Epoch[042] Batch [0199]/[3760]	Speed: 117.278137 samples/sec	 accuracy=29.179688	 loss=3.255596	 lr=0.099390
Epoch[042] Batch [0249]/[3760]	Speed: 117.620989 samples/sec	 accuracy=29.306250	 loss=3.246728	 lr=0.099388
Epoch[042] Batch [0299]/[3760]	Speed: 117.295335 samples/sec	 accuracy=29.244792	 loss=3.250373	 lr=0.099386
Epoch[042] Batch [0349]/[3760]	Speed: 117.637492 samples/sec	 accuracy=29.263393	 loss=3.246032	 lr=0.099384
Epoch[042] Batch [0399]/[3760]	Speed: 116.624366 samples/sec	 accuracy=29.589844	 loss=3.234256	 lr=0.099382
Epoch[042] Batch [0449]/[3760]	Speed: 117.737006 samples/sec	 accuracy=29.520833	 loss=3.236891	 lr=0.099380
Epoch[042] Batch [0499]/[3760]	Speed: 117.429128 samples/sec	 accuracy=29.393750	 loss=3.237390	 lr=0.099378
Epoch[042] Batch [0549]/[3760]	Speed: 117.014347 samples/sec	 accuracy=29.292614	 loss=3.243141	 lr=0.099376
Epoch[042] Batch [0599]/[3760]	Speed: 116.893446 samples/sec	 accuracy=29.315104	 loss=3.242911	 lr=0.099374
Epoch[042] Batch [0649]/[3760]	Speed: 117.192085 samples/sec	 accuracy=29.302885	 loss=3.247130	 lr=0.099372
Epoch[042] Batch [0699]/[3760]	Speed: 116.423734 samples/sec	 accuracy=29.325893	 loss=3.250222	 lr=0.099370
Epoch[042] Batch [0749]/[3760]	Speed: 117.931799 samples/sec	 accuracy=29.360417	 loss=3.248818	 lr=0.099368
Epoch[042] Batch [0799]/[3760]	Speed: 117.371493 samples/sec	 accuracy=29.285156	 loss=3.251846	 lr=0.099366
Epoch[042] Batch [0849]/[3760]	Speed: 117.593509 samples/sec	 accuracy=29.261029	 loss=3.254143	 lr=0.099364
Epoch[042] Batch [0899]/[3760]	Speed: 117.093629 samples/sec	 accuracy=29.208333	 loss=3.257148	 lr=0.099362
Epoch[042] Batch [0949]/[3760]	Speed: 116.810936 samples/sec	 accuracy=29.233553	 loss=3.255819	 lr=0.099360
Epoch[042] Batch [0999]/[3760]	Speed: 117.873442 samples/sec	 accuracy=29.182813	 loss=3.257953	 lr=0.099358
Epoch[042] Batch [1049]/[3760]	Speed: 117.294702 samples/sec	 accuracy=29.188988	 loss=3.259147	 lr=0.099356
Epoch[042] Batch [1099]/[3760]	Speed: 116.930811 samples/sec	 accuracy=29.198864	 loss=3.258726	 lr=0.099354
Epoch[042] Batch [1149]/[3760]	Speed: 117.143321 samples/sec	 accuracy=29.293478	 loss=3.256013	 lr=0.099352
Epoch[042] Batch [1199]/[3760]	Speed: 117.178957 samples/sec	 accuracy=29.329427	 loss=3.255910	 lr=0.099350
Epoch[042] Batch [1249]/[3760]	Speed: 117.641026 samples/sec	 accuracy=29.351250	 loss=3.255802	 lr=0.099348
Epoch[042] Batch [1299]/[3760]	Speed: 117.348758 samples/sec	 accuracy=29.337740	 loss=3.257244	 lr=0.099345
Epoch[042] Batch [1349]/[3760]	Speed: 117.301530 samples/sec	 accuracy=29.342593	 loss=3.257480	 lr=0.099343
Epoch[042] Batch [1399]/[3760]	Speed: 117.831541 samples/sec	 accuracy=29.310268	 loss=3.257958	 lr=0.099341
Epoch[042] Batch [1449]/[3760]	Speed: 116.389325 samples/sec	 accuracy=29.288793	 loss=3.258025	 lr=0.099339
Epoch[042] Batch [1499]/[3760]	Speed: 118.179240 samples/sec	 accuracy=29.281250	 loss=3.256390	 lr=0.099337
Epoch[042] Batch [1549]/[3760]	Speed: 117.570246 samples/sec	 accuracy=29.278226	 loss=3.256890	 lr=0.099335
Epoch[042] Batch [1599]/[3760]	Speed: 117.302665 samples/sec	 accuracy=29.303711	 loss=3.254134	 lr=0.099333
Epoch[042] Batch [1649]/[3760]	Speed: 117.590048 samples/sec	 accuracy=29.286932	 loss=3.254037	 lr=0.099331
Epoch[042] Batch [1699]/[3760]	Speed: 117.328259 samples/sec	 accuracy=29.307904	 loss=3.254110	 lr=0.099329
Epoch[042] Batch [1749]/[3760]	Speed: 117.627577 samples/sec	 accuracy=29.275893	 loss=3.257183	 lr=0.099327
Epoch[042] Batch [1799]/[3760]	Speed: 117.106275 samples/sec	 accuracy=29.268229	 loss=3.257332	 lr=0.099324
Epoch[042] Batch [1849]/[3760]	Speed: 116.890239 samples/sec	 accuracy=29.259291	 loss=3.257885	 lr=0.099322
Epoch[042] Batch [1899]/[3760]	Speed: 117.062964 samples/sec	 accuracy=29.286184	 loss=3.257292	 lr=0.099320
Epoch[042] Batch [1949]/[3760]	Speed: 117.009927 samples/sec	 accuracy=29.278846	 loss=3.257451	 lr=0.099318
Epoch[042] Batch [1999]/[3760]	Speed: 117.438075 samples/sec	 accuracy=29.244531	 loss=3.257768	 lr=0.099316
Epoch[042] Batch [2049]/[3760]	Speed: 117.543794 samples/sec	 accuracy=29.243902	 loss=3.258244	 lr=0.099314
Epoch[042] Batch [2099]/[3760]	Speed: 116.744164 samples/sec	 accuracy=29.222470	 loss=3.258455	 lr=0.099312
Epoch[042] Batch [2149]/[3760]	Speed: 117.139165 samples/sec	 accuracy=29.209302	 loss=3.259297	 lr=0.099310
Epoch[042] Batch [2199]/[3760]	Speed: 116.678438 samples/sec	 accuracy=29.199574	 loss=3.260327	 lr=0.099307
Epoch[042] Batch [2249]/[3760]	Speed: 116.706823 samples/sec	 accuracy=29.172917	 loss=3.261330	 lr=0.099305
Epoch[042] Batch [2299]/[3760]	Speed: 117.185208 samples/sec	 accuracy=29.170516	 loss=3.261281	 lr=0.099303
Epoch[042] Batch [2349]/[3760]	Speed: 116.936117 samples/sec	 accuracy=29.154255	 loss=3.261769	 lr=0.099301
Epoch[042] Batch [2399]/[3760]	Speed: 117.037762 samples/sec	 accuracy=29.163411	 loss=3.262127	 lr=0.099299
Epoch[042] Batch [2449]/[3760]	Speed: 117.004322 samples/sec	 accuracy=29.188776	 loss=3.261787	 lr=0.099297
Epoch[042] Batch [2499]/[3760]	Speed: 116.561321 samples/sec	 accuracy=29.213750	 loss=3.262365	 lr=0.099295
Epoch[042] Batch [2549]/[3760]	Speed: 117.586662 samples/sec	 accuracy=29.240196	 loss=3.262340	 lr=0.099292
Epoch[042] Batch [2599]/[3760]	Speed: 116.799358 samples/sec	 accuracy=29.254207	 loss=3.262139	 lr=0.099290
Epoch[042] Batch [2649]/[3760]	Speed: 117.338287 samples/sec	 accuracy=29.250000	 loss=3.262339	 lr=0.099288
Epoch[042] Batch [2699]/[3760]	Speed: 116.907303 samples/sec	 accuracy=29.255787	 loss=3.261857	 lr=0.099286
Epoch[042] Batch [2749]/[3760]	Speed: 116.760092 samples/sec	 accuracy=29.277841	 loss=3.261264	 lr=0.099284
Epoch[042] Batch [2799]/[3760]	Speed: 117.592746 samples/sec	 accuracy=29.251116	 loss=3.261799	 lr=0.099282
Epoch[042] Batch [2849]/[3760]	Speed: 118.185613 samples/sec	 accuracy=29.249452	 loss=3.262294	 lr=0.099279
Epoch[042] Batch [2899]/[3760]	Speed: 116.454824 samples/sec	 accuracy=29.247306	 loss=3.262391	 lr=0.099277
Epoch[042] Batch [2949]/[3760]	Speed: 116.949120 samples/sec	 accuracy=29.233051	 loss=3.263229	 lr=0.099275
Epoch[042] Batch [2999]/[3760]	Speed: 117.754984 samples/sec	 accuracy=29.233333	 loss=3.263275	 lr=0.099273
Epoch[042] Batch [3049]/[3760]	Speed: 117.397888 samples/sec	 accuracy=29.237193	 loss=3.263196	 lr=0.099271
Epoch[042] Batch [3099]/[3760]	Speed: 118.146111 samples/sec	 accuracy=29.223286	 loss=3.263800	 lr=0.099268
Epoch[042] Batch [3149]/[3760]	Speed: 117.304168 samples/sec	 accuracy=29.241071	 loss=3.263246	 lr=0.099266
Epoch[042] Batch [3199]/[3760]	Speed: 118.153831 samples/sec	 accuracy=29.245605	 loss=3.263288	 lr=0.099264
Epoch[042] Batch [3249]/[3760]	Speed: 117.441655 samples/sec	 accuracy=29.257692	 loss=3.262883	 lr=0.099262
Epoch[042] Batch [3299]/[3760]	Speed: 117.537729 samples/sec	 accuracy=29.245265	 loss=3.263594	 lr=0.099260
Epoch[042] Batch [3349]/[3760]	Speed: 116.477067 samples/sec	 accuracy=29.244403	 loss=3.263250	 lr=0.099257
Epoch[042] Batch [3399]/[3760]	Speed: 117.689675 samples/sec	 accuracy=29.255055	 loss=3.263723	 lr=0.099255
Epoch[042] Batch [3449]/[3760]	Speed: 116.996753 samples/sec	 accuracy=29.254529	 loss=3.263536	 lr=0.099253
Epoch[042] Batch [3499]/[3760]	Speed: 117.046872 samples/sec	 accuracy=29.231250	 loss=3.264044	 lr=0.099251
Epoch[042] Batch [3549]/[3760]	Speed: 118.041615 samples/sec	 accuracy=29.226673	 loss=3.263956	 lr=0.099249
Epoch[042] Batch [3599]/[3760]	Speed: 117.042581 samples/sec	 accuracy=29.221354	 loss=3.264272	 lr=0.099246
Epoch[042] Batch [3649]/[3760]	Speed: 117.030277 samples/sec	 accuracy=29.216610	 loss=3.264277	 lr=0.099244
Epoch[042] Batch [3699]/[3760]	Speed: 117.669862 samples/sec	 accuracy=29.250000	 loss=3.263248	 lr=0.099242
Epoch[042] Batch [3749]/[3760]	Speed: 121.127100 samples/sec	 accuracy=29.259583	 loss=3.263506	 lr=0.099240
Batch [0049]/[0303]: acc-top1=30.812500 acc-top5=58.562500
Batch [0099]/[0303]: acc-top1=28.828125 acc-top5=55.703125
Batch [0149]/[0303]: acc-top1=29.322917 acc-top5=56.458333
Batch [0199]/[0303]: acc-top1=31.921875 acc-top5=58.750000
Batch [0249]/[0303]: acc-top1=31.062500 acc-top5=58.287500
Batch [0299]/[0303]: acc-top1=30.635417 acc-top5=57.984375
[Epoch 042] training: accuracy=29.254488	 loss=3.263823
[Epoch 042] speed: 116 samples/sec	time cost: 2180.650544
[Epoch 042] validation: acc-top1=30.682756 acc-top5=57.967203 loss=3.300012
Epoch[043] Batch [0049]/[3759]	Speed: 86.837143 samples/sec	 accuracy=29.343750	 loss=3.232805	 lr=0.099237
Epoch[043] Batch [0099]/[3759]	Speed: 117.274467 samples/sec	 accuracy=29.265625	 loss=3.241923	 lr=0.099235
Epoch[043] Batch [0149]/[3759]	Speed: 117.611095 samples/sec	 accuracy=29.541667	 loss=3.223502	 lr=0.099232
Epoch[043] Batch [0199]/[3759]	Speed: 117.343388 samples/sec	 accuracy=29.437500	 loss=3.230448	 lr=0.099230
Epoch[043] Batch [0249]/[3759]	Speed: 117.858313 samples/sec	 accuracy=29.381250	 loss=3.235998	 lr=0.099228
Epoch[043] Batch [0299]/[3759]	Speed: 117.269312 samples/sec	 accuracy=29.744792	 loss=3.224212	 lr=0.099226
Epoch[043] Batch [0349]/[3759]	Speed: 117.862971 samples/sec	 accuracy=29.696429	 loss=3.230034	 lr=0.099223
Epoch[043] Batch [0399]/[3759]	Speed: 117.411911 samples/sec	 accuracy=29.757813	 loss=3.229291	 lr=0.099221
Epoch[043] Batch [0449]/[3759]	Speed: 116.847815 samples/sec	 accuracy=29.569444	 loss=3.232485	 lr=0.099219
Epoch[043] Batch [0499]/[3759]	Speed: 117.118403 samples/sec	 accuracy=29.559375	 loss=3.234400	 lr=0.099217
Epoch[043] Batch [0549]/[3759]	Speed: 117.363410 samples/sec	 accuracy=29.494318	 loss=3.239697	 lr=0.099214
Epoch[043] Batch [0599]/[3759]	Speed: 117.707776 samples/sec	 accuracy=29.473958	 loss=3.242519	 lr=0.099212
Epoch[043] Batch [0649]/[3759]	Speed: 117.004757 samples/sec	 accuracy=29.394231	 loss=3.246403	 lr=0.099210
Epoch[043] Batch [0699]/[3759]	Speed: 117.462827 samples/sec	 accuracy=29.395089	 loss=3.247203	 lr=0.099207
Epoch[043] Batch [0749]/[3759]	Speed: 116.569310 samples/sec	 accuracy=29.418750	 loss=3.246765	 lr=0.099205
Epoch[043] Batch [0799]/[3759]	Speed: 117.794816 samples/sec	 accuracy=29.402344	 loss=3.249147	 lr=0.099203
Epoch[043] Batch [0849]/[3759]	Speed: 117.001137 samples/sec	 accuracy=29.474265	 loss=3.247723	 lr=0.099201
Epoch[043] Batch [0899]/[3759]	Speed: 116.887241 samples/sec	 accuracy=29.434028	 loss=3.248901	 lr=0.099198
Epoch[043] Batch [0949]/[3759]	Speed: 117.158242 samples/sec	 accuracy=29.379934	 loss=3.252114	 lr=0.099196
Epoch[043] Batch [0999]/[3759]	Speed: 116.994941 samples/sec	 accuracy=29.340625	 loss=3.252570	 lr=0.099194
Epoch[043] Batch [1049]/[3759]	Speed: 117.534911 samples/sec	 accuracy=29.334821	 loss=3.254021	 lr=0.099191
Epoch[043] Batch [1099]/[3759]	Speed: 118.271280 samples/sec	 accuracy=29.319602	 loss=3.253265	 lr=0.099189
Epoch[043] Batch [1149]/[3759]	Speed: 117.924306 samples/sec	 accuracy=29.346467	 loss=3.252988	 lr=0.099187
Epoch[043] Batch [1199]/[3759]	Speed: 116.703982 samples/sec	 accuracy=29.403646	 loss=3.250660	 lr=0.099184
Epoch[043] Batch [1249]/[3759]	Speed: 117.730557 samples/sec	 accuracy=29.392500	 loss=3.251356	 lr=0.099182
Epoch[043] Batch [1299]/[3759]	Speed: 117.330737 samples/sec	 accuracy=29.439904	 loss=3.251952	 lr=0.099180
Epoch[043] Batch [1349]/[3759]	Speed: 117.011745 samples/sec	 accuracy=29.416667	 loss=3.253322	 lr=0.099177
Epoch[043] Batch [1399]/[3759]	Speed: 117.084475 samples/sec	 accuracy=29.416295	 loss=3.251662	 lr=0.099175
Epoch[043] Batch [1449]/[3759]	Speed: 117.309914 samples/sec	 accuracy=29.373922	 loss=3.253418	 lr=0.099173
Epoch[043] Batch [1499]/[3759]	Speed: 117.787770 samples/sec	 accuracy=29.376042	 loss=3.255478	 lr=0.099170
Epoch[043] Batch [1549]/[3759]	Speed: 118.117713 samples/sec	 accuracy=29.370968	 loss=3.257202	 lr=0.099168
Epoch[043] Batch [1599]/[3759]	Speed: 117.557874 samples/sec	 accuracy=29.404297	 loss=3.256163	 lr=0.099166
Epoch[043] Batch [1649]/[3759]	Speed: 117.749794 samples/sec	 accuracy=29.384470	 loss=3.256250	 lr=0.099163
Epoch[043] Batch [1699]/[3759]	Speed: 117.495378 samples/sec	 accuracy=29.401654	 loss=3.256174	 lr=0.099161
Epoch[043] Batch [1749]/[3759]	Speed: 117.700665 samples/sec	 accuracy=29.365179	 loss=3.258146	 lr=0.099159
Epoch[043] Batch [1799]/[3759]	Speed: 117.225691 samples/sec	 accuracy=29.352431	 loss=3.258770	 lr=0.099156
Epoch[043] Batch [1849]/[3759]	Speed: 116.964936 samples/sec	 accuracy=29.352196	 loss=3.258537	 lr=0.099154
Epoch[043] Batch [1899]/[3759]	Speed: 117.871935 samples/sec	 accuracy=29.358553	 loss=3.258748	 lr=0.099152
Epoch[043] Batch [1949]/[3759]	Speed: 117.332841 samples/sec	 accuracy=29.330929	 loss=3.259788	 lr=0.099149
Epoch[043] Batch [1999]/[3759]	Speed: 117.125771 samples/sec	 accuracy=29.323438	 loss=3.260034	 lr=0.099147
Epoch[043] Batch [2049]/[3759]	Speed: 117.314093 samples/sec	 accuracy=29.364329	 loss=3.258810	 lr=0.099144
Epoch[043] Batch [2099]/[3759]	Speed: 117.143049 samples/sec	 accuracy=29.362351	 loss=3.259337	 lr=0.099142
Epoch[043] Batch [2149]/[3759]	Speed: 116.728169 samples/sec	 accuracy=29.331395	 loss=3.259862	 lr=0.099140
Epoch[043] Batch [2199]/[3759]	Speed: 117.006172 samples/sec	 accuracy=29.345170	 loss=3.258466	 lr=0.099137
Epoch[043] Batch [2249]/[3759]	Speed: 116.854007 samples/sec	 accuracy=29.358333	 loss=3.258832	 lr=0.099135
Epoch[043] Batch [2299]/[3759]	Speed: 117.538455 samples/sec	 accuracy=29.340353	 loss=3.259370	 lr=0.099133
Epoch[043] Batch [2349]/[3759]	Speed: 117.229936 samples/sec	 accuracy=29.323138	 loss=3.260192	 lr=0.099130
Epoch[043] Batch [2399]/[3759]	Speed: 117.016379 samples/sec	 accuracy=29.305339	 loss=3.260393	 lr=0.099128
Epoch[043] Batch [2449]/[3759]	Speed: 117.388041 samples/sec	 accuracy=29.294643	 loss=3.260948	 lr=0.099125
Epoch[043] Batch [2499]/[3759]	Speed: 117.215310 samples/sec	 accuracy=29.276875	 loss=3.260748	 lr=0.099123
Epoch[043] Batch [2549]/[3759]	Speed: 117.115734 samples/sec	 accuracy=29.296569	 loss=3.260320	 lr=0.099121
Epoch[043] Batch [2599]/[3759]	Speed: 117.035988 samples/sec	 accuracy=29.273438	 loss=3.262036	 lr=0.099118
Epoch[043] Batch [2649]/[3759]	Speed: 116.547393 samples/sec	 accuracy=29.301297	 loss=3.260920	 lr=0.099116
Epoch[043] Batch [2699]/[3759]	Speed: 117.438104 samples/sec	 accuracy=29.285301	 loss=3.260653	 lr=0.099113
Epoch[043] Batch [2749]/[3759]	Speed: 117.441623 samples/sec	 accuracy=29.306818	 loss=3.260356	 lr=0.099111
Epoch[043] Batch [2799]/[3759]	Speed: 117.958153 samples/sec	 accuracy=29.292969	 loss=3.260500	 lr=0.099108
Epoch[043] Batch [2849]/[3759]	Speed: 117.071333 samples/sec	 accuracy=29.280702	 loss=3.261778	 lr=0.099106
Epoch[043] Batch [2899]/[3759]	Speed: 117.222097 samples/sec	 accuracy=29.280711	 loss=3.261456	 lr=0.099104
Epoch[043] Batch [2949]/[3759]	Speed: 117.663698 samples/sec	 accuracy=29.283369	 loss=3.261774	 lr=0.099101
Epoch[043] Batch [2999]/[3759]	Speed: 117.641520 samples/sec	 accuracy=29.277604	 loss=3.261709	 lr=0.099099
Epoch[043] Batch [3049]/[3759]	Speed: 117.498048 samples/sec	 accuracy=29.262295	 loss=3.262164	 lr=0.099096
Epoch[043] Batch [3099]/[3759]	Speed: 117.615048 samples/sec	 accuracy=29.302923	 loss=3.260785	 lr=0.099094
Epoch[043] Batch [3149]/[3759]	Speed: 116.848547 samples/sec	 accuracy=29.308036	 loss=3.260377	 lr=0.099091
Epoch[043] Batch [3199]/[3759]	Speed: 117.419167 samples/sec	 accuracy=29.318359	 loss=3.260150	 lr=0.099089
Epoch[043] Batch [3249]/[3759]	Speed: 116.956222 samples/sec	 accuracy=29.318750	 loss=3.259656	 lr=0.099087
Epoch[043] Batch [3299]/[3759]	Speed: 117.124670 samples/sec	 accuracy=29.306345	 loss=3.260402	 lr=0.099084
Epoch[043] Batch [3349]/[3759]	Speed: 116.873799 samples/sec	 accuracy=29.307369	 loss=3.260931	 lr=0.099082
Epoch[043] Batch [3399]/[3759]	Speed: 117.008170 samples/sec	 accuracy=29.292739	 loss=3.261003	 lr=0.099079
Epoch[043] Batch [3449]/[3759]	Speed: 117.407179 samples/sec	 accuracy=29.318388	 loss=3.260987	 lr=0.099077
Epoch[043] Batch [3499]/[3759]	Speed: 117.149467 samples/sec	 accuracy=29.318750	 loss=3.261265	 lr=0.099074
Epoch[043] Batch [3549]/[3759]	Speed: 117.467857 samples/sec	 accuracy=29.324384	 loss=3.261379	 lr=0.099072
Epoch[043] Batch [3599]/[3759]	Speed: 116.958632 samples/sec	 accuracy=29.317274	 loss=3.261725	 lr=0.099069
Epoch[043] Batch [3649]/[3759]	Speed: 116.794653 samples/sec	 accuracy=29.326199	 loss=3.261706	 lr=0.099067
Epoch[043] Batch [3699]/[3759]	Speed: 117.363645 samples/sec	 accuracy=29.313345	 loss=3.262322	 lr=0.099064
Epoch[043] Batch [3749]/[3759]	Speed: 120.201567 samples/sec	 accuracy=29.320417	 loss=3.262575	 lr=0.099062
Batch [0049]/[0303]: acc-top1=29.062500 acc-top5=55.593750
Batch [0099]/[0303]: acc-top1=27.000000 acc-top5=53.265625
Batch [0149]/[0303]: acc-top1=27.895833 acc-top5=54.031250
Batch [0199]/[0303]: acc-top1=30.351562 acc-top5=56.820312
Batch [0249]/[0303]: acc-top1=29.700000 acc-top5=56.512500
Batch [0299]/[0303]: acc-top1=29.260417 acc-top5=56.651042
[Epoch 043] training: accuracy=29.319217	 loss=3.262795
[Epoch 043] speed: 116 samples/sec	time cost: 2180.302156
[Epoch 043] validation: acc-top1=29.336840 acc-top5=56.708952 loss=3.289093
Epoch[044] Batch [0049]/[3760]	Speed: 87.953865 samples/sec	 accuracy=29.031250	 loss=3.277838	 lr=0.099059
Epoch[044] Batch [0099]/[3760]	Speed: 116.824991 samples/sec	 accuracy=29.187500	 loss=3.265145	 lr=0.099056
Epoch[044] Batch [0149]/[3760]	Speed: 116.713669 samples/sec	 accuracy=29.322917	 loss=3.246486	 lr=0.099054
Epoch[044] Batch [0199]/[3760]	Speed: 117.133182 samples/sec	 accuracy=29.320312	 loss=3.240424	 lr=0.099051
Epoch[044] Batch [0249]/[3760]	Speed: 118.484741 samples/sec	 accuracy=29.225000	 loss=3.259666	 lr=0.099049
Epoch[044] Batch [0299]/[3760]	Speed: 117.078148 samples/sec	 accuracy=29.536458	 loss=3.250139	 lr=0.099046
Epoch[044] Batch [0349]/[3760]	Speed: 117.204148 samples/sec	 accuracy=29.616071	 loss=3.249642	 lr=0.099044
Epoch[044] Batch [0399]/[3760]	Speed: 117.341484 samples/sec	 accuracy=29.515625	 loss=3.248635	 lr=0.099041
Epoch[044] Batch [0449]/[3760]	Speed: 116.959306 samples/sec	 accuracy=29.458333	 loss=3.249366	 lr=0.099039
Epoch[044] Batch [0499]/[3760]	Speed: 117.057455 samples/sec	 accuracy=29.703125	 loss=3.245801	 lr=0.099036
Epoch[044] Batch [0549]/[3760]	Speed: 117.356595 samples/sec	 accuracy=29.752841	 loss=3.245460	 lr=0.099034
Epoch[044] Batch [0599]/[3760]	Speed: 117.027243 samples/sec	 accuracy=29.750000	 loss=3.244120	 lr=0.099031
Epoch[044] Batch [0649]/[3760]	Speed: 116.279384 samples/sec	 accuracy=29.699519	 loss=3.243642	 lr=0.099029
Epoch[044] Batch [0699]/[3760]	Speed: 117.652342 samples/sec	 accuracy=29.582589	 loss=3.246520	 lr=0.099026
Epoch[044] Batch [0749]/[3760]	Speed: 116.813566 samples/sec	 accuracy=29.722917	 loss=3.244493	 lr=0.099024
Epoch[044] Batch [0799]/[3760]	Speed: 117.710048 samples/sec	 accuracy=29.666016	 loss=3.244790	 lr=0.099021
Epoch[044] Batch [0849]/[3760]	Speed: 117.373296 samples/sec	 accuracy=29.698529	 loss=3.241606	 lr=0.099019
Epoch[044] Batch [0899]/[3760]	Speed: 117.097726 samples/sec	 accuracy=29.796875	 loss=3.240702	 lr=0.099016
Epoch[044] Batch [0949]/[3760]	Speed: 117.717786 samples/sec	 accuracy=29.797697	 loss=3.241548	 lr=0.099014
Epoch[044] Batch [0999]/[3760]	Speed: 116.501449 samples/sec	 accuracy=29.776562	 loss=3.241900	 lr=0.099011
Epoch[044] Batch [1049]/[3760]	Speed: 117.476713 samples/sec	 accuracy=29.773810	 loss=3.241517	 lr=0.099008
Epoch[044] Batch [1099]/[3760]	Speed: 117.058355 samples/sec	 accuracy=29.775568	 loss=3.240879	 lr=0.099006
Epoch[044] Batch [1149]/[3760]	Speed: 118.141832 samples/sec	 accuracy=29.760870	 loss=3.242314	 lr=0.099003
Epoch[044] Batch [1199]/[3760]	Speed: 117.699396 samples/sec	 accuracy=29.688802	 loss=3.244691	 lr=0.099001
Epoch[044] Batch [1249]/[3760]	Speed: 117.314993 samples/sec	 accuracy=29.630000	 loss=3.247374	 lr=0.098998
Epoch[044] Batch [1299]/[3760]	Speed: 117.626815 samples/sec	 accuracy=29.614183	 loss=3.246222	 lr=0.098996
Epoch[044] Batch [1349]/[3760]	Speed: 117.041669 samples/sec	 accuracy=29.634259	 loss=3.247814	 lr=0.098993
Epoch[044] Batch [1399]/[3760]	Speed: 117.405878 samples/sec	 accuracy=29.645089	 loss=3.245299	 lr=0.098990
Epoch[044] Batch [1449]/[3760]	Speed: 117.713705 samples/sec	 accuracy=29.679957	 loss=3.242211	 lr=0.098988
Epoch[044] Batch [1499]/[3760]	Speed: 117.674500 samples/sec	 accuracy=29.614583	 loss=3.243217	 lr=0.098985
Epoch[044] Batch [1549]/[3760]	Speed: 118.012519 samples/sec	 accuracy=29.631048	 loss=3.243237	 lr=0.098983
Epoch[044] Batch [1599]/[3760]	Speed: 117.184478 samples/sec	 accuracy=29.642578	 loss=3.244439	 lr=0.098980
Epoch[044] Batch [1649]/[3760]	Speed: 117.387147 samples/sec	 accuracy=29.670455	 loss=3.242610	 lr=0.098978
Epoch[044] Batch [1699]/[3760]	Speed: 117.578617 samples/sec	 accuracy=29.638787	 loss=3.243943	 lr=0.098975
Epoch[044] Batch [1749]/[3760]	Speed: 117.628771 samples/sec	 accuracy=29.597321	 loss=3.245372	 lr=0.098972
Epoch[044] Batch [1799]/[3760]	Speed: 116.820194 samples/sec	 accuracy=29.596354	 loss=3.245428	 lr=0.098970
Epoch[044] Batch [1849]/[3760]	Speed: 117.992926 samples/sec	 accuracy=29.618243	 loss=3.245862	 lr=0.098967
Epoch[044] Batch [1899]/[3760]	Speed: 117.682860 samples/sec	 accuracy=29.621711	 loss=3.246353	 lr=0.098964
Epoch[044] Batch [1949]/[3760]	Speed: 117.955066 samples/sec	 accuracy=29.583333	 loss=3.248233	 lr=0.098962
Epoch[044] Batch [1999]/[3760]	Speed: 118.017104 samples/sec	 accuracy=29.582812	 loss=3.246742	 lr=0.098959
Epoch[044] Batch [2049]/[3760]	Speed: 117.058366 samples/sec	 accuracy=29.592988	 loss=3.247715	 lr=0.098957
Epoch[044] Batch [2099]/[3760]	Speed: 116.912228 samples/sec	 accuracy=29.581845	 loss=3.247720	 lr=0.098954
Epoch[044] Batch [2149]/[3760]	Speed: 117.771727 samples/sec	 accuracy=29.594477	 loss=3.248031	 lr=0.098951
Epoch[044] Batch [2199]/[3760]	Speed: 117.905980 samples/sec	 accuracy=29.591619	 loss=3.247308	 lr=0.098949
Epoch[044] Batch [2249]/[3760]	Speed: 117.653450 samples/sec	 accuracy=29.566667	 loss=3.246719	 lr=0.098946
Epoch[044] Batch [2299]/[3760]	Speed: 117.835379 samples/sec	 accuracy=29.561821	 loss=3.247881	 lr=0.098944
Epoch[044] Batch [2349]/[3760]	Speed: 116.697692 samples/sec	 accuracy=29.591755	 loss=3.246438	 lr=0.098941
Epoch[044] Batch [2399]/[3760]	Speed: 117.236971 samples/sec	 accuracy=29.623047	 loss=3.245444	 lr=0.098938
Epoch[044] Batch [2449]/[3760]	Speed: 117.616561 samples/sec	 accuracy=29.612883	 loss=3.245204	 lr=0.098936
Epoch[044] Batch [2499]/[3760]	Speed: 117.592469 samples/sec	 accuracy=29.599375	 loss=3.246628	 lr=0.098933
Epoch[044] Batch [2549]/[3760]	Speed: 117.860518 samples/sec	 accuracy=29.604779	 loss=3.246468	 lr=0.098930
Epoch[044] Batch [2599]/[3760]	Speed: 117.159116 samples/sec	 accuracy=29.589543	 loss=3.246755	 lr=0.098928
Epoch[044] Batch [2649]/[3760]	Speed: 116.554477 samples/sec	 accuracy=29.591392	 loss=3.246352	 lr=0.098925
Epoch[044] Batch [2699]/[3760]	Speed: 117.179271 samples/sec	 accuracy=29.604167	 loss=3.247238	 lr=0.098922
Epoch[044] Batch [2749]/[3760]	Speed: 117.383071 samples/sec	 accuracy=29.605114	 loss=3.247868	 lr=0.098920
Epoch[044] Batch [2799]/[3760]	Speed: 117.453582 samples/sec	 accuracy=29.601562	 loss=3.248346	 lr=0.098917
Epoch[044] Batch [2849]/[3760]	Speed: 117.165567 samples/sec	 accuracy=29.611842	 loss=3.247913	 lr=0.098914
Epoch[044] Batch [2899]/[3760]	Speed: 117.731342 samples/sec	 accuracy=29.591595	 loss=3.248939	 lr=0.098912
Epoch[044] Batch [2949]/[3760]	Speed: 117.349466 samples/sec	 accuracy=29.590572	 loss=3.249407	 lr=0.098909
Epoch[044] Batch [2999]/[3760]	Speed: 117.759666 samples/sec	 accuracy=29.607292	 loss=3.249537	 lr=0.098906
Epoch[044] Batch [3049]/[3760]	Speed: 117.533918 samples/sec	 accuracy=29.614242	 loss=3.249467	 lr=0.098904
Epoch[044] Batch [3099]/[3760]	Speed: 117.705561 samples/sec	 accuracy=29.617944	 loss=3.249477	 lr=0.098901
Epoch[044] Batch [3149]/[3760]	Speed: 117.174016 samples/sec	 accuracy=29.623512	 loss=3.249458	 lr=0.098898
Epoch[044] Batch [3199]/[3760]	Speed: 117.453695 samples/sec	 accuracy=29.606445	 loss=3.250818	 lr=0.098896
Epoch[044] Batch [3249]/[3760]	Speed: 117.653910 samples/sec	 accuracy=29.588942	 loss=3.251721	 lr=0.098893
Epoch[044] Batch [3299]/[3760]	Speed: 117.204671 samples/sec	 accuracy=29.566288	 loss=3.252253	 lr=0.098890
Epoch[044] Batch [3349]/[3760]	Speed: 117.205236 samples/sec	 accuracy=29.574160	 loss=3.251854	 lr=0.098887
Epoch[044] Batch [3399]/[3760]	Speed: 117.543425 samples/sec	 accuracy=29.581342	 loss=3.251648	 lr=0.098885
Epoch[044] Batch [3449]/[3760]	Speed: 117.561565 samples/sec	 accuracy=29.558877	 loss=3.253295	 lr=0.098882
Epoch[044] Batch [3499]/[3760]	Speed: 117.946611 samples/sec	 accuracy=29.563839	 loss=3.253533	 lr=0.098879
Epoch[044] Batch [3549]/[3760]	Speed: 117.409359 samples/sec	 accuracy=29.572183	 loss=3.254353	 lr=0.098877
Epoch[044] Batch [3599]/[3760]	Speed: 117.369348 samples/sec	 accuracy=29.564236	 loss=3.254808	 lr=0.098874
Epoch[044] Batch [3649]/[3760]	Speed: 117.454154 samples/sec	 accuracy=29.561216	 loss=3.255029	 lr=0.098871
Epoch[044] Batch [3699]/[3760]	Speed: 117.326883 samples/sec	 accuracy=29.575169	 loss=3.254285	 lr=0.098868
Epoch[044] Batch [3749]/[3760]	Speed: 119.975687 samples/sec	 accuracy=29.568333	 loss=3.254638	 lr=0.098866
Batch [0049]/[0303]: acc-top1=30.500000 acc-top5=56.562500
Batch [0099]/[0303]: acc-top1=27.453125 acc-top5=53.343750
Batch [0149]/[0303]: acc-top1=29.739583 acc-top5=56.052083
Batch [0199]/[0303]: acc-top1=31.906250 acc-top5=58.218750
Batch [0249]/[0303]: acc-top1=31.343750 acc-top5=57.912500
Batch [0299]/[0303]: acc-top1=30.651042 acc-top5=57.937500
[Epoch 044] training: accuracy=29.556599	 loss=3.255274
[Epoch 044] speed: 116 samples/sec	time cost: 2179.034347
[Epoch 044] validation: acc-top1=30.651815 acc-top5=57.941419 loss=3.304818
Epoch[045] Batch [0049]/[3760]	Speed: 87.090370 samples/sec	 accuracy=29.562500	 loss=3.237085	 lr=0.098862
Epoch[045] Batch [0099]/[3760]	Speed: 117.579928 samples/sec	 accuracy=29.781250	 loss=3.261583	 lr=0.098860
Epoch[045] Batch [0149]/[3760]	Speed: 117.449470 samples/sec	 accuracy=29.666667	 loss=3.251042	 lr=0.098857
Epoch[045] Batch [0199]/[3760]	Speed: 116.679971 samples/sec	 accuracy=29.429688	 loss=3.258834	 lr=0.098854
Epoch[045] Batch [0249]/[3760]	Speed: 116.915419 samples/sec	 accuracy=29.637500	 loss=3.249983	 lr=0.098851
Epoch[045] Batch [0299]/[3760]	Speed: 117.748203 samples/sec	 accuracy=29.348958	 loss=3.255299	 lr=0.098849
Epoch[045] Batch [0349]/[3760]	Speed: 117.174156 samples/sec	 accuracy=29.321429	 loss=3.256704	 lr=0.098846
Epoch[045] Batch [0399]/[3760]	Speed: 116.953121 samples/sec	 accuracy=29.445312	 loss=3.251346	 lr=0.098843
Epoch[045] Batch [0449]/[3760]	Speed: 117.350670 samples/sec	 accuracy=29.458333	 loss=3.246342	 lr=0.098840
Epoch[045] Batch [0499]/[3760]	Speed: 116.527815 samples/sec	 accuracy=29.396875	 loss=3.247886	 lr=0.098838
Epoch[045] Batch [0549]/[3760]	Speed: 116.624947 samples/sec	 accuracy=29.355114	 loss=3.249024	 lr=0.098835
Epoch[045] Batch [0599]/[3760]	Speed: 116.600500 samples/sec	 accuracy=29.302083	 loss=3.251588	 lr=0.098832
Epoch[045] Batch [0649]/[3760]	Speed: 116.399008 samples/sec	 accuracy=29.290865	 loss=3.253782	 lr=0.098829
Epoch[045] Batch [0699]/[3760]	Speed: 117.432893 samples/sec	 accuracy=29.236607	 loss=3.256309	 lr=0.098827
Epoch[045] Batch [0749]/[3760]	Speed: 116.673726 samples/sec	 accuracy=29.308333	 loss=3.254944	 lr=0.098824
Epoch[045] Batch [0799]/[3760]	Speed: 116.981006 samples/sec	 accuracy=29.392578	 loss=3.248523	 lr=0.098821
Epoch[045] Batch [0849]/[3760]	Speed: 117.032762 samples/sec	 accuracy=29.433824	 loss=3.247701	 lr=0.098818
Epoch[045] Batch [0899]/[3760]	Speed: 116.762081 samples/sec	 accuracy=29.468750	 loss=3.246859	 lr=0.098815
Epoch[045] Batch [0949]/[3760]	Speed: 117.446048 samples/sec	 accuracy=29.519737	 loss=3.245191	 lr=0.098813
Epoch[045] Batch [0999]/[3760]	Speed: 117.026700 samples/sec	 accuracy=29.543750	 loss=3.246060	 lr=0.098810
Epoch[045] Batch [1049]/[3760]	Speed: 117.622452 samples/sec	 accuracy=29.593750	 loss=3.246495	 lr=0.098807
Epoch[045] Batch [1099]/[3760]	Speed: 117.471722 samples/sec	 accuracy=29.683239	 loss=3.244230	 lr=0.098804
Epoch[045] Batch [1149]/[3760]	Speed: 117.415575 samples/sec	 accuracy=29.716033	 loss=3.243664	 lr=0.098801
Epoch[045] Batch [1199]/[3760]	Speed: 117.243306 samples/sec	 accuracy=29.640625	 loss=3.245911	 lr=0.098799
Epoch[045] Batch [1249]/[3760]	Speed: 117.459892 samples/sec	 accuracy=29.681250	 loss=3.245158	 lr=0.098796
Epoch[045] Batch [1299]/[3760]	Speed: 117.443358 samples/sec	 accuracy=29.657452	 loss=3.246774	 lr=0.098793
Epoch[045] Batch [1349]/[3760]	Speed: 117.407180 samples/sec	 accuracy=29.724537	 loss=3.244660	 lr=0.098790
Epoch[045] Batch [1399]/[3760]	Speed: 117.508073 samples/sec	 accuracy=29.743304	 loss=3.245127	 lr=0.098787
Epoch[045] Batch [1449]/[3760]	Speed: 117.214495 samples/sec	 accuracy=29.741379	 loss=3.246450	 lr=0.098785
Epoch[045] Batch [1499]/[3760]	Speed: 117.864016 samples/sec	 accuracy=29.757292	 loss=3.245377	 lr=0.098782
Epoch[045] Batch [1549]/[3760]	Speed: 117.235493 samples/sec	 accuracy=29.789315	 loss=3.244328	 lr=0.098779
Epoch[045] Batch [1599]/[3760]	Speed: 116.530559 samples/sec	 accuracy=29.765625	 loss=3.244727	 lr=0.098776
Epoch[045] Batch [1649]/[3760]	Speed: 116.452949 samples/sec	 accuracy=29.748106	 loss=3.244224	 lr=0.098773
Epoch[045] Batch [1699]/[3760]	Speed: 117.446864 samples/sec	 accuracy=29.738971	 loss=3.244874	 lr=0.098770
Epoch[045] Batch [1749]/[3760]	Speed: 117.611130 samples/sec	 accuracy=29.697321	 loss=3.245154	 lr=0.098767
Epoch[045] Batch [1799]/[3760]	Speed: 116.994867 samples/sec	 accuracy=29.710938	 loss=3.245340	 lr=0.098765
Epoch[045] Batch [1849]/[3760]	Speed: 116.956646 samples/sec	 accuracy=29.692568	 loss=3.246585	 lr=0.098762
Epoch[045] Batch [1899]/[3760]	Speed: 117.107471 samples/sec	 accuracy=29.688322	 loss=3.245289	 lr=0.098759
Epoch[045] Batch [1949]/[3760]	Speed: 117.394693 samples/sec	 accuracy=29.649038	 loss=3.246080	 lr=0.098756
Epoch[045] Batch [1999]/[3760]	Speed: 117.161155 samples/sec	 accuracy=29.624219	 loss=3.248612	 lr=0.098753
Epoch[045] Batch [2049]/[3760]	Speed: 117.526033 samples/sec	 accuracy=29.599085	 loss=3.249504	 lr=0.098750
Epoch[045] Batch [2099]/[3760]	Speed: 117.033859 samples/sec	 accuracy=29.571429	 loss=3.251108	 lr=0.098747
Epoch[045] Batch [2149]/[3760]	Speed: 116.908942 samples/sec	 accuracy=29.564680	 loss=3.250902	 lr=0.098745
Epoch[045] Batch [2199]/[3760]	Speed: 117.337534 samples/sec	 accuracy=29.535511	 loss=3.252275	 lr=0.098742
Epoch[045] Batch [2249]/[3760]	Speed: 116.883869 samples/sec	 accuracy=29.539583	 loss=3.252174	 lr=0.098739
Epoch[045] Batch [2299]/[3760]	Speed: 117.189665 samples/sec	 accuracy=29.551630	 loss=3.252572	 lr=0.098736
Epoch[045] Batch [2349]/[3760]	Speed: 117.446685 samples/sec	 accuracy=29.524601	 loss=3.253456	 lr=0.098733
Epoch[045] Batch [2399]/[3760]	Speed: 117.052296 samples/sec	 accuracy=29.507161	 loss=3.253373	 lr=0.098730
Epoch[045] Batch [2449]/[3760]	Speed: 117.640750 samples/sec	 accuracy=29.503189	 loss=3.253229	 lr=0.098727
Epoch[045] Batch [2499]/[3760]	Speed: 116.954190 samples/sec	 accuracy=29.515000	 loss=3.252812	 lr=0.098724
Epoch[045] Batch [2549]/[3760]	Speed: 116.908164 samples/sec	 accuracy=29.511029	 loss=3.253213	 lr=0.098722
Epoch[045] Batch [2599]/[3760]	Speed: 117.026116 samples/sec	 accuracy=29.525240	 loss=3.252530	 lr=0.098719
Epoch[045] Batch [2649]/[3760]	Speed: 117.200752 samples/sec	 accuracy=29.519458	 loss=3.252574	 lr=0.098716
Epoch[045] Batch [2699]/[3760]	Speed: 117.280667 samples/sec	 accuracy=29.543981	 loss=3.251623	 lr=0.098713
Epoch[045] Batch [2749]/[3760]	Speed: 117.304497 samples/sec	 accuracy=29.553977	 loss=3.251157	 lr=0.098710
Epoch[045] Batch [2799]/[3760]	Speed: 117.587719 samples/sec	 accuracy=29.554688	 loss=3.250539	 lr=0.098707
Epoch[045] Batch [2849]/[3760]	Speed: 117.295005 samples/sec	 accuracy=29.562500	 loss=3.250326	 lr=0.098704
Epoch[045] Batch [2899]/[3760]	Speed: 117.620813 samples/sec	 accuracy=29.579203	 loss=3.249668	 lr=0.098701
Epoch[045] Batch [2949]/[3760]	Speed: 118.015015 samples/sec	 accuracy=29.564619	 loss=3.250349	 lr=0.098698
Epoch[045] Batch [2999]/[3760]	Speed: 116.932302 samples/sec	 accuracy=29.571354	 loss=3.249133	 lr=0.098695
Epoch[045] Batch [3049]/[3760]	Speed: 116.906135 samples/sec	 accuracy=29.558402	 loss=3.249961	 lr=0.098692
Epoch[045] Batch [3099]/[3760]	Speed: 117.162959 samples/sec	 accuracy=29.554435	 loss=3.250034	 lr=0.098689
Epoch[045] Batch [3149]/[3760]	Speed: 116.963244 samples/sec	 accuracy=29.573413	 loss=3.249927	 lr=0.098687
Epoch[045] Batch [3199]/[3760]	Speed: 118.074666 samples/sec	 accuracy=29.569336	 loss=3.249863	 lr=0.098684
Epoch[045] Batch [3249]/[3760]	Speed: 117.958736 samples/sec	 accuracy=29.562019	 loss=3.249524	 lr=0.098681
Epoch[045] Batch [3299]/[3760]	Speed: 116.951881 samples/sec	 accuracy=29.552083	 loss=3.250067	 lr=0.098678
Epoch[045] Batch [3349]/[3760]	Speed: 117.419248 samples/sec	 accuracy=29.543377	 loss=3.251015	 lr=0.098675
Epoch[045] Batch [3399]/[3760]	Speed: 117.623493 samples/sec	 accuracy=29.533088	 loss=3.250583	 lr=0.098672
Epoch[045] Batch [3449]/[3760]	Speed: 117.195256 samples/sec	 accuracy=29.538043	 loss=3.249783	 lr=0.098669
Epoch[045] Batch [3499]/[3760]	Speed: 117.509141 samples/sec	 accuracy=29.551786	 loss=3.250029	 lr=0.098666
Epoch[045] Batch [3549]/[3760]	Speed: 117.540706 samples/sec	 accuracy=29.552817	 loss=3.249774	 lr=0.098663
Epoch[045] Batch [3599]/[3760]	Speed: 118.123877 samples/sec	 accuracy=29.552083	 loss=3.249694	 lr=0.098660
Epoch[045] Batch [3649]/[3760]	Speed: 117.659741 samples/sec	 accuracy=29.554366	 loss=3.249391	 lr=0.098657
Epoch[045] Batch [3699]/[3760]	Speed: 117.798462 samples/sec	 accuracy=29.565456	 loss=3.249770	 lr=0.098654
Epoch[045] Batch [3749]/[3760]	Speed: 120.342821 samples/sec	 accuracy=29.563333	 loss=3.250394	 lr=0.098651
Batch [0049]/[0303]: acc-top1=32.062500 acc-top5=58.500000
Batch [0099]/[0303]: acc-top1=29.750000 acc-top5=56.812500
Batch [0149]/[0303]: acc-top1=30.666667 acc-top5=57.781250
Batch [0199]/[0303]: acc-top1=33.101562 acc-top5=59.765625
Batch [0249]/[0303]: acc-top1=32.993750 acc-top5=59.906250
Batch [0299]/[0303]: acc-top1=32.015625 acc-top5=59.406250
[Epoch 045] training: accuracy=29.561586	 loss=3.250515
[Epoch 045] speed: 116 samples/sec	time cost: 2180.882872
[Epoch 045] validation: acc-top1=32.002888 acc-top5=59.442038 loss=3.111359
Epoch[046] Batch [0049]/[3759]	Speed: 88.973356 samples/sec	 accuracy=29.718750	 loss=3.211418	 lr=0.098647
Epoch[046] Batch [0099]/[3759]	Speed: 116.419428 samples/sec	 accuracy=29.406250	 loss=3.225780	 lr=0.098644
Epoch[046] Batch [0149]/[3759]	Speed: 117.535435 samples/sec	 accuracy=29.562500	 loss=3.232839	 lr=0.098642
Epoch[046] Batch [0199]/[3759]	Speed: 117.331677 samples/sec	 accuracy=29.835938	 loss=3.232783	 lr=0.098639
Epoch[046] Batch [0249]/[3759]	Speed: 117.435761 samples/sec	 accuracy=29.831250	 loss=3.236302	 lr=0.098636
Epoch[046] Batch [0299]/[3759]	Speed: 116.813809 samples/sec	 accuracy=29.703125	 loss=3.231057	 lr=0.098633
Epoch[046] Batch [0349]/[3759]	Speed: 117.538712 samples/sec	 accuracy=29.723214	 loss=3.230866	 lr=0.098630
Epoch[046] Batch [0399]/[3759]	Speed: 117.977712 samples/sec	 accuracy=29.804688	 loss=3.227875	 lr=0.098627
Epoch[046] Batch [0449]/[3759]	Speed: 117.697297 samples/sec	 accuracy=29.722222	 loss=3.233685	 lr=0.098624
Epoch[046] Batch [0499]/[3759]	Speed: 116.856249 samples/sec	 accuracy=29.650000	 loss=3.236906	 lr=0.098621
Epoch[046] Batch [0549]/[3759]	Speed: 117.484584 samples/sec	 accuracy=29.571023	 loss=3.242678	 lr=0.098618
Epoch[046] Batch [0599]/[3759]	Speed: 117.875116 samples/sec	 accuracy=29.583333	 loss=3.240345	 lr=0.098615
Epoch[046] Batch [0649]/[3759]	Speed: 117.719395 samples/sec	 accuracy=29.576923	 loss=3.241923	 lr=0.098611
Epoch[046] Batch [0699]/[3759]	Speed: 117.595301 samples/sec	 accuracy=29.656250	 loss=3.240061	 lr=0.098608
Epoch[046] Batch [0749]/[3759]	Speed: 116.706682 samples/sec	 accuracy=29.629167	 loss=3.241935	 lr=0.098605
Epoch[046] Batch [0799]/[3759]	Speed: 116.909585 samples/sec	 accuracy=29.695312	 loss=3.238467	 lr=0.098602
Epoch[046] Batch [0849]/[3759]	Speed: 117.234343 samples/sec	 accuracy=29.665441	 loss=3.241153	 lr=0.098599
Epoch[046] Batch [0899]/[3759]	Speed: 117.039425 samples/sec	 accuracy=29.633681	 loss=3.242412	 lr=0.098596
Epoch[046] Batch [0949]/[3759]	Speed: 117.479963 samples/sec	 accuracy=29.654605	 loss=3.242812	 lr=0.098593
Epoch[046] Batch [0999]/[3759]	Speed: 117.267738 samples/sec	 accuracy=29.675000	 loss=3.241242	 lr=0.098590
Epoch[046] Batch [1049]/[3759]	Speed: 117.661186 samples/sec	 accuracy=29.703869	 loss=3.241065	 lr=0.098587
Epoch[046] Batch [1099]/[3759]	Speed: 117.564248 samples/sec	 accuracy=29.663352	 loss=3.243126	 lr=0.098584
Epoch[046] Batch [1149]/[3759]	Speed: 117.961846 samples/sec	 accuracy=29.722826	 loss=3.240536	 lr=0.098581
Epoch[046] Batch [1199]/[3759]	Speed: 117.788614 samples/sec	 accuracy=29.721354	 loss=3.240568	 lr=0.098578
Epoch[046] Batch [1249]/[3759]	Speed: 117.732272 samples/sec	 accuracy=29.750000	 loss=3.238409	 lr=0.098575
Epoch[046] Batch [1299]/[3759]	Speed: 117.410586 samples/sec	 accuracy=29.717548	 loss=3.239410	 lr=0.098572
Epoch[046] Batch [1349]/[3759]	Speed: 118.198088 samples/sec	 accuracy=29.738426	 loss=3.239580	 lr=0.098569
Epoch[046] Batch [1399]/[3759]	Speed: 118.249449 samples/sec	 accuracy=29.734375	 loss=3.240124	 lr=0.098566
Epoch[046] Batch [1449]/[3759]	Speed: 117.789182 samples/sec	 accuracy=29.707974	 loss=3.241643	 lr=0.098563
Epoch[046] Batch [1499]/[3759]	Speed: 117.764431 samples/sec	 accuracy=29.691667	 loss=3.241175	 lr=0.098560
Epoch[046] Batch [1549]/[3759]	Speed: 117.415755 samples/sec	 accuracy=29.668347	 loss=3.242317	 lr=0.098557
Epoch[046] Batch [1599]/[3759]	Speed: 117.086265 samples/sec	 accuracy=29.708008	 loss=3.241042	 lr=0.098554
Epoch[046] Batch [1649]/[3759]	Speed: 117.736975 samples/sec	 accuracy=29.714962	 loss=3.242251	 lr=0.098550
Epoch[046] Batch [1699]/[3759]	Speed: 117.407792 samples/sec	 accuracy=29.771140	 loss=3.241058	 lr=0.098547
Epoch[046] Batch [1749]/[3759]	Speed: 117.897678 samples/sec	 accuracy=29.768750	 loss=3.242088	 lr=0.098544
Epoch[046] Batch [1799]/[3759]	Speed: 117.476750 samples/sec	 accuracy=29.756944	 loss=3.242540	 lr=0.098541
Epoch[046] Batch [1849]/[3759]	Speed: 116.558228 samples/sec	 accuracy=29.750000	 loss=3.243846	 lr=0.098538
Epoch[046] Batch [1899]/[3759]	Speed: 117.747443 samples/sec	 accuracy=29.736842	 loss=3.243381	 lr=0.098535
Epoch[046] Batch [1949]/[3759]	Speed: 118.073441 samples/sec	 accuracy=29.752404	 loss=3.242984	 lr=0.098532
Epoch[046] Batch [1999]/[3759]	Speed: 118.171154 samples/sec	 accuracy=29.753906	 loss=3.244206	 lr=0.098529
Epoch[046] Batch [2049]/[3759]	Speed: 117.252629 samples/sec	 accuracy=29.788110	 loss=3.244124	 lr=0.098526
Epoch[046] Batch [2099]/[3759]	Speed: 117.153894 samples/sec	 accuracy=29.793155	 loss=3.244560	 lr=0.098523
Epoch[046] Batch [2149]/[3759]	Speed: 118.274209 samples/sec	 accuracy=29.788517	 loss=3.246103	 lr=0.098519
Epoch[046] Batch [2199]/[3759]	Speed: 117.813690 samples/sec	 accuracy=29.784091	 loss=3.246706	 lr=0.098516
Epoch[046] Batch [2249]/[3759]	Speed: 117.172432 samples/sec	 accuracy=29.766667	 loss=3.247048	 lr=0.098513
Epoch[046] Batch [2299]/[3759]	Speed: 116.797883 samples/sec	 accuracy=29.754076	 loss=3.247987	 lr=0.098510
Epoch[046] Batch [2349]/[3759]	Speed: 116.953818 samples/sec	 accuracy=29.773271	 loss=3.247989	 lr=0.098507
Epoch[046] Batch [2399]/[3759]	Speed: 118.105923 samples/sec	 accuracy=29.785156	 loss=3.247691	 lr=0.098504
Epoch[046] Batch [2449]/[3759]	Speed: 117.634134 samples/sec	 accuracy=29.796556	 loss=3.247099	 lr=0.098501
Epoch[046] Batch [2499]/[3759]	Speed: 117.695862 samples/sec	 accuracy=29.784375	 loss=3.246813	 lr=0.098498
Epoch[046] Batch [2549]/[3759]	Speed: 117.492263 samples/sec	 accuracy=29.780637	 loss=3.247402	 lr=0.098494
Epoch[046] Batch [2599]/[3759]	Speed: 117.465746 samples/sec	 accuracy=29.768029	 loss=3.247301	 lr=0.098491
Epoch[046] Batch [2649]/[3759]	Speed: 117.554297 samples/sec	 accuracy=29.761203	 loss=3.247803	 lr=0.098488
Epoch[046] Batch [2699]/[3759]	Speed: 117.744676 samples/sec	 accuracy=29.726852	 loss=3.249037	 lr=0.098485
Epoch[046] Batch [2749]/[3759]	Speed: 117.443980 samples/sec	 accuracy=29.717614	 loss=3.248779	 lr=0.098482
Epoch[046] Batch [2799]/[3759]	Speed: 117.095277 samples/sec	 accuracy=29.718192	 loss=3.249411	 lr=0.098479
Epoch[046] Batch [2849]/[3759]	Speed: 116.684727 samples/sec	 accuracy=29.728070	 loss=3.248049	 lr=0.098476
Epoch[046] Batch [2899]/[3759]	Speed: 117.339686 samples/sec	 accuracy=29.745690	 loss=3.247965	 lr=0.098472
Epoch[046] Batch [2949]/[3759]	Speed: 117.954941 samples/sec	 accuracy=29.718750	 loss=3.249136	 lr=0.098469
Epoch[046] Batch [2999]/[3759]	Speed: 117.651568 samples/sec	 accuracy=29.708333	 loss=3.249049	 lr=0.098466
Epoch[046] Batch [3049]/[3759]	Speed: 116.694165 samples/sec	 accuracy=29.678279	 loss=3.249698	 lr=0.098463
Epoch[046] Batch [3099]/[3759]	Speed: 117.161939 samples/sec	 accuracy=29.674395	 loss=3.250362	 lr=0.098460
Epoch[046] Batch [3149]/[3759]	Speed: 117.921288 samples/sec	 accuracy=29.683532	 loss=3.249818	 lr=0.098457
Epoch[046] Batch [3199]/[3759]	Speed: 118.357198 samples/sec	 accuracy=29.702148	 loss=3.249176	 lr=0.098453
Epoch[046] Batch [3249]/[3759]	Speed: 117.853468 samples/sec	 accuracy=29.688462	 loss=3.249369	 lr=0.098450
Epoch[046] Batch [3299]/[3759]	Speed: 116.894004 samples/sec	 accuracy=29.678504	 loss=3.249462	 lr=0.098447
Epoch[046] Batch [3349]/[3759]	Speed: 117.245414 samples/sec	 accuracy=29.695896	 loss=3.249155	 lr=0.098444
Epoch[046] Batch [3399]/[3759]	Speed: 117.510941 samples/sec	 accuracy=29.701287	 loss=3.249127	 lr=0.098441
Epoch[046] Batch [3449]/[3759]	Speed: 117.253249 samples/sec	 accuracy=29.698370	 loss=3.249081	 lr=0.098437
Epoch[046] Batch [3499]/[3759]	Speed: 118.295410 samples/sec	 accuracy=29.702232	 loss=3.248477	 lr=0.098434
Epoch[046] Batch [3549]/[3759]	Speed: 117.429746 samples/sec	 accuracy=29.707746	 loss=3.248329	 lr=0.098431
Epoch[046] Batch [3599]/[3759]	Speed: 117.517885 samples/sec	 accuracy=29.707899	 loss=3.248168	 lr=0.098428
Epoch[046] Batch [3649]/[3759]	Speed: 117.485954 samples/sec	 accuracy=29.710616	 loss=3.248312	 lr=0.098425
Epoch[046] Batch [3699]/[3759]	Speed: 117.279401 samples/sec	 accuracy=29.693834	 loss=3.249114	 lr=0.098421
Epoch[046] Batch [3749]/[3759]	Speed: 121.092608 samples/sec	 accuracy=29.720000	 loss=3.248090	 lr=0.098418
Batch [0049]/[0303]: acc-top1=29.406250 acc-top5=54.437500
Batch [0099]/[0303]: acc-top1=27.328125 acc-top5=52.968750
Batch [0149]/[0303]: acc-top1=29.187500 acc-top5=54.697917
Batch [0199]/[0303]: acc-top1=31.804688 acc-top5=57.421875
Batch [0249]/[0303]: acc-top1=30.793750 acc-top5=56.768750
Batch [0299]/[0303]: acc-top1=30.177083 acc-top5=56.838542
[Epoch 046] training: accuracy=29.715765	 loss=3.248142
[Epoch 046] speed: 117 samples/sec	time cost: 2175.418072
[Epoch 046] validation: acc-top1=30.203177 acc-top5=56.879125 loss=3.322090
Epoch[047] Batch [0049]/[3760]	Speed: 86.056271 samples/sec	 accuracy=30.750000	 loss=3.163039	 lr=0.098414
Epoch[047] Batch [0099]/[3760]	Speed: 117.225589 samples/sec	 accuracy=30.421875	 loss=3.182868	 lr=0.098411
Epoch[047] Batch [0149]/[3760]	Speed: 118.226640 samples/sec	 accuracy=29.802083	 loss=3.224263	 lr=0.098408
Epoch[047] Batch [0199]/[3760]	Speed: 116.531914 samples/sec	 accuracy=30.015625	 loss=3.224379	 lr=0.098405
Epoch[047] Batch [0249]/[3760]	Speed: 116.871015 samples/sec	 accuracy=30.081250	 loss=3.226351	 lr=0.098401
Epoch[047] Batch [0299]/[3760]	Speed: 116.492148 samples/sec	 accuracy=29.958333	 loss=3.228457	 lr=0.098398
Epoch[047] Batch [0349]/[3760]	Speed: 116.700218 samples/sec	 accuracy=29.910714	 loss=3.232423	 lr=0.098395
Epoch[047] Batch [0399]/[3760]	Speed: 116.756028 samples/sec	 accuracy=29.988281	 loss=3.218682	 lr=0.098392
Epoch[047] Batch [0449]/[3760]	Speed: 117.489894 samples/sec	 accuracy=30.104167	 loss=3.215191	 lr=0.098388
Epoch[047] Batch [0499]/[3760]	Speed: 117.420625 samples/sec	 accuracy=29.984375	 loss=3.217301	 lr=0.098385
Epoch[047] Batch [0549]/[3760]	Speed: 117.398130 samples/sec	 accuracy=29.923295	 loss=3.221108	 lr=0.098382
Epoch[047] Batch [0599]/[3760]	Speed: 117.570723 samples/sec	 accuracy=29.820312	 loss=3.225732	 lr=0.098379
Epoch[047] Batch [0649]/[3760]	Speed: 117.377972 samples/sec	 accuracy=29.776442	 loss=3.230270	 lr=0.098375
Epoch[047] Batch [0699]/[3760]	Speed: 115.993939 samples/sec	 accuracy=29.691964	 loss=3.232900	 lr=0.098372
Epoch[047] Batch [0749]/[3760]	Speed: 116.917323 samples/sec	 accuracy=29.820833	 loss=3.229395	 lr=0.098369
Epoch[047] Batch [0799]/[3760]	Speed: 117.216814 samples/sec	 accuracy=29.863281	 loss=3.228405	 lr=0.098366
Epoch[047] Batch [0849]/[3760]	Speed: 116.830402 samples/sec	 accuracy=29.808824	 loss=3.229834	 lr=0.098362
Epoch[047] Batch [0899]/[3760]	Speed: 117.892894 samples/sec	 accuracy=29.730903	 loss=3.231098	 lr=0.098359
Epoch[047] Batch [0949]/[3760]	Speed: 116.953966 samples/sec	 accuracy=29.717105	 loss=3.230589	 lr=0.098356
Epoch[047] Batch [0999]/[3760]	Speed: 116.971695 samples/sec	 accuracy=29.718750	 loss=3.230295	 lr=0.098353
Epoch[047] Batch [1049]/[3760]	Speed: 117.838103 samples/sec	 accuracy=29.680060	 loss=3.230047	 lr=0.098349
Epoch[047] Batch [1099]/[3760]	Speed: 117.641929 samples/sec	 accuracy=29.669034	 loss=3.231969	 lr=0.098346
Epoch[047] Batch [1149]/[3760]	Speed: 117.603340 samples/sec	 accuracy=29.668478	 loss=3.234420	 lr=0.098343
Epoch[047] Batch [1199]/[3760]	Speed: 116.824492 samples/sec	 accuracy=29.704427	 loss=3.232065	 lr=0.098339
Epoch[047] Batch [1249]/[3760]	Speed: 117.505261 samples/sec	 accuracy=29.650000	 loss=3.231820	 lr=0.098336
Epoch[047] Batch [1299]/[3760]	Speed: 117.410672 samples/sec	 accuracy=29.590144	 loss=3.233451	 lr=0.098333
Epoch[047] Batch [1349]/[3760]	Speed: 116.714121 samples/sec	 accuracy=29.568287	 loss=3.236364	 lr=0.098329
Epoch[047] Batch [1399]/[3760]	Speed: 118.046588 samples/sec	 accuracy=29.580357	 loss=3.237684	 lr=0.098326
Epoch[047] Batch [1449]/[3760]	Speed: 117.031854 samples/sec	 accuracy=29.577586	 loss=3.237949	 lr=0.098323
Epoch[047] Batch [1499]/[3760]	Speed: 116.802437 samples/sec	 accuracy=29.566667	 loss=3.237989	 lr=0.098320
Epoch[047] Batch [1549]/[3760]	Speed: 117.969906 samples/sec	 accuracy=29.572581	 loss=3.238078	 lr=0.098316
Epoch[047] Batch [1599]/[3760]	Speed: 117.209530 samples/sec	 accuracy=29.579102	 loss=3.239706	 lr=0.098313
Epoch[047] Batch [1649]/[3760]	Speed: 117.925019 samples/sec	 accuracy=29.556818	 loss=3.240750	 lr=0.098310
Epoch[047] Batch [1699]/[3760]	Speed: 118.352656 samples/sec	 accuracy=29.568015	 loss=3.238787	 lr=0.098306
Epoch[047] Batch [1749]/[3760]	Speed: 117.026306 samples/sec	 accuracy=29.563393	 loss=3.238954	 lr=0.098303
Epoch[047] Batch [1799]/[3760]	Speed: 117.716935 samples/sec	 accuracy=29.572917	 loss=3.238797	 lr=0.098300
Epoch[047] Batch [1849]/[3760]	Speed: 117.376037 samples/sec	 accuracy=29.570946	 loss=3.238447	 lr=0.098296
Epoch[047] Batch [1899]/[3760]	Speed: 116.821596 samples/sec	 accuracy=29.569901	 loss=3.239261	 lr=0.098293
Epoch[047] Batch [1949]/[3760]	Speed: 116.844044 samples/sec	 accuracy=29.579327	 loss=3.239749	 lr=0.098290
Epoch[047] Batch [1999]/[3760]	Speed: 117.552207 samples/sec	 accuracy=29.601562	 loss=3.238841	 lr=0.098286
Epoch[047] Batch [2049]/[3760]	Speed: 117.032197 samples/sec	 accuracy=29.626524	 loss=3.237473	 lr=0.098283
Epoch[047] Batch [2099]/[3760]	Speed: 117.491226 samples/sec	 accuracy=29.670387	 loss=3.235474	 lr=0.098280
Epoch[047] Batch [2149]/[3760]	Speed: 117.642879 samples/sec	 accuracy=29.636628	 loss=3.235995	 lr=0.098276
Epoch[047] Batch [2199]/[3760]	Speed: 117.646221 samples/sec	 accuracy=29.644176	 loss=3.236075	 lr=0.098273
Epoch[047] Batch [2249]/[3760]	Speed: 117.180514 samples/sec	 accuracy=29.647917	 loss=3.236878	 lr=0.098269
Epoch[047] Batch [2299]/[3760]	Speed: 117.376798 samples/sec	 accuracy=29.639946	 loss=3.236932	 lr=0.098266
Epoch[047] Batch [2349]/[3760]	Speed: 117.930422 samples/sec	 accuracy=29.644947	 loss=3.237054	 lr=0.098263
Epoch[047] Batch [2399]/[3760]	Speed: 118.151813 samples/sec	 accuracy=29.658854	 loss=3.237667	 lr=0.098259
Epoch[047] Batch [2449]/[3760]	Speed: 117.227783 samples/sec	 accuracy=29.684311	 loss=3.236699	 lr=0.098256
Epoch[047] Batch [2499]/[3760]	Speed: 117.707692 samples/sec	 accuracy=29.695000	 loss=3.236661	 lr=0.098253
Epoch[047] Batch [2549]/[3760]	Speed: 117.769639 samples/sec	 accuracy=29.718750	 loss=3.236769	 lr=0.098249
Epoch[047] Batch [2599]/[3760]	Speed: 117.990037 samples/sec	 accuracy=29.720553	 loss=3.236639	 lr=0.098246
Epoch[047] Batch [2649]/[3760]	Speed: 116.948087 samples/sec	 accuracy=29.706368	 loss=3.238224	 lr=0.098242
Epoch[047] Batch [2699]/[3760]	Speed: 116.833750 samples/sec	 accuracy=29.710648	 loss=3.237861	 lr=0.098239
Epoch[047] Batch [2749]/[3760]	Speed: 117.431957 samples/sec	 accuracy=29.706818	 loss=3.238688	 lr=0.098236
Epoch[047] Batch [2799]/[3760]	Speed: 117.081296 samples/sec	 accuracy=29.713170	 loss=3.238713	 lr=0.098232
Epoch[047] Batch [2849]/[3760]	Speed: 117.353958 samples/sec	 accuracy=29.686952	 loss=3.239530	 lr=0.098229
Epoch[047] Batch [2899]/[3760]	Speed: 117.163017 samples/sec	 accuracy=29.705280	 loss=3.240039	 lr=0.098225
Epoch[047] Batch [2949]/[3760]	Speed: 117.511646 samples/sec	 accuracy=29.679555	 loss=3.241180	 lr=0.098222
Epoch[047] Batch [2999]/[3760]	Speed: 117.436692 samples/sec	 accuracy=29.653125	 loss=3.242609	 lr=0.098219
Epoch[047] Batch [3049]/[3760]	Speed: 117.093022 samples/sec	 accuracy=29.658811	 loss=3.242741	 lr=0.098215
Epoch[047] Batch [3099]/[3760]	Speed: 117.527993 samples/sec	 accuracy=29.647177	 loss=3.243237	 lr=0.098212
Epoch[047] Batch [3149]/[3760]	Speed: 117.702179 samples/sec	 accuracy=29.631448	 loss=3.243319	 lr=0.098208
Epoch[047] Batch [3199]/[3760]	Speed: 117.549067 samples/sec	 accuracy=29.645996	 loss=3.242901	 lr=0.098205
Epoch[047] Batch [3249]/[3760]	Speed: 117.223862 samples/sec	 accuracy=29.632212	 loss=3.243882	 lr=0.098202
Epoch[047] Batch [3299]/[3760]	Speed: 117.559528 samples/sec	 accuracy=29.618845	 loss=3.244682	 lr=0.098198
Epoch[047] Batch [3349]/[3760]	Speed: 118.296751 samples/sec	 accuracy=29.624534	 loss=3.244527	 lr=0.098195
Epoch[047] Batch [3399]/[3760]	Speed: 116.928811 samples/sec	 accuracy=29.601562	 loss=3.245192	 lr=0.098191
Epoch[047] Batch [3449]/[3760]	Speed: 117.534481 samples/sec	 accuracy=29.619112	 loss=3.244389	 lr=0.098188
Epoch[047] Batch [3499]/[3760]	Speed: 117.862140 samples/sec	 accuracy=29.627232	 loss=3.244315	 lr=0.098184
Epoch[047] Batch [3549]/[3760]	Speed: 117.446749 samples/sec	 accuracy=29.632482	 loss=3.244585	 lr=0.098181
Epoch[047] Batch [3599]/[3760]	Speed: 118.323644 samples/sec	 accuracy=29.647569	 loss=3.244017	 lr=0.098177
Epoch[047] Batch [3649]/[3760]	Speed: 117.335364 samples/sec	 accuracy=29.644264	 loss=3.244358	 lr=0.098174
Epoch[047] Batch [3699]/[3760]	Speed: 117.584753 samples/sec	 accuracy=29.635557	 loss=3.244357	 lr=0.098171
Epoch[047] Batch [3749]/[3760]	Speed: 121.361032 samples/sec	 accuracy=29.640000	 loss=3.244592	 lr=0.098167
Batch [0049]/[0303]: acc-top1=29.437500 acc-top5=56.656250
Batch [0099]/[0303]: acc-top1=27.078125 acc-top5=53.015625
Batch [0149]/[0303]: acc-top1=27.489583 acc-top5=53.541667
Batch [0199]/[0303]: acc-top1=29.882812 acc-top5=55.664062
Batch [0249]/[0303]: acc-top1=29.087500 acc-top5=55.537500
Batch [0299]/[0303]: acc-top1=28.718750 acc-top5=55.765625
[Epoch 047] training: accuracy=29.637217	 loss=3.244852
[Epoch 047] speed: 116 samples/sec	time cost: 2179.523314
[Epoch 047] validation: acc-top1=28.656147 acc-top5=55.770421 loss=3.499115
Epoch[048] Batch [0049]/[3760]	Speed: 86.989068 samples/sec	 accuracy=30.218750	 loss=3.203924	 lr=0.098163
Epoch[048] Batch [0099]/[3760]	Speed: 116.481391 samples/sec	 accuracy=29.703125	 loss=3.212839	 lr=0.098159
Epoch[048] Batch [0149]/[3760]	Speed: 117.554914 samples/sec	 accuracy=30.031250	 loss=3.188179	 lr=0.098156
Epoch[048] Batch [0199]/[3760]	Speed: 117.030223 samples/sec	 accuracy=29.906250	 loss=3.201264	 lr=0.098153
Epoch[048] Batch [0249]/[3760]	Speed: 117.346614 samples/sec	 accuracy=29.937500	 loss=3.202287	 lr=0.098149
Epoch[048] Batch [0299]/[3760]	Speed: 117.570994 samples/sec	 accuracy=29.848958	 loss=3.214519	 lr=0.098146
Epoch[048] Batch [0349]/[3760]	Speed: 118.641595 samples/sec	 accuracy=29.754464	 loss=3.226287	 lr=0.098142
Epoch[048] Batch [0399]/[3760]	Speed: 118.112206 samples/sec	 accuracy=29.660156	 loss=3.230113	 lr=0.098139
Epoch[048] Batch [0449]/[3760]	Speed: 117.954085 samples/sec	 accuracy=29.701389	 loss=3.225575	 lr=0.098135
Epoch[048] Batch [0499]/[3760]	Speed: 117.419666 samples/sec	 accuracy=29.850000	 loss=3.220070	 lr=0.098132
Epoch[048] Batch [0549]/[3760]	Speed: 117.393917 samples/sec	 accuracy=29.869318	 loss=3.217318	 lr=0.098128
Epoch[048] Batch [0599]/[3760]	Speed: 117.671091 samples/sec	 accuracy=29.882812	 loss=3.218223	 lr=0.098125
Epoch[048] Batch [0649]/[3760]	Speed: 117.549792 samples/sec	 accuracy=29.887019	 loss=3.219147	 lr=0.098121
Epoch[048] Batch [0699]/[3760]	Speed: 118.067928 samples/sec	 accuracy=29.892857	 loss=3.217788	 lr=0.098118
Epoch[048] Batch [0749]/[3760]	Speed: 117.231905 samples/sec	 accuracy=29.950000	 loss=3.217633	 lr=0.098114
Epoch[048] Batch [0799]/[3760]	Speed: 117.160566 samples/sec	 accuracy=29.955078	 loss=3.221269	 lr=0.098111
Epoch[048] Batch [0849]/[3760]	Speed: 117.076420 samples/sec	 accuracy=29.990809	 loss=3.225534	 lr=0.098107
Epoch[048] Batch [0899]/[3760]	Speed: 117.124840 samples/sec	 accuracy=29.920139	 loss=3.227242	 lr=0.098104
Epoch[048] Batch [0949]/[3760]	Speed: 117.766934 samples/sec	 accuracy=29.960526	 loss=3.227142	 lr=0.098100
Epoch[048] Batch [0999]/[3760]	Speed: 117.605331 samples/sec	 accuracy=29.967187	 loss=3.226319	 lr=0.098097
Epoch[048] Batch [1049]/[3760]	Speed: 117.375574 samples/sec	 accuracy=29.949405	 loss=3.226708	 lr=0.098093
Epoch[048] Batch [1099]/[3760]	Speed: 117.280114 samples/sec	 accuracy=29.990057	 loss=3.225894	 lr=0.098089
Epoch[048] Batch [1149]/[3760]	Speed: 117.435325 samples/sec	 accuracy=29.998641	 loss=3.224970	 lr=0.098086
Epoch[048] Batch [1199]/[3760]	Speed: 117.988573 samples/sec	 accuracy=29.957031	 loss=3.228665	 lr=0.098082
Epoch[048] Batch [1249]/[3760]	Speed: 117.315528 samples/sec	 accuracy=29.977500	 loss=3.227930	 lr=0.098079
Epoch[048] Batch [1299]/[3760]	Speed: 118.318413 samples/sec	 accuracy=29.998798	 loss=3.227804	 lr=0.098075
Epoch[048] Batch [1349]/[3760]	Speed: 116.959760 samples/sec	 accuracy=29.993056	 loss=3.228548	 lr=0.098072
Epoch[048] Batch [1399]/[3760]	Speed: 117.113440 samples/sec	 accuracy=30.006696	 loss=3.229539	 lr=0.098068
Epoch[048] Batch [1449]/[3760]	Speed: 117.723790 samples/sec	 accuracy=29.955819	 loss=3.231266	 lr=0.098065
Epoch[048] Batch [1499]/[3760]	Speed: 116.909832 samples/sec	 accuracy=29.927083	 loss=3.233112	 lr=0.098061
Epoch[048] Batch [1549]/[3760]	Speed: 117.050782 samples/sec	 accuracy=29.943548	 loss=3.233494	 lr=0.098058
Epoch[048] Batch [1599]/[3760]	Speed: 117.642603 samples/sec	 accuracy=29.974609	 loss=3.233076	 lr=0.098054
Epoch[048] Batch [1649]/[3760]	Speed: 117.588725 samples/sec	 accuracy=29.970644	 loss=3.233720	 lr=0.098050
Epoch[048] Batch [1699]/[3760]	Speed: 116.686733 samples/sec	 accuracy=29.977941	 loss=3.232678	 lr=0.098047
Epoch[048] Batch [1749]/[3760]	Speed: 117.461383 samples/sec	 accuracy=29.958929	 loss=3.233310	 lr=0.098043
Epoch[048] Batch [1799]/[3760]	Speed: 117.986053 samples/sec	 accuracy=29.955729	 loss=3.233619	 lr=0.098040
Epoch[048] Batch [1849]/[3760]	Speed: 117.058830 samples/sec	 accuracy=29.913851	 loss=3.234702	 lr=0.098036
Epoch[048] Batch [1899]/[3760]	Speed: 117.286175 samples/sec	 accuracy=29.887336	 loss=3.235946	 lr=0.098033
Epoch[048] Batch [1949]/[3760]	Speed: 117.473519 samples/sec	 accuracy=29.877404	 loss=3.236364	 lr=0.098029
Epoch[048] Batch [1999]/[3760]	Speed: 117.743157 samples/sec	 accuracy=29.893750	 loss=3.236737	 lr=0.098025
Epoch[048] Batch [2049]/[3760]	Speed: 117.774676 samples/sec	 accuracy=29.923018	 loss=3.235419	 lr=0.098022
Epoch[048] Batch [2099]/[3760]	Speed: 118.320212 samples/sec	 accuracy=29.905506	 loss=3.235490	 lr=0.098018
Epoch[048] Batch [2149]/[3760]	Speed: 117.106591 samples/sec	 accuracy=29.907703	 loss=3.235830	 lr=0.098015
Epoch[048] Batch [2199]/[3760]	Speed: 118.095986 samples/sec	 accuracy=29.894886	 loss=3.235713	 lr=0.098011
Epoch[048] Batch [2249]/[3760]	Speed: 117.681794 samples/sec	 accuracy=29.895833	 loss=3.234848	 lr=0.098007
Epoch[048] Batch [2299]/[3760]	Speed: 117.227844 samples/sec	 accuracy=29.899457	 loss=3.234806	 lr=0.098004
Epoch[048] Batch [2349]/[3760]	Speed: 117.464370 samples/sec	 accuracy=29.898936	 loss=3.234183	 lr=0.098000
Epoch[048] Batch [2399]/[3760]	Speed: 117.144243 samples/sec	 accuracy=29.895182	 loss=3.234158	 lr=0.097997
Epoch[048] Batch [2449]/[3760]	Speed: 117.186925 samples/sec	 accuracy=29.889031	 loss=3.234304	 lr=0.097993
Epoch[048] Batch [2499]/[3760]	Speed: 117.341759 samples/sec	 accuracy=29.875625	 loss=3.234308	 lr=0.097989
Epoch[048] Batch [2549]/[3760]	Speed: 116.628070 samples/sec	 accuracy=29.882353	 loss=3.234566	 lr=0.097986
Epoch[048] Batch [2599]/[3760]	Speed: 117.549559 samples/sec	 accuracy=29.879808	 loss=3.234704	 lr=0.097982
Epoch[048] Batch [2649]/[3760]	Speed: 117.480260 samples/sec	 accuracy=29.867925	 loss=3.235270	 lr=0.097978
Epoch[048] Batch [2699]/[3760]	Speed: 117.276939 samples/sec	 accuracy=29.865741	 loss=3.235498	 lr=0.097975
Epoch[048] Batch [2749]/[3760]	Speed: 117.268372 samples/sec	 accuracy=29.852841	 loss=3.235894	 lr=0.097971
Epoch[048] Batch [2799]/[3760]	Speed: 117.507038 samples/sec	 accuracy=29.877790	 loss=3.235435	 lr=0.097968
Epoch[048] Batch [2849]/[3760]	Speed: 117.073287 samples/sec	 accuracy=29.913377	 loss=3.234338	 lr=0.097964
Epoch[048] Batch [2899]/[3760]	Speed: 117.672734 samples/sec	 accuracy=29.898707	 loss=3.234404	 lr=0.097960
Epoch[048] Batch [2949]/[3760]	Speed: 117.708800 samples/sec	 accuracy=29.899894	 loss=3.233909	 lr=0.097957
Epoch[048] Batch [2999]/[3760]	Speed: 117.136046 samples/sec	 accuracy=29.902604	 loss=3.233383	 lr=0.097953
Epoch[048] Batch [3049]/[3760]	Speed: 117.539121 samples/sec	 accuracy=29.909836	 loss=3.232662	 lr=0.097949
Epoch[048] Batch [3099]/[3760]	Speed: 118.171656 samples/sec	 accuracy=29.896169	 loss=3.233239	 lr=0.097946
Epoch[048] Batch [3149]/[3760]	Speed: 117.444732 samples/sec	 accuracy=29.883929	 loss=3.233662	 lr=0.097942
Epoch[048] Batch [3199]/[3760]	Speed: 117.680243 samples/sec	 accuracy=29.892090	 loss=3.233837	 lr=0.097938
Epoch[048] Batch [3249]/[3760]	Speed: 118.238698 samples/sec	 accuracy=29.901923	 loss=3.233265	 lr=0.097935
Epoch[048] Batch [3299]/[3760]	Speed: 117.785780 samples/sec	 accuracy=29.912405	 loss=3.233059	 lr=0.097931
Epoch[048] Batch [3349]/[3760]	Speed: 117.782494 samples/sec	 accuracy=29.902052	 loss=3.233815	 lr=0.097927
Epoch[048] Batch [3399]/[3760]	Speed: 118.295941 samples/sec	 accuracy=29.897059	 loss=3.233719	 lr=0.097924
Epoch[048] Batch [3449]/[3760]	Speed: 117.303103 samples/sec	 accuracy=29.874547	 loss=3.234917	 lr=0.097920
Epoch[048] Batch [3499]/[3760]	Speed: 117.103252 samples/sec	 accuracy=29.880357	 loss=3.235255	 lr=0.097916
Epoch[048] Batch [3549]/[3760]	Speed: 117.032910 samples/sec	 accuracy=29.867077	 loss=3.236002	 lr=0.097913
Epoch[048] Batch [3599]/[3760]	Speed: 117.879782 samples/sec	 accuracy=29.851128	 loss=3.236873	 lr=0.097909
Epoch[048] Batch [3649]/[3760]	Speed: 117.265201 samples/sec	 accuracy=29.833904	 loss=3.237663	 lr=0.097905
Epoch[048] Batch [3699]/[3760]	Speed: 117.762721 samples/sec	 accuracy=29.854307	 loss=3.237647	 lr=0.097902
Epoch[048] Batch [3749]/[3760]	Speed: 120.861603 samples/sec	 accuracy=29.864167	 loss=3.236907	 lr=0.097898
Batch [0049]/[0303]: acc-top1=28.906250 acc-top5=57.875000
Batch [0099]/[0303]: acc-top1=28.296875 acc-top5=56.593750
Batch [0149]/[0303]: acc-top1=29.541667 acc-top5=56.510417
Batch [0199]/[0303]: acc-top1=31.929687 acc-top5=58.328125
Batch [0249]/[0303]: acc-top1=30.850000 acc-top5=58.137500
Batch [0299]/[0303]: acc-top1=30.192708 acc-top5=57.895833
[Epoch 048] training: accuracy=29.867852	 loss=3.236581
[Epoch 048] speed: 116 samples/sec	time cost: 2177.157781
[Epoch 048] validation: acc-top1=30.208333 acc-top5=57.987830 loss=3.244770
Epoch[049] Batch [0049]/[3759]	Speed: 88.578045 samples/sec	 accuracy=30.218750	 loss=3.228774	 lr=0.097893
Epoch[049] Batch [0099]/[3759]	Speed: 117.118221 samples/sec	 accuracy=30.000000	 loss=3.217381	 lr=0.097890
Epoch[049] Batch [0149]/[3759]	Speed: 117.747913 samples/sec	 accuracy=29.572917	 loss=3.229024	 lr=0.097886
Epoch[049] Batch [0199]/[3759]	Speed: 117.377044 samples/sec	 accuracy=30.023438	 loss=3.211369	 lr=0.097882
Epoch[049] Batch [0249]/[3759]	Speed: 117.363823 samples/sec	 accuracy=29.956250	 loss=3.208347	 lr=0.097879
Epoch[049] Batch [0299]/[3759]	Speed: 117.136808 samples/sec	 accuracy=29.869792	 loss=3.224848	 lr=0.097875
Epoch[049] Batch [0349]/[3759]	Speed: 118.008369 samples/sec	 accuracy=29.933036	 loss=3.215931	 lr=0.097871
Epoch[049] Batch [0399]/[3759]	Speed: 117.370613 samples/sec	 accuracy=29.992188	 loss=3.218702	 lr=0.097867
Epoch[049] Batch [0449]/[3759]	Speed: 117.366335 samples/sec	 accuracy=30.069444	 loss=3.217827	 lr=0.097864
Epoch[049] Batch [0499]/[3759]	Speed: 117.105911 samples/sec	 accuracy=30.050000	 loss=3.218102	 lr=0.097860
Epoch[049] Batch [0549]/[3759]	Speed: 117.715207 samples/sec	 accuracy=29.948864	 loss=3.221083	 lr=0.097856
Epoch[049] Batch [0599]/[3759]	Speed: 118.073227 samples/sec	 accuracy=29.835938	 loss=3.224271	 lr=0.097852
Epoch[049] Batch [0649]/[3759]	Speed: 117.172930 samples/sec	 accuracy=29.831731	 loss=3.223577	 lr=0.097849
Epoch[049] Batch [0699]/[3759]	Speed: 117.420621 samples/sec	 accuracy=29.915179	 loss=3.222777	 lr=0.097845
Epoch[049] Batch [0749]/[3759]	Speed: 117.808897 samples/sec	 accuracy=29.897917	 loss=3.226418	 lr=0.097841
Epoch[049] Batch [0799]/[3759]	Speed: 117.520792 samples/sec	 accuracy=29.880859	 loss=3.228108	 lr=0.097837
Epoch[049] Batch [0849]/[3759]	Speed: 117.449353 samples/sec	 accuracy=29.904412	 loss=3.224252	 lr=0.097834
Epoch[049] Batch [0899]/[3759]	Speed: 117.862905 samples/sec	 accuracy=29.873264	 loss=3.226623	 lr=0.097830
Epoch[049] Batch [0949]/[3759]	Speed: 117.117981 samples/sec	 accuracy=29.833882	 loss=3.225335	 lr=0.097826
Epoch[049] Batch [0999]/[3759]	Speed: 117.060091 samples/sec	 accuracy=29.848438	 loss=3.224171	 lr=0.097822
Epoch[049] Batch [1049]/[3759]	Speed: 117.057675 samples/sec	 accuracy=29.909226	 loss=3.221964	 lr=0.097819
Epoch[049] Batch [1099]/[3759]	Speed: 117.012788 samples/sec	 accuracy=29.867898	 loss=3.222496	 lr=0.097815
Epoch[049] Batch [1149]/[3759]	Speed: 117.077925 samples/sec	 accuracy=29.853261	 loss=3.221492	 lr=0.097811
Epoch[049] Batch [1199]/[3759]	Speed: 117.114831 samples/sec	 accuracy=29.846354	 loss=3.221191	 lr=0.097807
Epoch[049] Batch [1249]/[3759]	Speed: 117.443909 samples/sec	 accuracy=29.823750	 loss=3.221245	 lr=0.097804
Epoch[049] Batch [1299]/[3759]	Speed: 117.634315 samples/sec	 accuracy=29.812500	 loss=3.222996	 lr=0.097800
Epoch[049] Batch [1349]/[3759]	Speed: 117.630862 samples/sec	 accuracy=29.837963	 loss=3.223125	 lr=0.097796
Epoch[049] Batch [1399]/[3759]	Speed: 117.403469 samples/sec	 accuracy=29.827009	 loss=3.224509	 lr=0.097792
Epoch[049] Batch [1449]/[3759]	Speed: 116.953305 samples/sec	 accuracy=29.823276	 loss=3.224488	 lr=0.097788
Epoch[049] Batch [1499]/[3759]	Speed: 117.995743 samples/sec	 accuracy=29.862500	 loss=3.222793	 lr=0.097785
Epoch[049] Batch [1549]/[3759]	Speed: 117.477712 samples/sec	 accuracy=29.870968	 loss=3.223171	 lr=0.097781
Epoch[049] Batch [1599]/[3759]	Speed: 116.653352 samples/sec	 accuracy=29.880859	 loss=3.222801	 lr=0.097777
Epoch[049] Batch [1649]/[3759]	Speed: 117.513902 samples/sec	 accuracy=29.863636	 loss=3.222794	 lr=0.097773
Epoch[049] Batch [1699]/[3759]	Speed: 118.142507 samples/sec	 accuracy=29.876838	 loss=3.223772	 lr=0.097769
Epoch[049] Batch [1749]/[3759]	Speed: 117.176982 samples/sec	 accuracy=29.886607	 loss=3.222897	 lr=0.097766
Epoch[049] Batch [1799]/[3759]	Speed: 117.350721 samples/sec	 accuracy=29.845486	 loss=3.223228	 lr=0.097762
Epoch[049] Batch [1849]/[3759]	Speed: 118.020888 samples/sec	 accuracy=29.856419	 loss=3.224355	 lr=0.097758
Epoch[049] Batch [1899]/[3759]	Speed: 117.722673 samples/sec	 accuracy=29.865132	 loss=3.224907	 lr=0.097754
Epoch[049] Batch [1949]/[3759]	Speed: 117.588289 samples/sec	 accuracy=29.865385	 loss=3.224396	 lr=0.097750
Epoch[049] Batch [1999]/[3759]	Speed: 117.536079 samples/sec	 accuracy=29.847656	 loss=3.226300	 lr=0.097747
Epoch[049] Batch [2049]/[3759]	Speed: 117.275917 samples/sec	 accuracy=29.823933	 loss=3.226858	 lr=0.097743
Epoch[049] Batch [2099]/[3759]	Speed: 116.680869 samples/sec	 accuracy=29.837798	 loss=3.227632	 lr=0.097739
Epoch[049] Batch [2149]/[3759]	Speed: 117.875856 samples/sec	 accuracy=29.845203	 loss=3.227664	 lr=0.097735
Epoch[049] Batch [2199]/[3759]	Speed: 118.152658 samples/sec	 accuracy=29.857955	 loss=3.227631	 lr=0.097731
Epoch[049] Batch [2249]/[3759]	Speed: 117.827895 samples/sec	 accuracy=29.825694	 loss=3.228776	 lr=0.097727
Epoch[049] Batch [2299]/[3759]	Speed: 117.848203 samples/sec	 accuracy=29.810462	 loss=3.229808	 lr=0.097723
Epoch[049] Batch [2349]/[3759]	Speed: 117.857932 samples/sec	 accuracy=29.773936	 loss=3.231197	 lr=0.097720
Epoch[049] Batch [2399]/[3759]	Speed: 117.666947 samples/sec	 accuracy=29.811849	 loss=3.230077	 lr=0.097716
Epoch[049] Batch [2449]/[3759]	Speed: 117.443126 samples/sec	 accuracy=29.837372	 loss=3.230764	 lr=0.097712
Epoch[049] Batch [2499]/[3759]	Speed: 118.093144 samples/sec	 accuracy=29.848125	 loss=3.231547	 lr=0.097708
Epoch[049] Batch [2549]/[3759]	Speed: 117.319327 samples/sec	 accuracy=29.824755	 loss=3.232252	 lr=0.097704
Epoch[049] Batch [2599]/[3759]	Speed: 118.257861 samples/sec	 accuracy=29.851562	 loss=3.230951	 lr=0.097700
Epoch[049] Batch [2649]/[3759]	Speed: 117.799988 samples/sec	 accuracy=29.841392	 loss=3.231513	 lr=0.097696
Epoch[049] Batch [2699]/[3759]	Speed: 117.161172 samples/sec	 accuracy=29.843750	 loss=3.230041	 lr=0.097693
Epoch[049] Batch [2749]/[3759]	Speed: 117.066491 samples/sec	 accuracy=29.848295	 loss=3.229865	 lr=0.097689
Epoch[049] Batch [2799]/[3759]	Speed: 116.566639 samples/sec	 accuracy=29.861049	 loss=3.228730	 lr=0.097685
Epoch[049] Batch [2849]/[3759]	Speed: 117.278565 samples/sec	 accuracy=29.844846	 loss=3.229924	 lr=0.097681
Epoch[049] Batch [2899]/[3759]	Speed: 116.738554 samples/sec	 accuracy=29.848060	 loss=3.230272	 lr=0.097677
Epoch[049] Batch [2949]/[3759]	Speed: 118.041499 samples/sec	 accuracy=29.838983	 loss=3.230250	 lr=0.097673
Epoch[049] Batch [2999]/[3759]	Speed: 116.988820 samples/sec	 accuracy=29.821354	 loss=3.231758	 lr=0.097669
Epoch[049] Batch [3049]/[3759]	Speed: 116.984891 samples/sec	 accuracy=29.832992	 loss=3.230519	 lr=0.097665
Epoch[049] Batch [3099]/[3759]	Speed: 117.087764 samples/sec	 accuracy=29.837198	 loss=3.230724	 lr=0.097662
Epoch[049] Batch [3149]/[3759]	Speed: 117.542489 samples/sec	 accuracy=29.854167	 loss=3.230511	 lr=0.097658
Epoch[049] Batch [3199]/[3759]	Speed: 117.906978 samples/sec	 accuracy=29.838867	 loss=3.230846	 lr=0.097654
Epoch[049] Batch [3249]/[3759]	Speed: 117.308985 samples/sec	 accuracy=29.839423	 loss=3.230713	 lr=0.097650
Epoch[049] Batch [3299]/[3759]	Speed: 116.846136 samples/sec	 accuracy=29.857008	 loss=3.230384	 lr=0.097646
Epoch[049] Batch [3349]/[3759]	Speed: 117.427526 samples/sec	 accuracy=29.876866	 loss=3.229491	 lr=0.097642
Epoch[049] Batch [3399]/[3759]	Speed: 117.478653 samples/sec	 accuracy=29.864890	 loss=3.229742	 lr=0.097638
Epoch[049] Batch [3449]/[3759]	Speed: 116.837367 samples/sec	 accuracy=29.842844	 loss=3.231842	 lr=0.097634
Epoch[049] Batch [3499]/[3759]	Speed: 116.602563 samples/sec	 accuracy=29.826339	 loss=3.232515	 lr=0.097630
Epoch[049] Batch [3549]/[3759]	Speed: 116.731011 samples/sec	 accuracy=29.834067	 loss=3.232223	 lr=0.097626
Epoch[049] Batch [3599]/[3759]	Speed: 117.914188 samples/sec	 accuracy=29.829427	 loss=3.232231	 lr=0.097622
Epoch[049] Batch [3649]/[3759]	Speed: 117.873247 samples/sec	 accuracy=29.843322	 loss=3.231842	 lr=0.097618
Epoch[049] Batch [3699]/[3759]	Speed: 117.665503 samples/sec	 accuracy=29.837416	 loss=3.232518	 lr=0.097615
Epoch[049] Batch [3749]/[3759]	Speed: 120.618077 samples/sec	 accuracy=29.851250	 loss=3.232483	 lr=0.097611
Batch [0049]/[0303]: acc-top1=28.468750 acc-top5=55.875000
Batch [0099]/[0303]: acc-top1=26.921875 acc-top5=53.968750
Batch [0149]/[0303]: acc-top1=28.583333 acc-top5=55.364583
Batch [0199]/[0303]: acc-top1=31.085938 acc-top5=57.281250
Batch [0249]/[0303]: acc-top1=30.562500 acc-top5=57.387500
Batch [0299]/[0303]: acc-top1=29.494792 acc-top5=56.682292
[Epoch 049] training: accuracy=29.852105	 loss=3.232430
[Epoch 049] speed: 116 samples/sec	time cost: 2176.519056
[Epoch 049] validation: acc-top1=29.429662 acc-top5=56.678012 loss=3.296482
Epoch[050] Batch [0049]/[3760]	Speed: 86.783639 samples/sec	 accuracy=29.281250	 loss=3.236579	 lr=0.097606
Epoch[050] Batch [0099]/[3760]	Speed: 116.869107 samples/sec	 accuracy=29.750000	 loss=3.217723	 lr=0.097602
Epoch[050] Batch [0149]/[3760]	Speed: 117.865795 samples/sec	 accuracy=30.031250	 loss=3.204060	 lr=0.097598
Epoch[050] Batch [0199]/[3760]	Speed: 117.187145 samples/sec	 accuracy=30.187500	 loss=3.193575	 lr=0.097594
Epoch[050] Batch [0249]/[3760]	Speed: 117.766890 samples/sec	 accuracy=30.075000	 loss=3.201203	 lr=0.097590
Epoch[050] Batch [0299]/[3760]	Speed: 117.597150 samples/sec	 accuracy=30.166667	 loss=3.198946	 lr=0.097586
Epoch[050] Batch [0349]/[3760]	Speed: 116.742809 samples/sec	 accuracy=29.973214	 loss=3.208108	 lr=0.097582
Epoch[050] Batch [0399]/[3760]	Speed: 117.019714 samples/sec	 accuracy=29.953125	 loss=3.207283	 lr=0.097578
Epoch[050] Batch [0449]/[3760]	Speed: 117.152613 samples/sec	 accuracy=29.927083	 loss=3.214397	 lr=0.097574
Epoch[050] Batch [0499]/[3760]	Speed: 117.116304 samples/sec	 accuracy=29.837500	 loss=3.221171	 lr=0.097570
Epoch[050] Batch [0549]/[3760]	Speed: 117.769313 samples/sec	 accuracy=29.838068	 loss=3.223421	 lr=0.097566
Epoch[050] Batch [0599]/[3760]	Speed: 117.467905 samples/sec	 accuracy=29.854167	 loss=3.226921	 lr=0.097562
Epoch[050] Batch [0649]/[3760]	Speed: 116.519334 samples/sec	 accuracy=29.906250	 loss=3.226614	 lr=0.097558
Epoch[050] Batch [0699]/[3760]	Speed: 116.973369 samples/sec	 accuracy=29.888393	 loss=3.226796	 lr=0.097554
Epoch[050] Batch [0749]/[3760]	Speed: 117.499418 samples/sec	 accuracy=29.927083	 loss=3.229827	 lr=0.097550
Epoch[050] Batch [0799]/[3760]	Speed: 116.936442 samples/sec	 accuracy=29.960938	 loss=3.226029	 lr=0.097546
Epoch[050] Batch [0849]/[3760]	Speed: 117.339428 samples/sec	 accuracy=29.972426	 loss=3.224820	 lr=0.097542
Epoch[050] Batch [0899]/[3760]	Speed: 117.441588 samples/sec	 accuracy=30.012153	 loss=3.226217	 lr=0.097538
Epoch[050] Batch [0949]/[3760]	Speed: 116.786203 samples/sec	 accuracy=30.044408	 loss=3.224800	 lr=0.097534
Epoch[050] Batch [0999]/[3760]	Speed: 116.201166 samples/sec	 accuracy=30.156250	 loss=3.221395	 lr=0.097530
Epoch[050] Batch [1049]/[3760]	Speed: 117.658859 samples/sec	 accuracy=30.177083	 loss=3.221054	 lr=0.097526
Epoch[050] Batch [1099]/[3760]	Speed: 117.340074 samples/sec	 accuracy=30.180398	 loss=3.219326	 lr=0.097522
Epoch[050] Batch [1149]/[3760]	Speed: 118.291031 samples/sec	 accuracy=30.078804	 loss=3.224004	 lr=0.097518
Epoch[050] Batch [1199]/[3760]	Speed: 117.296628 samples/sec	 accuracy=30.065104	 loss=3.224189	 lr=0.097514
Epoch[050] Batch [1249]/[3760]	Speed: 117.475604 samples/sec	 accuracy=30.101250	 loss=3.221624	 lr=0.097510
Epoch[050] Batch [1299]/[3760]	Speed: 117.465632 samples/sec	 accuracy=30.093750	 loss=3.222447	 lr=0.097506
Epoch[050] Batch [1349]/[3760]	Speed: 118.222402 samples/sec	 accuracy=30.108796	 loss=3.221298	 lr=0.097502
Epoch[050] Batch [1399]/[3760]	Speed: 118.072875 samples/sec	 accuracy=30.095982	 loss=3.221999	 lr=0.097498
Epoch[050] Batch [1449]/[3760]	Speed: 117.104547 samples/sec	 accuracy=30.089440	 loss=3.222198	 lr=0.097494
Epoch[050] Batch [1499]/[3760]	Speed: 118.444837 samples/sec	 accuracy=30.082292	 loss=3.221965	 lr=0.097490
Epoch[050] Batch [1549]/[3760]	Speed: 117.940774 samples/sec	 accuracy=30.058468	 loss=3.223704	 lr=0.097486
Epoch[050] Batch [1599]/[3760]	Speed: 117.989482 samples/sec	 accuracy=30.012695	 loss=3.225684	 lr=0.097482
Epoch[050] Batch [1649]/[3760]	Speed: 117.237722 samples/sec	 accuracy=30.033144	 loss=3.224210	 lr=0.097478
Epoch[050] Batch [1699]/[3760]	Speed: 117.624404 samples/sec	 accuracy=30.009191	 loss=3.225555	 lr=0.097474
Epoch[050] Batch [1749]/[3760]	Speed: 117.774074 samples/sec	 accuracy=30.046429	 loss=3.224651	 lr=0.097470
Epoch[050] Batch [1799]/[3760]	Speed: 117.892200 samples/sec	 accuracy=30.028646	 loss=3.225144	 lr=0.097466
Epoch[050] Batch [1849]/[3760]	Speed: 117.462729 samples/sec	 accuracy=30.011824	 loss=3.225196	 lr=0.097462
Epoch[050] Batch [1899]/[3760]	Speed: 117.404128 samples/sec	 accuracy=30.013158	 loss=3.224901	 lr=0.097458
Epoch[050] Batch [1949]/[3760]	Speed: 117.968686 samples/sec	 accuracy=29.995994	 loss=3.225623	 lr=0.097454
Epoch[050] Batch [1999]/[3760]	Speed: 117.312538 samples/sec	 accuracy=29.989063	 loss=3.226780	 lr=0.097450
Epoch[050] Batch [2049]/[3760]	Speed: 117.455311 samples/sec	 accuracy=29.960366	 loss=3.227310	 lr=0.097446
Epoch[050] Batch [2099]/[3760]	Speed: 117.728537 samples/sec	 accuracy=29.989583	 loss=3.226591	 lr=0.097442
Epoch[050] Batch [2149]/[3760]	Speed: 116.614588 samples/sec	 accuracy=29.965116	 loss=3.229115	 lr=0.097438
Epoch[050] Batch [2199]/[3760]	Speed: 117.526597 samples/sec	 accuracy=29.969460	 loss=3.229131	 lr=0.097433
Epoch[050] Batch [2249]/[3760]	Speed: 117.038182 samples/sec	 accuracy=29.965972	 loss=3.229256	 lr=0.097429
Epoch[050] Batch [2299]/[3760]	Speed: 116.930955 samples/sec	 accuracy=29.966033	 loss=3.230240	 lr=0.097425
Epoch[050] Batch [2349]/[3760]	Speed: 117.155470 samples/sec	 accuracy=29.950133	 loss=3.230796	 lr=0.097421
Epoch[050] Batch [2399]/[3760]	Speed: 116.886408 samples/sec	 accuracy=29.975260	 loss=3.230903	 lr=0.097417
Epoch[050] Batch [2449]/[3760]	Speed: 117.942774 samples/sec	 accuracy=29.965561	 loss=3.230368	 lr=0.097413
Epoch[050] Batch [2499]/[3760]	Speed: 116.574576 samples/sec	 accuracy=29.975000	 loss=3.229067	 lr=0.097409
Epoch[050] Batch [2549]/[3760]	Speed: 116.283193 samples/sec	 accuracy=29.981005	 loss=3.228457	 lr=0.097405
Epoch[050] Batch [2599]/[3760]	Speed: 117.250724 samples/sec	 accuracy=29.962740	 loss=3.229014	 lr=0.097401
Epoch[050] Batch [2649]/[3760]	Speed: 117.141437 samples/sec	 accuracy=29.975825	 loss=3.228067	 lr=0.097397
Epoch[050] Batch [2699]/[3760]	Speed: 117.657473 samples/sec	 accuracy=29.962963	 loss=3.228508	 lr=0.097392
Epoch[050] Batch [2749]/[3760]	Speed: 117.393920 samples/sec	 accuracy=29.972159	 loss=3.228176	 lr=0.097388
Epoch[050] Batch [2799]/[3760]	Speed: 117.616643 samples/sec	 accuracy=29.982143	 loss=3.227328	 lr=0.097384
Epoch[050] Batch [2849]/[3760]	Speed: 117.747875 samples/sec	 accuracy=29.969846	 loss=3.228642	 lr=0.097380
Epoch[050] Batch [2899]/[3760]	Speed: 117.366282 samples/sec	 accuracy=29.937500	 loss=3.229541	 lr=0.097376
Epoch[050] Batch [2949]/[3760]	Speed: 117.989424 samples/sec	 accuracy=29.926907	 loss=3.229921	 lr=0.097372
Epoch[050] Batch [2999]/[3760]	Speed: 117.357926 samples/sec	 accuracy=29.915104	 loss=3.230295	 lr=0.097368
Epoch[050] Batch [3049]/[3760]	Speed: 117.889468 samples/sec	 accuracy=29.912398	 loss=3.230761	 lr=0.097364
Epoch[050] Batch [3099]/[3760]	Speed: 117.408721 samples/sec	 accuracy=29.924899	 loss=3.229700	 lr=0.097360
Epoch[050] Batch [3149]/[3760]	Speed: 117.262871 samples/sec	 accuracy=29.918651	 loss=3.229621	 lr=0.097355
Epoch[050] Batch [3199]/[3760]	Speed: 117.464028 samples/sec	 accuracy=29.898438	 loss=3.230440	 lr=0.097351
Epoch[050] Batch [3249]/[3760]	Speed: 117.473297 samples/sec	 accuracy=29.905288	 loss=3.230671	 lr=0.097347
Epoch[050] Batch [3299]/[3760]	Speed: 117.613160 samples/sec	 accuracy=29.900568	 loss=3.230650	 lr=0.097343
Epoch[050] Batch [3349]/[3760]	Speed: 117.489916 samples/sec	 accuracy=29.885261	 loss=3.230835	 lr=0.097339
Epoch[050] Batch [3399]/[3760]	Speed: 118.263370 samples/sec	 accuracy=29.883272	 loss=3.230048	 lr=0.097335
Epoch[050] Batch [3449]/[3760]	Speed: 117.171608 samples/sec	 accuracy=29.881341	 loss=3.230481	 lr=0.097330
Epoch[050] Batch [3499]/[3760]	Speed: 117.163397 samples/sec	 accuracy=29.900000	 loss=3.230098	 lr=0.097326
Epoch[050] Batch [3549]/[3760]	Speed: 117.232690 samples/sec	 accuracy=29.909331	 loss=3.229847	 lr=0.097322
Epoch[050] Batch [3599]/[3760]	Speed: 117.481369 samples/sec	 accuracy=29.917101	 loss=3.229031	 lr=0.097318
Epoch[050] Batch [3649]/[3760]	Speed: 118.089121 samples/sec	 accuracy=29.889127	 loss=3.229380	 lr=0.097314
Epoch[050] Batch [3699]/[3760]	Speed: 117.731503 samples/sec	 accuracy=29.904983	 loss=3.228991	 lr=0.097310
Epoch[050] Batch [3749]/[3760]	Speed: 120.796480 samples/sec	 accuracy=29.892083	 loss=3.229537	 lr=0.097305
Batch [0049]/[0303]: acc-top1=30.000000 acc-top5=56.250000
Batch [0099]/[0303]: acc-top1=27.828125 acc-top5=53.531250
Batch [0149]/[0303]: acc-top1=29.510417 acc-top5=55.614583
Batch [0199]/[0303]: acc-top1=31.781250 acc-top5=57.593750
Batch [0249]/[0303]: acc-top1=30.850000 acc-top5=57.387500
Batch [0299]/[0303]: acc-top1=30.661458 acc-top5=57.812500
[Epoch 050] training: accuracy=29.888630	 loss=3.229551
[Epoch 050] speed: 116 samples/sec	time cost: 2177.919614
[Epoch 050] validation: acc-top1=30.595091 acc-top5=57.766089 loss=3.363049
Epoch[051] Batch [0049]/[3760]	Speed: 87.741844 samples/sec	 accuracy=29.812500	 loss=3.238343	 lr=0.097300
Epoch[051] Batch [0099]/[3760]	Speed: 118.194783 samples/sec	 accuracy=30.781250	 loss=3.194965	 lr=0.097296
Epoch[051] Batch [0149]/[3760]	Speed: 117.796615 samples/sec	 accuracy=30.510417	 loss=3.201353	 lr=0.097292
Epoch[051] Batch [0199]/[3760]	Speed: 117.016701 samples/sec	 accuracy=30.445312	 loss=3.192475	 lr=0.097288
Epoch[051] Batch [0249]/[3760]	Speed: 117.184435 samples/sec	 accuracy=30.550000	 loss=3.196494	 lr=0.097284
Epoch[051] Batch [0299]/[3760]	Speed: 118.385861 samples/sec	 accuracy=30.546875	 loss=3.196445	 lr=0.097280
Epoch[051] Batch [0349]/[3760]	Speed: 117.216849 samples/sec	 accuracy=30.433036	 loss=3.199119	 lr=0.097275
Epoch[051] Batch [0399]/[3760]	Speed: 117.242737 samples/sec	 accuracy=30.523438	 loss=3.190423	 lr=0.097271
Epoch[051] Batch [0449]/[3760]	Speed: 117.499538 samples/sec	 accuracy=30.534722	 loss=3.183025	 lr=0.097267
Epoch[051] Batch [0499]/[3760]	Speed: 117.288050 samples/sec	 accuracy=30.415625	 loss=3.191744	 lr=0.097263
Epoch[051] Batch [0549]/[3760]	Speed: 118.037498 samples/sec	 accuracy=30.397727	 loss=3.189821	 lr=0.097259
Epoch[051] Batch [0599]/[3760]	Speed: 117.107314 samples/sec	 accuracy=30.375000	 loss=3.191386	 lr=0.097254
Epoch[051] Batch [0649]/[3760]	Speed: 116.975312 samples/sec	 accuracy=30.358173	 loss=3.193173	 lr=0.097250
Epoch[051] Batch [0699]/[3760]	Speed: 117.759688 samples/sec	 accuracy=30.517857	 loss=3.186933	 lr=0.097246
Epoch[051] Batch [0749]/[3760]	Speed: 116.885459 samples/sec	 accuracy=30.477083	 loss=3.185806	 lr=0.097242
Epoch[051] Batch [0799]/[3760]	Speed: 116.929798 samples/sec	 accuracy=30.500000	 loss=3.183632	 lr=0.097237
Epoch[051] Batch [0849]/[3760]	Speed: 117.743937 samples/sec	 accuracy=30.476103	 loss=3.185934	 lr=0.097233
Epoch[051] Batch [0899]/[3760]	Speed: 117.273145 samples/sec	 accuracy=30.473958	 loss=3.184380	 lr=0.097229
Epoch[051] Batch [0949]/[3760]	Speed: 116.837757 samples/sec	 accuracy=30.452303	 loss=3.186582	 lr=0.097225
Epoch[051] Batch [0999]/[3760]	Speed: 117.990963 samples/sec	 accuracy=30.407812	 loss=3.190270	 lr=0.097220
Epoch[051] Batch [1049]/[3760]	Speed: 117.428049 samples/sec	 accuracy=30.366071	 loss=3.192926	 lr=0.097216
Epoch[051] Batch [1099]/[3760]	Speed: 117.870456 samples/sec	 accuracy=30.305398	 loss=3.196064	 lr=0.097212
Epoch[051] Batch [1149]/[3760]	Speed: 117.798695 samples/sec	 accuracy=30.237772	 loss=3.199176	 lr=0.097208
Epoch[051] Batch [1199]/[3760]	Speed: 117.241108 samples/sec	 accuracy=30.187500	 loss=3.201269	 lr=0.097203
Epoch[051] Batch [1249]/[3760]	Speed: 117.374631 samples/sec	 accuracy=30.171250	 loss=3.202264	 lr=0.097199
Epoch[051] Batch [1299]/[3760]	Speed: 117.094388 samples/sec	 accuracy=30.114183	 loss=3.203881	 lr=0.097195
Epoch[051] Batch [1349]/[3760]	Speed: 117.303746 samples/sec	 accuracy=30.130787	 loss=3.203963	 lr=0.097191
Epoch[051] Batch [1399]/[3760]	Speed: 117.373619 samples/sec	 accuracy=30.095982	 loss=3.206385	 lr=0.097186
Epoch[051] Batch [1449]/[3760]	Speed: 116.504335 samples/sec	 accuracy=30.099138	 loss=3.207246	 lr=0.097182
Epoch[051] Batch [1499]/[3760]	Speed: 117.255763 samples/sec	 accuracy=30.090625	 loss=3.207820	 lr=0.097178
Epoch[051] Batch [1549]/[3760]	Speed: 116.897891 samples/sec	 accuracy=30.095766	 loss=3.208406	 lr=0.097174
Epoch[051] Batch [1599]/[3760]	Speed: 117.074636 samples/sec	 accuracy=30.123047	 loss=3.208373	 lr=0.097169
Epoch[051] Batch [1649]/[3760]	Speed: 117.219795 samples/sec	 accuracy=30.133523	 loss=3.208347	 lr=0.097165
Epoch[051] Batch [1699]/[3760]	Speed: 117.808118 samples/sec	 accuracy=30.165441	 loss=3.207243	 lr=0.097161
Epoch[051] Batch [1749]/[3760]	Speed: 117.819252 samples/sec	 accuracy=30.171429	 loss=3.207994	 lr=0.097157
Epoch[051] Batch [1799]/[3760]	Speed: 117.545969 samples/sec	 accuracy=30.171875	 loss=3.209292	 lr=0.097152
Epoch[051] Batch [1849]/[3760]	Speed: 117.870488 samples/sec	 accuracy=30.179054	 loss=3.209391	 lr=0.097148
Epoch[051] Batch [1899]/[3760]	Speed: 117.533086 samples/sec	 accuracy=30.208059	 loss=3.208558	 lr=0.097144
Epoch[051] Batch [1949]/[3760]	Speed: 116.824060 samples/sec	 accuracy=30.211538	 loss=3.208426	 lr=0.097139
Epoch[051] Batch [1999]/[3760]	Speed: 117.546860 samples/sec	 accuracy=30.229688	 loss=3.208936	 lr=0.097135
Epoch[051] Batch [2049]/[3760]	Speed: 117.652389 samples/sec	 accuracy=30.202744	 loss=3.212562	 lr=0.097131
Epoch[051] Batch [2099]/[3760]	Speed: 116.701123 samples/sec	 accuracy=30.216518	 loss=3.212154	 lr=0.097126
Epoch[051] Batch [2149]/[3760]	Speed: 117.271684 samples/sec	 accuracy=30.202035	 loss=3.214198	 lr=0.097122
Epoch[051] Batch [2199]/[3760]	Speed: 116.878062 samples/sec	 accuracy=30.206676	 loss=3.214654	 lr=0.097118
Epoch[051] Batch [2249]/[3760]	Speed: 117.159936 samples/sec	 accuracy=30.225000	 loss=3.213140	 lr=0.097113
Epoch[051] Batch [2299]/[3760]	Speed: 117.274194 samples/sec	 accuracy=30.201766	 loss=3.214410	 lr=0.097109
Epoch[051] Batch [2349]/[3760]	Speed: 116.752242 samples/sec	 accuracy=30.218085	 loss=3.213826	 lr=0.097105
Epoch[051] Batch [2399]/[3760]	Speed: 116.691636 samples/sec	 accuracy=30.226562	 loss=3.213855	 lr=0.097101
Epoch[051] Batch [2449]/[3760]	Speed: 116.776385 samples/sec	 accuracy=30.205995	 loss=3.214386	 lr=0.097096
Epoch[051] Batch [2499]/[3760]	Speed: 117.244630 samples/sec	 accuracy=30.189375	 loss=3.216249	 lr=0.097092
Epoch[051] Batch [2549]/[3760]	Speed: 116.387924 samples/sec	 accuracy=30.183211	 loss=3.217077	 lr=0.097088
Epoch[051] Batch [2599]/[3760]	Speed: 117.396579 samples/sec	 accuracy=30.174279	 loss=3.217622	 lr=0.097083
Epoch[051] Batch [2649]/[3760]	Speed: 117.103042 samples/sec	 accuracy=30.162146	 loss=3.218648	 lr=0.097079
Epoch[051] Batch [2699]/[3760]	Speed: 117.072130 samples/sec	 accuracy=30.156829	 loss=3.217923	 lr=0.097074
Epoch[051] Batch [2749]/[3760]	Speed: 117.349122 samples/sec	 accuracy=30.166477	 loss=3.217142	 lr=0.097070
Epoch[051] Batch [2799]/[3760]	Speed: 117.210670 samples/sec	 accuracy=30.161830	 loss=3.217271	 lr=0.097066
Epoch[051] Batch [2849]/[3760]	Speed: 117.758582 samples/sec	 accuracy=30.151864	 loss=3.217522	 lr=0.097061
Epoch[051] Batch [2899]/[3760]	Speed: 117.094491 samples/sec	 accuracy=30.143319	 loss=3.217781	 lr=0.097057
Epoch[051] Batch [2949]/[3760]	Speed: 117.355885 samples/sec	 accuracy=30.161017	 loss=3.216587	 lr=0.097053
Epoch[051] Batch [2999]/[3760]	Speed: 116.910003 samples/sec	 accuracy=30.164583	 loss=3.216831	 lr=0.097048
Epoch[051] Batch [3049]/[3760]	Speed: 117.284606 samples/sec	 accuracy=30.157275	 loss=3.217517	 lr=0.097044
Epoch[051] Batch [3099]/[3760]	Speed: 117.616556 samples/sec	 accuracy=30.150202	 loss=3.217786	 lr=0.097040
Epoch[051] Batch [3149]/[3760]	Speed: 118.111359 samples/sec	 accuracy=30.163690	 loss=3.218194	 lr=0.097035
Epoch[051] Batch [3199]/[3760]	Speed: 117.228124 samples/sec	 accuracy=30.163086	 loss=3.218394	 lr=0.097031
Epoch[051] Batch [3249]/[3760]	Speed: 117.019725 samples/sec	 accuracy=30.137981	 loss=3.219376	 lr=0.097026
Epoch[051] Batch [3299]/[3760]	Speed: 117.445511 samples/sec	 accuracy=30.144413	 loss=3.219248	 lr=0.097022
Epoch[051] Batch [3349]/[3760]	Speed: 116.716867 samples/sec	 accuracy=30.146455	 loss=3.220123	 lr=0.097018
Epoch[051] Batch [3399]/[3760]	Speed: 117.145366 samples/sec	 accuracy=30.153033	 loss=3.219483	 lr=0.097013
Epoch[051] Batch [3449]/[3760]	Speed: 117.452606 samples/sec	 accuracy=30.149457	 loss=3.220203	 lr=0.097009
Epoch[051] Batch [3499]/[3760]	Speed: 117.580308 samples/sec	 accuracy=30.149107	 loss=3.220489	 lr=0.097005
Epoch[051] Batch [3549]/[3760]	Speed: 118.067093 samples/sec	 accuracy=30.128961	 loss=3.220921	 lr=0.097000
Epoch[051] Batch [3599]/[3760]	Speed: 117.458104 samples/sec	 accuracy=30.143229	 loss=3.220484	 lr=0.096996
Epoch[051] Batch [3649]/[3760]	Speed: 117.780175 samples/sec	 accuracy=30.147688	 loss=3.220420	 lr=0.096991
Epoch[051] Batch [3699]/[3760]	Speed: 117.247840 samples/sec	 accuracy=30.145270	 loss=3.220337	 lr=0.096987
Epoch[051] Batch [3749]/[3760]	Speed: 120.079976 samples/sec	 accuracy=30.147083	 loss=3.220094	 lr=0.096983
Batch [0049]/[0303]: acc-top1=30.218750 acc-top5=55.125000
Batch [0099]/[0303]: acc-top1=28.062500 acc-top5=53.828125
Batch [0149]/[0303]: acc-top1=27.937500 acc-top5=54.270833
Batch [0199]/[0303]: acc-top1=31.125000 acc-top5=57.085937
Batch [0249]/[0303]: acc-top1=29.950000 acc-top5=56.150000
Batch [0299]/[0303]: acc-top1=29.348958 acc-top5=56.296875
[Epoch 051] training: accuracy=30.144199	 loss=3.220114
[Epoch 051] speed: 116 samples/sec	time cost: 2179.536654
[Epoch 051] validation: acc-top1=29.264645 acc-top5=56.265470 loss=3.370228
Epoch[052] Batch [0049]/[3759]	Speed: 87.757378 samples/sec	 accuracy=30.968750	 loss=3.208677	 lr=0.096977
Epoch[052] Batch [0099]/[3759]	Speed: 117.805265 samples/sec	 accuracy=30.906250	 loss=3.208585	 lr=0.096973
Epoch[052] Batch [0149]/[3759]	Speed: 117.178322 samples/sec	 accuracy=30.822917	 loss=3.181309	 lr=0.096968
Epoch[052] Batch [0199]/[3759]	Speed: 117.135324 samples/sec	 accuracy=30.515625	 loss=3.194022	 lr=0.096964
Epoch[052] Batch [0249]/[3759]	Speed: 117.731095 samples/sec	 accuracy=30.393750	 loss=3.187150	 lr=0.096960
Epoch[052] Batch [0299]/[3759]	Speed: 116.424020 samples/sec	 accuracy=30.270833	 loss=3.191473	 lr=0.096955
Epoch[052] Batch [0349]/[3759]	Speed: 117.998611 samples/sec	 accuracy=30.142857	 loss=3.199502	 lr=0.096951
Epoch[052] Batch [0399]/[3759]	Speed: 117.256032 samples/sec	 accuracy=30.207031	 loss=3.195856	 lr=0.096946
Epoch[052] Batch [0449]/[3759]	Speed: 117.611807 samples/sec	 accuracy=30.284722	 loss=3.197119	 lr=0.096942
Epoch[052] Batch [0499]/[3759]	Speed: 117.134826 samples/sec	 accuracy=30.340625	 loss=3.192392	 lr=0.096937
Epoch[052] Batch [0549]/[3759]	Speed: 116.691955 samples/sec	 accuracy=30.343750	 loss=3.192403	 lr=0.096933
Epoch[052] Batch [0599]/[3759]	Speed: 116.394647 samples/sec	 accuracy=30.434896	 loss=3.188906	 lr=0.096928
Epoch[052] Batch [0649]/[3759]	Speed: 116.843085 samples/sec	 accuracy=30.367788	 loss=3.189223	 lr=0.096924
Epoch[052] Batch [0699]/[3759]	Speed: 117.513470 samples/sec	 accuracy=30.488839	 loss=3.187207	 lr=0.096920
Epoch[052] Batch [0749]/[3759]	Speed: 116.967344 samples/sec	 accuracy=30.456250	 loss=3.188309	 lr=0.096915
Epoch[052] Batch [0799]/[3759]	Speed: 117.138984 samples/sec	 accuracy=30.435547	 loss=3.191542	 lr=0.096911
Epoch[052] Batch [0849]/[3759]	Speed: 117.120431 samples/sec	 accuracy=30.400735	 loss=3.197188	 lr=0.096906
Epoch[052] Batch [0899]/[3759]	Speed: 117.572491 samples/sec	 accuracy=30.416667	 loss=3.194856	 lr=0.096902
Epoch[052] Batch [0949]/[3759]	Speed: 118.071296 samples/sec	 accuracy=30.333882	 loss=3.199012	 lr=0.096897
Epoch[052] Batch [0999]/[3759]	Speed: 117.110926 samples/sec	 accuracy=30.331250	 loss=3.199049	 lr=0.096893
Epoch[052] Batch [1049]/[3759]	Speed: 118.074939 samples/sec	 accuracy=30.334821	 loss=3.197725	 lr=0.096888
Epoch[052] Batch [1099]/[3759]	Speed: 117.491012 samples/sec	 accuracy=30.407670	 loss=3.197095	 lr=0.096884
Epoch[052] Batch [1149]/[3759]	Speed: 116.970935 samples/sec	 accuracy=30.335598	 loss=3.200168	 lr=0.096879
Epoch[052] Batch [1199]/[3759]	Speed: 117.349679 samples/sec	 accuracy=30.278646	 loss=3.202200	 lr=0.096875
Epoch[052] Batch [1249]/[3759]	Speed: 117.812019 samples/sec	 accuracy=30.223750	 loss=3.205952	 lr=0.096870
Epoch[052] Batch [1299]/[3759]	Speed: 117.812734 samples/sec	 accuracy=30.170673	 loss=3.208648	 lr=0.096866
Epoch[052] Batch [1349]/[3759]	Speed: 117.516525 samples/sec	 accuracy=30.151620	 loss=3.210152	 lr=0.096861
Epoch[052] Batch [1399]/[3759]	Speed: 117.411877 samples/sec	 accuracy=30.145089	 loss=3.210351	 lr=0.096857
Epoch[052] Batch [1449]/[3759]	Speed: 116.787888 samples/sec	 accuracy=30.160560	 loss=3.208340	 lr=0.096852
Epoch[052] Batch [1499]/[3759]	Speed: 117.147670 samples/sec	 accuracy=30.162500	 loss=3.209737	 lr=0.096848
Epoch[052] Batch [1549]/[3759]	Speed: 117.760721 samples/sec	 accuracy=30.146169	 loss=3.209637	 lr=0.096843
Epoch[052] Batch [1599]/[3759]	Speed: 117.570954 samples/sec	 accuracy=30.144531	 loss=3.210276	 lr=0.096839
Epoch[052] Batch [1649]/[3759]	Speed: 116.409403 samples/sec	 accuracy=30.131629	 loss=3.211713	 lr=0.096834
Epoch[052] Batch [1699]/[3759]	Speed: 117.382873 samples/sec	 accuracy=30.131434	 loss=3.211182	 lr=0.096830
Epoch[052] Batch [1749]/[3759]	Speed: 117.850874 samples/sec	 accuracy=30.147321	 loss=3.211188	 lr=0.096825
Epoch[052] Batch [1799]/[3759]	Speed: 117.640674 samples/sec	 accuracy=30.165799	 loss=3.209245	 lr=0.096821
Epoch[052] Batch [1849]/[3759]	Speed: 117.375261 samples/sec	 accuracy=30.182432	 loss=3.207884	 lr=0.096816
Epoch[052] Batch [1899]/[3759]	Speed: 116.873599 samples/sec	 accuracy=30.202303	 loss=3.207489	 lr=0.096812
Epoch[052] Batch [1949]/[3759]	Speed: 116.563964 samples/sec	 accuracy=30.129808	 loss=3.209449	 lr=0.096807
Epoch[052] Batch [1999]/[3759]	Speed: 117.510360 samples/sec	 accuracy=30.141406	 loss=3.208297	 lr=0.096803
Epoch[052] Batch [2049]/[3759]	Speed: 117.709427 samples/sec	 accuracy=30.166921	 loss=3.207928	 lr=0.096798
Epoch[052] Batch [2099]/[3759]	Speed: 117.662826 samples/sec	 accuracy=30.143601	 loss=3.209815	 lr=0.096794
Epoch[052] Batch [2149]/[3759]	Speed: 116.958119 samples/sec	 accuracy=30.145349	 loss=3.211249	 lr=0.096789
Epoch[052] Batch [2199]/[3759]	Speed: 116.636817 samples/sec	 accuracy=30.124290	 loss=3.212805	 lr=0.096784
Epoch[052] Batch [2249]/[3759]	Speed: 117.068699 samples/sec	 accuracy=30.103472	 loss=3.214360	 lr=0.096780
Epoch[052] Batch [2299]/[3759]	Speed: 117.232371 samples/sec	 accuracy=30.098505	 loss=3.214689	 lr=0.096775
Epoch[052] Batch [2349]/[3759]	Speed: 117.889450 samples/sec	 accuracy=30.107048	 loss=3.214493	 lr=0.096771
Epoch[052] Batch [2399]/[3759]	Speed: 118.239366 samples/sec	 accuracy=30.125000	 loss=3.213602	 lr=0.096766
Epoch[052] Batch [2449]/[3759]	Speed: 116.419746 samples/sec	 accuracy=30.090561	 loss=3.214900	 lr=0.096762
Epoch[052] Batch [2499]/[3759]	Speed: 117.154005 samples/sec	 accuracy=30.073750	 loss=3.215541	 lr=0.096757
Epoch[052] Batch [2549]/[3759]	Speed: 116.561329 samples/sec	 accuracy=30.069240	 loss=3.217064	 lr=0.096752
Epoch[052] Batch [2599]/[3759]	Speed: 117.368123 samples/sec	 accuracy=30.054087	 loss=3.218460	 lr=0.096748
Epoch[052] Batch [2649]/[3759]	Speed: 118.072093 samples/sec	 accuracy=30.061910	 loss=3.218335	 lr=0.096743
Epoch[052] Batch [2699]/[3759]	Speed: 116.948055 samples/sec	 accuracy=30.075231	 loss=3.218240	 lr=0.096739
Epoch[052] Batch [2749]/[3759]	Speed: 117.411933 samples/sec	 accuracy=30.061364	 loss=3.218216	 lr=0.096734
Epoch[052] Batch [2799]/[3759]	Speed: 117.744572 samples/sec	 accuracy=30.084263	 loss=3.217752	 lr=0.096730
Epoch[052] Batch [2849]/[3759]	Speed: 117.161070 samples/sec	 accuracy=30.073465	 loss=3.218579	 lr=0.096725
Epoch[052] Batch [2899]/[3759]	Speed: 116.221583 samples/sec	 accuracy=30.077047	 loss=3.218200	 lr=0.096720
Epoch[052] Batch [2949]/[3759]	Speed: 117.583170 samples/sec	 accuracy=30.072034	 loss=3.218408	 lr=0.096716
Epoch[052] Batch [2999]/[3759]	Speed: 117.017476 samples/sec	 accuracy=30.088021	 loss=3.218053	 lr=0.096711
Epoch[052] Batch [3049]/[3759]	Speed: 117.496879 samples/sec	 accuracy=30.091701	 loss=3.218344	 lr=0.096707
Epoch[052] Batch [3099]/[3759]	Speed: 117.249434 samples/sec	 accuracy=30.082661	 loss=3.218634	 lr=0.096702
Epoch[052] Batch [3149]/[3759]	Speed: 118.115357 samples/sec	 accuracy=30.078869	 loss=3.218600	 lr=0.096697
Epoch[052] Batch [3199]/[3759]	Speed: 118.151969 samples/sec	 accuracy=30.102539	 loss=3.217150	 lr=0.096693
Epoch[052] Batch [3249]/[3759]	Speed: 117.266640 samples/sec	 accuracy=30.108173	 loss=3.216517	 lr=0.096688
Epoch[052] Batch [3299]/[3759]	Speed: 117.659847 samples/sec	 accuracy=30.085701	 loss=3.217765	 lr=0.096684
Epoch[052] Batch [3349]/[3759]	Speed: 117.839377 samples/sec	 accuracy=30.078825	 loss=3.218650	 lr=0.096679
Epoch[052] Batch [3399]/[3759]	Speed: 117.568300 samples/sec	 accuracy=30.066176	 loss=3.218264	 lr=0.096674
Epoch[052] Batch [3449]/[3759]	Speed: 116.849195 samples/sec	 accuracy=30.062047	 loss=3.218212	 lr=0.096670
Epoch[052] Batch [3499]/[3759]	Speed: 116.970824 samples/sec	 accuracy=30.061607	 loss=3.218315	 lr=0.096665
Epoch[052] Batch [3549]/[3759]	Speed: 116.833243 samples/sec	 accuracy=30.046215	 loss=3.219105	 lr=0.096660
Epoch[052] Batch [3599]/[3759]	Speed: 117.161440 samples/sec	 accuracy=30.065538	 loss=3.219046	 lr=0.096656
Epoch[052] Batch [3649]/[3759]	Speed: 116.765719 samples/sec	 accuracy=30.072346	 loss=3.218819	 lr=0.096651
Epoch[052] Batch [3699]/[3759]	Speed: 117.394634 samples/sec	 accuracy=30.060389	 loss=3.219218	 lr=0.096646
Epoch[052] Batch [3749]/[3759]	Speed: 120.110121 samples/sec	 accuracy=30.066667	 loss=3.219092	 lr=0.096642
Batch [0049]/[0303]: acc-top1=28.687500 acc-top5=55.500000
Batch [0099]/[0303]: acc-top1=26.812500 acc-top5=53.500000
Batch [0149]/[0303]: acc-top1=28.208333 acc-top5=54.666667
Batch [0199]/[0303]: acc-top1=31.343750 acc-top5=57.515625
Batch [0249]/[0303]: acc-top1=30.656250 acc-top5=57.468750
Batch [0299]/[0303]: acc-top1=30.052083 acc-top5=57.536458
[Epoch 052] training: accuracy=30.069500	 loss=3.219362
[Epoch 052] speed: 116 samples/sec	time cost: 2179.447895
[Epoch 052] validation: acc-top1=30.120668 acc-top5=57.693894 loss=3.274667
Epoch[053] Batch [0049]/[3760]	Speed: 87.525934 samples/sec	 accuracy=30.031250	 loss=3.173634	 lr=0.096636
Epoch[053] Batch [0099]/[3760]	Speed: 117.383949 samples/sec	 accuracy=30.312500	 loss=3.189294	 lr=0.096632
Epoch[053] Batch [0149]/[3760]	Speed: 117.318976 samples/sec	 accuracy=30.593750	 loss=3.182497	 lr=0.096627
Epoch[053] Batch [0199]/[3760]	Speed: 116.825102 samples/sec	 accuracy=30.703125	 loss=3.189834	 lr=0.096622
Epoch[053] Batch [0249]/[3760]	Speed: 117.423421 samples/sec	 accuracy=30.437500	 loss=3.206951	 lr=0.096618
Epoch[053] Batch [0299]/[3760]	Speed: 117.217129 samples/sec	 accuracy=30.505208	 loss=3.211528	 lr=0.096613
Epoch[053] Batch [0349]/[3760]	Speed: 116.878857 samples/sec	 accuracy=30.441964	 loss=3.207364	 lr=0.096608
Epoch[053] Batch [0399]/[3760]	Speed: 116.856307 samples/sec	 accuracy=30.433594	 loss=3.210097	 lr=0.096604
Epoch[053] Batch [0449]/[3760]	Speed: 117.246408 samples/sec	 accuracy=30.472222	 loss=3.205218	 lr=0.096599
Epoch[053] Batch [0499]/[3760]	Speed: 116.809756 samples/sec	 accuracy=30.450000	 loss=3.205349	 lr=0.096594
Epoch[053] Batch [0549]/[3760]	Speed: 117.251001 samples/sec	 accuracy=30.389205	 loss=3.211529	 lr=0.096590
Epoch[053] Batch [0599]/[3760]	Speed: 117.116333 samples/sec	 accuracy=30.375000	 loss=3.209098	 lr=0.096585
Epoch[053] Batch [0649]/[3760]	Speed: 117.503759 samples/sec	 accuracy=30.384615	 loss=3.210334	 lr=0.096580
Epoch[053] Batch [0699]/[3760]	Speed: 116.953439 samples/sec	 accuracy=30.428571	 loss=3.211723	 lr=0.096576
Epoch[053] Batch [0749]/[3760]	Speed: 117.005207 samples/sec	 accuracy=30.293750	 loss=3.215115	 lr=0.096571
Epoch[053] Batch [0799]/[3760]	Speed: 116.968639 samples/sec	 accuracy=30.212891	 loss=3.219270	 lr=0.096566
Epoch[053] Batch [0849]/[3760]	Speed: 116.675770 samples/sec	 accuracy=30.290441	 loss=3.216272	 lr=0.096562
Epoch[053] Batch [0899]/[3760]	Speed: 118.359463 samples/sec	 accuracy=30.319444	 loss=3.215235	 lr=0.096557
Epoch[053] Batch [0949]/[3760]	Speed: 117.200127 samples/sec	 accuracy=30.287829	 loss=3.215096	 lr=0.096552
Epoch[053] Batch [0999]/[3760]	Speed: 116.781428 samples/sec	 accuracy=30.251563	 loss=3.216496	 lr=0.096547
Epoch[053] Batch [1049]/[3760]	Speed: 117.250307 samples/sec	 accuracy=30.288690	 loss=3.217232	 lr=0.096543
Epoch[053] Batch [1099]/[3760]	Speed: 117.158503 samples/sec	 accuracy=30.289773	 loss=3.217009	 lr=0.096538
Epoch[053] Batch [1149]/[3760]	Speed: 118.045919 samples/sec	 accuracy=30.267663	 loss=3.217147	 lr=0.096533
Epoch[053] Batch [1199]/[3760]	Speed: 116.791835 samples/sec	 accuracy=30.250000	 loss=3.217366	 lr=0.096529
Epoch[053] Batch [1249]/[3760]	Speed: 116.820109 samples/sec	 accuracy=30.218750	 loss=3.216483	 lr=0.096524
Epoch[053] Batch [1299]/[3760]	Speed: 117.073827 samples/sec	 accuracy=30.216346	 loss=3.215775	 lr=0.096519
Epoch[053] Batch [1349]/[3760]	Speed: 117.195685 samples/sec	 accuracy=30.201389	 loss=3.215843	 lr=0.096514
Epoch[053] Batch [1399]/[3760]	Speed: 117.470997 samples/sec	 accuracy=30.183036	 loss=3.217586	 lr=0.096510
Epoch[053] Batch [1449]/[3760]	Speed: 116.951377 samples/sec	 accuracy=30.215517	 loss=3.216018	 lr=0.096505
Epoch[053] Batch [1499]/[3760]	Speed: 116.748052 samples/sec	 accuracy=30.220833	 loss=3.217312	 lr=0.096500
Epoch[053] Batch [1549]/[3760]	Speed: 116.928264 samples/sec	 accuracy=30.248992	 loss=3.215350	 lr=0.096495
Epoch[053] Batch [1599]/[3760]	Speed: 117.717483 samples/sec	 accuracy=30.245117	 loss=3.215208	 lr=0.096491
Epoch[053] Batch [1649]/[3760]	Speed: 116.761273 samples/sec	 accuracy=30.267992	 loss=3.214272	 lr=0.096486
Epoch[053] Batch [1699]/[3760]	Speed: 117.562025 samples/sec	 accuracy=30.285846	 loss=3.214305	 lr=0.096481
Epoch[053] Batch [1749]/[3760]	Speed: 116.992063 samples/sec	 accuracy=30.282143	 loss=3.213311	 lr=0.096476
Epoch[053] Batch [1799]/[3760]	Speed: 117.620508 samples/sec	 accuracy=30.289931	 loss=3.211787	 lr=0.096472
Epoch[053] Batch [1849]/[3760]	Speed: 117.776330 samples/sec	 accuracy=30.292230	 loss=3.211771	 lr=0.096467
Epoch[053] Batch [1899]/[3760]	Speed: 117.576408 samples/sec	 accuracy=30.282895	 loss=3.211139	 lr=0.096462
Epoch[053] Batch [1949]/[3760]	Speed: 117.320991 samples/sec	 accuracy=30.293269	 loss=3.211832	 lr=0.096457
Epoch[053] Batch [1999]/[3760]	Speed: 116.988894 samples/sec	 accuracy=30.300000	 loss=3.211620	 lr=0.096453
Epoch[053] Batch [2049]/[3760]	Speed: 118.130099 samples/sec	 accuracy=30.301829	 loss=3.212182	 lr=0.096448
Epoch[053] Batch [2099]/[3760]	Speed: 117.612110 samples/sec	 accuracy=30.313988	 loss=3.211539	 lr=0.096443
Epoch[053] Batch [2149]/[3760]	Speed: 117.298111 samples/sec	 accuracy=30.312500	 loss=3.211786	 lr=0.096438
Epoch[053] Batch [2199]/[3760]	Speed: 117.476331 samples/sec	 accuracy=30.292614	 loss=3.211735	 lr=0.096434
Epoch[053] Batch [2249]/[3760]	Speed: 117.221813 samples/sec	 accuracy=30.281250	 loss=3.212810	 lr=0.096429
Epoch[053] Batch [2299]/[3760]	Speed: 117.453192 samples/sec	 accuracy=30.326766	 loss=3.212260	 lr=0.096424
Epoch[053] Batch [2349]/[3760]	Speed: 117.069303 samples/sec	 accuracy=30.305851	 loss=3.213683	 lr=0.096419
Epoch[053] Batch [2399]/[3760]	Speed: 117.120018 samples/sec	 accuracy=30.279948	 loss=3.215390	 lr=0.096414
Epoch[053] Batch [2449]/[3760]	Speed: 117.508567 samples/sec	 accuracy=30.242347	 loss=3.216588	 lr=0.096410
Epoch[053] Batch [2499]/[3760]	Speed: 117.419678 samples/sec	 accuracy=30.232500	 loss=3.217158	 lr=0.096405
Epoch[053] Batch [2549]/[3760]	Speed: 117.505991 samples/sec	 accuracy=30.208333	 loss=3.218160	 lr=0.096400
Epoch[053] Batch [2599]/[3760]	Speed: 117.207827 samples/sec	 accuracy=30.210337	 loss=3.218706	 lr=0.096395
Epoch[053] Batch [2649]/[3760]	Speed: 117.199660 samples/sec	 accuracy=30.199882	 loss=3.219335	 lr=0.096390
Epoch[053] Batch [2699]/[3760]	Speed: 117.930774 samples/sec	 accuracy=30.196759	 loss=3.218852	 lr=0.096386
Epoch[053] Batch [2749]/[3760]	Speed: 117.150178 samples/sec	 accuracy=30.231250	 loss=3.217735	 lr=0.096381
Epoch[053] Batch [2799]/[3760]	Speed: 117.588238 samples/sec	 accuracy=30.231027	 loss=3.218014	 lr=0.096376
Epoch[053] Batch [2849]/[3760]	Speed: 117.136947 samples/sec	 accuracy=30.225329	 loss=3.219258	 lr=0.096371
Epoch[053] Batch [2899]/[3760]	Speed: 117.552009 samples/sec	 accuracy=30.230065	 loss=3.219924	 lr=0.096366
Epoch[053] Batch [2949]/[3760]	Speed: 117.390125 samples/sec	 accuracy=30.231992	 loss=3.220622	 lr=0.096361
Epoch[053] Batch [2999]/[3760]	Speed: 117.971541 samples/sec	 accuracy=30.232813	 loss=3.220316	 lr=0.096357
Epoch[053] Batch [3049]/[3760]	Speed: 118.020626 samples/sec	 accuracy=30.240779	 loss=3.219977	 lr=0.096352
Epoch[053] Batch [3099]/[3760]	Speed: 117.781524 samples/sec	 accuracy=30.218246	 loss=3.221199	 lr=0.096347
Epoch[053] Batch [3149]/[3760]	Speed: 117.141341 samples/sec	 accuracy=30.210317	 loss=3.220814	 lr=0.096342
Epoch[053] Batch [3199]/[3760]	Speed: 117.541580 samples/sec	 accuracy=30.218262	 loss=3.220427	 lr=0.096337
Epoch[053] Batch [3249]/[3760]	Speed: 117.581017 samples/sec	 accuracy=30.240865	 loss=3.218884	 lr=0.096332
Epoch[053] Batch [3299]/[3760]	Speed: 118.219333 samples/sec	 accuracy=30.229167	 loss=3.219908	 lr=0.096328
Epoch[053] Batch [3349]/[3760]	Speed: 117.301480 samples/sec	 accuracy=30.234608	 loss=3.219803	 lr=0.096323
Epoch[053] Batch [3399]/[3760]	Speed: 117.918940 samples/sec	 accuracy=30.237592	 loss=3.220137	 lr=0.096318
Epoch[053] Batch [3449]/[3760]	Speed: 117.394124 samples/sec	 accuracy=30.229167	 loss=3.220570	 lr=0.096313
Epoch[053] Batch [3499]/[3760]	Speed: 118.201028 samples/sec	 accuracy=30.234821	 loss=3.221065	 lr=0.096308
Epoch[053] Batch [3549]/[3760]	Speed: 116.844659 samples/sec	 accuracy=30.225792	 loss=3.221148	 lr=0.096303
Epoch[053] Batch [3599]/[3760]	Speed: 118.197207 samples/sec	 accuracy=30.244358	 loss=3.220323	 lr=0.096298
Epoch[053] Batch [3649]/[3760]	Speed: 116.992874 samples/sec	 accuracy=30.248288	 loss=3.220069	 lr=0.096293
Epoch[053] Batch [3699]/[3760]	Speed: 117.246917 samples/sec	 accuracy=30.254645	 loss=3.218498	 lr=0.096289
Epoch[053] Batch [3749]/[3760]	Speed: 120.848031 samples/sec	 accuracy=30.262083	 loss=3.219203	 lr=0.096284
Batch [0049]/[0303]: acc-top1=27.906250 acc-top5=52.656250
Batch [0099]/[0303]: acc-top1=26.546875 acc-top5=52.812500
Batch [0149]/[0303]: acc-top1=27.677083 acc-top5=53.875000
Batch [0199]/[0303]: acc-top1=30.640625 acc-top5=56.703125
Batch [0249]/[0303]: acc-top1=29.837500 acc-top5=56.268750
Batch [0299]/[0303]: acc-top1=29.536458 acc-top5=56.203125
[Epoch 053] training: accuracy=30.263464	 loss=3.219197
[Epoch 053] speed: 116 samples/sec	time cost: 2178.987655
[Epoch 053] validation: acc-top1=29.553424 acc-top5=56.136551 loss=3.370646
Epoch[054] Batch [0049]/[3760]	Speed: 87.492034 samples/sec	 accuracy=30.781250	 loss=3.185823	 lr=0.096278
Epoch[054] Batch [0099]/[3760]	Speed: 117.014377 samples/sec	 accuracy=30.421875	 loss=3.200772	 lr=0.096273
Epoch[054] Batch [0149]/[3760]	Speed: 118.007439 samples/sec	 accuracy=30.572917	 loss=3.186760	 lr=0.096268
Epoch[054] Batch [0199]/[3760]	Speed: 117.886361 samples/sec	 accuracy=30.343750	 loss=3.195824	 lr=0.096263
Epoch[054] Batch [0249]/[3760]	Speed: 117.486813 samples/sec	 accuracy=30.412500	 loss=3.202737	 lr=0.096258
Epoch[054] Batch [0299]/[3760]	Speed: 117.841070 samples/sec	 accuracy=30.270833	 loss=3.204284	 lr=0.096253
Epoch[054] Batch [0349]/[3760]	Speed: 118.007641 samples/sec	 accuracy=30.491071	 loss=3.195272	 lr=0.096249
Epoch[054] Batch [0399]/[3760]	Speed: 117.216705 samples/sec	 accuracy=30.410156	 loss=3.194889	 lr=0.096244
Epoch[054] Batch [0449]/[3760]	Speed: 117.918428 samples/sec	 accuracy=30.222222	 loss=3.203893	 lr=0.096239
Epoch[054] Batch [0499]/[3760]	Speed: 117.377999 samples/sec	 accuracy=30.190625	 loss=3.201697	 lr=0.096234
Epoch[054] Batch [0549]/[3760]	Speed: 118.216177 samples/sec	 accuracy=30.142045	 loss=3.202766	 lr=0.096229
Epoch[054] Batch [0599]/[3760]	Speed: 117.752943 samples/sec	 accuracy=30.270833	 loss=3.202082	 lr=0.096224
Epoch[054] Batch [0649]/[3760]	Speed: 116.694601 samples/sec	 accuracy=30.302885	 loss=3.202005	 lr=0.096219
Epoch[054] Batch [0699]/[3760]	Speed: 117.241440 samples/sec	 accuracy=30.165179	 loss=3.203688	 lr=0.096214
Epoch[054] Batch [0749]/[3760]	Speed: 117.369349 samples/sec	 accuracy=30.231250	 loss=3.204725	 lr=0.096209
Epoch[054] Batch [0799]/[3760]	Speed: 118.018987 samples/sec	 accuracy=30.240234	 loss=3.206712	 lr=0.096204
Epoch[054] Batch [0849]/[3760]	Speed: 117.517400 samples/sec	 accuracy=30.308824	 loss=3.205059	 lr=0.096199
Epoch[054] Batch [0899]/[3760]	Speed: 117.330690 samples/sec	 accuracy=30.241319	 loss=3.207768	 lr=0.096194
Epoch[054] Batch [0949]/[3760]	Speed: 117.121885 samples/sec	 accuracy=30.222039	 loss=3.207254	 lr=0.096189
Epoch[054] Batch [0999]/[3760]	Speed: 117.120485 samples/sec	 accuracy=30.320312	 loss=3.204238	 lr=0.096185
Epoch[054] Batch [1049]/[3760]	Speed: 116.914391 samples/sec	 accuracy=30.238095	 loss=3.206819	 lr=0.096180
Epoch[054] Batch [1099]/[3760]	Speed: 118.395120 samples/sec	 accuracy=30.187500	 loss=3.210024	 lr=0.096175
Epoch[054] Batch [1149]/[3760]	Speed: 116.469124 samples/sec	 accuracy=30.182065	 loss=3.210246	 lr=0.096170
Epoch[054] Batch [1199]/[3760]	Speed: 117.312693 samples/sec	 accuracy=30.209635	 loss=3.208224	 lr=0.096165
Epoch[054] Batch [1249]/[3760]	Speed: 117.523998 samples/sec	 accuracy=30.227500	 loss=3.206512	 lr=0.096160
Epoch[054] Batch [1299]/[3760]	Speed: 117.483254 samples/sec	 accuracy=30.218750	 loss=3.207531	 lr=0.096155
Epoch[054] Batch [1349]/[3760]	Speed: 117.346965 samples/sec	 accuracy=30.206019	 loss=3.208210	 lr=0.096150
Epoch[054] Batch [1399]/[3760]	Speed: 117.089954 samples/sec	 accuracy=30.263393	 loss=3.206298	 lr=0.096145
Epoch[054] Batch [1449]/[3760]	Speed: 117.229612 samples/sec	 accuracy=30.297414	 loss=3.205481	 lr=0.096140
Epoch[054] Batch [1499]/[3760]	Speed: 117.058817 samples/sec	 accuracy=30.279167	 loss=3.205149	 lr=0.096135
Epoch[054] Batch [1549]/[3760]	Speed: 117.065936 samples/sec	 accuracy=30.313508	 loss=3.205200	 lr=0.096130
Epoch[054] Batch [1599]/[3760]	Speed: 117.167990 samples/sec	 accuracy=30.333984	 loss=3.204946	 lr=0.096125
Epoch[054] Batch [1649]/[3760]	Speed: 117.466759 samples/sec	 accuracy=30.336174	 loss=3.205016	 lr=0.096120
Epoch[054] Batch [1699]/[3760]	Speed: 118.091646 samples/sec	 accuracy=30.322610	 loss=3.206459	 lr=0.096115
Epoch[054] Batch [1749]/[3760]	Speed: 117.452911 samples/sec	 accuracy=30.304464	 loss=3.207008	 lr=0.096110
Epoch[054] Batch [1799]/[3760]	Speed: 118.119451 samples/sec	 accuracy=30.283854	 loss=3.207384	 lr=0.096105
Epoch[054] Batch [1849]/[3760]	Speed: 117.591864 samples/sec	 accuracy=30.301520	 loss=3.206496	 lr=0.096100
Epoch[054] Batch [1899]/[3760]	Speed: 117.563651 samples/sec	 accuracy=30.300164	 loss=3.207047	 lr=0.096095
Epoch[054] Batch [1949]/[3760]	Speed: 117.809034 samples/sec	 accuracy=30.315705	 loss=3.206711	 lr=0.096090
Epoch[054] Batch [1999]/[3760]	Speed: 118.162386 samples/sec	 accuracy=30.289844	 loss=3.207195	 lr=0.096085
Epoch[054] Batch [2049]/[3760]	Speed: 117.040989 samples/sec	 accuracy=30.283537	 loss=3.206923	 lr=0.096080
Epoch[054] Batch [2099]/[3760]	Speed: 117.058286 samples/sec	 accuracy=30.312500	 loss=3.205391	 lr=0.096075
Epoch[054] Batch [2149]/[3760]	Speed: 117.715313 samples/sec	 accuracy=30.295058	 loss=3.206274	 lr=0.096070
Epoch[054] Batch [2199]/[3760]	Speed: 117.598623 samples/sec	 accuracy=30.306818	 loss=3.206010	 lr=0.096065
Epoch[054] Batch [2249]/[3760]	Speed: 117.370521 samples/sec	 accuracy=30.313889	 loss=3.206667	 lr=0.096060
Epoch[054] Batch [2299]/[3760]	Speed: 117.728424 samples/sec	 accuracy=30.336277	 loss=3.205343	 lr=0.096055
Epoch[054] Batch [2349]/[3760]	Speed: 117.815279 samples/sec	 accuracy=30.305851	 loss=3.206873	 lr=0.096050
Epoch[054] Batch [2399]/[3760]	Speed: 116.954297 samples/sec	 accuracy=30.327474	 loss=3.206246	 lr=0.096045
Epoch[054] Batch [2449]/[3760]	Speed: 117.398031 samples/sec	 accuracy=30.344388	 loss=3.205329	 lr=0.096040
Epoch[054] Batch [2499]/[3760]	Speed: 117.586842 samples/sec	 accuracy=30.351250	 loss=3.204432	 lr=0.096035
Epoch[054] Batch [2549]/[3760]	Speed: 117.349218 samples/sec	 accuracy=30.335784	 loss=3.205200	 lr=0.096030
Epoch[054] Batch [2599]/[3760]	Speed: 117.517716 samples/sec	 accuracy=30.320312	 loss=3.205822	 lr=0.096025
Epoch[054] Batch [2649]/[3760]	Speed: 118.821584 samples/sec	 accuracy=30.343160	 loss=3.205723	 lr=0.096020
Epoch[054] Batch [2699]/[3760]	Speed: 117.916699 samples/sec	 accuracy=30.307870	 loss=3.206831	 lr=0.096015
Epoch[054] Batch [2749]/[3760]	Speed: 117.695397 samples/sec	 accuracy=30.293182	 loss=3.206955	 lr=0.096010
Epoch[054] Batch [2799]/[3760]	Speed: 117.585164 samples/sec	 accuracy=30.277902	 loss=3.207616	 lr=0.096005
Epoch[054] Batch [2849]/[3760]	Speed: 117.336023 samples/sec	 accuracy=30.295504	 loss=3.206740	 lr=0.096000
Epoch[054] Batch [2899]/[3760]	Speed: 117.555406 samples/sec	 accuracy=30.317349	 loss=3.205653	 lr=0.095995
Epoch[054] Batch [2949]/[3760]	Speed: 117.178261 samples/sec	 accuracy=30.310911	 loss=3.205669	 lr=0.095990
Epoch[054] Batch [2999]/[3760]	Speed: 116.768777 samples/sec	 accuracy=30.314583	 loss=3.205469	 lr=0.095984
Epoch[054] Batch [3049]/[3760]	Speed: 117.244793 samples/sec	 accuracy=30.307889	 loss=3.205848	 lr=0.095979
Epoch[054] Batch [3099]/[3760]	Speed: 117.232182 samples/sec	 accuracy=30.300403	 loss=3.205651	 lr=0.095974
Epoch[054] Batch [3149]/[3760]	Speed: 117.737993 samples/sec	 accuracy=30.279266	 loss=3.206839	 lr=0.095969
Epoch[054] Batch [3199]/[3760]	Speed: 117.337291 samples/sec	 accuracy=30.289551	 loss=3.206073	 lr=0.095964
Epoch[054] Batch [3249]/[3760]	Speed: 116.920779 samples/sec	 accuracy=30.277885	 loss=3.207168	 lr=0.095959
Epoch[054] Batch [3299]/[3760]	Speed: 118.150603 samples/sec	 accuracy=30.282197	 loss=3.206564	 lr=0.095954
Epoch[054] Batch [3349]/[3760]	Speed: 116.896990 samples/sec	 accuracy=30.283116	 loss=3.206513	 lr=0.095949
Epoch[054] Batch [3399]/[3760]	Speed: 117.000439 samples/sec	 accuracy=30.275276	 loss=3.207294	 lr=0.095944
Epoch[054] Batch [3449]/[3760]	Speed: 117.264696 samples/sec	 accuracy=30.286685	 loss=3.206698	 lr=0.095939
Epoch[054] Batch [3499]/[3760]	Speed: 117.195709 samples/sec	 accuracy=30.284821	 loss=3.206685	 lr=0.095934
Epoch[054] Batch [3549]/[3760]	Speed: 117.424829 samples/sec	 accuracy=30.277289	 loss=3.207042	 lr=0.095929
Epoch[054] Batch [3599]/[3760]	Speed: 116.927946 samples/sec	 accuracy=30.273438	 loss=3.207204	 lr=0.095923
Epoch[054] Batch [3649]/[3760]	Speed: 117.392353 samples/sec	 accuracy=30.269264	 loss=3.207722	 lr=0.095918
Epoch[054] Batch [3699]/[3760]	Speed: 117.498441 samples/sec	 accuracy=30.288851	 loss=3.206688	 lr=0.095913
Epoch[054] Batch [3749]/[3760]	Speed: 120.713436 samples/sec	 accuracy=30.285833	 loss=3.206910	 lr=0.095908
Batch [0049]/[0303]: acc-top1=28.437500 acc-top5=55.718750
Batch [0099]/[0303]: acc-top1=27.265625 acc-top5=54.609375
Batch [0149]/[0303]: acc-top1=28.989583 acc-top5=55.937500
Batch [0199]/[0303]: acc-top1=32.421875 acc-top5=58.617188
Batch [0249]/[0303]: acc-top1=32.050000 acc-top5=58.931250
Batch [0299]/[0303]: acc-top1=31.770833 acc-top5=59.208333
[Epoch 054] training: accuracy=30.281749	 loss=3.207139
[Epoch 054] speed: 116 samples/sec	time cost: 2178.190730
[Epoch 054] validation: acc-top1=31.801774 acc-top5=59.179043 loss=3.136613
Epoch[055] Batch [0049]/[3759]	Speed: 87.437399 samples/sec	 accuracy=30.281250	 loss=3.242035	 lr=0.095902
Epoch[055] Batch [0099]/[3759]	Speed: 118.039160 samples/sec	 accuracy=30.296875	 loss=3.214525	 lr=0.095897
Epoch[055] Batch [0149]/[3759]	Speed: 118.028295 samples/sec	 accuracy=30.062500	 loss=3.214886	 lr=0.095892
Epoch[055] Batch [0199]/[3759]	Speed: 117.848327 samples/sec	 accuracy=30.164062	 loss=3.203406	 lr=0.095887
Epoch[055] Batch [0249]/[3759]	Speed: 117.376425 samples/sec	 accuracy=30.100000	 loss=3.204382	 lr=0.095882
Epoch[055] Batch [0299]/[3759]	Speed: 117.161814 samples/sec	 accuracy=30.031250	 loss=3.213926	 lr=0.095876
Epoch[055] Batch [0349]/[3759]	Speed: 117.692777 samples/sec	 accuracy=30.013393	 loss=3.212533	 lr=0.095871
Epoch[055] Batch [0399]/[3759]	Speed: 116.888383 samples/sec	 accuracy=29.941406	 loss=3.216678	 lr=0.095866
Epoch[055] Batch [0449]/[3759]	Speed: 117.222141 samples/sec	 accuracy=29.829861	 loss=3.225335	 lr=0.095861
Epoch[055] Batch [0499]/[3759]	Speed: 117.706635 samples/sec	 accuracy=29.843750	 loss=3.220528	 lr=0.095856
Epoch[055] Batch [0549]/[3759]	Speed: 118.918543 samples/sec	 accuracy=29.769886	 loss=3.219233	 lr=0.095851
Epoch[055] Batch [0599]/[3759]	Speed: 117.003515 samples/sec	 accuracy=29.929688	 loss=3.213803	 lr=0.095846
Epoch[055] Batch [0649]/[3759]	Speed: 117.680337 samples/sec	 accuracy=29.884615	 loss=3.215159	 lr=0.095840
Epoch[055] Batch [0699]/[3759]	Speed: 117.357186 samples/sec	 accuracy=29.832589	 loss=3.216662	 lr=0.095835
Epoch[055] Batch [0749]/[3759]	Speed: 116.578837 samples/sec	 accuracy=30.022917	 loss=3.207168	 lr=0.095830
Epoch[055] Batch [0799]/[3759]	Speed: 118.270918 samples/sec	 accuracy=30.033203	 loss=3.209222	 lr=0.095825
Epoch[055] Batch [0849]/[3759]	Speed: 117.952046 samples/sec	 accuracy=30.062500	 loss=3.208593	 lr=0.095820
Epoch[055] Batch [0899]/[3759]	Speed: 117.348892 samples/sec	 accuracy=30.076389	 loss=3.208395	 lr=0.095815
Epoch[055] Batch [0949]/[3759]	Speed: 116.305544 samples/sec	 accuracy=30.105263	 loss=3.207732	 lr=0.095809
Epoch[055] Batch [0999]/[3759]	Speed: 116.903133 samples/sec	 accuracy=30.123438	 loss=3.207948	 lr=0.095804
Epoch[055] Batch [1049]/[3759]	Speed: 117.546975 samples/sec	 accuracy=30.142857	 loss=3.208922	 lr=0.095799
Epoch[055] Batch [1099]/[3759]	Speed: 117.165826 samples/sec	 accuracy=30.127841	 loss=3.208016	 lr=0.095794
Epoch[055] Batch [1149]/[3759]	Speed: 117.220557 samples/sec	 accuracy=30.127717	 loss=3.209432	 lr=0.095789
Epoch[055] Batch [1199]/[3759]	Speed: 117.643547 samples/sec	 accuracy=30.156250	 loss=3.208506	 lr=0.095784
Epoch[055] Batch [1249]/[3759]	Speed: 117.615894 samples/sec	 accuracy=30.170000	 loss=3.209591	 lr=0.095778
Epoch[055] Batch [1299]/[3759]	Speed: 117.456358 samples/sec	 accuracy=30.163462	 loss=3.207388	 lr=0.095773
Epoch[055] Batch [1349]/[3759]	Speed: 117.338654 samples/sec	 accuracy=30.210648	 loss=3.206126	 lr=0.095768
Epoch[055] Batch [1399]/[3759]	Speed: 118.044378 samples/sec	 accuracy=30.224330	 loss=3.205142	 lr=0.095763
Epoch[055] Batch [1449]/[3759]	Speed: 117.475290 samples/sec	 accuracy=30.229526	 loss=3.205095	 lr=0.095758
Epoch[055] Batch [1499]/[3759]	Speed: 116.907241 samples/sec	 accuracy=30.236458	 loss=3.203870	 lr=0.095752
Epoch[055] Batch [1549]/[3759]	Speed: 117.457100 samples/sec	 accuracy=30.245968	 loss=3.203235	 lr=0.095747
Epoch[055] Batch [1599]/[3759]	Speed: 116.922670 samples/sec	 accuracy=30.265625	 loss=3.204269	 lr=0.095742
Epoch[055] Batch [1649]/[3759]	Speed: 117.551098 samples/sec	 accuracy=30.294508	 loss=3.202705	 lr=0.095737
Epoch[055] Batch [1699]/[3759]	Speed: 117.076113 samples/sec	 accuracy=30.299632	 loss=3.202028	 lr=0.095732
Epoch[055] Batch [1749]/[3759]	Speed: 117.597624 samples/sec	 accuracy=30.265179	 loss=3.202426	 lr=0.095726
Epoch[055] Batch [1799]/[3759]	Speed: 117.795086 samples/sec	 accuracy=30.248264	 loss=3.203699	 lr=0.095721
Epoch[055] Batch [1849]/[3759]	Speed: 118.152909 samples/sec	 accuracy=30.240709	 loss=3.203562	 lr=0.095716
Epoch[055] Batch [1899]/[3759]	Speed: 117.750259 samples/sec	 accuracy=30.291118	 loss=3.201829	 lr=0.095711
Epoch[055] Batch [1949]/[3759]	Speed: 116.852675 samples/sec	 accuracy=30.282853	 loss=3.201620	 lr=0.095705
Epoch[055] Batch [1999]/[3759]	Speed: 116.933074 samples/sec	 accuracy=30.280469	 loss=3.202167	 lr=0.095700
Epoch[055] Batch [2049]/[3759]	Speed: 116.952003 samples/sec	 accuracy=30.288872	 loss=3.201408	 lr=0.095695
Epoch[055] Batch [2099]/[3759]	Speed: 115.900531 samples/sec	 accuracy=30.325893	 loss=3.199961	 lr=0.095690
Epoch[055] Batch [2149]/[3759]	Speed: 117.040785 samples/sec	 accuracy=30.313953	 loss=3.200915	 lr=0.095685
Epoch[055] Batch [2199]/[3759]	Speed: 116.802920 samples/sec	 accuracy=30.295455	 loss=3.201586	 lr=0.095679
Epoch[055] Batch [2249]/[3759]	Speed: 116.924179 samples/sec	 accuracy=30.286111	 loss=3.201334	 lr=0.095674
Epoch[055] Batch [2299]/[3759]	Speed: 117.878855 samples/sec	 accuracy=30.292120	 loss=3.201226	 lr=0.095669
Epoch[055] Batch [2349]/[3759]	Speed: 118.014568 samples/sec	 accuracy=30.264628	 loss=3.203152	 lr=0.095664
Epoch[055] Batch [2399]/[3759]	Speed: 117.462728 samples/sec	 accuracy=30.238281	 loss=3.203396	 lr=0.095658
Epoch[055] Batch [2449]/[3759]	Speed: 117.474369 samples/sec	 accuracy=30.240434	 loss=3.203726	 lr=0.095653
Epoch[055] Batch [2499]/[3759]	Speed: 116.399973 samples/sec	 accuracy=30.260625	 loss=3.202961	 lr=0.095648
Epoch[055] Batch [2549]/[3759]	Speed: 117.910091 samples/sec	 accuracy=30.235294	 loss=3.203325	 lr=0.095643
Epoch[055] Batch [2599]/[3759]	Speed: 117.658175 samples/sec	 accuracy=30.241587	 loss=3.203773	 lr=0.095637
Epoch[055] Batch [2649]/[3759]	Speed: 116.637143 samples/sec	 accuracy=30.225825	 loss=3.204986	 lr=0.095632
Epoch[055] Batch [2699]/[3759]	Speed: 116.833923 samples/sec	 accuracy=30.233796	 loss=3.205082	 lr=0.095627
Epoch[055] Batch [2749]/[3759]	Speed: 117.125928 samples/sec	 accuracy=30.236364	 loss=3.205161	 lr=0.095621
Epoch[055] Batch [2799]/[3759]	Speed: 117.664000 samples/sec	 accuracy=30.241071	 loss=3.204847	 lr=0.095616
Epoch[055] Batch [2849]/[3759]	Speed: 118.008635 samples/sec	 accuracy=30.234101	 loss=3.205191	 lr=0.095611
Epoch[055] Batch [2899]/[3759]	Speed: 117.755609 samples/sec	 accuracy=30.238685	 loss=3.204462	 lr=0.095606
Epoch[055] Batch [2949]/[3759]	Speed: 117.090747 samples/sec	 accuracy=30.227754	 loss=3.204607	 lr=0.095600
Epoch[055] Batch [2999]/[3759]	Speed: 116.563136 samples/sec	 accuracy=30.240625	 loss=3.205047	 lr=0.095595
Epoch[055] Batch [3049]/[3759]	Speed: 117.226922 samples/sec	 accuracy=30.271004	 loss=3.204341	 lr=0.095590
Epoch[055] Batch [3099]/[3759]	Speed: 116.798927 samples/sec	 accuracy=30.281250	 loss=3.204989	 lr=0.095584
Epoch[055] Batch [3149]/[3759]	Speed: 117.176757 samples/sec	 accuracy=30.287202	 loss=3.204514	 lr=0.095579
Epoch[055] Batch [3199]/[3759]	Speed: 117.056872 samples/sec	 accuracy=30.292480	 loss=3.205154	 lr=0.095574
Epoch[055] Batch [3249]/[3759]	Speed: 117.526230 samples/sec	 accuracy=30.288462	 loss=3.204779	 lr=0.095568
Epoch[055] Batch [3299]/[3759]	Speed: 117.034876 samples/sec	 accuracy=30.284564	 loss=3.205221	 lr=0.095563
Epoch[055] Batch [3349]/[3759]	Speed: 117.634614 samples/sec	 accuracy=30.295709	 loss=3.204934	 lr=0.095558
Epoch[055] Batch [3399]/[3759]	Speed: 117.479496 samples/sec	 accuracy=30.281710	 loss=3.205402	 lr=0.095553
Epoch[055] Batch [3449]/[3759]	Speed: 117.935803 samples/sec	 accuracy=30.289402	 loss=3.205240	 lr=0.095547
Epoch[055] Batch [3499]/[3759]	Speed: 117.902735 samples/sec	 accuracy=30.283036	 loss=3.205843	 lr=0.095542
Epoch[055] Batch [3549]/[3759]	Speed: 117.309570 samples/sec	 accuracy=30.299296	 loss=3.205341	 lr=0.095537
Epoch[055] Batch [3599]/[3759]	Speed: 116.773056 samples/sec	 accuracy=30.303819	 loss=3.205179	 lr=0.095531
Epoch[055] Batch [3649]/[3759]	Speed: 118.162881 samples/sec	 accuracy=30.282962	 loss=3.205695	 lr=0.095526
Epoch[055] Batch [3699]/[3759]	Speed: 117.385096 samples/sec	 accuracy=30.297720	 loss=3.205854	 lr=0.095521
Epoch[055] Batch [3749]/[3759]	Speed: 121.027785 samples/sec	 accuracy=30.287917	 loss=3.205933	 lr=0.095515
Batch [0049]/[0303]: acc-top1=31.562500 acc-top5=54.718750
Batch [0099]/[0303]: acc-top1=28.546875 acc-top5=53.343750
Batch [0149]/[0303]: acc-top1=30.135417 acc-top5=55.114583
Batch [0199]/[0303]: acc-top1=32.796875 acc-top5=57.507812
Batch [0249]/[0303]: acc-top1=32.106250 acc-top5=57.462500
Batch [0299]/[0303]: acc-top1=31.489583 acc-top5=57.901042
[Epoch 055] training: accuracy=30.284816	 loss=3.206532
[Epoch 055] speed: 116 samples/sec	time cost: 2177.608003
[Epoch 055] validation: acc-top1=31.502682 acc-top5=57.956889 loss=3.213295
Epoch[056] Batch [0049]/[3760]	Speed: 86.148681 samples/sec	 accuracy=31.781250	 loss=3.194268	 lr=0.095509
Epoch[056] Batch [0099]/[3760]	Speed: 117.507658 samples/sec	 accuracy=30.562500	 loss=3.212286	 lr=0.095504
Epoch[056] Batch [0149]/[3760]	Speed: 116.931242 samples/sec	 accuracy=30.781250	 loss=3.177249	 lr=0.095498
Epoch[056] Batch [0199]/[3760]	Speed: 117.265826 samples/sec	 accuracy=30.562500	 loss=3.169023	 lr=0.095493
Epoch[056] Batch [0249]/[3760]	Speed: 117.710658 samples/sec	 accuracy=30.893750	 loss=3.166767	 lr=0.095488
Epoch[056] Batch [0299]/[3760]	Speed: 116.375747 samples/sec	 accuracy=31.015625	 loss=3.161761	 lr=0.095482
Epoch[056] Batch [0349]/[3760]	Speed: 117.993414 samples/sec	 accuracy=30.986607	 loss=3.160586	 lr=0.095477
Epoch[056] Batch [0399]/[3760]	Speed: 117.315316 samples/sec	 accuracy=30.820313	 loss=3.168185	 lr=0.095471
Epoch[056] Batch [0449]/[3760]	Speed: 117.368875 samples/sec	 accuracy=30.760417	 loss=3.171626	 lr=0.095466
Epoch[056] Batch [0499]/[3760]	Speed: 117.292931 samples/sec	 accuracy=30.675000	 loss=3.179723	 lr=0.095461
Epoch[056] Batch [0549]/[3760]	Speed: 116.733771 samples/sec	 accuracy=30.667614	 loss=3.184934	 lr=0.095455
Epoch[056] Batch [0599]/[3760]	Speed: 117.638637 samples/sec	 accuracy=30.635417	 loss=3.188588	 lr=0.095450
Epoch[056] Batch [0649]/[3760]	Speed: 117.369418 samples/sec	 accuracy=30.644231	 loss=3.190446	 lr=0.095445
Epoch[056] Batch [0699]/[3760]	Speed: 116.110917 samples/sec	 accuracy=30.607143	 loss=3.185990	 lr=0.095439
Epoch[056] Batch [0749]/[3760]	Speed: 117.385477 samples/sec	 accuracy=30.652083	 loss=3.184376	 lr=0.095434
Epoch[056] Batch [0799]/[3760]	Speed: 117.017112 samples/sec	 accuracy=30.582031	 loss=3.185889	 lr=0.095428
Epoch[056] Batch [0849]/[3760]	Speed: 116.171412 samples/sec	 accuracy=30.636029	 loss=3.183745	 lr=0.095423
Epoch[056] Batch [0899]/[3760]	Speed: 117.954269 samples/sec	 accuracy=30.645833	 loss=3.184226	 lr=0.095418
Epoch[056] Batch [0949]/[3760]	Speed: 117.983192 samples/sec	 accuracy=30.565789	 loss=3.185109	 lr=0.095412
Epoch[056] Batch [0999]/[3760]	Speed: 117.130991 samples/sec	 accuracy=30.582812	 loss=3.185261	 lr=0.095407
Epoch[056] Batch [1049]/[3760]	Speed: 117.875416 samples/sec	 accuracy=30.547619	 loss=3.187396	 lr=0.095401
Epoch[056] Batch [1099]/[3760]	Speed: 117.789359 samples/sec	 accuracy=30.555398	 loss=3.188305	 lr=0.095396
Epoch[056] Batch [1149]/[3760]	Speed: 117.609187 samples/sec	 accuracy=30.485054	 loss=3.192846	 lr=0.095391
Epoch[056] Batch [1199]/[3760]	Speed: 117.697733 samples/sec	 accuracy=30.548177	 loss=3.190033	 lr=0.095385
Epoch[056] Batch [1249]/[3760]	Speed: 117.209542 samples/sec	 accuracy=30.608750	 loss=3.188433	 lr=0.095380
Epoch[056] Batch [1299]/[3760]	Speed: 117.292702 samples/sec	 accuracy=30.575721	 loss=3.189574	 lr=0.095374
Epoch[056] Batch [1349]/[3760]	Speed: 117.293633 samples/sec	 accuracy=30.578704	 loss=3.190747	 lr=0.095369
Epoch[056] Batch [1399]/[3760]	Speed: 116.917801 samples/sec	 accuracy=30.560268	 loss=3.192908	 lr=0.095364
Epoch[056] Batch [1449]/[3760]	Speed: 117.202540 samples/sec	 accuracy=30.526940	 loss=3.193779	 lr=0.095358
Epoch[056] Batch [1499]/[3760]	Speed: 117.625129 samples/sec	 accuracy=30.513542	 loss=3.194363	 lr=0.095353
Epoch[056] Batch [1549]/[3760]	Speed: 117.535367 samples/sec	 accuracy=30.518145	 loss=3.193322	 lr=0.095347
Epoch[056] Batch [1599]/[3760]	Speed: 117.262057 samples/sec	 accuracy=30.554688	 loss=3.191812	 lr=0.095342
Epoch[056] Batch [1649]/[3760]	Speed: 116.361369 samples/sec	 accuracy=30.548295	 loss=3.191623	 lr=0.095336
Epoch[056] Batch [1699]/[3760]	Speed: 117.600357 samples/sec	 accuracy=30.533088	 loss=3.192074	 lr=0.095331
Epoch[056] Batch [1749]/[3760]	Speed: 117.978087 samples/sec	 accuracy=30.522321	 loss=3.193097	 lr=0.095326
Epoch[056] Batch [1799]/[3760]	Speed: 117.935260 samples/sec	 accuracy=30.524306	 loss=3.194504	 lr=0.095320
Epoch[056] Batch [1849]/[3760]	Speed: 116.948277 samples/sec	 accuracy=30.474662	 loss=3.196193	 lr=0.095315
Epoch[056] Batch [1899]/[3760]	Speed: 117.880603 samples/sec	 accuracy=30.500000	 loss=3.197049	 lr=0.095309
Epoch[056] Batch [1949]/[3760]	Speed: 117.409582 samples/sec	 accuracy=30.464744	 loss=3.197621	 lr=0.095304
Epoch[056] Batch [1999]/[3760]	Speed: 117.628696 samples/sec	 accuracy=30.465625	 loss=3.198242	 lr=0.095298
Epoch[056] Batch [2049]/[3760]	Speed: 117.518907 samples/sec	 accuracy=30.443598	 loss=3.198987	 lr=0.095293
Epoch[056] Batch [2099]/[3760]	Speed: 118.222413 samples/sec	 accuracy=30.422619	 loss=3.199403	 lr=0.095287
Epoch[056] Batch [2149]/[3760]	Speed: 116.435099 samples/sec	 accuracy=30.396076	 loss=3.201649	 lr=0.095282
Epoch[056] Batch [2199]/[3760]	Speed: 117.279105 samples/sec	 accuracy=30.411222	 loss=3.200666	 lr=0.095276
Epoch[056] Batch [2249]/[3760]	Speed: 117.428362 samples/sec	 accuracy=30.427778	 loss=3.200374	 lr=0.095271
Epoch[056] Batch [2299]/[3760]	Speed: 117.991973 samples/sec	 accuracy=30.461957	 loss=3.199374	 lr=0.095265
Epoch[056] Batch [2349]/[3760]	Speed: 116.501055 samples/sec	 accuracy=30.456782	 loss=3.200238	 lr=0.095260
Epoch[056] Batch [2399]/[3760]	Speed: 117.830641 samples/sec	 accuracy=30.424479	 loss=3.200902	 lr=0.095255
Epoch[056] Batch [2449]/[3760]	Speed: 117.106984 samples/sec	 accuracy=30.449617	 loss=3.199618	 lr=0.095249
Epoch[056] Batch [2499]/[3760]	Speed: 116.646481 samples/sec	 accuracy=30.445625	 loss=3.199788	 lr=0.095244
Epoch[056] Batch [2549]/[3760]	Speed: 117.140444 samples/sec	 accuracy=30.459559	 loss=3.198599	 lr=0.095238
Epoch[056] Batch [2599]/[3760]	Speed: 117.421441 samples/sec	 accuracy=30.475361	 loss=3.197805	 lr=0.095233
Epoch[056] Batch [2649]/[3760]	Speed: 117.023546 samples/sec	 accuracy=30.492335	 loss=3.197455	 lr=0.095227
Epoch[056] Batch [2699]/[3760]	Speed: 117.034168 samples/sec	 accuracy=30.495370	 loss=3.197261	 lr=0.095222
Epoch[056] Batch [2749]/[3760]	Speed: 117.673708 samples/sec	 accuracy=30.484659	 loss=3.197630	 lr=0.095216
Epoch[056] Batch [2799]/[3760]	Speed: 117.845686 samples/sec	 accuracy=30.501116	 loss=3.198039	 lr=0.095211
Epoch[056] Batch [2849]/[3760]	Speed: 117.952760 samples/sec	 accuracy=30.498355	 loss=3.198573	 lr=0.095205
Epoch[056] Batch [2899]/[3760]	Speed: 117.126654 samples/sec	 accuracy=30.483836	 loss=3.198824	 lr=0.095200
Epoch[056] Batch [2949]/[3760]	Speed: 117.379495 samples/sec	 accuracy=30.499470	 loss=3.197696	 lr=0.095194
Epoch[056] Batch [2999]/[3760]	Speed: 117.149516 samples/sec	 accuracy=30.498958	 loss=3.198091	 lr=0.095188
Epoch[056] Batch [3049]/[3760]	Speed: 116.942796 samples/sec	 accuracy=30.490779	 loss=3.198819	 lr=0.095183
Epoch[056] Batch [3099]/[3760]	Speed: 117.234475 samples/sec	 accuracy=30.483871	 loss=3.199988	 lr=0.095177
Epoch[056] Batch [3149]/[3760]	Speed: 117.930416 samples/sec	 accuracy=30.469246	 loss=3.200865	 lr=0.095172
Epoch[056] Batch [3199]/[3760]	Speed: 116.813134 samples/sec	 accuracy=30.485352	 loss=3.200564	 lr=0.095166
Epoch[056] Batch [3249]/[3760]	Speed: 117.512519 samples/sec	 accuracy=30.489423	 loss=3.200153	 lr=0.095161
Epoch[056] Batch [3299]/[3760]	Speed: 117.627514 samples/sec	 accuracy=30.482955	 loss=3.200560	 lr=0.095155
Epoch[056] Batch [3349]/[3760]	Speed: 117.531655 samples/sec	 accuracy=30.481343	 loss=3.200184	 lr=0.095150
Epoch[056] Batch [3399]/[3760]	Speed: 117.439523 samples/sec	 accuracy=30.475643	 loss=3.199921	 lr=0.095144
Epoch[056] Batch [3449]/[3760]	Speed: 117.284815 samples/sec	 accuracy=30.471467	 loss=3.200527	 lr=0.095139
Epoch[056] Batch [3499]/[3760]	Speed: 117.388347 samples/sec	 accuracy=30.465625	 loss=3.200452	 lr=0.095133
Epoch[056] Batch [3549]/[3760]	Speed: 116.788239 samples/sec	 accuracy=30.455546	 loss=3.200545	 lr=0.095128
Epoch[056] Batch [3599]/[3760]	Speed: 118.249560 samples/sec	 accuracy=30.462674	 loss=3.199826	 lr=0.095122
Epoch[056] Batch [3649]/[3760]	Speed: 117.321661 samples/sec	 accuracy=30.451199	 loss=3.200127	 lr=0.095116
Epoch[056] Batch [3699]/[3760]	Speed: 117.043391 samples/sec	 accuracy=30.448902	 loss=3.200416	 lr=0.095111
Epoch[056] Batch [3749]/[3760]	Speed: 120.005660 samples/sec	 accuracy=30.447083	 loss=3.201484	 lr=0.095105
Batch [0049]/[0303]: acc-top1=30.187500 acc-top5=57.281250
Batch [0099]/[0303]: acc-top1=28.406250 acc-top5=55.921875
Batch [0149]/[0303]: acc-top1=30.541667 acc-top5=57.427083
Batch [0199]/[0303]: acc-top1=32.625000 acc-top5=59.414062
Batch [0249]/[0303]: acc-top1=31.981250 acc-top5=58.818750
Batch [0299]/[0303]: acc-top1=31.515625 acc-top5=58.453125
[Epoch 056] training: accuracy=30.443401	 loss=3.201488
[Epoch 056] speed: 116 samples/sec	time cost: 2180.067435
[Epoch 056] validation: acc-top1=31.580033 acc-top5=58.482880 loss=3.224316
Epoch[057] Batch [0049]/[3760]	Speed: 87.557685 samples/sec	 accuracy=28.968750	 loss=3.258819	 lr=0.095099
Epoch[057] Batch [0099]/[3760]	Speed: 117.800941 samples/sec	 accuracy=29.453125	 loss=3.220325	 lr=0.095093
Epoch[057] Batch [0149]/[3760]	Speed: 116.773473 samples/sec	 accuracy=30.135417	 loss=3.186969	 lr=0.095088
Epoch[057] Batch [0199]/[3760]	Speed: 116.919761 samples/sec	 accuracy=30.234375	 loss=3.187864	 lr=0.095082
Epoch[057] Batch [0249]/[3760]	Speed: 118.161491 samples/sec	 accuracy=30.631250	 loss=3.179562	 lr=0.095076
Epoch[057] Batch [0299]/[3760]	Speed: 117.062585 samples/sec	 accuracy=30.640625	 loss=3.173419	 lr=0.095071
Epoch[057] Batch [0349]/[3760]	Speed: 117.198530 samples/sec	 accuracy=30.705357	 loss=3.171131	 lr=0.095065
Epoch[057] Batch [0399]/[3760]	Speed: 116.869993 samples/sec	 accuracy=30.687500	 loss=3.174958	 lr=0.095060
Epoch[057] Batch [0449]/[3760]	Speed: 117.496244 samples/sec	 accuracy=30.604167	 loss=3.181474	 lr=0.095054
Epoch[057] Batch [0499]/[3760]	Speed: 117.580091 samples/sec	 accuracy=30.600000	 loss=3.185885	 lr=0.095048
Epoch[057] Batch [0549]/[3760]	Speed: 116.631715 samples/sec	 accuracy=30.568182	 loss=3.186543	 lr=0.095043
Epoch[057] Batch [0599]/[3760]	Speed: 117.355645 samples/sec	 accuracy=30.562500	 loss=3.183290	 lr=0.095037
Epoch[057] Batch [0649]/[3760]	Speed: 116.688534 samples/sec	 accuracy=30.521635	 loss=3.183733	 lr=0.095032
Epoch[057] Batch [0699]/[3760]	Speed: 117.126400 samples/sec	 accuracy=30.466518	 loss=3.184835	 lr=0.095026
Epoch[057] Batch [0749]/[3760]	Speed: 116.982670 samples/sec	 accuracy=30.479167	 loss=3.184110	 lr=0.095020
Epoch[057] Batch [0799]/[3760]	Speed: 116.924245 samples/sec	 accuracy=30.406250	 loss=3.184687	 lr=0.095015
Epoch[057] Batch [0849]/[3760]	Speed: 116.647082 samples/sec	 accuracy=30.371324	 loss=3.186311	 lr=0.095009
Epoch[057] Batch [0899]/[3760]	Speed: 117.201658 samples/sec	 accuracy=30.411458	 loss=3.184712	 lr=0.095004
Epoch[057] Batch [0949]/[3760]	Speed: 117.262889 samples/sec	 accuracy=30.407895	 loss=3.184859	 lr=0.094998
Epoch[057] Batch [0999]/[3760]	Speed: 116.852598 samples/sec	 accuracy=30.410938	 loss=3.182862	 lr=0.094992
Epoch[057] Batch [1049]/[3760]	Speed: 117.732100 samples/sec	 accuracy=30.436012	 loss=3.181978	 lr=0.094987
Epoch[057] Batch [1099]/[3760]	Speed: 117.455395 samples/sec	 accuracy=30.478693	 loss=3.180344	 lr=0.094981
Epoch[057] Batch [1149]/[3760]	Speed: 117.736149 samples/sec	 accuracy=30.445652	 loss=3.183396	 lr=0.094975
Epoch[057] Batch [1199]/[3760]	Speed: 116.844140 samples/sec	 accuracy=30.427083	 loss=3.184563	 lr=0.094970
Epoch[057] Batch [1249]/[3760]	Speed: 117.918453 samples/sec	 accuracy=30.356250	 loss=3.185900	 lr=0.094964
Epoch[057] Batch [1299]/[3760]	Speed: 116.975450 samples/sec	 accuracy=30.330529	 loss=3.186963	 lr=0.094959
Epoch[057] Batch [1349]/[3760]	Speed: 117.559611 samples/sec	 accuracy=30.336806	 loss=3.185225	 lr=0.094953
Epoch[057] Batch [1399]/[3760]	Speed: 117.488837 samples/sec	 accuracy=30.357143	 loss=3.186527	 lr=0.094947
Epoch[057] Batch [1449]/[3760]	Speed: 117.534034 samples/sec	 accuracy=30.369612	 loss=3.187329	 lr=0.094942
Epoch[057] Batch [1499]/[3760]	Speed: 117.336630 samples/sec	 accuracy=30.409375	 loss=3.186541	 lr=0.094936
Epoch[057] Batch [1549]/[3760]	Speed: 116.846403 samples/sec	 accuracy=30.439516	 loss=3.187184	 lr=0.094930
Epoch[057] Batch [1599]/[3760]	Speed: 117.410946 samples/sec	 accuracy=30.472656	 loss=3.185699	 lr=0.094925
Epoch[057] Batch [1649]/[3760]	Speed: 116.980718 samples/sec	 accuracy=30.513258	 loss=3.184893	 lr=0.094919
Epoch[057] Batch [1699]/[3760]	Speed: 117.416096 samples/sec	 accuracy=30.544118	 loss=3.184904	 lr=0.094913
Epoch[057] Batch [1749]/[3760]	Speed: 117.145031 samples/sec	 accuracy=30.560714	 loss=3.185829	 lr=0.094908
Epoch[057] Batch [1799]/[3760]	Speed: 117.745823 samples/sec	 accuracy=30.559028	 loss=3.185637	 lr=0.094902
Epoch[057] Batch [1849]/[3760]	Speed: 117.242053 samples/sec	 accuracy=30.550676	 loss=3.186664	 lr=0.094896
Epoch[057] Batch [1899]/[3760]	Speed: 117.547613 samples/sec	 accuracy=30.550164	 loss=3.187224	 lr=0.094891
Epoch[057] Batch [1949]/[3760]	Speed: 117.074473 samples/sec	 accuracy=30.519231	 loss=3.188023	 lr=0.094885
Epoch[057] Batch [1999]/[3760]	Speed: 117.125473 samples/sec	 accuracy=30.503906	 loss=3.188977	 lr=0.094879
Epoch[057] Batch [2049]/[3760]	Speed: 117.174373 samples/sec	 accuracy=30.469512	 loss=3.190576	 lr=0.094874
Epoch[057] Batch [2099]/[3760]	Speed: 117.817564 samples/sec	 accuracy=30.492560	 loss=3.188971	 lr=0.094868
Epoch[057] Batch [2149]/[3760]	Speed: 117.394983 samples/sec	 accuracy=30.470930	 loss=3.188832	 lr=0.094862
Epoch[057] Batch [2199]/[3760]	Speed: 116.755092 samples/sec	 accuracy=30.470881	 loss=3.189234	 lr=0.094856
Epoch[057] Batch [2249]/[3760]	Speed: 117.298338 samples/sec	 accuracy=30.456250	 loss=3.188711	 lr=0.094851
Epoch[057] Batch [2299]/[3760]	Speed: 117.964012 samples/sec	 accuracy=30.470109	 loss=3.188532	 lr=0.094845
Epoch[057] Batch [2349]/[3760]	Speed: 116.423459 samples/sec	 accuracy=30.465426	 loss=3.188444	 lr=0.094839
Epoch[057] Batch [2399]/[3760]	Speed: 117.933076 samples/sec	 accuracy=30.479818	 loss=3.188188	 lr=0.094834
Epoch[057] Batch [2449]/[3760]	Speed: 117.636947 samples/sec	 accuracy=30.474490	 loss=3.188995	 lr=0.094828
Epoch[057] Batch [2499]/[3760]	Speed: 117.848713 samples/sec	 accuracy=30.492500	 loss=3.189895	 lr=0.094822
Epoch[057] Batch [2549]/[3760]	Speed: 116.875302 samples/sec	 accuracy=30.497549	 loss=3.190406	 lr=0.094816
Epoch[057] Batch [2599]/[3760]	Speed: 117.560432 samples/sec	 accuracy=30.500601	 loss=3.190055	 lr=0.094811
Epoch[057] Batch [2649]/[3760]	Speed: 117.422916 samples/sec	 accuracy=30.528892	 loss=3.190402	 lr=0.094805
Epoch[057] Batch [2699]/[3760]	Speed: 117.415889 samples/sec	 accuracy=30.521412	 loss=3.190617	 lr=0.094799
Epoch[057] Batch [2749]/[3760]	Speed: 117.862084 samples/sec	 accuracy=30.514205	 loss=3.191305	 lr=0.094794
Epoch[057] Batch [2799]/[3760]	Speed: 116.862112 samples/sec	 accuracy=30.495536	 loss=3.192412	 lr=0.094788
Epoch[057] Batch [2849]/[3760]	Speed: 116.791331 samples/sec	 accuracy=30.491228	 loss=3.192515	 lr=0.094782
Epoch[057] Batch [2899]/[3760]	Speed: 117.391421 samples/sec	 accuracy=30.476293	 loss=3.192736	 lr=0.094776
Epoch[057] Batch [2949]/[3760]	Speed: 117.429738 samples/sec	 accuracy=30.462924	 loss=3.193278	 lr=0.094771
Epoch[057] Batch [2999]/[3760]	Speed: 117.696151 samples/sec	 accuracy=30.475521	 loss=3.192476	 lr=0.094765
Epoch[057] Batch [3049]/[3760]	Speed: 117.264126 samples/sec	 accuracy=30.476434	 loss=3.192483	 lr=0.094759
Epoch[057] Batch [3099]/[3760]	Speed: 117.015335 samples/sec	 accuracy=30.493448	 loss=3.192277	 lr=0.094753
Epoch[057] Batch [3149]/[3760]	Speed: 117.079960 samples/sec	 accuracy=30.468254	 loss=3.193593	 lr=0.094748
Epoch[057] Batch [3199]/[3760]	Speed: 118.073979 samples/sec	 accuracy=30.455078	 loss=3.194512	 lr=0.094742
Epoch[057] Batch [3249]/[3760]	Speed: 117.622687 samples/sec	 accuracy=30.441346	 loss=3.194993	 lr=0.094736
Epoch[057] Batch [3299]/[3760]	Speed: 117.305267 samples/sec	 accuracy=30.443182	 loss=3.195200	 lr=0.094730
Epoch[057] Batch [3349]/[3760]	Speed: 117.767590 samples/sec	 accuracy=30.438899	 loss=3.194741	 lr=0.094725
Epoch[057] Batch [3399]/[3760]	Speed: 117.938940 samples/sec	 accuracy=30.427849	 loss=3.195034	 lr=0.094719
Epoch[057] Batch [3449]/[3760]	Speed: 117.529031 samples/sec	 accuracy=30.429348	 loss=3.194205	 lr=0.094713
Epoch[057] Batch [3499]/[3760]	Speed: 116.782173 samples/sec	 accuracy=30.432143	 loss=3.194149	 lr=0.094707
Epoch[057] Batch [3549]/[3760]	Speed: 117.773600 samples/sec	 accuracy=30.437500	 loss=3.194449	 lr=0.094702
Epoch[057] Batch [3599]/[3760]	Speed: 117.760148 samples/sec	 accuracy=30.470920	 loss=3.193188	 lr=0.094696
Epoch[057] Batch [3649]/[3760]	Speed: 117.748508 samples/sec	 accuracy=30.447346	 loss=3.194993	 lr=0.094690
Epoch[057] Batch [3699]/[3760]	Speed: 117.384411 samples/sec	 accuracy=30.425253	 loss=3.196165	 lr=0.094684
Epoch[057] Batch [3749]/[3760]	Speed: 120.419841 samples/sec	 accuracy=30.411667	 loss=3.196408	 lr=0.094678
Batch [0049]/[0303]: acc-top1=29.843750 acc-top5=58.156250
Batch [0099]/[0303]: acc-top1=28.218750 acc-top5=56.250000
Batch [0149]/[0303]: acc-top1=29.020833 acc-top5=55.677083
Batch [0199]/[0303]: acc-top1=31.500000 acc-top5=57.945312
Batch [0249]/[0303]: acc-top1=30.681250 acc-top5=57.431250
Batch [0299]/[0303]: acc-top1=30.296875 acc-top5=57.411458
[Epoch 057] training: accuracy=30.408910	 loss=3.196608
[Epoch 057] speed: 116 samples/sec	time cost: 2180.156543
[Epoch 057] validation: acc-top1=30.311469 acc-top5=57.492781 loss=3.295557
Epoch[058] Batch [0049]/[3759]	Speed: 87.659895 samples/sec	 accuracy=30.812500	 loss=3.150746	 lr=0.094671
Epoch[058] Batch [0099]/[3759]	Speed: 117.082172 samples/sec	 accuracy=31.359375	 loss=3.140244	 lr=0.094666
Epoch[058] Batch [0149]/[3759]	Speed: 117.029758 samples/sec	 accuracy=31.166667	 loss=3.137804	 lr=0.094660
Epoch[058] Batch [0199]/[3759]	Speed: 116.911192 samples/sec	 accuracy=31.210938	 loss=3.137418	 lr=0.094654
Epoch[058] Batch [0249]/[3759]	Speed: 117.077270 samples/sec	 accuracy=31.237500	 loss=3.137390	 lr=0.094648
Epoch[058] Batch [0299]/[3759]	Speed: 116.142081 samples/sec	 accuracy=31.239583	 loss=3.143634	 lr=0.094642
Epoch[058] Batch [0349]/[3759]	Speed: 117.552113 samples/sec	 accuracy=31.035714	 loss=3.153933	 lr=0.094637
Epoch[058] Batch [0399]/[3759]	Speed: 117.182889 samples/sec	 accuracy=31.070313	 loss=3.151719	 lr=0.094631
Epoch[058] Batch [0449]/[3759]	Speed: 116.319371 samples/sec	 accuracy=31.038194	 loss=3.157589	 lr=0.094625
Epoch[058] Batch [0499]/[3759]	Speed: 116.809970 samples/sec	 accuracy=31.031250	 loss=3.161204	 lr=0.094619
Epoch[058] Batch [0549]/[3759]	Speed: 117.231704 samples/sec	 accuracy=30.963068	 loss=3.162797	 lr=0.094613
Epoch[058] Batch [0599]/[3759]	Speed: 116.553712 samples/sec	 accuracy=30.953125	 loss=3.163614	 lr=0.094608
Epoch[058] Batch [0649]/[3759]	Speed: 116.700133 samples/sec	 accuracy=31.086538	 loss=3.162597	 lr=0.094602
Epoch[058] Batch [0699]/[3759]	Speed: 118.061965 samples/sec	 accuracy=30.950893	 loss=3.167927	 lr=0.094596
Epoch[058] Batch [0749]/[3759]	Speed: 116.667965 samples/sec	 accuracy=30.950000	 loss=3.167566	 lr=0.094590
Epoch[058] Batch [0799]/[3759]	Speed: 117.338318 samples/sec	 accuracy=30.923828	 loss=3.167548	 lr=0.094584
Epoch[058] Batch [0849]/[3759]	Speed: 116.989748 samples/sec	 accuracy=30.887868	 loss=3.170227	 lr=0.094578
Epoch[058] Batch [0899]/[3759]	Speed: 117.774168 samples/sec	 accuracy=30.876736	 loss=3.170182	 lr=0.094573
Epoch[058] Batch [0949]/[3759]	Speed: 117.328233 samples/sec	 accuracy=30.922697	 loss=3.169492	 lr=0.094567
Epoch[058] Batch [0999]/[3759]	Speed: 117.109681 samples/sec	 accuracy=30.898437	 loss=3.171511	 lr=0.094561
Epoch[058] Batch [1049]/[3759]	Speed: 117.144984 samples/sec	 accuracy=30.894345	 loss=3.173680	 lr=0.094555
Epoch[058] Batch [1099]/[3759]	Speed: 116.887933 samples/sec	 accuracy=30.825284	 loss=3.176132	 lr=0.094549
Epoch[058] Batch [1149]/[3759]	Speed: 117.305669 samples/sec	 accuracy=30.819293	 loss=3.176555	 lr=0.094543
Epoch[058] Batch [1199]/[3759]	Speed: 116.546495 samples/sec	 accuracy=30.795573	 loss=3.176109	 lr=0.094537
Epoch[058] Batch [1249]/[3759]	Speed: 116.919172 samples/sec	 accuracy=30.830000	 loss=3.175260	 lr=0.094532
Epoch[058] Batch [1299]/[3759]	Speed: 117.085942 samples/sec	 accuracy=30.822115	 loss=3.176084	 lr=0.094526
Epoch[058] Batch [1349]/[3759]	Speed: 117.136029 samples/sec	 accuracy=30.806713	 loss=3.177157	 lr=0.094520
Epoch[058] Batch [1399]/[3759]	Speed: 116.485230 samples/sec	 accuracy=30.797991	 loss=3.176254	 lr=0.094514
Epoch[058] Batch [1449]/[3759]	Speed: 117.493248 samples/sec	 accuracy=30.782328	 loss=3.175980	 lr=0.094508
Epoch[058] Batch [1499]/[3759]	Speed: 116.364436 samples/sec	 accuracy=30.751042	 loss=3.177886	 lr=0.094502
Epoch[058] Batch [1549]/[3759]	Speed: 117.551914 samples/sec	 accuracy=30.706653	 loss=3.179672	 lr=0.094496
Epoch[058] Batch [1599]/[3759]	Speed: 117.898646 samples/sec	 accuracy=30.674805	 loss=3.180546	 lr=0.094490
Epoch[058] Batch [1649]/[3759]	Speed: 116.745645 samples/sec	 accuracy=30.681818	 loss=3.179920	 lr=0.094485
Epoch[058] Batch [1699]/[3759]	Speed: 117.260642 samples/sec	 accuracy=30.670037	 loss=3.181513	 lr=0.094479
Epoch[058] Batch [1749]/[3759]	Speed: 117.476329 samples/sec	 accuracy=30.644643	 loss=3.185790	 lr=0.094473
Epoch[058] Batch [1799]/[3759]	Speed: 116.961602 samples/sec	 accuracy=30.637153	 loss=3.185738	 lr=0.094467
Epoch[058] Batch [1849]/[3759]	Speed: 117.318217 samples/sec	 accuracy=30.682432	 loss=3.184377	 lr=0.094461
Epoch[058] Batch [1899]/[3759]	Speed: 117.877718 samples/sec	 accuracy=30.658717	 loss=3.186416	 lr=0.094455
Epoch[058] Batch [1949]/[3759]	Speed: 117.339510 samples/sec	 accuracy=30.637019	 loss=3.187709	 lr=0.094449
Epoch[058] Batch [1999]/[3759]	Speed: 117.844310 samples/sec	 accuracy=30.600781	 loss=3.189078	 lr=0.094443
Epoch[058] Batch [2049]/[3759]	Speed: 117.805484 samples/sec	 accuracy=30.623476	 loss=3.188036	 lr=0.094437
Epoch[058] Batch [2099]/[3759]	Speed: 116.952529 samples/sec	 accuracy=30.643601	 loss=3.186382	 lr=0.094431
Epoch[058] Batch [2149]/[3759]	Speed: 116.995538 samples/sec	 accuracy=30.657703	 loss=3.186036	 lr=0.094425
Epoch[058] Batch [2199]/[3759]	Speed: 116.590307 samples/sec	 accuracy=30.671165	 loss=3.186445	 lr=0.094420
Epoch[058] Batch [2249]/[3759]	Speed: 117.346946 samples/sec	 accuracy=30.670139	 loss=3.185710	 lr=0.094414
Epoch[058] Batch [2299]/[3759]	Speed: 116.894806 samples/sec	 accuracy=30.703125	 loss=3.184536	 lr=0.094408
Epoch[058] Batch [2349]/[3759]	Speed: 116.738997 samples/sec	 accuracy=30.732048	 loss=3.184676	 lr=0.094402
Epoch[058] Batch [2399]/[3759]	Speed: 117.937915 samples/sec	 accuracy=30.703125	 loss=3.186520	 lr=0.094396
Epoch[058] Batch [2449]/[3759]	Speed: 117.516095 samples/sec	 accuracy=30.717474	 loss=3.186907	 lr=0.094390
Epoch[058] Batch [2499]/[3759]	Speed: 117.198785 samples/sec	 accuracy=30.717500	 loss=3.187553	 lr=0.094384
Epoch[058] Batch [2549]/[3759]	Speed: 116.657321 samples/sec	 accuracy=30.717525	 loss=3.187303	 lr=0.094378
Epoch[058] Batch [2599]/[3759]	Speed: 116.782752 samples/sec	 accuracy=30.734375	 loss=3.185508	 lr=0.094372
Epoch[058] Batch [2649]/[3759]	Speed: 116.458608 samples/sec	 accuracy=30.745283	 loss=3.185863	 lr=0.094366
Epoch[058] Batch [2699]/[3759]	Speed: 116.136897 samples/sec	 accuracy=30.736690	 loss=3.186967	 lr=0.094360
Epoch[058] Batch [2749]/[3759]	Speed: 117.866100 samples/sec	 accuracy=30.709659	 loss=3.188293	 lr=0.094354
Epoch[058] Batch [2799]/[3759]	Speed: 117.195783 samples/sec	 accuracy=30.692522	 loss=3.189068	 lr=0.094348
Epoch[058] Batch [2849]/[3759]	Speed: 117.029450 samples/sec	 accuracy=30.702303	 loss=3.188928	 lr=0.094342
Epoch[058] Batch [2899]/[3759]	Speed: 117.785558 samples/sec	 accuracy=30.684806	 loss=3.189372	 lr=0.094336
Epoch[058] Batch [2949]/[3759]	Speed: 118.011812 samples/sec	 accuracy=30.673199	 loss=3.189850	 lr=0.094330
Epoch[058] Batch [2999]/[3759]	Speed: 117.801557 samples/sec	 accuracy=30.686458	 loss=3.189395	 lr=0.094324
Epoch[058] Batch [3049]/[3759]	Speed: 117.394653 samples/sec	 accuracy=30.714652	 loss=3.188639	 lr=0.094318
Epoch[058] Batch [3099]/[3759]	Speed: 117.238433 samples/sec	 accuracy=30.697077	 loss=3.188863	 lr=0.094312
Epoch[058] Batch [3149]/[3759]	Speed: 117.464505 samples/sec	 accuracy=30.700397	 loss=3.189229	 lr=0.094307
Epoch[058] Batch [3199]/[3759]	Speed: 116.633727 samples/sec	 accuracy=30.700195	 loss=3.190836	 lr=0.094301
Epoch[058] Batch [3249]/[3759]	Speed: 116.974392 samples/sec	 accuracy=30.689423	 loss=3.190916	 lr=0.094295
Epoch[058] Batch [3299]/[3759]	Speed: 117.256056 samples/sec	 accuracy=30.687973	 loss=3.191344	 lr=0.094289
Epoch[058] Batch [3349]/[3759]	Speed: 117.205655 samples/sec	 accuracy=30.683302	 loss=3.191786	 lr=0.094283
Epoch[058] Batch [3399]/[3759]	Speed: 116.879665 samples/sec	 accuracy=30.660386	 loss=3.192800	 lr=0.094277
Epoch[058] Batch [3449]/[3759]	Speed: 117.311452 samples/sec	 accuracy=30.653080	 loss=3.193058	 lr=0.094271
Epoch[058] Batch [3499]/[3759]	Speed: 118.326180 samples/sec	 accuracy=30.658482	 loss=3.192835	 lr=0.094265
Epoch[058] Batch [3549]/[3759]	Speed: 116.451598 samples/sec	 accuracy=30.632923	 loss=3.193598	 lr=0.094259
Epoch[058] Batch [3599]/[3759]	Speed: 117.423889 samples/sec	 accuracy=30.618490	 loss=3.194516	 lr=0.094253
Epoch[058] Batch [3649]/[3759]	Speed: 117.011801 samples/sec	 accuracy=30.608733	 loss=3.194517	 lr=0.094247
Epoch[058] Batch [3699]/[3759]	Speed: 116.968765 samples/sec	 accuracy=30.602618	 loss=3.194493	 lr=0.094241
Epoch[058] Batch [3749]/[3759]	Speed: 121.002506 samples/sec	 accuracy=30.588333	 loss=3.195398	 lr=0.094235
Batch [0049]/[0303]: acc-top1=33.187500 acc-top5=59.062500
Batch [0099]/[0303]: acc-top1=29.968750 acc-top5=56.234375
Batch [0149]/[0303]: acc-top1=31.468750 acc-top5=57.447917
Batch [0199]/[0303]: acc-top1=34.046875 acc-top5=59.960938
Batch [0249]/[0303]: acc-top1=33.643750 acc-top5=59.837500
Batch [0299]/[0303]: acc-top1=33.166667 acc-top5=59.984375
[Epoch 058] training: accuracy=30.590333	 loss=3.195310
[Epoch 058] speed: 116 samples/sec	time cost: 2182.524387
[Epoch 058] validation: acc-top1=33.111592 acc-top5=59.988655 loss=3.124831
Epoch[059] Batch [0049]/[3760]	Speed: 88.681670 samples/sec	 accuracy=31.250000	 loss=3.152421	 lr=0.094227
Epoch[059] Batch [0099]/[3760]	Speed: 116.497917 samples/sec	 accuracy=31.671875	 loss=3.142973	 lr=0.094221
Epoch[059] Batch [0149]/[3760]	Speed: 116.741670 samples/sec	 accuracy=31.656250	 loss=3.138755	 lr=0.094215
Epoch[059] Batch [0199]/[3760]	Speed: 117.500151 samples/sec	 accuracy=31.546875	 loss=3.140151	 lr=0.094209
Epoch[059] Batch [0249]/[3760]	Speed: 117.277863 samples/sec	 accuracy=31.368750	 loss=3.144168	 lr=0.094203
Epoch[059] Batch [0299]/[3760]	Speed: 117.302726 samples/sec	 accuracy=31.244792	 loss=3.148479	 lr=0.094197
Epoch[059] Batch [0349]/[3760]	Speed: 117.372554 samples/sec	 accuracy=31.111607	 loss=3.155719	 lr=0.094191
Epoch[059] Batch [0399]/[3760]	Speed: 116.459335 samples/sec	 accuracy=31.136719	 loss=3.152367	 lr=0.094185
Epoch[059] Batch [0449]/[3760]	Speed: 117.719470 samples/sec	 accuracy=31.048611	 loss=3.156512	 lr=0.094179
Epoch[059] Batch [0499]/[3760]	Speed: 117.202715 samples/sec	 accuracy=31.078125	 loss=3.161538	 lr=0.094173
Epoch[059] Batch [0549]/[3760]	Speed: 117.088093 samples/sec	 accuracy=31.116477	 loss=3.164294	 lr=0.094167
Epoch[059] Batch [0599]/[3760]	Speed: 117.207748 samples/sec	 accuracy=31.197917	 loss=3.163199	 lr=0.094161
Epoch[059] Batch [0649]/[3760]	Speed: 116.621020 samples/sec	 accuracy=31.120192	 loss=3.169134	 lr=0.094155
Epoch[059] Batch [0699]/[3760]	Speed: 117.583736 samples/sec	 accuracy=31.191964	 loss=3.169999	 lr=0.094149
Epoch[059] Batch [0749]/[3760]	Speed: 118.235127 samples/sec	 accuracy=31.154167	 loss=3.171668	 lr=0.094143
Epoch[059] Batch [0799]/[3760]	Speed: 117.712672 samples/sec	 accuracy=31.144531	 loss=3.171175	 lr=0.094137
Epoch[059] Batch [0849]/[3760]	Speed: 116.776855 samples/sec	 accuracy=31.132353	 loss=3.172560	 lr=0.094131
Epoch[059] Batch [0899]/[3760]	Speed: 117.361414 samples/sec	 accuracy=31.052083	 loss=3.177017	 lr=0.094125
Epoch[059] Batch [0949]/[3760]	Speed: 117.768625 samples/sec	 accuracy=30.995066	 loss=3.177901	 lr=0.094119
Epoch[059] Batch [0999]/[3760]	Speed: 116.988594 samples/sec	 accuracy=30.984375	 loss=3.179072	 lr=0.094113
Epoch[059] Batch [1049]/[3760]	Speed: 117.290804 samples/sec	 accuracy=30.953869	 loss=3.180558	 lr=0.094107
Epoch[059] Batch [1099]/[3760]	Speed: 115.981388 samples/sec	 accuracy=30.978693	 loss=3.180252	 lr=0.094101
Epoch[059] Batch [1149]/[3760]	Speed: 116.816258 samples/sec	 accuracy=31.006793	 loss=3.179666	 lr=0.094094
Epoch[059] Batch [1199]/[3760]	Speed: 116.299523 samples/sec	 accuracy=31.011719	 loss=3.178102	 lr=0.094088
Epoch[059] Batch [1249]/[3760]	Speed: 117.883872 samples/sec	 accuracy=30.982500	 loss=3.179359	 lr=0.094082
Epoch[059] Batch [1299]/[3760]	Speed: 117.479830 samples/sec	 accuracy=30.956731	 loss=3.180960	 lr=0.094076
Epoch[059] Batch [1349]/[3760]	Speed: 117.678463 samples/sec	 accuracy=30.921296	 loss=3.182140	 lr=0.094070
Epoch[059] Batch [1399]/[3760]	Speed: 117.424199 samples/sec	 accuracy=30.943080	 loss=3.182704	 lr=0.094064
Epoch[059] Batch [1449]/[3760]	Speed: 116.769398 samples/sec	 accuracy=30.954741	 loss=3.182835	 lr=0.094058
Epoch[059] Batch [1499]/[3760]	Speed: 116.736077 samples/sec	 accuracy=30.956250	 loss=3.183963	 lr=0.094052
Epoch[059] Batch [1549]/[3760]	Speed: 116.843491 samples/sec	 accuracy=30.939516	 loss=3.184667	 lr=0.094046
Epoch[059] Batch [1599]/[3760]	Speed: 117.455861 samples/sec	 accuracy=30.912109	 loss=3.187256	 lr=0.094040
Epoch[059] Batch [1649]/[3760]	Speed: 117.540543 samples/sec	 accuracy=30.906250	 loss=3.186085	 lr=0.094033
Epoch[059] Batch [1699]/[3760]	Speed: 117.950075 samples/sec	 accuracy=30.862132	 loss=3.187927	 lr=0.094027
Epoch[059] Batch [1749]/[3760]	Speed: 117.884111 samples/sec	 accuracy=30.891071	 loss=3.186382	 lr=0.094021
Epoch[059] Batch [1799]/[3760]	Speed: 118.680659 samples/sec	 accuracy=30.901910	 loss=3.186344	 lr=0.094015
Epoch[059] Batch [1849]/[3760]	Speed: 117.138823 samples/sec	 accuracy=30.861486	 loss=3.187912	 lr=0.094009
Epoch[059] Batch [1899]/[3760]	Speed: 117.777562 samples/sec	 accuracy=30.847862	 loss=3.188570	 lr=0.094003
Epoch[059] Batch [1949]/[3760]	Speed: 117.499129 samples/sec	 accuracy=30.862179	 loss=3.187713	 lr=0.093997
Epoch[059] Batch [1999]/[3760]	Speed: 117.726508 samples/sec	 accuracy=30.842969	 loss=3.187769	 lr=0.093991
Epoch[059] Batch [2049]/[3760]	Speed: 116.942329 samples/sec	 accuracy=30.811738	 loss=3.189398	 lr=0.093984
Epoch[059] Batch [2099]/[3760]	Speed: 117.056054 samples/sec	 accuracy=30.775298	 loss=3.190571	 lr=0.093978
Epoch[059] Batch [2149]/[3760]	Speed: 117.983218 samples/sec	 accuracy=30.747820	 loss=3.191885	 lr=0.093972
Epoch[059] Batch [2199]/[3760]	Speed: 118.002780 samples/sec	 accuracy=30.749290	 loss=3.192130	 lr=0.093966
Epoch[059] Batch [2249]/[3760]	Speed: 117.338537 samples/sec	 accuracy=30.756250	 loss=3.192660	 lr=0.093960
Epoch[059] Batch [2299]/[3760]	Speed: 117.869727 samples/sec	 accuracy=30.746603	 loss=3.193192	 lr=0.093954
Epoch[059] Batch [2349]/[3760]	Speed: 116.775038 samples/sec	 accuracy=30.739362	 loss=3.193844	 lr=0.093948
Epoch[059] Batch [2399]/[3760]	Speed: 118.071822 samples/sec	 accuracy=30.749349	 loss=3.193508	 lr=0.093941
Epoch[059] Batch [2449]/[3760]	Speed: 116.840211 samples/sec	 accuracy=30.721301	 loss=3.193840	 lr=0.093935
Epoch[059] Batch [2499]/[3760]	Speed: 117.304556 samples/sec	 accuracy=30.728125	 loss=3.193899	 lr=0.093929
Epoch[059] Batch [2549]/[3760]	Speed: 117.656789 samples/sec	 accuracy=30.705270	 loss=3.195170	 lr=0.093923
Epoch[059] Batch [2599]/[3760]	Speed: 116.810054 samples/sec	 accuracy=30.701923	 loss=3.195042	 lr=0.093917
Epoch[059] Batch [2649]/[3760]	Speed: 116.479102 samples/sec	 accuracy=30.692217	 loss=3.195228	 lr=0.093911
Epoch[059] Batch [2699]/[3760]	Speed: 116.920308 samples/sec	 accuracy=30.704282	 loss=3.194473	 lr=0.093905
Epoch[059] Batch [2749]/[3760]	Speed: 117.779292 samples/sec	 accuracy=30.701136	 loss=3.194334	 lr=0.093898
Epoch[059] Batch [2799]/[3760]	Speed: 117.253874 samples/sec	 accuracy=30.711496	 loss=3.193990	 lr=0.093892
Epoch[059] Batch [2849]/[3760]	Speed: 117.732639 samples/sec	 accuracy=30.713268	 loss=3.193304	 lr=0.093886
Epoch[059] Batch [2899]/[3760]	Speed: 116.828932 samples/sec	 accuracy=30.726832	 loss=3.192839	 lr=0.093880
Epoch[059] Batch [2949]/[3760]	Speed: 117.369386 samples/sec	 accuracy=30.724047	 loss=3.193432	 lr=0.093874
Epoch[059] Batch [2999]/[3760]	Speed: 117.378514 samples/sec	 accuracy=30.746354	 loss=3.192573	 lr=0.093867
Epoch[059] Batch [3049]/[3760]	Speed: 117.122247 samples/sec	 accuracy=30.746926	 loss=3.192915	 lr=0.093861
Epoch[059] Batch [3099]/[3760]	Speed: 117.492502 samples/sec	 accuracy=30.779738	 loss=3.191643	 lr=0.093855
Epoch[059] Batch [3149]/[3760]	Speed: 117.662370 samples/sec	 accuracy=30.766369	 loss=3.192392	 lr=0.093849
Epoch[059] Batch [3199]/[3760]	Speed: 117.917699 samples/sec	 accuracy=30.779785	 loss=3.191481	 lr=0.093843
Epoch[059] Batch [3249]/[3760]	Speed: 117.556636 samples/sec	 accuracy=30.796635	 loss=3.191194	 lr=0.093836
Epoch[059] Batch [3299]/[3760]	Speed: 117.891382 samples/sec	 accuracy=30.787879	 loss=3.191697	 lr=0.093830
Epoch[059] Batch [3349]/[3760]	Speed: 117.558717 samples/sec	 accuracy=30.776586	 loss=3.191647	 lr=0.093824
Epoch[059] Batch [3399]/[3760]	Speed: 117.307093 samples/sec	 accuracy=30.771599	 loss=3.192526	 lr=0.093818
Epoch[059] Batch [3449]/[3760]	Speed: 118.053603 samples/sec	 accuracy=30.777627	 loss=3.192469	 lr=0.093812
Epoch[059] Batch [3499]/[3760]	Speed: 116.674714 samples/sec	 accuracy=30.786161	 loss=3.191980	 lr=0.093805
Epoch[059] Batch [3549]/[3760]	Speed: 117.724224 samples/sec	 accuracy=30.772447	 loss=3.192314	 lr=0.093799
Epoch[059] Batch [3599]/[3760]	Speed: 116.863328 samples/sec	 accuracy=30.763455	 loss=3.192963	 lr=0.093793
Epoch[059] Batch [3649]/[3760]	Speed: 118.263121 samples/sec	 accuracy=30.755565	 loss=3.193403	 lr=0.093787
Epoch[059] Batch [3699]/[3760]	Speed: 117.427116 samples/sec	 accuracy=30.751689	 loss=3.193457	 lr=0.093781
Epoch[059] Batch [3749]/[3760]	Speed: 120.510035 samples/sec	 accuracy=30.746250	 loss=3.193722	 lr=0.093774
Batch [0049]/[0303]: acc-top1=30.968750 acc-top5=55.625000
Batch [0099]/[0303]: acc-top1=28.203125 acc-top5=53.578125
Batch [0149]/[0303]: acc-top1=29.979167 acc-top5=55.354167
Batch [0199]/[0303]: acc-top1=32.742188 acc-top5=58.117188
Batch [0249]/[0303]: acc-top1=31.987500 acc-top5=58.156250
Batch [0299]/[0303]: acc-top1=31.354167 acc-top5=58.223958
[Epoch 059] training: accuracy=30.752992	 loss=3.193573
[Epoch 059] speed: 116 samples/sec	time cost: 2179.091674
[Epoch 059] validation: acc-top1=31.265470 acc-top5=58.152847 loss=3.218123
Epoch[060] Batch [0049]/[3760]	Speed: 86.781486 samples/sec	 accuracy=30.968750	 loss=3.176050	 lr=0.093767
Epoch[060] Batch [0099]/[3760]	Speed: 118.007601 samples/sec	 accuracy=30.500000	 loss=3.190623	 lr=0.093761
Epoch[060] Batch [0149]/[3760]	Speed: 117.761980 samples/sec	 accuracy=31.062500	 loss=3.163156	 lr=0.093754
Epoch[060] Batch [0199]/[3760]	Speed: 117.934004 samples/sec	 accuracy=31.390625	 loss=3.154419	 lr=0.093748
Epoch[060] Batch [0249]/[3760]	Speed: 117.577138 samples/sec	 accuracy=31.456250	 loss=3.152674	 lr=0.093742
Epoch[060] Batch [0299]/[3760]	Speed: 117.159694 samples/sec	 accuracy=31.427083	 loss=3.140638	 lr=0.093736
Epoch[060] Batch [0349]/[3760]	Speed: 117.350753 samples/sec	 accuracy=31.290179	 loss=3.151720	 lr=0.093729
Epoch[060] Batch [0399]/[3760]	Speed: 117.557179 samples/sec	 accuracy=31.222656	 loss=3.157710	 lr=0.093723
Epoch[060] Batch [0449]/[3760]	Speed: 117.328353 samples/sec	 accuracy=31.225694	 loss=3.160482	 lr=0.093717
Epoch[060] Batch [0499]/[3760]	Speed: 117.763860 samples/sec	 accuracy=31.181250	 loss=3.163039	 lr=0.093711
Epoch[060] Batch [0549]/[3760]	Speed: 117.874908 samples/sec	 accuracy=31.127841	 loss=3.164782	 lr=0.093704
Epoch[060] Batch [0599]/[3760]	Speed: 117.695378 samples/sec	 accuracy=31.065104	 loss=3.167692	 lr=0.093698
Epoch[060] Batch [0649]/[3760]	Speed: 118.086816 samples/sec	 accuracy=31.134615	 loss=3.165847	 lr=0.093692
Epoch[060] Batch [0699]/[3760]	Speed: 116.891641 samples/sec	 accuracy=31.037946	 loss=3.170961	 lr=0.093685
Epoch[060] Batch [0749]/[3760]	Speed: 117.056992 samples/sec	 accuracy=30.983333	 loss=3.170918	 lr=0.093679
Epoch[060] Batch [0799]/[3760]	Speed: 116.751353 samples/sec	 accuracy=31.000000	 loss=3.167359	 lr=0.093673
Epoch[060] Batch [0849]/[3760]	Speed: 117.369287 samples/sec	 accuracy=31.005515	 loss=3.167700	 lr=0.093667
Epoch[060] Batch [0899]/[3760]	Speed: 117.812436 samples/sec	 accuracy=30.986111	 loss=3.168212	 lr=0.093660
Epoch[060] Batch [0949]/[3760]	Speed: 117.282448 samples/sec	 accuracy=31.036184	 loss=3.165949	 lr=0.093654
Epoch[060] Batch [0999]/[3760]	Speed: 117.247325 samples/sec	 accuracy=30.995313	 loss=3.168845	 lr=0.093648
Epoch[060] Batch [1049]/[3760]	Speed: 117.403388 samples/sec	 accuracy=30.992560	 loss=3.168701	 lr=0.093641
Epoch[060] Batch [1099]/[3760]	Speed: 117.048473 samples/sec	 accuracy=30.960227	 loss=3.169436	 lr=0.093635
Epoch[060] Batch [1149]/[3760]	Speed: 117.576529 samples/sec	 accuracy=30.918478	 loss=3.170708	 lr=0.093629
Epoch[060] Batch [1199]/[3760]	Speed: 117.160503 samples/sec	 accuracy=30.888021	 loss=3.170822	 lr=0.093623
Epoch[060] Batch [1249]/[3760]	Speed: 118.195491 samples/sec	 accuracy=30.918750	 loss=3.169538	 lr=0.093616
Epoch[060] Batch [1299]/[3760]	Speed: 116.959645 samples/sec	 accuracy=30.911058	 loss=3.169424	 lr=0.093610
Epoch[060] Batch [1349]/[3760]	Speed: 117.767951 samples/sec	 accuracy=30.875000	 loss=3.170863	 lr=0.093604
Epoch[060] Batch [1399]/[3760]	Speed: 117.129860 samples/sec	 accuracy=30.832589	 loss=3.173330	 lr=0.093597
Epoch[060] Batch [1449]/[3760]	Speed: 117.295529 samples/sec	 accuracy=30.821121	 loss=3.173067	 lr=0.093591
Epoch[060] Batch [1499]/[3760]	Speed: 117.747447 samples/sec	 accuracy=30.795833	 loss=3.175658	 lr=0.093585
Epoch[060] Batch [1549]/[3760]	Speed: 117.712777 samples/sec	 accuracy=30.741935	 loss=3.178578	 lr=0.093578
Epoch[060] Batch [1599]/[3760]	Speed: 117.077917 samples/sec	 accuracy=30.731445	 loss=3.178728	 lr=0.093572
Epoch[060] Batch [1649]/[3760]	Speed: 117.369832 samples/sec	 accuracy=30.720644	 loss=3.179105	 lr=0.093566
Epoch[060] Batch [1699]/[3760]	Speed: 116.628137 samples/sec	 accuracy=30.713235	 loss=3.179899	 lr=0.093559
Epoch[060] Batch [1749]/[3760]	Speed: 117.282747 samples/sec	 accuracy=30.720536	 loss=3.179376	 lr=0.093553
Epoch[060] Batch [1799]/[3760]	Speed: 116.880373 samples/sec	 accuracy=30.705729	 loss=3.179893	 lr=0.093547
Epoch[060] Batch [1849]/[3760]	Speed: 117.844661 samples/sec	 accuracy=30.674831	 loss=3.181195	 lr=0.093540
Epoch[060] Batch [1899]/[3760]	Speed: 117.550587 samples/sec	 accuracy=30.670230	 loss=3.180232	 lr=0.093534
Epoch[060] Batch [1949]/[3760]	Speed: 116.906787 samples/sec	 accuracy=30.669872	 loss=3.180020	 lr=0.093528
Epoch[060] Batch [1999]/[3760]	Speed: 117.937129 samples/sec	 accuracy=30.650781	 loss=3.180115	 lr=0.093521
Epoch[060] Batch [2049]/[3760]	Speed: 117.231786 samples/sec	 accuracy=30.640244	 loss=3.179447	 lr=0.093515
Epoch[060] Batch [2099]/[3760]	Speed: 117.346275 samples/sec	 accuracy=30.662946	 loss=3.179658	 lr=0.093509
Epoch[060] Batch [2149]/[3760]	Speed: 117.971491 samples/sec	 accuracy=30.686047	 loss=3.179774	 lr=0.093502
Epoch[060] Batch [2199]/[3760]	Speed: 117.874213 samples/sec	 accuracy=30.722301	 loss=3.178561	 lr=0.093496
Epoch[060] Batch [2249]/[3760]	Speed: 117.230090 samples/sec	 accuracy=30.761806	 loss=3.177194	 lr=0.093490
Epoch[060] Batch [2299]/[3760]	Speed: 117.715595 samples/sec	 accuracy=30.770380	 loss=3.177147	 lr=0.093483
Epoch[060] Batch [2349]/[3760]	Speed: 117.619459 samples/sec	 accuracy=30.724734	 loss=3.178565	 lr=0.093477
Epoch[060] Batch [2399]/[3760]	Speed: 117.896002 samples/sec	 accuracy=30.725260	 loss=3.178478	 lr=0.093470
Epoch[060] Batch [2449]/[3760]	Speed: 118.132467 samples/sec	 accuracy=30.705357	 loss=3.179392	 lr=0.093464
Epoch[060] Batch [2499]/[3760]	Speed: 116.864656 samples/sec	 accuracy=30.711250	 loss=3.179373	 lr=0.093458
Epoch[060] Batch [2549]/[3760]	Speed: 116.917163 samples/sec	 accuracy=30.738971	 loss=3.178074	 lr=0.093451
Epoch[060] Batch [2599]/[3760]	Speed: 117.586432 samples/sec	 accuracy=30.740986	 loss=3.178161	 lr=0.093445
Epoch[060] Batch [2649]/[3760]	Speed: 117.680693 samples/sec	 accuracy=30.747052	 loss=3.178007	 lr=0.093439
Epoch[060] Batch [2699]/[3760]	Speed: 117.542445 samples/sec	 accuracy=30.724537	 loss=3.178640	 lr=0.093432
Epoch[060] Batch [2749]/[3760]	Speed: 117.119525 samples/sec	 accuracy=30.747159	 loss=3.177923	 lr=0.093426
Epoch[060] Batch [2799]/[3760]	Speed: 118.236979 samples/sec	 accuracy=30.756138	 loss=3.177774	 lr=0.093419
Epoch[060] Batch [2849]/[3760]	Speed: 116.072222 samples/sec	 accuracy=30.746162	 loss=3.179526	 lr=0.093413
Epoch[060] Batch [2899]/[3760]	Speed: 118.207385 samples/sec	 accuracy=30.772091	 loss=3.178764	 lr=0.093407
Epoch[060] Batch [2949]/[3760]	Speed: 117.921451 samples/sec	 accuracy=30.784958	 loss=3.178947	 lr=0.093400
Epoch[060] Batch [2999]/[3760]	Speed: 117.031318 samples/sec	 accuracy=30.773438	 loss=3.179019	 lr=0.093394
Epoch[060] Batch [3049]/[3760]	Speed: 116.729170 samples/sec	 accuracy=30.777152	 loss=3.179417	 lr=0.093387
Epoch[060] Batch [3099]/[3760]	Speed: 118.118324 samples/sec	 accuracy=30.768649	 loss=3.179564	 lr=0.093381
Epoch[060] Batch [3149]/[3760]	Speed: 117.801496 samples/sec	 accuracy=30.748512	 loss=3.179470	 lr=0.093375
Epoch[060] Batch [3199]/[3760]	Speed: 117.688541 samples/sec	 accuracy=30.732910	 loss=3.179983	 lr=0.093368
Epoch[060] Batch [3249]/[3760]	Speed: 117.609578 samples/sec	 accuracy=30.721154	 loss=3.180338	 lr=0.093362
Epoch[060] Batch [3299]/[3760]	Speed: 118.184592 samples/sec	 accuracy=30.715436	 loss=3.181039	 lr=0.093355
Epoch[060] Batch [3349]/[3760]	Speed: 117.222398 samples/sec	 accuracy=30.701493	 loss=3.182052	 lr=0.093349
Epoch[060] Batch [3399]/[3760]	Speed: 117.186820 samples/sec	 accuracy=30.687040	 loss=3.182810	 lr=0.093342
Epoch[060] Batch [3449]/[3760]	Speed: 116.383524 samples/sec	 accuracy=30.695199	 loss=3.183053	 lr=0.093336
Epoch[060] Batch [3499]/[3760]	Speed: 116.986117 samples/sec	 accuracy=30.695982	 loss=3.183608	 lr=0.093330
Epoch[060] Batch [3549]/[3760]	Speed: 117.293562 samples/sec	 accuracy=30.690581	 loss=3.183938	 lr=0.093323
Epoch[060] Batch [3599]/[3760]	Speed: 117.700223 samples/sec	 accuracy=30.701823	 loss=3.183117	 lr=0.093317
Epoch[060] Batch [3649]/[3760]	Speed: 117.328566 samples/sec	 accuracy=30.699486	 loss=3.183116	 lr=0.093310
Epoch[060] Batch [3699]/[3760]	Speed: 117.324446 samples/sec	 accuracy=30.699747	 loss=3.183551	 lr=0.093304
Epoch[060] Batch [3749]/[3760]	Speed: 120.386288 samples/sec	 accuracy=30.696250	 loss=3.184585	 lr=0.093297
Batch [0049]/[0303]: acc-top1=29.343750 acc-top5=56.562500
Batch [0099]/[0303]: acc-top1=27.187500 acc-top5=54.453125
Batch [0149]/[0303]: acc-top1=29.114583 acc-top5=55.687500
Batch [0199]/[0303]: acc-top1=31.398438 acc-top5=57.695312
Batch [0249]/[0303]: acc-top1=30.925000 acc-top5=57.550000
Batch [0299]/[0303]: acc-top1=30.479167 acc-top5=57.682292
[Epoch 060] training: accuracy=30.693152	 loss=3.184501
[Epoch 060] speed: 116 samples/sec	time cost: 2177.953830
[Epoch 060] validation: acc-top1=30.543523 acc-top5=57.724835 loss=3.312907
Epoch[061] Batch [0049]/[3759]	Speed: 87.629579 samples/sec	 accuracy=31.812500	 loss=3.106753	 lr=0.093290
Epoch[061] Batch [0099]/[3759]	Speed: 117.056094 samples/sec	 accuracy=32.234375	 loss=3.085887	 lr=0.093283
Epoch[061] Batch [0149]/[3759]	Speed: 117.264390 samples/sec	 accuracy=31.822917	 loss=3.108979	 lr=0.093277
Epoch[061] Batch [0199]/[3759]	Speed: 117.420324 samples/sec	 accuracy=31.570312	 loss=3.128393	 lr=0.093270
Epoch[061] Batch [0249]/[3759]	Speed: 117.038546 samples/sec	 accuracy=31.487500	 loss=3.140351	 lr=0.093264
Epoch[061] Batch [0299]/[3759]	Speed: 117.115127 samples/sec	 accuracy=31.416667	 loss=3.150438	 lr=0.093257
Epoch[061] Batch [0349]/[3759]	Speed: 117.504085 samples/sec	 accuracy=31.535714	 loss=3.143127	 lr=0.093251
Epoch[061] Batch [0399]/[3759]	Speed: 117.270886 samples/sec	 accuracy=31.460938	 loss=3.141397	 lr=0.093244
Epoch[061] Batch [0449]/[3759]	Speed: 117.624989 samples/sec	 accuracy=31.364583	 loss=3.144623	 lr=0.093238
Epoch[061] Batch [0499]/[3759]	Speed: 116.910502 samples/sec	 accuracy=31.328125	 loss=3.146277	 lr=0.093231
Epoch[061] Batch [0549]/[3759]	Speed: 117.179059 samples/sec	 accuracy=31.261364	 loss=3.146891	 lr=0.093225
Epoch[061] Batch [0599]/[3759]	Speed: 117.437939 samples/sec	 accuracy=31.312500	 loss=3.146732	 lr=0.093218
Epoch[061] Batch [0649]/[3759]	Speed: 117.784143 samples/sec	 accuracy=31.360577	 loss=3.146886	 lr=0.093212
Epoch[061] Batch [0699]/[3759]	Speed: 117.234662 samples/sec	 accuracy=31.229911	 loss=3.156336	 lr=0.093205
Epoch[061] Batch [0749]/[3759]	Speed: 117.674026 samples/sec	 accuracy=31.289583	 loss=3.153167	 lr=0.093199
Epoch[061] Batch [0799]/[3759]	Speed: 116.737811 samples/sec	 accuracy=31.263672	 loss=3.153220	 lr=0.093193
Epoch[061] Batch [0849]/[3759]	Speed: 116.038987 samples/sec	 accuracy=31.283088	 loss=3.152199	 lr=0.093186
Epoch[061] Batch [0899]/[3759]	Speed: 117.228249 samples/sec	 accuracy=31.237847	 loss=3.155437	 lr=0.093180
Epoch[061] Batch [0949]/[3759]	Speed: 117.259942 samples/sec	 accuracy=31.180921	 loss=3.158289	 lr=0.093173
Epoch[061] Batch [0999]/[3759]	Speed: 117.046932 samples/sec	 accuracy=31.134375	 loss=3.162176	 lr=0.093166
Epoch[061] Batch [1049]/[3759]	Speed: 117.400099 samples/sec	 accuracy=31.114583	 loss=3.162456	 lr=0.093160
Epoch[061] Batch [1099]/[3759]	Speed: 117.544617 samples/sec	 accuracy=31.146307	 loss=3.164169	 lr=0.093153
Epoch[061] Batch [1149]/[3759]	Speed: 117.841160 samples/sec	 accuracy=31.099185	 loss=3.167082	 lr=0.093147
Epoch[061] Batch [1199]/[3759]	Speed: 116.845140 samples/sec	 accuracy=31.040365	 loss=3.168517	 lr=0.093140
Epoch[061] Batch [1249]/[3759]	Speed: 117.183697 samples/sec	 accuracy=30.983750	 loss=3.170397	 lr=0.093134
Epoch[061] Batch [1299]/[3759]	Speed: 116.967346 samples/sec	 accuracy=30.973558	 loss=3.171168	 lr=0.093127
Epoch[061] Batch [1349]/[3759]	Speed: 116.373990 samples/sec	 accuracy=30.966435	 loss=3.172133	 lr=0.093121
Epoch[061] Batch [1399]/[3759]	Speed: 116.906116 samples/sec	 accuracy=30.968750	 loss=3.172154	 lr=0.093114
Epoch[061] Batch [1449]/[3759]	Speed: 116.840794 samples/sec	 accuracy=31.025862	 loss=3.171253	 lr=0.093108
Epoch[061] Batch [1499]/[3759]	Speed: 116.709166 samples/sec	 accuracy=31.026042	 loss=3.171455	 lr=0.093101
Epoch[061] Batch [1549]/[3759]	Speed: 117.074896 samples/sec	 accuracy=30.981855	 loss=3.173075	 lr=0.093095
Epoch[061] Batch [1599]/[3759]	Speed: 117.347745 samples/sec	 accuracy=30.945312	 loss=3.174369	 lr=0.093088
Epoch[061] Batch [1649]/[3759]	Speed: 118.071874 samples/sec	 accuracy=30.978220	 loss=3.173755	 lr=0.093082
Epoch[061] Batch [1699]/[3759]	Speed: 117.973758 samples/sec	 accuracy=31.006434	 loss=3.172046	 lr=0.093075
Epoch[061] Batch [1749]/[3759]	Speed: 117.592471 samples/sec	 accuracy=31.006250	 loss=3.172953	 lr=0.093069
Epoch[061] Batch [1799]/[3759]	Speed: 116.842028 samples/sec	 accuracy=31.032986	 loss=3.172516	 lr=0.093062
Epoch[061] Batch [1849]/[3759]	Speed: 116.831421 samples/sec	 accuracy=31.042230	 loss=3.172164	 lr=0.093055
Epoch[061] Batch [1899]/[3759]	Speed: 117.016908 samples/sec	 accuracy=31.046875	 loss=3.172330	 lr=0.093049
Epoch[061] Batch [1949]/[3759]	Speed: 116.921932 samples/sec	 accuracy=31.047276	 loss=3.172400	 lr=0.093042
Epoch[061] Batch [1999]/[3759]	Speed: 117.035642 samples/sec	 accuracy=31.053125	 loss=3.173201	 lr=0.093036
Epoch[061] Batch [2049]/[3759]	Speed: 117.885792 samples/sec	 accuracy=31.032774	 loss=3.174166	 lr=0.093029
Epoch[061] Batch [2099]/[3759]	Speed: 117.727745 samples/sec	 accuracy=31.065476	 loss=3.172957	 lr=0.093023
Epoch[061] Batch [2149]/[3759]	Speed: 117.087497 samples/sec	 accuracy=31.067587	 loss=3.172555	 lr=0.093016
Epoch[061] Batch [2199]/[3759]	Speed: 117.128167 samples/sec	 accuracy=31.071023	 loss=3.172217	 lr=0.093009
Epoch[061] Batch [2249]/[3759]	Speed: 116.879864 samples/sec	 accuracy=31.045139	 loss=3.172152	 lr=0.093003
Epoch[061] Batch [2299]/[3759]	Speed: 117.480026 samples/sec	 accuracy=31.046875	 loss=3.170223	 lr=0.092996
Epoch[061] Batch [2349]/[3759]	Speed: 117.160115 samples/sec	 accuracy=31.058511	 loss=3.170223	 lr=0.092990
Epoch[061] Batch [2399]/[3759]	Speed: 117.865014 samples/sec	 accuracy=31.085286	 loss=3.169958	 lr=0.092983
Epoch[061] Batch [2449]/[3759]	Speed: 117.280848 samples/sec	 accuracy=31.070153	 loss=3.170215	 lr=0.092977
Epoch[061] Batch [2499]/[3759]	Speed: 116.980929 samples/sec	 accuracy=31.065625	 loss=3.169347	 lr=0.092970
Epoch[061] Batch [2549]/[3759]	Speed: 116.552093 samples/sec	 accuracy=31.047794	 loss=3.169723	 lr=0.092963
Epoch[061] Batch [2599]/[3759]	Speed: 116.845018 samples/sec	 accuracy=31.039062	 loss=3.169509	 lr=0.092957
Epoch[061] Batch [2649]/[3759]	Speed: 117.465067 samples/sec	 accuracy=31.002358	 loss=3.170483	 lr=0.092950
Epoch[061] Batch [2699]/[3759]	Speed: 116.747947 samples/sec	 accuracy=30.989005	 loss=3.171228	 lr=0.092944
Epoch[061] Batch [2749]/[3759]	Speed: 117.119574 samples/sec	 accuracy=31.002273	 loss=3.170660	 lr=0.092937
Epoch[061] Batch [2799]/[3759]	Speed: 117.816926 samples/sec	 accuracy=31.001116	 loss=3.170443	 lr=0.092930
Epoch[061] Batch [2849]/[3759]	Speed: 117.080496 samples/sec	 accuracy=30.968750	 loss=3.172092	 lr=0.092924
Epoch[061] Batch [2899]/[3759]	Speed: 117.389781 samples/sec	 accuracy=30.956897	 loss=3.172928	 lr=0.092917
Epoch[061] Batch [2949]/[3759]	Speed: 116.536845 samples/sec	 accuracy=30.995763	 loss=3.171364	 lr=0.092910
Epoch[061] Batch [2999]/[3759]	Speed: 117.227280 samples/sec	 accuracy=30.998958	 loss=3.171773	 lr=0.092904
Epoch[061] Batch [3049]/[3759]	Speed: 117.657278 samples/sec	 accuracy=30.987705	 loss=3.171784	 lr=0.092897
Epoch[061] Batch [3099]/[3759]	Speed: 116.913165 samples/sec	 accuracy=30.966230	 loss=3.172603	 lr=0.092891
Epoch[061] Batch [3149]/[3759]	Speed: 117.600563 samples/sec	 accuracy=30.954365	 loss=3.173566	 lr=0.092884
Epoch[061] Batch [3199]/[3759]	Speed: 117.550062 samples/sec	 accuracy=30.950195	 loss=3.174259	 lr=0.092877
Epoch[061] Batch [3249]/[3759]	Speed: 117.700844 samples/sec	 accuracy=30.958654	 loss=3.174407	 lr=0.092871
Epoch[061] Batch [3299]/[3759]	Speed: 117.189324 samples/sec	 accuracy=30.959280	 loss=3.174771	 lr=0.092864
Epoch[061] Batch [3349]/[3759]	Speed: 117.138919 samples/sec	 accuracy=30.971549	 loss=3.174700	 lr=0.092857
Epoch[061] Batch [3399]/[3759]	Speed: 117.374031 samples/sec	 accuracy=30.959099	 loss=3.174722	 lr=0.092851
Epoch[061] Batch [3449]/[3759]	Speed: 117.714018 samples/sec	 accuracy=30.960145	 loss=3.174811	 lr=0.092844
Epoch[061] Batch [3499]/[3759]	Speed: 117.350485 samples/sec	 accuracy=30.954911	 loss=3.175109	 lr=0.092837
Epoch[061] Batch [3549]/[3759]	Speed: 117.607696 samples/sec	 accuracy=30.961268	 loss=3.174649	 lr=0.092831
Epoch[061] Batch [3599]/[3759]	Speed: 117.471388 samples/sec	 accuracy=30.973090	 loss=3.174212	 lr=0.092824
Epoch[061] Batch [3649]/[3759]	Speed: 117.522102 samples/sec	 accuracy=30.955479	 loss=3.175068	 lr=0.092818
Epoch[061] Batch [3699]/[3759]	Speed: 116.234118 samples/sec	 accuracy=30.953970	 loss=3.175007	 lr=0.092811
Epoch[061] Batch [3749]/[3759]	Speed: 120.557568 samples/sec	 accuracy=30.969167	 loss=3.174627	 lr=0.092804
Batch [0049]/[0303]: acc-top1=30.906250 acc-top5=58.937500
Batch [0099]/[0303]: acc-top1=29.296875 acc-top5=56.734375
Batch [0149]/[0303]: acc-top1=30.541667 acc-top5=57.156250
Batch [0199]/[0303]: acc-top1=33.273438 acc-top5=59.515625
Batch [0249]/[0303]: acc-top1=31.937500 acc-top5=59.000000
Batch [0299]/[0303]: acc-top1=31.348958 acc-top5=58.432292
[Epoch 061] training: accuracy=30.966514	 loss=3.174573
[Epoch 061] speed: 116 samples/sec	time cost: 2180.808583
[Epoch 061] validation: acc-top1=31.275784 acc-top5=58.441625 loss=3.245693
Epoch[062] Batch [0049]/[3760]	Speed: 87.184689 samples/sec	 accuracy=32.593750	 loss=3.138767	 lr=0.092796
Epoch[062] Batch [0099]/[3760]	Speed: 117.745345 samples/sec	 accuracy=32.234375	 loss=3.156952	 lr=0.092790
Epoch[062] Batch [0149]/[3760]	Speed: 118.305480 samples/sec	 accuracy=31.937500	 loss=3.156517	 lr=0.092783
Epoch[062] Batch [0199]/[3760]	Speed: 116.842639 samples/sec	 accuracy=31.609375	 loss=3.165958	 lr=0.092776
Epoch[062] Batch [0249]/[3760]	Speed: 117.847150 samples/sec	 accuracy=31.518750	 loss=3.160701	 lr=0.092770
Epoch[062] Batch [0299]/[3760]	Speed: 117.665898 samples/sec	 accuracy=31.432292	 loss=3.150609	 lr=0.092763
Epoch[062] Batch [0349]/[3760]	Speed: 117.595782 samples/sec	 accuracy=31.517857	 loss=3.143266	 lr=0.092756
Epoch[062] Batch [0399]/[3760]	Speed: 117.096090 samples/sec	 accuracy=31.394531	 loss=3.148810	 lr=0.092750
Epoch[062] Batch [0449]/[3760]	Speed: 117.842403 samples/sec	 accuracy=31.368056	 loss=3.152594	 lr=0.092743
Epoch[062] Batch [0499]/[3760]	Speed: 116.953380 samples/sec	 accuracy=31.456250	 loss=3.145408	 lr=0.092736
Epoch[062] Batch [0549]/[3760]	Speed: 116.367564 samples/sec	 accuracy=31.423295	 loss=3.148349	 lr=0.092729
Epoch[062] Batch [0599]/[3760]	Speed: 117.495217 samples/sec	 accuracy=31.552083	 loss=3.142554	 lr=0.092723
Epoch[062] Batch [0649]/[3760]	Speed: 117.237477 samples/sec	 accuracy=31.600962	 loss=3.138226	 lr=0.092716
Epoch[062] Batch [0699]/[3760]	Speed: 116.938011 samples/sec	 accuracy=31.473214	 loss=3.144837	 lr=0.092709
Epoch[062] Batch [0749]/[3760]	Speed: 117.587190 samples/sec	 accuracy=31.429167	 loss=3.146366	 lr=0.092703
Epoch[062] Batch [0799]/[3760]	Speed: 117.854429 samples/sec	 accuracy=31.501953	 loss=3.144370	 lr=0.092696
Epoch[062] Batch [0849]/[3760]	Speed: 117.635641 samples/sec	 accuracy=31.419118	 loss=3.147407	 lr=0.092689
Epoch[062] Batch [0899]/[3760]	Speed: 117.802318 samples/sec	 accuracy=31.460069	 loss=3.146926	 lr=0.092683
Epoch[062] Batch [0949]/[3760]	Speed: 117.193774 samples/sec	 accuracy=31.394737	 loss=3.147601	 lr=0.092676
Epoch[062] Batch [0999]/[3760]	Speed: 117.273384 samples/sec	 accuracy=31.378125	 loss=3.146713	 lr=0.092669
Epoch[062] Batch [1049]/[3760]	Speed: 117.487917 samples/sec	 accuracy=31.431548	 loss=3.144630	 lr=0.092662
Epoch[062] Batch [1099]/[3760]	Speed: 114.722177 samples/sec	 accuracy=31.420455	 loss=3.146390	 lr=0.092656
Epoch[062] Batch [1149]/[3760]	Speed: 116.988030 samples/sec	 accuracy=31.377717	 loss=3.148315	 lr=0.092649
Epoch[062] Batch [1199]/[3760]	Speed: 117.148134 samples/sec	 accuracy=31.337240	 loss=3.151441	 lr=0.092642
Epoch[062] Batch [1249]/[3760]	Speed: 117.065007 samples/sec	 accuracy=31.338750	 loss=3.150921	 lr=0.092635
Epoch[062] Batch [1299]/[3760]	Speed: 117.295370 samples/sec	 accuracy=31.361779	 loss=3.151138	 lr=0.092629
Epoch[062] Batch [1349]/[3760]	Speed: 117.641315 samples/sec	 accuracy=31.341435	 loss=3.153664	 lr=0.092622
Epoch[062] Batch [1399]/[3760]	Speed: 116.789939 samples/sec	 accuracy=31.296875	 loss=3.153727	 lr=0.092615
Epoch[062] Batch [1449]/[3760]	Speed: 117.812335 samples/sec	 accuracy=31.301724	 loss=3.152958	 lr=0.092608
Epoch[062] Batch [1499]/[3760]	Speed: 118.101267 samples/sec	 accuracy=31.311458	 loss=3.153627	 lr=0.092602
Epoch[062] Batch [1549]/[3760]	Speed: 117.166325 samples/sec	 accuracy=31.311492	 loss=3.152751	 lr=0.092595
Epoch[062] Batch [1599]/[3760]	Speed: 116.965862 samples/sec	 accuracy=31.312500	 loss=3.153374	 lr=0.092588
Epoch[062] Batch [1649]/[3760]	Speed: 117.236099 samples/sec	 accuracy=31.320076	 loss=3.153019	 lr=0.092581
Epoch[062] Batch [1699]/[3760]	Speed: 116.527201 samples/sec	 accuracy=31.307904	 loss=3.153292	 lr=0.092575
Epoch[062] Batch [1749]/[3760]	Speed: 116.685006 samples/sec	 accuracy=31.266964	 loss=3.154270	 lr=0.092568
Epoch[062] Batch [1799]/[3760]	Speed: 117.933839 samples/sec	 accuracy=31.219618	 loss=3.156990	 lr=0.092561
Epoch[062] Batch [1849]/[3760]	Speed: 117.603913 samples/sec	 accuracy=31.221284	 loss=3.157903	 lr=0.092554
Epoch[062] Batch [1899]/[3760]	Speed: 117.809872 samples/sec	 accuracy=31.237664	 loss=3.158401	 lr=0.092548
Epoch[062] Batch [1949]/[3760]	Speed: 117.458092 samples/sec	 accuracy=31.218750	 loss=3.158512	 lr=0.092541
Epoch[062] Batch [1999]/[3760]	Speed: 117.772619 samples/sec	 accuracy=31.242969	 loss=3.156401	 lr=0.092534
Epoch[062] Batch [2049]/[3760]	Speed: 116.992200 samples/sec	 accuracy=31.227134	 loss=3.156563	 lr=0.092527
Epoch[062] Batch [2099]/[3760]	Speed: 117.136438 samples/sec	 accuracy=31.200893	 loss=3.158364	 lr=0.092520
Epoch[062] Batch [2149]/[3760]	Speed: 117.996345 samples/sec	 accuracy=31.182413	 loss=3.158376	 lr=0.092514
Epoch[062] Batch [2199]/[3760]	Speed: 117.159037 samples/sec	 accuracy=31.198864	 loss=3.157871	 lr=0.092507
Epoch[062] Batch [2249]/[3760]	Speed: 116.842552 samples/sec	 accuracy=31.184722	 loss=3.158967	 lr=0.092500
Epoch[062] Batch [2299]/[3760]	Speed: 117.987820 samples/sec	 accuracy=31.164402	 loss=3.159351	 lr=0.092493
Epoch[062] Batch [2349]/[3760]	Speed: 117.169621 samples/sec	 accuracy=31.139628	 loss=3.160819	 lr=0.092487
Epoch[062] Batch [2399]/[3760]	Speed: 117.186463 samples/sec	 accuracy=31.124349	 loss=3.161847	 lr=0.092480
Epoch[062] Batch [2449]/[3760]	Speed: 117.331244 samples/sec	 accuracy=31.109056	 loss=3.163373	 lr=0.092473
Epoch[062] Batch [2499]/[3760]	Speed: 116.709637 samples/sec	 accuracy=31.104375	 loss=3.163520	 lr=0.092466
Epoch[062] Batch [2549]/[3760]	Speed: 118.000129 samples/sec	 accuracy=31.109681	 loss=3.163661	 lr=0.092459
Epoch[062] Batch [2599]/[3760]	Speed: 117.969927 samples/sec	 accuracy=31.106971	 loss=3.163495	 lr=0.092452
Epoch[062] Batch [2649]/[3760]	Speed: 117.247909 samples/sec	 accuracy=31.130896	 loss=3.162113	 lr=0.092446
Epoch[062] Batch [2699]/[3760]	Speed: 117.460440 samples/sec	 accuracy=31.097801	 loss=3.163906	 lr=0.092439
Epoch[062] Batch [2749]/[3760]	Speed: 117.165653 samples/sec	 accuracy=31.096591	 loss=3.164352	 lr=0.092432
Epoch[062] Batch [2799]/[3760]	Speed: 117.220428 samples/sec	 accuracy=31.082031	 loss=3.165500	 lr=0.092425
Epoch[062] Batch [2849]/[3760]	Speed: 117.182120 samples/sec	 accuracy=31.096491	 loss=3.166614	 lr=0.092418
Epoch[062] Batch [2899]/[3760]	Speed: 116.581663 samples/sec	 accuracy=31.099677	 loss=3.166184	 lr=0.092412
Epoch[062] Batch [2949]/[3760]	Speed: 117.859308 samples/sec	 accuracy=31.101165	 loss=3.165883	 lr=0.092405
Epoch[062] Batch [2999]/[3760]	Speed: 117.454213 samples/sec	 accuracy=31.083854	 loss=3.167438	 lr=0.092398
Epoch[062] Batch [3049]/[3760]	Speed: 117.258253 samples/sec	 accuracy=31.074795	 loss=3.168896	 lr=0.092391
Epoch[062] Batch [3099]/[3760]	Speed: 117.431358 samples/sec	 accuracy=31.075101	 loss=3.169234	 lr=0.092384
Epoch[062] Batch [3149]/[3760]	Speed: 117.388844 samples/sec	 accuracy=31.067460	 loss=3.169644	 lr=0.092377
Epoch[062] Batch [3199]/[3760]	Speed: 117.460720 samples/sec	 accuracy=31.079102	 loss=3.169391	 lr=0.092371
Epoch[062] Batch [3249]/[3760]	Speed: 116.945315 samples/sec	 accuracy=31.095192	 loss=3.169172	 lr=0.092364
Epoch[062] Batch [3299]/[3760]	Speed: 117.948715 samples/sec	 accuracy=31.095170	 loss=3.169241	 lr=0.092357
Epoch[062] Batch [3349]/[3760]	Speed: 117.367612 samples/sec	 accuracy=31.103078	 loss=3.169068	 lr=0.092350
Epoch[062] Batch [3399]/[3760]	Speed: 118.282202 samples/sec	 accuracy=31.075368	 loss=3.169339	 lr=0.092343
Epoch[062] Batch [3449]/[3760]	Speed: 116.793071 samples/sec	 accuracy=31.081975	 loss=3.169259	 lr=0.092336
Epoch[062] Batch [3499]/[3760]	Speed: 117.287209 samples/sec	 accuracy=31.079464	 loss=3.169567	 lr=0.092329
Epoch[062] Batch [3549]/[3760]	Speed: 117.237077 samples/sec	 accuracy=31.073944	 loss=3.169922	 lr=0.092323
Epoch[062] Batch [3599]/[3760]	Speed: 117.389756 samples/sec	 accuracy=31.077691	 loss=3.170345	 lr=0.092316
Epoch[062] Batch [3649]/[3760]	Speed: 117.796630 samples/sec	 accuracy=31.080479	 loss=3.169627	 lr=0.092309
Epoch[062] Batch [3699]/[3760]	Speed: 117.284451 samples/sec	 accuracy=31.085304	 loss=3.169332	 lr=0.092302
Epoch[062] Batch [3749]/[3760]	Speed: 120.837751 samples/sec	 accuracy=31.078333	 loss=3.169766	 lr=0.092295
Batch [0049]/[0303]: acc-top1=33.437500 acc-top5=58.843750
Batch [0099]/[0303]: acc-top1=31.078125 acc-top5=56.500000
Batch [0149]/[0303]: acc-top1=31.968750 acc-top5=57.583333
Batch [0199]/[0303]: acc-top1=34.578125 acc-top5=60.078125
Batch [0249]/[0303]: acc-top1=33.743750 acc-top5=59.637500
Batch [0299]/[0303]: acc-top1=33.104167 acc-top5=59.494792
[Epoch 062] training: accuracy=31.086686	 loss=3.169585
[Epoch 062] speed: 116 samples/sec	time cost: 2179.013447
[Epoch 062] validation: acc-top1=33.090965 acc-top5=59.529703 loss=3.125779
Epoch[063] Batch [0049]/[3760]	Speed: 86.520339 samples/sec	 accuracy=31.500000	 loss=3.127000	 lr=0.092287
Epoch[063] Batch [0099]/[3760]	Speed: 118.120958 samples/sec	 accuracy=31.765625	 loss=3.121675	 lr=0.092280
Epoch[063] Batch [0149]/[3760]	Speed: 117.015314 samples/sec	 accuracy=31.822917	 loss=3.131784	 lr=0.092273
Epoch[063] Batch [0199]/[3760]	Speed: 116.960522 samples/sec	 accuracy=31.625000	 loss=3.153011	 lr=0.092266
Epoch[063] Batch [0249]/[3760]	Speed: 117.222893 samples/sec	 accuracy=31.306250	 loss=3.165575	 lr=0.092259
Epoch[063] Batch [0299]/[3760]	Speed: 117.674630 samples/sec	 accuracy=31.187500	 loss=3.168579	 lr=0.092252
Epoch[063] Batch [0349]/[3760]	Speed: 117.257743 samples/sec	 accuracy=31.303571	 loss=3.169182	 lr=0.092245
Epoch[063] Batch [0399]/[3760]	Speed: 117.333283 samples/sec	 accuracy=31.269531	 loss=3.163971	 lr=0.092239
Epoch[063] Batch [0449]/[3760]	Speed: 117.297112 samples/sec	 accuracy=31.267361	 loss=3.165201	 lr=0.092232
Epoch[063] Batch [0499]/[3760]	Speed: 117.487395 samples/sec	 accuracy=31.171875	 loss=3.166916	 lr=0.092225
Epoch[063] Batch [0549]/[3760]	Speed: 117.930276 samples/sec	 accuracy=31.042614	 loss=3.168934	 lr=0.092218
Epoch[063] Batch [0599]/[3760]	Speed: 117.178255 samples/sec	 accuracy=30.846354	 loss=3.172275	 lr=0.092211
Epoch[063] Batch [0649]/[3760]	Speed: 117.950019 samples/sec	 accuracy=30.826923	 loss=3.172745	 lr=0.092204
Epoch[063] Batch [0699]/[3760]	Speed: 117.354225 samples/sec	 accuracy=30.857143	 loss=3.171459	 lr=0.092197
Epoch[063] Batch [0749]/[3760]	Speed: 117.306060 samples/sec	 accuracy=30.933333	 loss=3.167722	 lr=0.092190
Epoch[063] Batch [0799]/[3760]	Speed: 115.460924 samples/sec	 accuracy=30.945312	 loss=3.167148	 lr=0.092183
Epoch[063] Batch [0849]/[3760]	Speed: 117.581790 samples/sec	 accuracy=31.007353	 loss=3.165098	 lr=0.092176
Epoch[063] Batch [0899]/[3760]	Speed: 117.150710 samples/sec	 accuracy=30.979167	 loss=3.164999	 lr=0.092169
Epoch[063] Batch [0949]/[3760]	Speed: 117.560862 samples/sec	 accuracy=30.958882	 loss=3.165584	 lr=0.092162
Epoch[063] Batch [0999]/[3760]	Speed: 116.904273 samples/sec	 accuracy=31.042188	 loss=3.165670	 lr=0.092156
Epoch[063] Batch [1049]/[3760]	Speed: 117.891736 samples/sec	 accuracy=31.087798	 loss=3.162631	 lr=0.092149
Epoch[063] Batch [1099]/[3760]	Speed: 117.878052 samples/sec	 accuracy=31.069602	 loss=3.164402	 lr=0.092142
Epoch[063] Batch [1149]/[3760]	Speed: 116.146327 samples/sec	 accuracy=31.061141	 loss=3.165567	 lr=0.092135
Epoch[063] Batch [1199]/[3760]	Speed: 118.335020 samples/sec	 accuracy=31.058594	 loss=3.168569	 lr=0.092128
Epoch[063] Batch [1249]/[3760]	Speed: 116.569394 samples/sec	 accuracy=31.060000	 loss=3.167007	 lr=0.092121
Epoch[063] Batch [1299]/[3760]	Speed: 118.063440 samples/sec	 accuracy=31.098558	 loss=3.165066	 lr=0.092114
Epoch[063] Batch [1349]/[3760]	Speed: 117.484759 samples/sec	 accuracy=31.140046	 loss=3.162932	 lr=0.092107
Epoch[063] Batch [1399]/[3760]	Speed: 117.489964 samples/sec	 accuracy=31.190848	 loss=3.161146	 lr=0.092100
Epoch[063] Batch [1449]/[3760]	Speed: 116.680861 samples/sec	 accuracy=31.140086	 loss=3.162462	 lr=0.092093
Epoch[063] Batch [1499]/[3760]	Speed: 117.687108 samples/sec	 accuracy=31.097917	 loss=3.162006	 lr=0.092086
Epoch[063] Batch [1549]/[3760]	Speed: 117.707420 samples/sec	 accuracy=31.117944	 loss=3.162073	 lr=0.092079
Epoch[063] Batch [1599]/[3760]	Speed: 117.393695 samples/sec	 accuracy=31.101562	 loss=3.162710	 lr=0.092072
Epoch[063] Batch [1649]/[3760]	Speed: 117.361563 samples/sec	 accuracy=31.085227	 loss=3.163725	 lr=0.092065
Epoch[063] Batch [1699]/[3760]	Speed: 116.930489 samples/sec	 accuracy=31.042279	 loss=3.166119	 lr=0.092058
Epoch[063] Batch [1749]/[3760]	Speed: 117.295360 samples/sec	 accuracy=31.040179	 loss=3.166733	 lr=0.092051
Epoch[063] Batch [1799]/[3760]	Speed: 117.204275 samples/sec	 accuracy=31.053819	 loss=3.165607	 lr=0.092044
Epoch[063] Batch [1849]/[3760]	Speed: 117.382057 samples/sec	 accuracy=31.069257	 loss=3.166051	 lr=0.092037
Epoch[063] Batch [1899]/[3760]	Speed: 117.290860 samples/sec	 accuracy=31.055099	 loss=3.167159	 lr=0.092030
Epoch[063] Batch [1949]/[3760]	Speed: 116.815979 samples/sec	 accuracy=31.049679	 loss=3.167532	 lr=0.092023
Epoch[063] Batch [1999]/[3760]	Speed: 116.978054 samples/sec	 accuracy=31.059375	 loss=3.167329	 lr=0.092016
Epoch[063] Batch [2049]/[3760]	Speed: 116.346376 samples/sec	 accuracy=31.064787	 loss=3.167300	 lr=0.092009
Epoch[063] Batch [2099]/[3760]	Speed: 118.823559 samples/sec	 accuracy=31.121280	 loss=3.165804	 lr=0.092002
Epoch[063] Batch [2149]/[3760]	Speed: 117.213227 samples/sec	 accuracy=31.079215	 loss=3.167564	 lr=0.091995
Epoch[063] Batch [2199]/[3760]	Speed: 117.345877 samples/sec	 accuracy=31.052557	 loss=3.168414	 lr=0.091988
Epoch[063] Batch [2249]/[3760]	Speed: 117.202411 samples/sec	 accuracy=31.065278	 loss=3.167738	 lr=0.091981
Epoch[063] Batch [2299]/[3760]	Speed: 117.469164 samples/sec	 accuracy=31.061141	 loss=3.168405	 lr=0.091974
Epoch[063] Batch [2349]/[3760]	Speed: 117.574930 samples/sec	 accuracy=31.063830	 loss=3.168152	 lr=0.091967
Epoch[063] Batch [2399]/[3760]	Speed: 117.841242 samples/sec	 accuracy=31.067057	 loss=3.168003	 lr=0.091960
Epoch[063] Batch [2449]/[3760]	Speed: 117.336373 samples/sec	 accuracy=31.042092	 loss=3.167835	 lr=0.091953
Epoch[063] Batch [2499]/[3760]	Speed: 117.183718 samples/sec	 accuracy=31.028750	 loss=3.168317	 lr=0.091946
Epoch[063] Batch [2549]/[3760]	Speed: 117.606711 samples/sec	 accuracy=31.010417	 loss=3.169887	 lr=0.091939
Epoch[063] Batch [2599]/[3760]	Speed: 117.567145 samples/sec	 accuracy=31.003005	 loss=3.169625	 lr=0.091932
Epoch[063] Batch [2649]/[3760]	Speed: 117.081650 samples/sec	 accuracy=30.995283	 loss=3.170052	 lr=0.091925
Epoch[063] Batch [2699]/[3760]	Speed: 117.202419 samples/sec	 accuracy=30.968750	 loss=3.170123	 lr=0.091918
Epoch[063] Batch [2749]/[3760]	Speed: 117.475021 samples/sec	 accuracy=30.961932	 loss=3.170986	 lr=0.091911
Epoch[063] Batch [2799]/[3760]	Speed: 117.479141 samples/sec	 accuracy=30.993862	 loss=3.171145	 lr=0.091904
Epoch[063] Batch [2849]/[3760]	Speed: 117.458544 samples/sec	 accuracy=31.000000	 loss=3.170548	 lr=0.091897
Epoch[063] Batch [2899]/[3760]	Speed: 116.612383 samples/sec	 accuracy=30.981681	 loss=3.171338	 lr=0.091890
Epoch[063] Batch [2949]/[3760]	Speed: 118.167455 samples/sec	 accuracy=30.989407	 loss=3.171241	 lr=0.091883
Epoch[063] Batch [2999]/[3760]	Speed: 116.872193 samples/sec	 accuracy=30.978646	 loss=3.171863	 lr=0.091876
Epoch[063] Batch [3049]/[3760]	Speed: 117.453716 samples/sec	 accuracy=30.995389	 loss=3.171142	 lr=0.091869
Epoch[063] Batch [3099]/[3760]	Speed: 116.587635 samples/sec	 accuracy=30.980343	 loss=3.170784	 lr=0.091862
Epoch[063] Batch [3149]/[3760]	Speed: 116.914651 samples/sec	 accuracy=30.975694	 loss=3.171000	 lr=0.091855
Epoch[063] Batch [3199]/[3760]	Speed: 118.149367 samples/sec	 accuracy=30.991699	 loss=3.169988	 lr=0.091848
Epoch[063] Batch [3249]/[3760]	Speed: 117.212205 samples/sec	 accuracy=30.976923	 loss=3.170777	 lr=0.091841
Epoch[063] Batch [3299]/[3760]	Speed: 117.627522 samples/sec	 accuracy=30.978693	 loss=3.171271	 lr=0.091834
Epoch[063] Batch [3349]/[3760]	Speed: 117.424711 samples/sec	 accuracy=30.969683	 loss=3.171431	 lr=0.091826
Epoch[063] Batch [3399]/[3760]	Speed: 117.192333 samples/sec	 accuracy=30.966452	 loss=3.171997	 lr=0.091819
Epoch[063] Batch [3449]/[3760]	Speed: 117.703501 samples/sec	 accuracy=30.977808	 loss=3.171701	 lr=0.091812
Epoch[063] Batch [3499]/[3760]	Speed: 117.613039 samples/sec	 accuracy=30.969196	 loss=3.172028	 lr=0.091805
Epoch[063] Batch [3549]/[3760]	Speed: 117.165163 samples/sec	 accuracy=30.995158	 loss=3.171924	 lr=0.091798
Epoch[063] Batch [3599]/[3760]	Speed: 118.227917 samples/sec	 accuracy=30.996528	 loss=3.172698	 lr=0.091791
Epoch[063] Batch [3649]/[3760]	Speed: 117.417137 samples/sec	 accuracy=31.005137	 loss=3.172241	 lr=0.091784
Epoch[063] Batch [3699]/[3760]	Speed: 117.304963 samples/sec	 accuracy=31.005490	 loss=3.172662	 lr=0.091777
Epoch[063] Batch [3749]/[3760]	Speed: 119.961272 samples/sec	 accuracy=31.009167	 loss=3.172341	 lr=0.091770
Batch [0049]/[0303]: acc-top1=31.843750 acc-top5=56.843750
Batch [0099]/[0303]: acc-top1=29.140625 acc-top5=54.984375
Batch [0149]/[0303]: acc-top1=30.447917 acc-top5=55.791667
Batch [0199]/[0303]: acc-top1=33.500000 acc-top5=58.570312
Batch [0249]/[0303]: acc-top1=32.187500 acc-top5=58.031250
Batch [0299]/[0303]: acc-top1=30.854167 acc-top5=57.296875
[Epoch 063] training: accuracy=31.008145	 loss=3.172211
[Epoch 063] speed: 116 samples/sec	time cost: 2179.407454
[Epoch 063] validation: acc-top1=30.801361 acc-top5=57.245256 loss=3.334008
Epoch[064] Batch [0049]/[3759]	Speed: 87.635865 samples/sec	 accuracy=32.437500	 loss=3.117713	 lr=0.091761
Epoch[064] Batch [0099]/[3759]	Speed: 117.727398 samples/sec	 accuracy=32.406250	 loss=3.111479	 lr=0.091754
Epoch[064] Batch [0149]/[3759]	Speed: 117.229624 samples/sec	 accuracy=31.833333	 loss=3.149492	 lr=0.091747
Epoch[064] Batch [0199]/[3759]	Speed: 116.555544 samples/sec	 accuracy=31.796875	 loss=3.151365	 lr=0.091740
Epoch[064] Batch [0249]/[3759]	Speed: 117.463302 samples/sec	 accuracy=31.975000	 loss=3.130682	 lr=0.091733
Epoch[064] Batch [0299]/[3759]	Speed: 117.195510 samples/sec	 accuracy=31.895833	 loss=3.137338	 lr=0.091726
Epoch[064] Batch [0349]/[3759]	Speed: 117.104247 samples/sec	 accuracy=32.004464	 loss=3.136636	 lr=0.091719
Epoch[064] Batch [0399]/[3759]	Speed: 117.763102 samples/sec	 accuracy=31.781250	 loss=3.145338	 lr=0.091712
Epoch[064] Batch [0449]/[3759]	Speed: 116.820172 samples/sec	 accuracy=31.833333	 loss=3.146091	 lr=0.091704
Epoch[064] Batch [0499]/[3759]	Speed: 116.049774 samples/sec	 accuracy=31.696875	 loss=3.152817	 lr=0.091697
Epoch[064] Batch [0549]/[3759]	Speed: 117.411421 samples/sec	 accuracy=31.616477	 loss=3.149517	 lr=0.091690
Epoch[064] Batch [0599]/[3759]	Speed: 117.812246 samples/sec	 accuracy=31.570312	 loss=3.150565	 lr=0.091683
Epoch[064] Batch [0649]/[3759]	Speed: 117.646742 samples/sec	 accuracy=31.490385	 loss=3.152119	 lr=0.091676
Epoch[064] Batch [0699]/[3759]	Speed: 117.592781 samples/sec	 accuracy=31.421875	 loss=3.152066	 lr=0.091669
Epoch[064] Batch [0749]/[3759]	Speed: 117.558913 samples/sec	 accuracy=31.454167	 loss=3.153935	 lr=0.091662
Epoch[064] Batch [0799]/[3759]	Speed: 117.128808 samples/sec	 accuracy=31.300781	 loss=3.156636	 lr=0.091655
Epoch[064] Batch [0849]/[3759]	Speed: 116.789596 samples/sec	 accuracy=31.319853	 loss=3.153610	 lr=0.091647
Epoch[064] Batch [0899]/[3759]	Speed: 116.821471 samples/sec	 accuracy=31.402778	 loss=3.150496	 lr=0.091640
Epoch[064] Batch [0949]/[3759]	Speed: 117.790891 samples/sec	 accuracy=31.416118	 loss=3.149676	 lr=0.091633
Epoch[064] Batch [0999]/[3759]	Speed: 117.318870 samples/sec	 accuracy=31.410938	 loss=3.148988	 lr=0.091626
Epoch[064] Batch [1049]/[3759]	Speed: 118.139124 samples/sec	 accuracy=31.377976	 loss=3.150365	 lr=0.091619
Epoch[064] Batch [1099]/[3759]	Speed: 114.569095 samples/sec	 accuracy=31.329545	 loss=3.151317	 lr=0.091612
Epoch[064] Batch [1149]/[3759]	Speed: 117.400212 samples/sec	 accuracy=31.309783	 loss=3.151096	 lr=0.091605
Epoch[064] Batch [1199]/[3759]	Speed: 116.569673 samples/sec	 accuracy=31.328125	 loss=3.152650	 lr=0.091597
Epoch[064] Batch [1249]/[3759]	Speed: 117.944597 samples/sec	 accuracy=31.348750	 loss=3.151874	 lr=0.091590
Epoch[064] Batch [1299]/[3759]	Speed: 117.235683 samples/sec	 accuracy=31.308894	 loss=3.151825	 lr=0.091583
Epoch[064] Batch [1349]/[3759]	Speed: 117.176188 samples/sec	 accuracy=31.288194	 loss=3.151631	 lr=0.091576
Epoch[064] Batch [1399]/[3759]	Speed: 117.255291 samples/sec	 accuracy=31.281250	 loss=3.151055	 lr=0.091569
Epoch[064] Batch [1449]/[3759]	Speed: 117.319103 samples/sec	 accuracy=31.290948	 loss=3.151496	 lr=0.091562
Epoch[064] Batch [1499]/[3759]	Speed: 116.485041 samples/sec	 accuracy=31.312500	 loss=3.151461	 lr=0.091554
Epoch[064] Batch [1549]/[3759]	Speed: 117.209968 samples/sec	 accuracy=31.276210	 loss=3.152602	 lr=0.091547
Epoch[064] Batch [1599]/[3759]	Speed: 117.431941 samples/sec	 accuracy=31.321289	 loss=3.151426	 lr=0.091540
Epoch[064] Batch [1649]/[3759]	Speed: 117.403187 samples/sec	 accuracy=31.329545	 loss=3.152758	 lr=0.091533
Epoch[064] Batch [1699]/[3759]	Speed: 116.720935 samples/sec	 accuracy=31.272978	 loss=3.155510	 lr=0.091526
Epoch[064] Batch [1749]/[3759]	Speed: 117.699448 samples/sec	 accuracy=31.311607	 loss=3.153724	 lr=0.091519
Epoch[064] Batch [1799]/[3759]	Speed: 117.812927 samples/sec	 accuracy=31.329861	 loss=3.154716	 lr=0.091511
Epoch[064] Batch [1849]/[3759]	Speed: 117.307146 samples/sec	 accuracy=31.293919	 loss=3.156195	 lr=0.091504
Epoch[064] Batch [1899]/[3759]	Speed: 117.672875 samples/sec	 accuracy=31.295230	 loss=3.155879	 lr=0.091497
Epoch[064] Batch [1949]/[3759]	Speed: 117.821791 samples/sec	 accuracy=31.306891	 loss=3.155004	 lr=0.091490
Epoch[064] Batch [1999]/[3759]	Speed: 118.432415 samples/sec	 accuracy=31.283594	 loss=3.156198	 lr=0.091483
Epoch[064] Batch [2049]/[3759]	Speed: 117.058547 samples/sec	 accuracy=31.278963	 loss=3.156009	 lr=0.091475
Epoch[064] Batch [2099]/[3759]	Speed: 117.743560 samples/sec	 accuracy=31.295387	 loss=3.154643	 lr=0.091468
Epoch[064] Batch [2149]/[3759]	Speed: 117.451977 samples/sec	 accuracy=31.263081	 loss=3.154436	 lr=0.091461
Epoch[064] Batch [2199]/[3759]	Speed: 117.388352 samples/sec	 accuracy=31.250710	 loss=3.155640	 lr=0.091454
Epoch[064] Batch [2249]/[3759]	Speed: 117.507769 samples/sec	 accuracy=31.235417	 loss=3.155490	 lr=0.091447
Epoch[064] Batch [2299]/[3759]	Speed: 116.666847 samples/sec	 accuracy=31.243207	 loss=3.155547	 lr=0.091439
Epoch[064] Batch [2349]/[3759]	Speed: 117.911108 samples/sec	 accuracy=31.232713	 loss=3.157157	 lr=0.091432
Epoch[064] Batch [2399]/[3759]	Speed: 117.934415 samples/sec	 accuracy=31.235677	 loss=3.156987	 lr=0.091425
Epoch[064] Batch [2449]/[3759]	Speed: 117.753794 samples/sec	 accuracy=31.247449	 loss=3.156877	 lr=0.091418
Epoch[064] Batch [2499]/[3759]	Speed: 117.813749 samples/sec	 accuracy=31.236875	 loss=3.157517	 lr=0.091410
Epoch[064] Batch [2549]/[3759]	Speed: 117.140077 samples/sec	 accuracy=31.215686	 loss=3.158304	 lr=0.091403
Epoch[064] Batch [2599]/[3759]	Speed: 117.112868 samples/sec	 accuracy=31.225361	 loss=3.158877	 lr=0.091396
Epoch[064] Batch [2649]/[3759]	Speed: 117.489646 samples/sec	 accuracy=31.237028	 loss=3.158006	 lr=0.091389
Epoch[064] Batch [2699]/[3759]	Speed: 117.390236 samples/sec	 accuracy=31.206019	 loss=3.159231	 lr=0.091382
Epoch[064] Batch [2749]/[3759]	Speed: 117.690037 samples/sec	 accuracy=31.197159	 loss=3.159381	 lr=0.091374
Epoch[064] Batch [2799]/[3759]	Speed: 117.329461 samples/sec	 accuracy=31.186942	 loss=3.160025	 lr=0.091367
Epoch[064] Batch [2849]/[3759]	Speed: 117.756741 samples/sec	 accuracy=31.180921	 loss=3.160656	 lr=0.091360
Epoch[064] Batch [2899]/[3759]	Speed: 117.138259 samples/sec	 accuracy=31.163793	 loss=3.162330	 lr=0.091353
Epoch[064] Batch [2949]/[3759]	Speed: 117.531408 samples/sec	 accuracy=31.185911	 loss=3.161415	 lr=0.091345
Epoch[064] Batch [2999]/[3759]	Speed: 117.675304 samples/sec	 accuracy=31.184375	 loss=3.161289	 lr=0.091338
Epoch[064] Batch [3049]/[3759]	Speed: 117.293951 samples/sec	 accuracy=31.195184	 loss=3.161049	 lr=0.091331
Epoch[064] Batch [3099]/[3759]	Speed: 117.904465 samples/sec	 accuracy=31.199597	 loss=3.161124	 lr=0.091324
Epoch[064] Batch [3149]/[3759]	Speed: 117.808036 samples/sec	 accuracy=31.212798	 loss=3.161296	 lr=0.091316
Epoch[064] Batch [3199]/[3759]	Speed: 116.600335 samples/sec	 accuracy=31.233887	 loss=3.161153	 lr=0.091309
Epoch[064] Batch [3249]/[3759]	Speed: 117.225335 samples/sec	 accuracy=31.229808	 loss=3.161440	 lr=0.091302
Epoch[064] Batch [3299]/[3759]	Speed: 117.925710 samples/sec	 accuracy=31.204545	 loss=3.161605	 lr=0.091294
Epoch[064] Batch [3349]/[3759]	Speed: 117.266451 samples/sec	 accuracy=31.200560	 loss=3.162382	 lr=0.091287
Epoch[064] Batch [3399]/[3759]	Speed: 117.487044 samples/sec	 accuracy=31.192555	 loss=3.162918	 lr=0.091280
Epoch[064] Batch [3449]/[3759]	Speed: 116.728062 samples/sec	 accuracy=31.170290	 loss=3.163345	 lr=0.091273
Epoch[064] Batch [3499]/[3759]	Speed: 116.984787 samples/sec	 accuracy=31.197321	 loss=3.162338	 lr=0.091265
Epoch[064] Batch [3549]/[3759]	Speed: 117.950629 samples/sec	 accuracy=31.208627	 loss=3.162136	 lr=0.091258
Epoch[064] Batch [3599]/[3759]	Speed: 117.360194 samples/sec	 accuracy=31.190104	 loss=3.162591	 lr=0.091251
Epoch[064] Batch [3649]/[3759]	Speed: 117.688408 samples/sec	 accuracy=31.192209	 loss=3.162279	 lr=0.091243
Epoch[064] Batch [3699]/[3759]	Speed: 117.025315 samples/sec	 accuracy=31.167652	 loss=3.162979	 lr=0.091236
Epoch[064] Batch [3749]/[3759]	Speed: 119.862142 samples/sec	 accuracy=31.162500	 loss=3.163155	 lr=0.091229
Batch [0049]/[0303]: acc-top1=29.843750 acc-top5=54.718750
Batch [0099]/[0303]: acc-top1=27.140625 acc-top5=52.875000
Batch [0149]/[0303]: acc-top1=29.083333 acc-top5=54.458333
Batch [0199]/[0303]: acc-top1=31.992188 acc-top5=57.273438
Batch [0249]/[0303]: acc-top1=30.912500 acc-top5=56.737500
Batch [0299]/[0303]: acc-top1=30.572917 acc-top5=56.723958
[Epoch 064] training: accuracy=31.156474	 loss=3.163458
[Epoch 064] speed: 116 samples/sec	time cost: 2179.332755
[Epoch 064] validation: acc-top1=30.682756 acc-top5=56.775990 loss=3.338909
Epoch[065] Batch [0049]/[3760]	Speed: 88.293470 samples/sec	 accuracy=31.468750	 loss=3.160828	 lr=0.091220
Epoch[065] Batch [0099]/[3760]	Speed: 117.662023 samples/sec	 accuracy=31.234375	 loss=3.155056	 lr=0.091213
Epoch[065] Batch [0149]/[3760]	Speed: 117.661051 samples/sec	 accuracy=31.781250	 loss=3.130167	 lr=0.091206
Epoch[065] Batch [0199]/[3760]	Speed: 117.384316 samples/sec	 accuracy=31.390625	 loss=3.136351	 lr=0.091198
Epoch[065] Batch [0249]/[3760]	Speed: 117.477135 samples/sec	 accuracy=31.356250	 loss=3.141222	 lr=0.091191
Epoch[065] Batch [0299]/[3760]	Speed: 116.533527 samples/sec	 accuracy=31.364583	 loss=3.138576	 lr=0.091184
Epoch[065] Batch [0349]/[3760]	Speed: 116.708131 samples/sec	 accuracy=31.419643	 loss=3.137765	 lr=0.091176
Epoch[065] Batch [0399]/[3760]	Speed: 116.672454 samples/sec	 accuracy=31.460938	 loss=3.141703	 lr=0.091169
Epoch[065] Batch [0449]/[3760]	Speed: 117.446976 samples/sec	 accuracy=31.430556	 loss=3.146087	 lr=0.091162
Epoch[065] Batch [0499]/[3760]	Speed: 117.649443 samples/sec	 accuracy=31.562500	 loss=3.143309	 lr=0.091154
Epoch[065] Batch [0549]/[3760]	Speed: 117.198737 samples/sec	 accuracy=31.531250	 loss=3.145046	 lr=0.091147
Epoch[065] Batch [0599]/[3760]	Speed: 118.066087 samples/sec	 accuracy=31.502604	 loss=3.146252	 lr=0.091140
Epoch[065] Batch [0649]/[3760]	Speed: 117.439077 samples/sec	 accuracy=31.454327	 loss=3.142879	 lr=0.091132
Epoch[065] Batch [0699]/[3760]	Speed: 117.634200 samples/sec	 accuracy=31.479911	 loss=3.140064	 lr=0.091125
Epoch[065] Batch [0749]/[3760]	Speed: 116.928628 samples/sec	 accuracy=31.433333	 loss=3.142661	 lr=0.091118
Epoch[065] Batch [0799]/[3760]	Speed: 117.933661 samples/sec	 accuracy=31.431641	 loss=3.140584	 lr=0.091110
Epoch[065] Batch [0849]/[3760]	Speed: 117.164059 samples/sec	 accuracy=31.373162	 loss=3.142894	 lr=0.091103
Epoch[065] Batch [0899]/[3760]	Speed: 117.008097 samples/sec	 accuracy=31.277778	 loss=3.143794	 lr=0.091096
Epoch[065] Batch [0949]/[3760]	Speed: 118.034323 samples/sec	 accuracy=31.271382	 loss=3.143811	 lr=0.091088
Epoch[065] Batch [0999]/[3760]	Speed: 117.830413 samples/sec	 accuracy=31.309375	 loss=3.144837	 lr=0.091081
Epoch[065] Batch [1049]/[3760]	Speed: 117.360079 samples/sec	 accuracy=31.319940	 loss=3.147752	 lr=0.091074
Epoch[065] Batch [1099]/[3760]	Speed: 117.367437 samples/sec	 accuracy=31.284091	 loss=3.150669	 lr=0.091066
Epoch[065] Batch [1149]/[3760]	Speed: 117.192786 samples/sec	 accuracy=31.255435	 loss=3.151128	 lr=0.091059
Epoch[065] Batch [1199]/[3760]	Speed: 117.567784 samples/sec	 accuracy=31.289062	 loss=3.150381	 lr=0.091052
Epoch[065] Batch [1249]/[3760]	Speed: 117.592055 samples/sec	 accuracy=31.258750	 loss=3.152718	 lr=0.091044
Epoch[065] Batch [1299]/[3760]	Speed: 117.388932 samples/sec	 accuracy=31.247596	 loss=3.153548	 lr=0.091037
Epoch[065] Batch [1349]/[3760]	Speed: 117.078964 samples/sec	 accuracy=31.231481	 loss=3.154669	 lr=0.091030
Epoch[065] Batch [1399]/[3760]	Speed: 116.874433 samples/sec	 accuracy=31.255580	 loss=3.155564	 lr=0.091022
Epoch[065] Batch [1449]/[3760]	Speed: 117.286123 samples/sec	 accuracy=31.257543	 loss=3.153900	 lr=0.091015
Epoch[065] Batch [1499]/[3760]	Speed: 116.921763 samples/sec	 accuracy=31.279167	 loss=3.153551	 lr=0.091007
Epoch[065] Batch [1549]/[3760]	Speed: 116.974528 samples/sec	 accuracy=31.252016	 loss=3.154092	 lr=0.091000
Epoch[065] Batch [1599]/[3760]	Speed: 117.243060 samples/sec	 accuracy=31.280273	 loss=3.153554	 lr=0.090993
Epoch[065] Batch [1649]/[3760]	Speed: 116.745987 samples/sec	 accuracy=31.261364	 loss=3.154096	 lr=0.090985
Epoch[065] Batch [1699]/[3760]	Speed: 118.053671 samples/sec	 accuracy=31.299632	 loss=3.152541	 lr=0.090978
Epoch[065] Batch [1749]/[3760]	Speed: 116.539730 samples/sec	 accuracy=31.297321	 loss=3.154921	 lr=0.090971
Epoch[065] Batch [1799]/[3760]	Speed: 117.479363 samples/sec	 accuracy=31.263021	 loss=3.157062	 lr=0.090963
Epoch[065] Batch [1849]/[3760]	Speed: 118.107020 samples/sec	 accuracy=31.286318	 loss=3.157114	 lr=0.090956
Epoch[065] Batch [1899]/[3760]	Speed: 116.542158 samples/sec	 accuracy=31.274671	 loss=3.156926	 lr=0.090948
Epoch[065] Batch [1949]/[3760]	Speed: 117.580552 samples/sec	 accuracy=31.286859	 loss=3.156495	 lr=0.090941
Epoch[065] Batch [1999]/[3760]	Speed: 117.811977 samples/sec	 accuracy=31.310156	 loss=3.155416	 lr=0.090934
Epoch[065] Batch [2049]/[3760]	Speed: 118.253750 samples/sec	 accuracy=31.340701	 loss=3.153014	 lr=0.090926
Epoch[065] Batch [2099]/[3760]	Speed: 117.361904 samples/sec	 accuracy=31.331101	 loss=3.152715	 lr=0.090919
Epoch[065] Batch [2149]/[3760]	Speed: 116.775200 samples/sec	 accuracy=31.314680	 loss=3.153306	 lr=0.090911
Epoch[065] Batch [2199]/[3760]	Speed: 117.482846 samples/sec	 accuracy=31.307528	 loss=3.153400	 lr=0.090904
Epoch[065] Batch [2249]/[3760]	Speed: 117.184561 samples/sec	 accuracy=31.313889	 loss=3.152262	 lr=0.090896
Epoch[065] Batch [2299]/[3760]	Speed: 117.351468 samples/sec	 accuracy=31.303668	 loss=3.152672	 lr=0.090889
Epoch[065] Batch [2349]/[3760]	Speed: 117.776096 samples/sec	 accuracy=31.285239	 loss=3.153659	 lr=0.090882
Epoch[065] Batch [2399]/[3760]	Speed: 117.370194 samples/sec	 accuracy=31.279948	 loss=3.153688	 lr=0.090874
Epoch[065] Batch [2449]/[3760]	Speed: 116.547147 samples/sec	 accuracy=31.271046	 loss=3.154923	 lr=0.090867
Epoch[065] Batch [2499]/[3760]	Speed: 117.465761 samples/sec	 accuracy=31.268125	 loss=3.155172	 lr=0.090859
Epoch[065] Batch [2549]/[3760]	Speed: 117.246564 samples/sec	 accuracy=31.297794	 loss=3.153967	 lr=0.090852
Epoch[065] Batch [2599]/[3760]	Speed: 117.319709 samples/sec	 accuracy=31.302284	 loss=3.153516	 lr=0.090844
Epoch[065] Batch [2649]/[3760]	Speed: 117.374793 samples/sec	 accuracy=31.300118	 loss=3.153495	 lr=0.090837
Epoch[065] Batch [2699]/[3760]	Speed: 117.550330 samples/sec	 accuracy=31.311921	 loss=3.152488	 lr=0.090830
Epoch[065] Batch [2749]/[3760]	Speed: 117.009820 samples/sec	 accuracy=31.337500	 loss=3.151932	 lr=0.090822
Epoch[065] Batch [2799]/[3760]	Speed: 117.000994 samples/sec	 accuracy=31.342634	 loss=3.152504	 lr=0.090815
Epoch[065] Batch [2849]/[3760]	Speed: 116.768956 samples/sec	 accuracy=31.317982	 loss=3.154079	 lr=0.090807
Epoch[065] Batch [2899]/[3760]	Speed: 117.291774 samples/sec	 accuracy=31.306034	 loss=3.154556	 lr=0.090800
Epoch[065] Batch [2949]/[3760]	Speed: 118.020016 samples/sec	 accuracy=31.299788	 loss=3.155075	 lr=0.090792
Epoch[065] Batch [2999]/[3760]	Speed: 117.827842 samples/sec	 accuracy=31.302083	 loss=3.154925	 lr=0.090785
Epoch[065] Batch [3049]/[3760]	Speed: 117.515522 samples/sec	 accuracy=31.294570	 loss=3.155123	 lr=0.090777
Epoch[065] Batch [3099]/[3760]	Speed: 117.065799 samples/sec	 accuracy=31.318548	 loss=3.155263	 lr=0.090770
Epoch[065] Batch [3149]/[3760]	Speed: 116.261838 samples/sec	 accuracy=31.334821	 loss=3.154426	 lr=0.090762
Epoch[065] Batch [3199]/[3760]	Speed: 117.670667 samples/sec	 accuracy=31.335449	 loss=3.155209	 lr=0.090755
Epoch[065] Batch [3249]/[3760]	Speed: 116.172620 samples/sec	 accuracy=31.333654	 loss=3.155573	 lr=0.090748
Epoch[065] Batch [3299]/[3760]	Speed: 117.214417 samples/sec	 accuracy=31.341383	 loss=3.155135	 lr=0.090740
Epoch[065] Batch [3349]/[3760]	Speed: 116.631859 samples/sec	 accuracy=31.354478	 loss=3.154623	 lr=0.090733
Epoch[065] Batch [3399]/[3760]	Speed: 117.161116 samples/sec	 accuracy=31.354779	 loss=3.154790	 lr=0.090725
Epoch[065] Batch [3449]/[3760]	Speed: 117.543782 samples/sec	 accuracy=31.340127	 loss=3.154782	 lr=0.090718
Epoch[065] Batch [3499]/[3760]	Speed: 117.834391 samples/sec	 accuracy=31.328571	 loss=3.155100	 lr=0.090710
Epoch[065] Batch [3549]/[3760]	Speed: 117.258667 samples/sec	 accuracy=31.332746	 loss=3.155413	 lr=0.090703
Epoch[065] Batch [3599]/[3760]	Speed: 117.529873 samples/sec	 accuracy=31.317708	 loss=3.156228	 lr=0.090695
Epoch[065] Batch [3649]/[3760]	Speed: 117.242718 samples/sec	 accuracy=31.304795	 loss=3.156442	 lr=0.090688
Epoch[065] Batch [3699]/[3760]	Speed: 117.256688 samples/sec	 accuracy=31.308277	 loss=3.156227	 lr=0.090680
Epoch[065] Batch [3749]/[3760]	Speed: 120.201415 samples/sec	 accuracy=31.300417	 loss=3.156724	 lr=0.090673
Batch [0049]/[0303]: acc-top1=33.718750 acc-top5=58.312500
Batch [0099]/[0303]: acc-top1=30.515625 acc-top5=56.265625
Batch [0149]/[0303]: acc-top1=31.833333 acc-top5=57.729167
Batch [0199]/[0303]: acc-top1=34.554688 acc-top5=59.953125
Batch [0249]/[0303]: acc-top1=33.925000 acc-top5=59.768750
Batch [0299]/[0303]: acc-top1=33.505208 acc-top5=59.515625
[Epoch 065] training: accuracy=31.299036	 loss=3.156735
[Epoch 065] speed: 116 samples/sec	time cost: 2180.076907
[Epoch 065] validation: acc-top1=33.426155 acc-top5=59.467822 loss=3.144279
Epoch[066] Batch [0049]/[3759]	Speed: 86.531457 samples/sec	 accuracy=31.281250	 loss=3.127585	 lr=0.090664
Epoch[066] Batch [0099]/[3759]	Speed: 116.902923 samples/sec	 accuracy=31.171875	 loss=3.157555	 lr=0.090656
Epoch[066] Batch [0149]/[3759]	Speed: 117.903709 samples/sec	 accuracy=31.562500	 loss=3.140514	 lr=0.090649
Epoch[066] Batch [0199]/[3759]	Speed: 117.647749 samples/sec	 accuracy=31.492188	 loss=3.143474	 lr=0.090641
Epoch[066] Batch [0249]/[3759]	Speed: 116.216409 samples/sec	 accuracy=31.493750	 loss=3.142791	 lr=0.090634
Epoch[066] Batch [0299]/[3759]	Speed: 117.696177 samples/sec	 accuracy=31.567708	 loss=3.139933	 lr=0.090626
Epoch[066] Batch [0349]/[3759]	Speed: 117.800669 samples/sec	 accuracy=31.816964	 loss=3.130070	 lr=0.090619
Epoch[066] Batch [0399]/[3759]	Speed: 117.812874 samples/sec	 accuracy=31.773438	 loss=3.135990	 lr=0.090611
Epoch[066] Batch [0449]/[3759]	Speed: 118.230180 samples/sec	 accuracy=31.729167	 loss=3.136786	 lr=0.090603
Epoch[066] Batch [0499]/[3759]	Speed: 117.522111 samples/sec	 accuracy=31.668750	 loss=3.136423	 lr=0.090596
Epoch[066] Batch [0549]/[3759]	Speed: 117.437026 samples/sec	 accuracy=31.792614	 loss=3.130117	 lr=0.090588
Epoch[066] Batch [0599]/[3759]	Speed: 117.768837 samples/sec	 accuracy=31.570312	 loss=3.138579	 lr=0.090581
Epoch[066] Batch [0649]/[3759]	Speed: 116.778072 samples/sec	 accuracy=31.560096	 loss=3.140537	 lr=0.090573
Epoch[066] Batch [0699]/[3759]	Speed: 117.136847 samples/sec	 accuracy=31.598214	 loss=3.138771	 lr=0.090566
Epoch[066] Batch [0749]/[3759]	Speed: 117.592347 samples/sec	 accuracy=31.522917	 loss=3.142065	 lr=0.090558
Epoch[066] Batch [0799]/[3759]	Speed: 116.718739 samples/sec	 accuracy=31.517578	 loss=3.141189	 lr=0.090551
Epoch[066] Batch [0849]/[3759]	Speed: 116.981326 samples/sec	 accuracy=31.520221	 loss=3.142716	 lr=0.090543
Epoch[066] Batch [0899]/[3759]	Speed: 117.311253 samples/sec	 accuracy=31.559028	 loss=3.140731	 lr=0.090536
Epoch[066] Batch [0949]/[3759]	Speed: 116.930647 samples/sec	 accuracy=31.524671	 loss=3.139877	 lr=0.090528
Epoch[066] Batch [0999]/[3759]	Speed: 117.466392 samples/sec	 accuracy=31.526563	 loss=3.139225	 lr=0.090521
Epoch[066] Batch [1049]/[3759]	Speed: 117.966940 samples/sec	 accuracy=31.534226	 loss=3.139773	 lr=0.090513
Epoch[066] Batch [1099]/[3759]	Speed: 117.511864 samples/sec	 accuracy=31.522727	 loss=3.141839	 lr=0.090505
Epoch[066] Batch [1149]/[3759]	Speed: 116.851014 samples/sec	 accuracy=31.517663	 loss=3.141182	 lr=0.090498
Epoch[066] Batch [1199]/[3759]	Speed: 117.929848 samples/sec	 accuracy=31.470052	 loss=3.143810	 lr=0.090490
Epoch[066] Batch [1249]/[3759]	Speed: 116.976416 samples/sec	 accuracy=31.457500	 loss=3.142389	 lr=0.090483
Epoch[066] Batch [1299]/[3759]	Speed: 117.697288 samples/sec	 accuracy=31.466346	 loss=3.140818	 lr=0.090475
Epoch[066] Batch [1349]/[3759]	Speed: 116.962429 samples/sec	 accuracy=31.506944	 loss=3.141411	 lr=0.090468
Epoch[066] Batch [1399]/[3759]	Speed: 117.400201 samples/sec	 accuracy=31.505580	 loss=3.142926	 lr=0.090460
Epoch[066] Batch [1449]/[3759]	Speed: 117.244923 samples/sec	 accuracy=31.518319	 loss=3.141159	 lr=0.090452
Epoch[066] Batch [1499]/[3759]	Speed: 117.321770 samples/sec	 accuracy=31.560417	 loss=3.140146	 lr=0.090445
Epoch[066] Batch [1549]/[3759]	Speed: 117.967422 samples/sec	 accuracy=31.551411	 loss=3.139577	 lr=0.090437
Epoch[066] Batch [1599]/[3759]	Speed: 116.504033 samples/sec	 accuracy=31.565430	 loss=3.136872	 lr=0.090430
Epoch[066] Batch [1649]/[3759]	Speed: 118.067389 samples/sec	 accuracy=31.514205	 loss=3.138801	 lr=0.090422
Epoch[066] Batch [1699]/[3759]	Speed: 118.198409 samples/sec	 accuracy=31.511949	 loss=3.138863	 lr=0.090414
Epoch[066] Batch [1749]/[3759]	Speed: 117.496703 samples/sec	 accuracy=31.532143	 loss=3.137413	 lr=0.090407
Epoch[066] Batch [1799]/[3759]	Speed: 117.244862 samples/sec	 accuracy=31.535590	 loss=3.137280	 lr=0.090399
Epoch[066] Batch [1849]/[3759]	Speed: 117.732443 samples/sec	 accuracy=31.537162	 loss=3.139143	 lr=0.090392
Epoch[066] Batch [1899]/[3759]	Speed: 117.254943 samples/sec	 accuracy=31.510691	 loss=3.140767	 lr=0.090384
Epoch[066] Batch [1949]/[3759]	Speed: 117.395642 samples/sec	 accuracy=31.509615	 loss=3.139301	 lr=0.090376
Epoch[066] Batch [1999]/[3759]	Speed: 117.696331 samples/sec	 accuracy=31.510156	 loss=3.140593	 lr=0.090369
Epoch[066] Batch [2049]/[3759]	Speed: 116.741836 samples/sec	 accuracy=31.495427	 loss=3.140749	 lr=0.090361
Epoch[066] Batch [2099]/[3759]	Speed: 116.766208 samples/sec	 accuracy=31.478423	 loss=3.141891	 lr=0.090354
Epoch[066] Batch [2149]/[3759]	Speed: 117.046666 samples/sec	 accuracy=31.472384	 loss=3.141936	 lr=0.090346
Epoch[066] Batch [2199]/[3759]	Speed: 117.671037 samples/sec	 accuracy=31.453835	 loss=3.144028	 lr=0.090338
Epoch[066] Batch [2249]/[3759]	Speed: 118.157890 samples/sec	 accuracy=31.465972	 loss=3.143440	 lr=0.090331
Epoch[066] Batch [2299]/[3759]	Speed: 117.811176 samples/sec	 accuracy=31.417120	 loss=3.144049	 lr=0.090323
Epoch[066] Batch [2349]/[3759]	Speed: 117.825759 samples/sec	 accuracy=31.403590	 loss=3.143970	 lr=0.090316
Epoch[066] Batch [2399]/[3759]	Speed: 117.533363 samples/sec	 accuracy=31.390625	 loss=3.144409	 lr=0.090308
Epoch[066] Batch [2449]/[3759]	Speed: 117.429783 samples/sec	 accuracy=31.427934	 loss=3.143554	 lr=0.090300
Epoch[066] Batch [2499]/[3759]	Speed: 117.435191 samples/sec	 accuracy=31.426875	 loss=3.144114	 lr=0.090293
Epoch[066] Batch [2549]/[3759]	Speed: 117.153231 samples/sec	 accuracy=31.441789	 loss=3.142907	 lr=0.090285
Epoch[066] Batch [2599]/[3759]	Speed: 117.897292 samples/sec	 accuracy=31.445913	 loss=3.143286	 lr=0.090277
Epoch[066] Batch [2649]/[3759]	Speed: 117.368392 samples/sec	 accuracy=31.430425	 loss=3.144462	 lr=0.090270
Epoch[066] Batch [2699]/[3759]	Speed: 116.815888 samples/sec	 accuracy=31.436343	 loss=3.144443	 lr=0.090262
Epoch[066] Batch [2749]/[3759]	Speed: 117.728907 samples/sec	 accuracy=31.409659	 loss=3.145490	 lr=0.090254
Epoch[066] Batch [2799]/[3759]	Speed: 116.935389 samples/sec	 accuracy=31.415737	 loss=3.145353	 lr=0.090247
Epoch[066] Batch [2849]/[3759]	Speed: 118.326572 samples/sec	 accuracy=31.389803	 loss=3.146299	 lr=0.090239
Epoch[066] Batch [2899]/[3759]	Speed: 116.624164 samples/sec	 accuracy=31.375000	 loss=3.146705	 lr=0.090231
Epoch[066] Batch [2949]/[3759]	Speed: 117.937832 samples/sec	 accuracy=31.354873	 loss=3.147734	 lr=0.090224
Epoch[066] Batch [2999]/[3759]	Speed: 117.844357 samples/sec	 accuracy=31.360417	 loss=3.147955	 lr=0.090216
Epoch[066] Batch [3049]/[3759]	Speed: 118.107879 samples/sec	 accuracy=31.351434	 loss=3.147729	 lr=0.090208
Epoch[066] Batch [3099]/[3759]	Speed: 117.340236 samples/sec	 accuracy=31.356351	 loss=3.147720	 lr=0.090201
Epoch[066] Batch [3149]/[3759]	Speed: 117.289995 samples/sec	 accuracy=31.343750	 loss=3.148573	 lr=0.090193
Epoch[066] Batch [3199]/[3759]	Speed: 116.687826 samples/sec	 accuracy=31.349121	 loss=3.147981	 lr=0.090185
Epoch[066] Batch [3249]/[3759]	Speed: 117.303031 samples/sec	 accuracy=31.346154	 loss=3.147878	 lr=0.090178
Epoch[066] Batch [3299]/[3759]	Speed: 116.323878 samples/sec	 accuracy=31.368845	 loss=3.147100	 lr=0.090170
Epoch[066] Batch [3349]/[3759]	Speed: 117.070016 samples/sec	 accuracy=31.341884	 loss=3.148481	 lr=0.090162
Epoch[066] Batch [3399]/[3759]	Speed: 117.285222 samples/sec	 accuracy=31.332721	 loss=3.149076	 lr=0.090155
Epoch[066] Batch [3449]/[3759]	Speed: 117.841176 samples/sec	 accuracy=31.344656	 loss=3.148845	 lr=0.090147
Epoch[066] Batch [3499]/[3759]	Speed: 118.048665 samples/sec	 accuracy=31.347321	 loss=3.148840	 lr=0.090139
Epoch[066] Batch [3549]/[3759]	Speed: 117.419994 samples/sec	 accuracy=31.343750	 loss=3.148976	 lr=0.090132
Epoch[066] Batch [3599]/[3759]	Speed: 117.404744 samples/sec	 accuracy=31.334635	 loss=3.148906	 lr=0.090124
Epoch[066] Batch [3649]/[3759]	Speed: 117.236181 samples/sec	 accuracy=31.303510	 loss=3.150093	 lr=0.090116
Epoch[066] Batch [3699]/[3759]	Speed: 117.319288 samples/sec	 accuracy=31.299831	 loss=3.149717	 lr=0.090109
Epoch[066] Batch [3749]/[3759]	Speed: 121.017463 samples/sec	 accuracy=31.303750	 loss=3.149628	 lr=0.090101
Batch [0049]/[0303]: acc-top1=30.312500 acc-top5=56.468750
Batch [0099]/[0303]: acc-top1=28.250000 acc-top5=54.375000
Batch [0149]/[0303]: acc-top1=30.208333 acc-top5=55.812500
Batch [0199]/[0303]: acc-top1=32.609375 acc-top5=58.218750
Batch [0249]/[0303]: acc-top1=31.762500 acc-top5=58.175000
Batch [0299]/[0303]: acc-top1=30.593750 acc-top5=57.697917
[Epoch 066] training: accuracy=31.304868	 loss=3.149529
[Epoch 066] speed: 116 samples/sec	time cost: 2177.704697
[Epoch 066] validation: acc-top1=30.543523 acc-top5=57.621700 loss=3.456051
Epoch[067] Batch [0049]/[3760]	Speed: 86.166407 samples/sec	 accuracy=31.531250	 loss=3.151569	 lr=0.090092
Epoch[067] Batch [0099]/[3760]	Speed: 117.857539 samples/sec	 accuracy=31.531250	 loss=3.133765	 lr=0.090084
Epoch[067] Batch [0149]/[3760]	Speed: 116.715715 samples/sec	 accuracy=31.958333	 loss=3.140656	 lr=0.090076
Epoch[067] Batch [0199]/[3760]	Speed: 117.385494 samples/sec	 accuracy=31.726562	 loss=3.151991	 lr=0.090069
Epoch[067] Batch [0249]/[3760]	Speed: 116.669205 samples/sec	 accuracy=31.850000	 loss=3.143692	 lr=0.090061
Epoch[067] Batch [0299]/[3760]	Speed: 117.567885 samples/sec	 accuracy=31.442708	 loss=3.150380	 lr=0.090053
Epoch[067] Batch [0349]/[3760]	Speed: 117.077421 samples/sec	 accuracy=31.714286	 loss=3.143321	 lr=0.090046
Epoch[067] Batch [0399]/[3760]	Speed: 117.494105 samples/sec	 accuracy=31.648438	 loss=3.146302	 lr=0.090038
Epoch[067] Batch [0449]/[3760]	Speed: 117.634338 samples/sec	 accuracy=31.673611	 loss=3.145751	 lr=0.090030
Epoch[067] Batch [0499]/[3760]	Speed: 116.914525 samples/sec	 accuracy=31.603125	 loss=3.146289	 lr=0.090022
Epoch[067] Batch [0549]/[3760]	Speed: 117.594275 samples/sec	 accuracy=31.500000	 loss=3.151772	 lr=0.090015
Epoch[067] Batch [0599]/[3760]	Speed: 116.162705 samples/sec	 accuracy=31.544271	 loss=3.149614	 lr=0.090007
Epoch[067] Batch [0649]/[3760]	Speed: 117.080732 samples/sec	 accuracy=31.550481	 loss=3.150565	 lr=0.089999
Epoch[067] Batch [0699]/[3760]	Speed: 117.471182 samples/sec	 accuracy=31.531250	 loss=3.152831	 lr=0.089991
Epoch[067] Batch [0749]/[3760]	Speed: 117.592521 samples/sec	 accuracy=31.610417	 loss=3.144966	 lr=0.089984
Epoch[067] Batch [0799]/[3760]	Speed: 117.952740 samples/sec	 accuracy=31.623047	 loss=3.144484	 lr=0.089976
Epoch[067] Batch [0849]/[3760]	Speed: 117.164384 samples/sec	 accuracy=31.650735	 loss=3.145767	 lr=0.089968
Epoch[067] Batch [0899]/[3760]	Speed: 117.658785 samples/sec	 accuracy=31.682292	 loss=3.143644	 lr=0.089960
Epoch[067] Batch [0949]/[3760]	Speed: 117.413594 samples/sec	 accuracy=31.633224	 loss=3.144981	 lr=0.089953
Epoch[067] Batch [0999]/[3760]	Speed: 117.428155 samples/sec	 accuracy=31.590625	 loss=3.145555	 lr=0.089945
Epoch[067] Batch [1049]/[3760]	Speed: 117.218594 samples/sec	 accuracy=31.589286	 loss=3.147764	 lr=0.089937
Epoch[067] Batch [1099]/[3760]	Speed: 117.514880 samples/sec	 accuracy=31.612216	 loss=3.145781	 lr=0.089929
Epoch[067] Batch [1149]/[3760]	Speed: 117.210257 samples/sec	 accuracy=31.641304	 loss=3.143242	 lr=0.089922
Epoch[067] Batch [1199]/[3760]	Speed: 117.069386 samples/sec	 accuracy=31.660156	 loss=3.142162	 lr=0.089914
Epoch[067] Batch [1249]/[3760]	Speed: 116.997105 samples/sec	 accuracy=31.642500	 loss=3.143267	 lr=0.089906
Epoch[067] Batch [1299]/[3760]	Speed: 117.593652 samples/sec	 accuracy=31.710337	 loss=3.142408	 lr=0.089898
Epoch[067] Batch [1349]/[3760]	Speed: 117.263033 samples/sec	 accuracy=31.704861	 loss=3.141679	 lr=0.089891
Epoch[067] Batch [1399]/[3760]	Speed: 116.978784 samples/sec	 accuracy=31.657366	 loss=3.144882	 lr=0.089883
Epoch[067] Batch [1449]/[3760]	Speed: 117.855024 samples/sec	 accuracy=31.651940	 loss=3.145450	 lr=0.089875
Epoch[067] Batch [1499]/[3760]	Speed: 117.639287 samples/sec	 accuracy=31.634375	 loss=3.145049	 lr=0.089867
Epoch[067] Batch [1549]/[3760]	Speed: 117.211654 samples/sec	 accuracy=31.632056	 loss=3.144485	 lr=0.089859
Epoch[067] Batch [1599]/[3760]	Speed: 117.459610 samples/sec	 accuracy=31.664062	 loss=3.143748	 lr=0.089852
Epoch[067] Batch [1649]/[3760]	Speed: 117.608782 samples/sec	 accuracy=31.607955	 loss=3.146071	 lr=0.089844
Epoch[067] Batch [1699]/[3760]	Speed: 117.434845 samples/sec	 accuracy=31.573529	 loss=3.146117	 lr=0.089836
Epoch[067] Batch [1749]/[3760]	Speed: 116.505313 samples/sec	 accuracy=31.582143	 loss=3.144202	 lr=0.089828
Epoch[067] Batch [1799]/[3760]	Speed: 116.808084 samples/sec	 accuracy=31.571181	 loss=3.143738	 lr=0.089820
Epoch[067] Batch [1849]/[3760]	Speed: 117.537861 samples/sec	 accuracy=31.582770	 loss=3.142706	 lr=0.089813
Epoch[067] Batch [1899]/[3760]	Speed: 118.011394 samples/sec	 accuracy=31.605263	 loss=3.140412	 lr=0.089805
Epoch[067] Batch [1949]/[3760]	Speed: 116.488329 samples/sec	 accuracy=31.619391	 loss=3.140943	 lr=0.089797
Epoch[067] Batch [1999]/[3760]	Speed: 117.383879 samples/sec	 accuracy=31.625781	 loss=3.140624	 lr=0.089789
Epoch[067] Batch [2049]/[3760]	Speed: 116.805872 samples/sec	 accuracy=31.589939	 loss=3.140533	 lr=0.089781
Epoch[067] Batch [2099]/[3760]	Speed: 116.873584 samples/sec	 accuracy=31.551339	 loss=3.142645	 lr=0.089774
Epoch[067] Batch [2149]/[3760]	Speed: 117.079516 samples/sec	 accuracy=31.575581	 loss=3.141789	 lr=0.089766
Epoch[067] Batch [2199]/[3760]	Speed: 117.677906 samples/sec	 accuracy=31.587358	 loss=3.141082	 lr=0.089758
Epoch[067] Batch [2249]/[3760]	Speed: 117.497401 samples/sec	 accuracy=31.575694	 loss=3.141066	 lr=0.089750
Epoch[067] Batch [2299]/[3760]	Speed: 117.609994 samples/sec	 accuracy=31.584239	 loss=3.140498	 lr=0.089742
Epoch[067] Batch [2349]/[3760]	Speed: 116.938057 samples/sec	 accuracy=31.568484	 loss=3.141124	 lr=0.089734
Epoch[067] Batch [2399]/[3760]	Speed: 117.099604 samples/sec	 accuracy=31.567057	 loss=3.141341	 lr=0.089727
Epoch[067] Batch [2449]/[3760]	Speed: 117.472003 samples/sec	 accuracy=31.552934	 loss=3.141894	 lr=0.089719
Epoch[067] Batch [2499]/[3760]	Speed: 118.139723 samples/sec	 accuracy=31.527500	 loss=3.143652	 lr=0.089711
Epoch[067] Batch [2549]/[3760]	Speed: 117.542095 samples/sec	 accuracy=31.547181	 loss=3.143338	 lr=0.089703
Epoch[067] Batch [2599]/[3760]	Speed: 116.912523 samples/sec	 accuracy=31.532452	 loss=3.143756	 lr=0.089695
Epoch[067] Batch [2649]/[3760]	Speed: 116.592272 samples/sec	 accuracy=31.525354	 loss=3.143858	 lr=0.089687
Epoch[067] Batch [2699]/[3760]	Speed: 117.219959 samples/sec	 accuracy=31.519676	 loss=3.144200	 lr=0.089680
Epoch[067] Batch [2749]/[3760]	Speed: 116.827297 samples/sec	 accuracy=31.535795	 loss=3.143660	 lr=0.089672
Epoch[067] Batch [2799]/[3760]	Speed: 117.969764 samples/sec	 accuracy=31.516741	 loss=3.143938	 lr=0.089664
Epoch[067] Batch [2849]/[3760]	Speed: 117.286326 samples/sec	 accuracy=31.509320	 loss=3.144794	 lr=0.089656
Epoch[067] Batch [2899]/[3760]	Speed: 116.690017 samples/sec	 accuracy=31.500000	 loss=3.145118	 lr=0.089648
Epoch[067] Batch [2949]/[3760]	Speed: 116.693984 samples/sec	 accuracy=31.474576	 loss=3.146197	 lr=0.089640
Epoch[067] Batch [2999]/[3760]	Speed: 117.193814 samples/sec	 accuracy=31.469792	 loss=3.146618	 lr=0.089632
Epoch[067] Batch [3049]/[3760]	Speed: 116.705184 samples/sec	 accuracy=31.473361	 loss=3.146466	 lr=0.089625
Epoch[067] Batch [3099]/[3760]	Speed: 117.896491 samples/sec	 accuracy=31.452117	 loss=3.147168	 lr=0.089617
Epoch[067] Batch [3149]/[3760]	Speed: 117.407747 samples/sec	 accuracy=31.430556	 loss=3.147409	 lr=0.089609
Epoch[067] Batch [3199]/[3760]	Speed: 117.907397 samples/sec	 accuracy=31.420410	 loss=3.147766	 lr=0.089601
Epoch[067] Batch [3249]/[3760]	Speed: 117.720383 samples/sec	 accuracy=31.411058	 loss=3.147578	 lr=0.089593
Epoch[067] Batch [3299]/[3760]	Speed: 117.092623 samples/sec	 accuracy=31.406723	 loss=3.147950	 lr=0.089585
Epoch[067] Batch [3349]/[3760]	Speed: 117.594903 samples/sec	 accuracy=31.391325	 loss=3.148814	 lr=0.089577
Epoch[067] Batch [3399]/[3760]	Speed: 117.883636 samples/sec	 accuracy=31.404412	 loss=3.148294	 lr=0.089569
Epoch[067] Batch [3449]/[3760]	Speed: 117.736334 samples/sec	 accuracy=31.400815	 loss=3.148089	 lr=0.089562
Epoch[067] Batch [3499]/[3760]	Speed: 117.946275 samples/sec	 accuracy=31.408036	 loss=3.148165	 lr=0.089554
Epoch[067] Batch [3549]/[3760]	Speed: 117.335548 samples/sec	 accuracy=31.399648	 loss=3.148830	 lr=0.089546
Epoch[067] Batch [3599]/[3760]	Speed: 117.638242 samples/sec	 accuracy=31.409722	 loss=3.149331	 lr=0.089538
Epoch[067] Batch [3649]/[3760]	Speed: 117.775417 samples/sec	 accuracy=31.398116	 loss=3.149415	 lr=0.089530
Epoch[067] Batch [3699]/[3760]	Speed: 117.003468 samples/sec	 accuracy=31.387247	 loss=3.149865	 lr=0.089522
Epoch[067] Batch [3749]/[3760]	Speed: 120.923638 samples/sec	 accuracy=31.389167	 loss=3.149965	 lr=0.089514
Batch [0049]/[0303]: acc-top1=31.593750 acc-top5=56.468750
Batch [0099]/[0303]: acc-top1=27.984375 acc-top5=53.375000
Batch [0149]/[0303]: acc-top1=29.031250 acc-top5=54.572917
Batch [0199]/[0303]: acc-top1=32.335938 acc-top5=57.953125
Batch [0249]/[0303]: acc-top1=31.518750 acc-top5=57.818750
Batch [0299]/[0303]: acc-top1=31.093750 acc-top5=57.760417
[Epoch 067] training: accuracy=31.394614	 loss=3.149568
[Epoch 067] speed: 116 samples/sec	time cost: 2180.257988
[Epoch 067] validation: acc-top1=31.048886 acc-top5=57.657797 loss=3.259352
Epoch[068] Batch [0049]/[3760]	Speed: 88.178913 samples/sec	 accuracy=32.218750	 loss=3.089386	 lr=0.089505
Epoch[068] Batch [0099]/[3760]	Speed: 117.539255 samples/sec	 accuracy=32.406250	 loss=3.100994	 lr=0.089497
Epoch[068] Batch [0149]/[3760]	Speed: 117.360164 samples/sec	 accuracy=32.187500	 loss=3.104518	 lr=0.089489
Epoch[068] Batch [0199]/[3760]	Speed: 117.230259 samples/sec	 accuracy=32.265625	 loss=3.102957	 lr=0.089481
Epoch[068] Batch [0249]/[3760]	Speed: 116.780403 samples/sec	 accuracy=32.106250	 loss=3.107344	 lr=0.089473
Epoch[068] Batch [0299]/[3760]	Speed: 117.291237 samples/sec	 accuracy=32.062500	 loss=3.114713	 lr=0.089465
Epoch[068] Batch [0349]/[3760]	Speed: 116.994615 samples/sec	 accuracy=32.178571	 loss=3.111254	 lr=0.089457
Epoch[068] Batch [0399]/[3760]	Speed: 117.097502 samples/sec	 accuracy=32.195312	 loss=3.114860	 lr=0.089449
Epoch[068] Batch [0449]/[3760]	Speed: 117.007353 samples/sec	 accuracy=32.065972	 loss=3.121331	 lr=0.089441
Epoch[068] Batch [0499]/[3760]	Speed: 116.597924 samples/sec	 accuracy=31.956250	 loss=3.123435	 lr=0.089433
Epoch[068] Batch [0549]/[3760]	Speed: 117.276702 samples/sec	 accuracy=31.846591	 loss=3.128874	 lr=0.089426
Epoch[068] Batch [0599]/[3760]	Speed: 117.308059 samples/sec	 accuracy=31.825521	 loss=3.130601	 lr=0.089418
Epoch[068] Batch [0649]/[3760]	Speed: 116.813917 samples/sec	 accuracy=31.826923	 loss=3.125862	 lr=0.089410
Epoch[068] Batch [0699]/[3760]	Speed: 117.600492 samples/sec	 accuracy=31.725446	 loss=3.130281	 lr=0.089402
Epoch[068] Batch [0749]/[3760]	Speed: 117.797157 samples/sec	 accuracy=31.702083	 loss=3.132976	 lr=0.089394
Epoch[068] Batch [0799]/[3760]	Speed: 115.884645 samples/sec	 accuracy=31.687500	 loss=3.131943	 lr=0.089386
Epoch[068] Batch [0849]/[3760]	Speed: 117.019228 samples/sec	 accuracy=31.669118	 loss=3.132180	 lr=0.089378
Epoch[068] Batch [0899]/[3760]	Speed: 117.556332 samples/sec	 accuracy=31.736111	 loss=3.128762	 lr=0.089370
Epoch[068] Batch [0949]/[3760]	Speed: 116.673538 samples/sec	 accuracy=31.735197	 loss=3.129475	 lr=0.089362
Epoch[068] Batch [0999]/[3760]	Speed: 117.395485 samples/sec	 accuracy=31.745313	 loss=3.128057	 lr=0.089354
Epoch[068] Batch [1049]/[3760]	Speed: 117.436540 samples/sec	 accuracy=31.751488	 loss=3.126730	 lr=0.089346
Epoch[068] Batch [1099]/[3760]	Speed: 116.835995 samples/sec	 accuracy=31.775568	 loss=3.124262	 lr=0.089338
Epoch[068] Batch [1149]/[3760]	Speed: 117.618693 samples/sec	 accuracy=31.792120	 loss=3.123925	 lr=0.089330
Epoch[068] Batch [1199]/[3760]	Speed: 116.791810 samples/sec	 accuracy=31.717448	 loss=3.124906	 lr=0.089322
Epoch[068] Batch [1249]/[3760]	Speed: 117.588842 samples/sec	 accuracy=31.690000	 loss=3.126413	 lr=0.089314
Epoch[068] Batch [1299]/[3760]	Speed: 118.682159 samples/sec	 accuracy=31.750000	 loss=3.124634	 lr=0.089306
Epoch[068] Batch [1349]/[3760]	Speed: 117.766597 samples/sec	 accuracy=31.765046	 loss=3.124316	 lr=0.089298
Epoch[068] Batch [1399]/[3760]	Speed: 116.823996 samples/sec	 accuracy=31.733259	 loss=3.126217	 lr=0.089290
Epoch[068] Batch [1449]/[3760]	Speed: 117.004236 samples/sec	 accuracy=31.715517	 loss=3.126350	 lr=0.089282
Epoch[068] Batch [1499]/[3760]	Speed: 117.639166 samples/sec	 accuracy=31.687500	 loss=3.127422	 lr=0.089274
Epoch[068] Batch [1549]/[3760]	Speed: 117.429858 samples/sec	 accuracy=31.637097	 loss=3.128378	 lr=0.089266
Epoch[068] Batch [1599]/[3760]	Speed: 118.086595 samples/sec	 accuracy=31.634766	 loss=3.128756	 lr=0.089258
Epoch[068] Batch [1649]/[3760]	Speed: 117.841847 samples/sec	 accuracy=31.637311	 loss=3.129229	 lr=0.089250
Epoch[068] Batch [1699]/[3760]	Speed: 117.368729 samples/sec	 accuracy=31.619485	 loss=3.131195	 lr=0.089242
Epoch[068] Batch [1749]/[3760]	Speed: 116.777643 samples/sec	 accuracy=31.608929	 loss=3.132659	 lr=0.089234
Epoch[068] Batch [1799]/[3760]	Speed: 117.404151 samples/sec	 accuracy=31.625000	 loss=3.132234	 lr=0.089226
Epoch[068] Batch [1849]/[3760]	Speed: 117.351376 samples/sec	 accuracy=31.646959	 loss=3.133698	 lr=0.089218
Epoch[068] Batch [1899]/[3760]	Speed: 117.124417 samples/sec	 accuracy=31.639803	 loss=3.134516	 lr=0.089210
Epoch[068] Batch [1949]/[3760]	Speed: 116.806807 samples/sec	 accuracy=31.645032	 loss=3.134185	 lr=0.089202
Epoch[068] Batch [1999]/[3760]	Speed: 117.623865 samples/sec	 accuracy=31.673438	 loss=3.132237	 lr=0.089194
Epoch[068] Batch [2049]/[3760]	Speed: 116.806877 samples/sec	 accuracy=31.636433	 loss=3.134865	 lr=0.089186
Epoch[068] Batch [2099]/[3760]	Speed: 117.317962 samples/sec	 accuracy=31.633185	 loss=3.135196	 lr=0.089178
Epoch[068] Batch [2149]/[3760]	Speed: 116.230631 samples/sec	 accuracy=31.611919	 loss=3.135612	 lr=0.089170
Epoch[068] Batch [2199]/[3760]	Speed: 117.262481 samples/sec	 accuracy=31.617188	 loss=3.134709	 lr=0.089162
Epoch[068] Batch [2249]/[3760]	Speed: 117.225558 samples/sec	 accuracy=31.593056	 loss=3.137222	 lr=0.089154
Epoch[068] Batch [2299]/[3760]	Speed: 118.096179 samples/sec	 accuracy=31.580842	 loss=3.138253	 lr=0.089146
Epoch[068] Batch [2349]/[3760]	Speed: 116.832168 samples/sec	 accuracy=31.593085	 loss=3.138655	 lr=0.089138
Epoch[068] Batch [2399]/[3760]	Speed: 117.584194 samples/sec	 accuracy=31.569661	 loss=3.139234	 lr=0.089130
Epoch[068] Batch [2449]/[3760]	Speed: 117.283867 samples/sec	 accuracy=31.548469	 loss=3.139776	 lr=0.089122
Epoch[068] Batch [2499]/[3760]	Speed: 117.947699 samples/sec	 accuracy=31.526250	 loss=3.140402	 lr=0.089114
Epoch[068] Batch [2549]/[3760]	Speed: 117.353766 samples/sec	 accuracy=31.498162	 loss=3.140541	 lr=0.089106
Epoch[068] Batch [2599]/[3760]	Speed: 117.354363 samples/sec	 accuracy=31.497596	 loss=3.140187	 lr=0.089098
Epoch[068] Batch [2649]/[3760]	Speed: 116.401831 samples/sec	 accuracy=31.484670	 loss=3.139738	 lr=0.089090
Epoch[068] Batch [2699]/[3760]	Speed: 117.596177 samples/sec	 accuracy=31.478588	 loss=3.139884	 lr=0.089082
Epoch[068] Batch [2749]/[3760]	Speed: 117.491855 samples/sec	 accuracy=31.483523	 loss=3.140089	 lr=0.089074
Epoch[068] Batch [2799]/[3760]	Speed: 117.288569 samples/sec	 accuracy=31.474888	 loss=3.140388	 lr=0.089066
Epoch[068] Batch [2849]/[3760]	Speed: 117.155218 samples/sec	 accuracy=31.453399	 loss=3.141281	 lr=0.089058
Epoch[068] Batch [2899]/[3760]	Speed: 116.761525 samples/sec	 accuracy=31.488147	 loss=3.140421	 lr=0.089050
Epoch[068] Batch [2949]/[3760]	Speed: 116.798549 samples/sec	 accuracy=31.480403	 loss=3.141294	 lr=0.089042
Epoch[068] Batch [2999]/[3760]	Speed: 117.639854 samples/sec	 accuracy=31.466146	 loss=3.142255	 lr=0.089034
Epoch[068] Batch [3049]/[3760]	Speed: 117.043475 samples/sec	 accuracy=31.442623	 loss=3.142937	 lr=0.089026
Epoch[068] Batch [3099]/[3760]	Speed: 117.370905 samples/sec	 accuracy=31.441532	 loss=3.143525	 lr=0.089018
Epoch[068] Batch [3149]/[3760]	Speed: 117.248031 samples/sec	 accuracy=31.427083	 loss=3.144935	 lr=0.089010
Epoch[068] Batch [3199]/[3760]	Speed: 117.605310 samples/sec	 accuracy=31.411621	 loss=3.145644	 lr=0.089001
Epoch[068] Batch [3249]/[3760]	Speed: 117.584161 samples/sec	 accuracy=31.422115	 loss=3.145665	 lr=0.088993
Epoch[068] Batch [3299]/[3760]	Speed: 116.058494 samples/sec	 accuracy=31.412405	 loss=3.146243	 lr=0.088985
Epoch[068] Batch [3349]/[3760]	Speed: 117.500215 samples/sec	 accuracy=31.419310	 loss=3.146765	 lr=0.088977
Epoch[068] Batch [3399]/[3760]	Speed: 117.875470 samples/sec	 accuracy=31.409007	 loss=3.147070	 lr=0.088969
Epoch[068] Batch [3449]/[3760]	Speed: 117.103660 samples/sec	 accuracy=31.392663	 loss=3.147329	 lr=0.088961
Epoch[068] Batch [3499]/[3760]	Speed: 117.265527 samples/sec	 accuracy=31.370982	 loss=3.147855	 lr=0.088953
Epoch[068] Batch [3549]/[3760]	Speed: 117.996114 samples/sec	 accuracy=31.379401	 loss=3.147915	 lr=0.088945
Epoch[068] Batch [3599]/[3760]	Speed: 116.674985 samples/sec	 accuracy=31.382378	 loss=3.148102	 lr=0.088937
Epoch[068] Batch [3649]/[3760]	Speed: 117.026634 samples/sec	 accuracy=31.390839	 loss=3.147910	 lr=0.088929
Epoch[068] Batch [3699]/[3760]	Speed: 117.703368 samples/sec	 accuracy=31.417230	 loss=3.146378	 lr=0.088921
Epoch[068] Batch [3749]/[3760]	Speed: 120.242374 samples/sec	 accuracy=31.420417	 loss=3.146480	 lr=0.088913
Batch [0049]/[0303]: acc-top1=32.125000 acc-top5=58.156250
Batch [0099]/[0303]: acc-top1=29.921875 acc-top5=55.906250
Batch [0149]/[0303]: acc-top1=31.250000 acc-top5=56.979167
Batch [0199]/[0303]: acc-top1=33.921875 acc-top5=59.312500
Batch [0249]/[0303]: acc-top1=33.250000 acc-top5=59.231250
Batch [0299]/[0303]: acc-top1=32.588542 acc-top5=59.708333
[Epoch 068] training: accuracy=31.424535	 loss=3.146289
[Epoch 068] speed: 116 samples/sec	time cost: 2180.664332
[Epoch 068] validation: acc-top1=32.668111 acc-top5=59.808168 loss=3.179518
Epoch[069] Batch [0049]/[3759]	Speed: 87.862430 samples/sec	 accuracy=30.406250	 loss=3.162589	 lr=0.088903
Epoch[069] Batch [0099]/[3759]	Speed: 117.495867 samples/sec	 accuracy=31.062500	 loss=3.138244	 lr=0.088895
Epoch[069] Batch [0149]/[3759]	Speed: 116.653600 samples/sec	 accuracy=31.489583	 loss=3.125867	 lr=0.088887
Epoch[069] Batch [0199]/[3759]	Speed: 116.841107 samples/sec	 accuracy=31.492188	 loss=3.130766	 lr=0.088879
Epoch[069] Batch [0249]/[3759]	Speed: 117.671082 samples/sec	 accuracy=31.687500	 loss=3.126210	 lr=0.088870
Epoch[069] Batch [0299]/[3759]	Speed: 116.922443 samples/sec	 accuracy=31.713542	 loss=3.121954	 lr=0.088862
Epoch[069] Batch [0349]/[3759]	Speed: 116.653196 samples/sec	 accuracy=31.669643	 loss=3.132788	 lr=0.088854
Epoch[069] Batch [0399]/[3759]	Speed: 117.058172 samples/sec	 accuracy=31.730469	 loss=3.129963	 lr=0.088846
Epoch[069] Batch [0449]/[3759]	Speed: 117.374378 samples/sec	 accuracy=31.798611	 loss=3.130489	 lr=0.088838
Epoch[069] Batch [0499]/[3759]	Speed: 117.135635 samples/sec	 accuracy=31.850000	 loss=3.126387	 lr=0.088830
Epoch[069] Batch [0549]/[3759]	Speed: 116.861708 samples/sec	 accuracy=31.721591	 loss=3.130323	 lr=0.088822
Epoch[069] Batch [0599]/[3759]	Speed: 117.187385 samples/sec	 accuracy=31.622396	 loss=3.131605	 lr=0.088814
Epoch[069] Batch [0649]/[3759]	Speed: 116.547587 samples/sec	 accuracy=31.721154	 loss=3.129349	 lr=0.088805
Epoch[069] Batch [0699]/[3759]	Speed: 117.218797 samples/sec	 accuracy=31.816964	 loss=3.130238	 lr=0.088797
Epoch[069] Batch [0749]/[3759]	Speed: 117.220532 samples/sec	 accuracy=31.877083	 loss=3.124113	 lr=0.088789
Epoch[069] Batch [0799]/[3759]	Speed: 116.940719 samples/sec	 accuracy=31.853516	 loss=3.121854	 lr=0.088781
Epoch[069] Batch [0849]/[3759]	Speed: 117.049176 samples/sec	 accuracy=31.863971	 loss=3.120064	 lr=0.088773
Epoch[069] Batch [0899]/[3759]	Speed: 117.373350 samples/sec	 accuracy=31.871528	 loss=3.121666	 lr=0.088765
Epoch[069] Batch [0949]/[3759]	Speed: 116.910806 samples/sec	 accuracy=31.810855	 loss=3.123046	 lr=0.088757
Epoch[069] Batch [0999]/[3759]	Speed: 116.842530 samples/sec	 accuracy=31.803125	 loss=3.122503	 lr=0.088748
Epoch[069] Batch [1049]/[3759]	Speed: 116.736358 samples/sec	 accuracy=31.735119	 loss=3.126502	 lr=0.088740
Epoch[069] Batch [1099]/[3759]	Speed: 117.333902 samples/sec	 accuracy=31.740057	 loss=3.126229	 lr=0.088732
Epoch[069] Batch [1149]/[3759]	Speed: 117.315917 samples/sec	 accuracy=31.790761	 loss=3.125680	 lr=0.088724
Epoch[069] Batch [1199]/[3759]	Speed: 117.192911 samples/sec	 accuracy=31.805990	 loss=3.123957	 lr=0.088716
Epoch[069] Batch [1249]/[3759]	Speed: 117.504073 samples/sec	 accuracy=31.828750	 loss=3.124188	 lr=0.088708
Epoch[069] Batch [1299]/[3759]	Speed: 116.528394 samples/sec	 accuracy=31.802885	 loss=3.125419	 lr=0.088699
Epoch[069] Batch [1349]/[3759]	Speed: 117.345532 samples/sec	 accuracy=31.829861	 loss=3.124361	 lr=0.088691
Epoch[069] Batch [1399]/[3759]	Speed: 116.954487 samples/sec	 accuracy=31.840402	 loss=3.123078	 lr=0.088683
Epoch[069] Batch [1449]/[3759]	Speed: 116.904966 samples/sec	 accuracy=31.788793	 loss=3.126934	 lr=0.088675
Epoch[069] Batch [1499]/[3759]	Speed: 116.753058 samples/sec	 accuracy=31.818750	 loss=3.125613	 lr=0.088667
Epoch[069] Batch [1549]/[3759]	Speed: 117.366366 samples/sec	 accuracy=31.820565	 loss=3.125260	 lr=0.088659
Epoch[069] Batch [1599]/[3759]	Speed: 117.501516 samples/sec	 accuracy=31.822266	 loss=3.124304	 lr=0.088650
Epoch[069] Batch [1649]/[3759]	Speed: 116.920502 samples/sec	 accuracy=31.756629	 loss=3.127351	 lr=0.088642
Epoch[069] Batch [1699]/[3759]	Speed: 117.836681 samples/sec	 accuracy=31.738051	 loss=3.128289	 lr=0.088634
Epoch[069] Batch [1749]/[3759]	Speed: 116.948110 samples/sec	 accuracy=31.729464	 loss=3.128888	 lr=0.088626
Epoch[069] Batch [1799]/[3759]	Speed: 116.708321 samples/sec	 accuracy=31.723958	 loss=3.129526	 lr=0.088618
Epoch[069] Batch [1849]/[3759]	Speed: 117.541265 samples/sec	 accuracy=31.722973	 loss=3.129538	 lr=0.088609
Epoch[069] Batch [1899]/[3759]	Speed: 116.322079 samples/sec	 accuracy=31.711349	 loss=3.129428	 lr=0.088601
Epoch[069] Batch [1949]/[3759]	Speed: 117.380777 samples/sec	 accuracy=31.714744	 loss=3.130040	 lr=0.088593
Epoch[069] Batch [1999]/[3759]	Speed: 117.164780 samples/sec	 accuracy=31.701563	 loss=3.129790	 lr=0.088585
Epoch[069] Batch [2049]/[3759]	Speed: 116.543886 samples/sec	 accuracy=31.709604	 loss=3.130586	 lr=0.088577
Epoch[069] Batch [2099]/[3759]	Speed: 117.256181 samples/sec	 accuracy=31.703125	 loss=3.131017	 lr=0.088568
Epoch[069] Batch [2149]/[3759]	Speed: 117.218815 samples/sec	 accuracy=31.696221	 loss=3.131365	 lr=0.088560
Epoch[069] Batch [2199]/[3759]	Speed: 118.040455 samples/sec	 accuracy=31.720170	 loss=3.131622	 lr=0.088552
Epoch[069] Batch [2249]/[3759]	Speed: 117.413471 samples/sec	 accuracy=31.722222	 loss=3.130879	 lr=0.088544
Epoch[069] Batch [2299]/[3759]	Speed: 117.430217 samples/sec	 accuracy=31.732337	 loss=3.130546	 lr=0.088536
Epoch[069] Batch [2349]/[3759]	Speed: 117.686859 samples/sec	 accuracy=31.695479	 loss=3.131986	 lr=0.088527
Epoch[069] Batch [2399]/[3759]	Speed: 117.599987 samples/sec	 accuracy=31.731120	 loss=3.131186	 lr=0.088519
Epoch[069] Batch [2449]/[3759]	Speed: 117.811264 samples/sec	 accuracy=31.718750	 loss=3.130859	 lr=0.088511
Epoch[069] Batch [2499]/[3759]	Speed: 117.146005 samples/sec	 accuracy=31.745625	 loss=3.130954	 lr=0.088503
Epoch[069] Batch [2549]/[3759]	Speed: 117.317632 samples/sec	 accuracy=31.770833	 loss=3.130354	 lr=0.088494
Epoch[069] Batch [2599]/[3759]	Speed: 116.927898 samples/sec	 accuracy=31.737981	 loss=3.132177	 lr=0.088486
Epoch[069] Batch [2649]/[3759]	Speed: 117.700435 samples/sec	 accuracy=31.728774	 loss=3.132190	 lr=0.088478
Epoch[069] Batch [2699]/[3759]	Speed: 115.750950 samples/sec	 accuracy=31.717593	 loss=3.131934	 lr=0.088470
Epoch[069] Batch [2749]/[3759]	Speed: 117.880562 samples/sec	 accuracy=31.711364	 loss=3.132226	 lr=0.088462
Epoch[069] Batch [2799]/[3759]	Speed: 117.190989 samples/sec	 accuracy=31.707589	 loss=3.131480	 lr=0.088453
Epoch[069] Batch [2849]/[3759]	Speed: 117.262033 samples/sec	 accuracy=31.706140	 loss=3.131892	 lr=0.088445
Epoch[069] Batch [2899]/[3759]	Speed: 117.603418 samples/sec	 accuracy=31.697737	 loss=3.133315	 lr=0.088437
Epoch[069] Batch [2949]/[3759]	Speed: 117.265937 samples/sec	 accuracy=31.697564	 loss=3.133963	 lr=0.088429
Epoch[069] Batch [2999]/[3759]	Speed: 117.419373 samples/sec	 accuracy=31.706771	 loss=3.133252	 lr=0.088420
Epoch[069] Batch [3049]/[3759]	Speed: 117.952163 samples/sec	 accuracy=31.721824	 loss=3.132326	 lr=0.088412
Epoch[069] Batch [3099]/[3759]	Speed: 116.876665 samples/sec	 accuracy=31.719254	 loss=3.131919	 lr=0.088404
Epoch[069] Batch [3149]/[3759]	Speed: 117.984691 samples/sec	 accuracy=31.754960	 loss=3.130935	 lr=0.088396
Epoch[069] Batch [3199]/[3759]	Speed: 118.072784 samples/sec	 accuracy=31.758301	 loss=3.131009	 lr=0.088387
Epoch[069] Batch [3249]/[3759]	Speed: 116.489189 samples/sec	 accuracy=31.750962	 loss=3.131076	 lr=0.088379
Epoch[069] Batch [3299]/[3759]	Speed: 117.032365 samples/sec	 accuracy=31.747633	 loss=3.131186	 lr=0.088371
Epoch[069] Batch [3349]/[3759]	Speed: 116.447551 samples/sec	 accuracy=31.756996	 loss=3.130531	 lr=0.088362
Epoch[069] Batch [3399]/[3759]	Speed: 117.393880 samples/sec	 accuracy=31.746324	 loss=3.131370	 lr=0.088354
Epoch[069] Batch [3449]/[3759]	Speed: 116.993703 samples/sec	 accuracy=31.731431	 loss=3.131682	 lr=0.088346
Epoch[069] Batch [3499]/[3759]	Speed: 117.672533 samples/sec	 accuracy=31.716518	 loss=3.132386	 lr=0.088338
Epoch[069] Batch [3549]/[3759]	Speed: 117.634146 samples/sec	 accuracy=31.699824	 loss=3.132718	 lr=0.088329
Epoch[069] Batch [3599]/[3759]	Speed: 117.303395 samples/sec	 accuracy=31.681424	 loss=3.133364	 lr=0.088321
Epoch[069] Batch [3649]/[3759]	Speed: 117.478743 samples/sec	 accuracy=31.674229	 loss=3.133869	 lr=0.088313
Epoch[069] Batch [3699]/[3759]	Speed: 117.321694 samples/sec	 accuracy=31.667230	 loss=3.134270	 lr=0.088305
Epoch[069] Batch [3749]/[3759]	Speed: 121.030583 samples/sec	 accuracy=31.650417	 loss=3.134998	 lr=0.088296
Batch [0049]/[0303]: acc-top1=31.406250 acc-top5=58.125000
Batch [0099]/[0303]: acc-top1=29.250000 acc-top5=56.562500
Batch [0149]/[0303]: acc-top1=31.125000 acc-top5=57.520833
Batch [0199]/[0303]: acc-top1=34.164062 acc-top5=60.203125
Batch [0249]/[0303]: acc-top1=33.475000 acc-top5=59.893750
Batch [0299]/[0303]: acc-top1=32.515625 acc-top5=59.802083
[Epoch 069] training: accuracy=31.654446	 loss=3.135055
[Epoch 069] speed: 116 samples/sec	time cost: 2180.445763
[Epoch 069] validation: acc-top1=32.497937 acc-top5=59.803012 loss=3.152101
Epoch[070] Batch [0049]/[3760]	Speed: 86.216556 samples/sec	 accuracy=32.187500	 loss=3.078918	 lr=0.088286
Epoch[070] Batch [0099]/[3760]	Speed: 116.814512 samples/sec	 accuracy=31.812500	 loss=3.092532	 lr=0.088278
Epoch[070] Batch [0149]/[3760]	Speed: 117.277949 samples/sec	 accuracy=32.093750	 loss=3.096837	 lr=0.088270
Epoch[070] Batch [0199]/[3760]	Speed: 116.951294 samples/sec	 accuracy=32.304688	 loss=3.080233	 lr=0.088262
Epoch[070] Batch [0249]/[3760]	Speed: 117.315764 samples/sec	 accuracy=32.400000	 loss=3.081897	 lr=0.088253
Epoch[070] Batch [0299]/[3760]	Speed: 116.199797 samples/sec	 accuracy=32.328125	 loss=3.088089	 lr=0.088245
Epoch[070] Batch [0349]/[3760]	Speed: 117.316932 samples/sec	 accuracy=32.258929	 loss=3.085527	 lr=0.088237
Epoch[070] Batch [0399]/[3760]	Speed: 117.337998 samples/sec	 accuracy=32.203125	 loss=3.093646	 lr=0.088228
Epoch[070] Batch [0449]/[3760]	Speed: 117.560381 samples/sec	 accuracy=32.145833	 loss=3.094704	 lr=0.088220
Epoch[070] Batch [0499]/[3760]	Speed: 117.175136 samples/sec	 accuracy=32.050000	 loss=3.096745	 lr=0.088212
Epoch[070] Batch [0549]/[3760]	Speed: 117.490860 samples/sec	 accuracy=31.974432	 loss=3.099169	 lr=0.088203
Epoch[070] Batch [0599]/[3760]	Speed: 117.694379 samples/sec	 accuracy=32.028646	 loss=3.098342	 lr=0.088195
Epoch[070] Batch [0649]/[3760]	Speed: 116.988530 samples/sec	 accuracy=31.990385	 loss=3.102362	 lr=0.088187
Epoch[070] Batch [0699]/[3760]	Speed: 117.052604 samples/sec	 accuracy=31.982143	 loss=3.103613	 lr=0.088178
Epoch[070] Batch [0749]/[3760]	Speed: 116.834076 samples/sec	 accuracy=32.066667	 loss=3.103853	 lr=0.088170
Epoch[070] Batch [0799]/[3760]	Speed: 117.701974 samples/sec	 accuracy=32.140625	 loss=3.100654	 lr=0.088162
Epoch[070] Batch [0849]/[3760]	Speed: 116.927464 samples/sec	 accuracy=32.095588	 loss=3.102293	 lr=0.088153
Epoch[070] Batch [0899]/[3760]	Speed: 116.969624 samples/sec	 accuracy=32.029514	 loss=3.105641	 lr=0.088145
Epoch[070] Batch [0949]/[3760]	Speed: 117.697917 samples/sec	 accuracy=32.000000	 loss=3.108545	 lr=0.088137
Epoch[070] Batch [0999]/[3760]	Speed: 116.855505 samples/sec	 accuracy=32.017187	 loss=3.107782	 lr=0.088128
Epoch[070] Batch [1049]/[3760]	Speed: 117.060556 samples/sec	 accuracy=31.971726	 loss=3.109724	 lr=0.088120
Epoch[070] Batch [1099]/[3760]	Speed: 117.012502 samples/sec	 accuracy=31.943182	 loss=3.109244	 lr=0.088112
Epoch[070] Batch [1149]/[3760]	Speed: 117.725878 samples/sec	 accuracy=31.903533	 loss=3.110017	 lr=0.088103
Epoch[070] Batch [1199]/[3760]	Speed: 116.214842 samples/sec	 accuracy=31.902344	 loss=3.110428	 lr=0.088095
Epoch[070] Batch [1249]/[3760]	Speed: 116.841695 samples/sec	 accuracy=31.865000	 loss=3.111855	 lr=0.088087
Epoch[070] Batch [1299]/[3760]	Speed: 116.741074 samples/sec	 accuracy=31.873798	 loss=3.111603	 lr=0.088078
Epoch[070] Batch [1349]/[3760]	Speed: 116.717492 samples/sec	 accuracy=31.866898	 loss=3.112514	 lr=0.088070
Epoch[070] Batch [1399]/[3760]	Speed: 117.075004 samples/sec	 accuracy=31.850446	 loss=3.113351	 lr=0.088062
Epoch[070] Batch [1449]/[3760]	Speed: 117.488346 samples/sec	 accuracy=31.894397	 loss=3.111986	 lr=0.088053
Epoch[070] Batch [1499]/[3760]	Speed: 116.895985 samples/sec	 accuracy=31.894792	 loss=3.112902	 lr=0.088045
Epoch[070] Batch [1549]/[3760]	Speed: 117.253216 samples/sec	 accuracy=31.967742	 loss=3.109591	 lr=0.088036
Epoch[070] Batch [1599]/[3760]	Speed: 116.985283 samples/sec	 accuracy=31.942383	 loss=3.110290	 lr=0.088028
Epoch[070] Batch [1649]/[3760]	Speed: 117.712332 samples/sec	 accuracy=31.923295	 loss=3.111432	 lr=0.088020
Epoch[070] Batch [1699]/[3760]	Speed: 117.761890 samples/sec	 accuracy=31.920956	 loss=3.111702	 lr=0.088011
Epoch[070] Batch [1749]/[3760]	Speed: 116.902925 samples/sec	 accuracy=31.898214	 loss=3.113010	 lr=0.088003
Epoch[070] Batch [1799]/[3760]	Speed: 117.446887 samples/sec	 accuracy=31.863715	 loss=3.114480	 lr=0.087995
Epoch[070] Batch [1849]/[3760]	Speed: 117.281991 samples/sec	 accuracy=31.831081	 loss=3.115654	 lr=0.087986
Epoch[070] Batch [1899]/[3760]	Speed: 117.560753 samples/sec	 accuracy=31.851974	 loss=3.115581	 lr=0.087978
Epoch[070] Batch [1949]/[3760]	Speed: 117.071626 samples/sec	 accuracy=31.818910	 loss=3.116170	 lr=0.087969
Epoch[070] Batch [1999]/[3760]	Speed: 117.142812 samples/sec	 accuracy=31.820312	 loss=3.116462	 lr=0.087961
Epoch[070] Batch [2049]/[3760]	Speed: 117.869163 samples/sec	 accuracy=31.794970	 loss=3.119275	 lr=0.087953
Epoch[070] Batch [2099]/[3760]	Speed: 117.729183 samples/sec	 accuracy=31.763393	 loss=3.120882	 lr=0.087944
Epoch[070] Batch [2149]/[3760]	Speed: 117.190451 samples/sec	 accuracy=31.756541	 loss=3.121198	 lr=0.087936
Epoch[070] Batch [2199]/[3760]	Speed: 116.895329 samples/sec	 accuracy=31.788352	 loss=3.120326	 lr=0.087927
Epoch[070] Batch [2249]/[3760]	Speed: 116.012263 samples/sec	 accuracy=31.773611	 loss=3.121477	 lr=0.087919
Epoch[070] Batch [2299]/[3760]	Speed: 116.808086 samples/sec	 accuracy=31.731658	 loss=3.122576	 lr=0.087911
Epoch[070] Batch [2349]/[3760]	Speed: 117.133015 samples/sec	 accuracy=31.739362	 loss=3.122931	 lr=0.087902
Epoch[070] Batch [2399]/[3760]	Speed: 117.804115 samples/sec	 accuracy=31.753906	 loss=3.123155	 lr=0.087894
Epoch[070] Batch [2449]/[3760]	Speed: 116.516028 samples/sec	 accuracy=31.741071	 loss=3.123895	 lr=0.087885
Epoch[070] Batch [2499]/[3760]	Speed: 117.276221 samples/sec	 accuracy=31.721875	 loss=3.125684	 lr=0.087877
Epoch[070] Batch [2549]/[3760]	Speed: 117.220823 samples/sec	 accuracy=31.727941	 loss=3.124960	 lr=0.087868
Epoch[070] Batch [2599]/[3760]	Speed: 117.221121 samples/sec	 accuracy=31.711538	 loss=3.124973	 lr=0.087860
Epoch[070] Batch [2649]/[3760]	Speed: 116.868039 samples/sec	 accuracy=31.706958	 loss=3.124752	 lr=0.087852
Epoch[070] Batch [2699]/[3760]	Speed: 117.321837 samples/sec	 accuracy=31.717014	 loss=3.124864	 lr=0.087843
Epoch[070] Batch [2749]/[3760]	Speed: 117.566901 samples/sec	 accuracy=31.717614	 loss=3.126029	 lr=0.087835
Epoch[070] Batch [2799]/[3760]	Speed: 117.121504 samples/sec	 accuracy=31.714844	 loss=3.126844	 lr=0.087826
Epoch[070] Batch [2849]/[3760]	Speed: 117.373720 samples/sec	 accuracy=31.735197	 loss=3.126572	 lr=0.087818
Epoch[070] Batch [2899]/[3760]	Speed: 116.323609 samples/sec	 accuracy=31.730065	 loss=3.127105	 lr=0.087809
Epoch[070] Batch [2949]/[3760]	Speed: 117.236179 samples/sec	 accuracy=31.740466	 loss=3.127393	 lr=0.087801
Epoch[070] Batch [2999]/[3760]	Speed: 116.992567 samples/sec	 accuracy=31.742188	 loss=3.127575	 lr=0.087793
Epoch[070] Batch [3049]/[3760]	Speed: 117.996696 samples/sec	 accuracy=31.750512	 loss=3.127337	 lr=0.087784
Epoch[070] Batch [3099]/[3760]	Speed: 117.101374 samples/sec	 accuracy=31.744456	 loss=3.127560	 lr=0.087776
Epoch[070] Batch [3149]/[3760]	Speed: 116.759280 samples/sec	 accuracy=31.730159	 loss=3.127605	 lr=0.087767
Epoch[070] Batch [3199]/[3760]	Speed: 117.241560 samples/sec	 accuracy=31.724609	 loss=3.127710	 lr=0.087759
Epoch[070] Batch [3249]/[3760]	Speed: 116.706823 samples/sec	 accuracy=31.742788	 loss=3.128109	 lr=0.087750
Epoch[070] Batch [3299]/[3760]	Speed: 117.169502 samples/sec	 accuracy=31.741004	 loss=3.127386	 lr=0.087742
Epoch[070] Batch [3349]/[3760]	Speed: 117.254517 samples/sec	 accuracy=31.715019	 loss=3.128407	 lr=0.087733
Epoch[070] Batch [3399]/[3760]	Speed: 117.055364 samples/sec	 accuracy=31.722886	 loss=3.128537	 lr=0.087725
Epoch[070] Batch [3449]/[3760]	Speed: 116.745706 samples/sec	 accuracy=31.713315	 loss=3.129186	 lr=0.087716
Epoch[070] Batch [3499]/[3760]	Speed: 117.096880 samples/sec	 accuracy=31.729018	 loss=3.128795	 lr=0.087708
Epoch[070] Batch [3549]/[3760]	Speed: 117.131738 samples/sec	 accuracy=31.704665	 loss=3.129571	 lr=0.087700
Epoch[070] Batch [3599]/[3760]	Speed: 116.168284 samples/sec	 accuracy=31.733073	 loss=3.128350	 lr=0.087691
Epoch[070] Batch [3649]/[3760]	Speed: 118.044561 samples/sec	 accuracy=31.726455	 loss=3.128601	 lr=0.087683
Epoch[070] Batch [3699]/[3760]	Speed: 117.169752 samples/sec	 accuracy=31.721706	 loss=3.129230	 lr=0.087674
Epoch[070] Batch [3749]/[3760]	Speed: 120.705223 samples/sec	 accuracy=31.711250	 loss=3.129218	 lr=0.087666
Batch [0049]/[0303]: acc-top1=33.625000 acc-top5=58.812500
Batch [0099]/[0303]: acc-top1=29.906250 acc-top5=57.890625
Batch [0149]/[0303]: acc-top1=31.739583 acc-top5=58.833333
Batch [0199]/[0303]: acc-top1=34.562500 acc-top5=60.812500
Batch [0249]/[0303]: acc-top1=33.743750 acc-top5=60.756250
Batch [0299]/[0303]: acc-top1=33.026042 acc-top5=60.526042
[Epoch 070] training: accuracy=31.715010	 loss=3.129274
[Epoch 070] speed: 116 samples/sec	time cost: 2183.313633
[Epoch 070] validation: acc-top1=33.065182 acc-top5=60.586840 loss=3.102715
Epoch[071] Batch [0049]/[3760]	Speed: 87.345177 samples/sec	 accuracy=31.718750	 loss=3.118978	 lr=0.087655
Epoch[071] Batch [0099]/[3760]	Speed: 116.357895 samples/sec	 accuracy=31.687500	 loss=3.111519	 lr=0.087647
Epoch[071] Batch [0149]/[3760]	Speed: 117.080321 samples/sec	 accuracy=31.979167	 loss=3.096516	 lr=0.087638
Epoch[071] Batch [0199]/[3760]	Speed: 116.235929 samples/sec	 accuracy=32.335938	 loss=3.079195	 lr=0.087630
Epoch[071] Batch [0249]/[3760]	Speed: 117.391792 samples/sec	 accuracy=32.287500	 loss=3.080674	 lr=0.087622
Epoch[071] Batch [0299]/[3760]	Speed: 117.678326 samples/sec	 accuracy=32.328125	 loss=3.084460	 lr=0.087613
Epoch[071] Batch [0349]/[3760]	Speed: 118.076313 samples/sec	 accuracy=32.339286	 loss=3.085387	 lr=0.087605
Epoch[071] Batch [0399]/[3760]	Speed: 117.398889 samples/sec	 accuracy=32.199219	 loss=3.097248	 lr=0.087596
Epoch[071] Batch [0449]/[3760]	Speed: 117.339812 samples/sec	 accuracy=32.260417	 loss=3.096677	 lr=0.087588
Epoch[071] Batch [0499]/[3760]	Speed: 117.057836 samples/sec	 accuracy=32.218750	 loss=3.101629	 lr=0.087579
Epoch[071] Batch [0549]/[3760]	Speed: 117.140858 samples/sec	 accuracy=32.204545	 loss=3.102000	 lr=0.087570
Epoch[071] Batch [0599]/[3760]	Speed: 117.553255 samples/sec	 accuracy=32.169271	 loss=3.103198	 lr=0.087562
Epoch[071] Batch [0649]/[3760]	Speed: 117.133973 samples/sec	 accuracy=32.156250	 loss=3.100461	 lr=0.087553
Epoch[071] Batch [0699]/[3760]	Speed: 117.060240 samples/sec	 accuracy=32.218750	 loss=3.100998	 lr=0.087545
Epoch[071] Batch [0749]/[3760]	Speed: 117.192102 samples/sec	 accuracy=32.222917	 loss=3.100406	 lr=0.087536
Epoch[071] Batch [0799]/[3760]	Speed: 117.187993 samples/sec	 accuracy=32.216797	 loss=3.101078	 lr=0.087528
Epoch[071] Batch [0849]/[3760]	Speed: 117.152191 samples/sec	 accuracy=32.202206	 loss=3.099846	 lr=0.087519
Epoch[071] Batch [0899]/[3760]	Speed: 117.527002 samples/sec	 accuracy=32.137153	 loss=3.101435	 lr=0.087511
Epoch[071] Batch [0949]/[3760]	Speed: 117.833895 samples/sec	 accuracy=32.110197	 loss=3.103063	 lr=0.087502
Epoch[071] Batch [0999]/[3760]	Speed: 116.436327 samples/sec	 accuracy=32.123438	 loss=3.101586	 lr=0.087494
Epoch[071] Batch [1049]/[3760]	Speed: 117.093357 samples/sec	 accuracy=32.127976	 loss=3.101220	 lr=0.087485
Epoch[071] Batch [1099]/[3760]	Speed: 117.395263 samples/sec	 accuracy=32.190341	 loss=3.101308	 lr=0.087477
Epoch[071] Batch [1149]/[3760]	Speed: 116.751587 samples/sec	 accuracy=32.206522	 loss=3.100409	 lr=0.087468
Epoch[071] Batch [1199]/[3760]	Speed: 117.852113 samples/sec	 accuracy=32.143229	 loss=3.101158	 lr=0.087460
Epoch[071] Batch [1249]/[3760]	Speed: 116.948845 samples/sec	 accuracy=32.152500	 loss=3.102765	 lr=0.087451
Epoch[071] Batch [1299]/[3760]	Speed: 115.535905 samples/sec	 accuracy=32.063702	 loss=3.108324	 lr=0.087443
Epoch[071] Batch [1349]/[3760]	Speed: 117.497711 samples/sec	 accuracy=32.076389	 loss=3.109419	 lr=0.087434
Epoch[071] Batch [1399]/[3760]	Speed: 117.335754 samples/sec	 accuracy=32.064732	 loss=3.109618	 lr=0.087425
Epoch[071] Batch [1449]/[3760]	Speed: 117.223904 samples/sec	 accuracy=32.031250	 loss=3.111550	 lr=0.087417
Epoch[071] Batch [1499]/[3760]	Speed: 117.390089 samples/sec	 accuracy=32.001042	 loss=3.113099	 lr=0.087408
Epoch[071] Batch [1549]/[3760]	Speed: 116.771608 samples/sec	 accuracy=31.969758	 loss=3.114613	 lr=0.087400
Epoch[071] Batch [1599]/[3760]	Speed: 117.345083 samples/sec	 accuracy=31.950195	 loss=3.115532	 lr=0.087391
Epoch[071] Batch [1649]/[3760]	Speed: 117.215527 samples/sec	 accuracy=31.949811	 loss=3.115477	 lr=0.087383
Epoch[071] Batch [1699]/[3760]	Speed: 117.722325 samples/sec	 accuracy=31.924632	 loss=3.116786	 lr=0.087374
Epoch[071] Batch [1749]/[3760]	Speed: 117.203837 samples/sec	 accuracy=31.901786	 loss=3.118973	 lr=0.087366
Epoch[071] Batch [1799]/[3760]	Speed: 117.865669 samples/sec	 accuracy=31.907986	 loss=3.119597	 lr=0.087357
Epoch[071] Batch [1849]/[3760]	Speed: 117.506902 samples/sec	 accuracy=31.878378	 loss=3.121239	 lr=0.087348
Epoch[071] Batch [1899]/[3760]	Speed: 118.314105 samples/sec	 accuracy=31.853618	 loss=3.121808	 lr=0.087340
Epoch[071] Batch [1949]/[3760]	Speed: 117.527746 samples/sec	 accuracy=31.848558	 loss=3.122801	 lr=0.087331
Epoch[071] Batch [1999]/[3760]	Speed: 117.748034 samples/sec	 accuracy=31.859375	 loss=3.123035	 lr=0.087323
Epoch[071] Batch [2049]/[3760]	Speed: 116.467247 samples/sec	 accuracy=31.885671	 loss=3.122471	 lr=0.087314
Epoch[071] Batch [2099]/[3760]	Speed: 117.068294 samples/sec	 accuracy=31.849702	 loss=3.123277	 lr=0.087305
Epoch[071] Batch [2149]/[3760]	Speed: 117.054498 samples/sec	 accuracy=31.859738	 loss=3.123237	 lr=0.087297
Epoch[071] Batch [2199]/[3760]	Speed: 117.432132 samples/sec	 accuracy=31.840199	 loss=3.123006	 lr=0.087288
Epoch[071] Batch [2249]/[3760]	Speed: 117.731589 samples/sec	 accuracy=31.840972	 loss=3.123639	 lr=0.087280
Epoch[071] Batch [2299]/[3760]	Speed: 117.023383 samples/sec	 accuracy=31.850543	 loss=3.124469	 lr=0.087271
Epoch[071] Batch [2349]/[3760]	Speed: 117.047658 samples/sec	 accuracy=31.855053	 loss=3.124210	 lr=0.087263
Epoch[071] Batch [2399]/[3760]	Speed: 117.323021 samples/sec	 accuracy=31.863932	 loss=3.124927	 lr=0.087254
Epoch[071] Batch [2449]/[3760]	Speed: 116.461802 samples/sec	 accuracy=31.848214	 loss=3.126171	 lr=0.087245
Epoch[071] Batch [2499]/[3760]	Speed: 117.107484 samples/sec	 accuracy=31.850625	 loss=3.126167	 lr=0.087237
Epoch[071] Batch [2549]/[3760]	Speed: 117.113028 samples/sec	 accuracy=31.853554	 loss=3.126661	 lr=0.087228
Epoch[071] Batch [2599]/[3760]	Speed: 117.435708 samples/sec	 accuracy=31.855769	 loss=3.125780	 lr=0.087219
Epoch[071] Batch [2649]/[3760]	Speed: 117.260189 samples/sec	 accuracy=31.844340	 loss=3.126326	 lr=0.087211
Epoch[071] Batch [2699]/[3760]	Speed: 118.383248 samples/sec	 accuracy=31.821181	 loss=3.127336	 lr=0.087202
Epoch[071] Batch [2749]/[3760]	Speed: 117.102019 samples/sec	 accuracy=31.810795	 loss=3.127606	 lr=0.087194
Epoch[071] Batch [2799]/[3760]	Speed: 117.313229 samples/sec	 accuracy=31.823103	 loss=3.126678	 lr=0.087185
Epoch[071] Batch [2849]/[3760]	Speed: 118.109417 samples/sec	 accuracy=31.849781	 loss=3.125557	 lr=0.087176
Epoch[071] Batch [2899]/[3760]	Speed: 117.251503 samples/sec	 accuracy=31.851293	 loss=3.126155	 lr=0.087168
Epoch[071] Batch [2949]/[3760]	Speed: 117.648407 samples/sec	 accuracy=31.852225	 loss=3.126951	 lr=0.087159
Epoch[071] Batch [2999]/[3760]	Speed: 116.813348 samples/sec	 accuracy=31.874479	 loss=3.125607	 lr=0.087151
Epoch[071] Batch [3049]/[3760]	Speed: 117.598132 samples/sec	 accuracy=31.877561	 loss=3.125637	 lr=0.087142
Epoch[071] Batch [3099]/[3760]	Speed: 117.303395 samples/sec	 accuracy=31.893649	 loss=3.125329	 lr=0.087133
Epoch[071] Batch [3149]/[3760]	Speed: 116.228098 samples/sec	 accuracy=31.882440	 loss=3.126875	 lr=0.087125
Epoch[071] Batch [3199]/[3760]	Speed: 116.884169 samples/sec	 accuracy=31.881836	 loss=3.126490	 lr=0.087116
Epoch[071] Batch [3249]/[3760]	Speed: 117.549251 samples/sec	 accuracy=31.875481	 loss=3.126460	 lr=0.087107
Epoch[071] Batch [3299]/[3760]	Speed: 117.631355 samples/sec	 accuracy=31.868371	 loss=3.126381	 lr=0.087099
Epoch[071] Batch [3349]/[3760]	Speed: 117.397385 samples/sec	 accuracy=31.862407	 loss=3.127128	 lr=0.087090
Epoch[071] Batch [3399]/[3760]	Speed: 118.117094 samples/sec	 accuracy=31.849724	 loss=3.127523	 lr=0.087081
Epoch[071] Batch [3449]/[3760]	Speed: 117.784028 samples/sec	 accuracy=31.837409	 loss=3.128340	 lr=0.087073
Epoch[071] Batch [3499]/[3760]	Speed: 117.595005 samples/sec	 accuracy=31.839286	 loss=3.128186	 lr=0.087064
Epoch[071] Batch [3549]/[3760]	Speed: 117.329010 samples/sec	 accuracy=31.829665	 loss=3.128264	 lr=0.087055
Epoch[071] Batch [3599]/[3760]	Speed: 117.488261 samples/sec	 accuracy=31.845052	 loss=3.127824	 lr=0.087047
Epoch[071] Batch [3649]/[3760]	Speed: 116.779872 samples/sec	 accuracy=31.826627	 loss=3.128386	 lr=0.087038
Epoch[071] Batch [3699]/[3760]	Speed: 117.295079 samples/sec	 accuracy=31.833193	 loss=3.127709	 lr=0.087029
Epoch[071] Batch [3749]/[3760]	Speed: 119.990473 samples/sec	 accuracy=31.847500	 loss=3.126710	 lr=0.087021
Batch [0049]/[0303]: acc-top1=32.250000 acc-top5=57.656250
Batch [0099]/[0303]: acc-top1=29.625000 acc-top5=55.625000
Batch [0149]/[0303]: acc-top1=31.354167 acc-top5=56.645833
Batch [0199]/[0303]: acc-top1=33.335938 acc-top5=58.562500
Batch [0249]/[0303]: acc-top1=32.368750 acc-top5=58.406250
Batch [0299]/[0303]: acc-top1=32.125000 acc-top5=58.911458
[Epoch 071] training: accuracy=31.841340	 loss=3.126981
[Epoch 071] speed: 116 samples/sec	time cost: 2179.877415
[Epoch 071] validation: acc-top1=32.126650 acc-top5=58.910891 loss=3.237991
Epoch[072] Batch [0049]/[3759]	Speed: 86.793679 samples/sec	 accuracy=31.718750	 loss=3.148294	 lr=0.087010
Epoch[072] Batch [0099]/[3759]	Speed: 117.508464 samples/sec	 accuracy=32.000000	 loss=3.110144	 lr=0.087002
Epoch[072] Batch [0149]/[3759]	Speed: 117.453717 samples/sec	 accuracy=32.239583	 loss=3.093681	 lr=0.086993
Epoch[072] Batch [0199]/[3759]	Speed: 117.067670 samples/sec	 accuracy=32.132812	 loss=3.099944	 lr=0.086984
Epoch[072] Batch [0249]/[3759]	Speed: 117.172356 samples/sec	 accuracy=32.512500	 loss=3.091021	 lr=0.086976
Epoch[072] Batch [0299]/[3759]	Speed: 116.491029 samples/sec	 accuracy=32.520833	 loss=3.093623	 lr=0.086967
Epoch[072] Batch [0349]/[3759]	Speed: 117.394321 samples/sec	 accuracy=32.410714	 loss=3.093324	 lr=0.086958
Epoch[072] Batch [0399]/[3759]	Speed: 116.766724 samples/sec	 accuracy=32.406250	 loss=3.091796	 lr=0.086950
Epoch[072] Batch [0449]/[3759]	Speed: 117.573431 samples/sec	 accuracy=32.295139	 loss=3.096434	 lr=0.086941
Epoch[072] Batch [0499]/[3759]	Speed: 116.990939 samples/sec	 accuracy=32.343750	 loss=3.098453	 lr=0.086932
Epoch[072] Batch [0549]/[3759]	Speed: 117.795380 samples/sec	 accuracy=32.363636	 loss=3.095683	 lr=0.086923
Epoch[072] Batch [0599]/[3759]	Speed: 117.747884 samples/sec	 accuracy=32.302083	 loss=3.102476	 lr=0.086915
Epoch[072] Batch [0649]/[3759]	Speed: 117.347012 samples/sec	 accuracy=32.389423	 loss=3.100231	 lr=0.086906
Epoch[072] Batch [0699]/[3759]	Speed: 116.943206 samples/sec	 accuracy=32.343750	 loss=3.102069	 lr=0.086897
Epoch[072] Batch [0749]/[3759]	Speed: 117.666490 samples/sec	 accuracy=32.385417	 loss=3.099249	 lr=0.086889
Epoch[072] Batch [0799]/[3759]	Speed: 116.978622 samples/sec	 accuracy=32.437500	 loss=3.094094	 lr=0.086880
Epoch[072] Batch [0849]/[3759]	Speed: 117.071452 samples/sec	 accuracy=32.393382	 loss=3.096402	 lr=0.086871
Epoch[072] Batch [0899]/[3759]	Speed: 117.330916 samples/sec	 accuracy=32.315972	 loss=3.100407	 lr=0.086863
Epoch[072] Batch [0949]/[3759]	Speed: 117.348278 samples/sec	 accuracy=32.302632	 loss=3.101770	 lr=0.086854
Epoch[072] Batch [0999]/[3759]	Speed: 117.040639 samples/sec	 accuracy=32.306250	 loss=3.099572	 lr=0.086845
Epoch[072] Batch [1049]/[3759]	Speed: 117.972952 samples/sec	 accuracy=32.297619	 loss=3.100733	 lr=0.086836
Epoch[072] Batch [1099]/[3759]	Speed: 116.662210 samples/sec	 accuracy=32.267045	 loss=3.102330	 lr=0.086828
Epoch[072] Batch [1149]/[3759]	Speed: 118.136750 samples/sec	 accuracy=32.244565	 loss=3.102204	 lr=0.086819
Epoch[072] Batch [1199]/[3759]	Speed: 117.066295 samples/sec	 accuracy=32.248698	 loss=3.100994	 lr=0.086810
Epoch[072] Batch [1249]/[3759]	Speed: 117.638358 samples/sec	 accuracy=32.233750	 loss=3.101867	 lr=0.086802
Epoch[072] Batch [1299]/[3759]	Speed: 117.106323 samples/sec	 accuracy=32.182692	 loss=3.103499	 lr=0.086793
Epoch[072] Batch [1349]/[3759]	Speed: 116.789515 samples/sec	 accuracy=32.185185	 loss=3.103761	 lr=0.086784
Epoch[072] Batch [1399]/[3759]	Speed: 117.912097 samples/sec	 accuracy=32.217634	 loss=3.101909	 lr=0.086775
Epoch[072] Batch [1449]/[3759]	Speed: 117.707748 samples/sec	 accuracy=32.240302	 loss=3.101475	 lr=0.086767
Epoch[072] Batch [1499]/[3759]	Speed: 116.935120 samples/sec	 accuracy=32.216667	 loss=3.102520	 lr=0.086758
Epoch[072] Batch [1549]/[3759]	Speed: 117.226068 samples/sec	 accuracy=32.183468	 loss=3.105010	 lr=0.086749
Epoch[072] Batch [1599]/[3759]	Speed: 117.307782 samples/sec	 accuracy=32.189453	 loss=3.105791	 lr=0.086740
Epoch[072] Batch [1649]/[3759]	Speed: 116.564890 samples/sec	 accuracy=32.195076	 loss=3.104856	 lr=0.086732
Epoch[072] Batch [1699]/[3759]	Speed: 117.571023 samples/sec	 accuracy=32.192096	 loss=3.105308	 lr=0.086723
Epoch[072] Batch [1749]/[3759]	Speed: 116.277732 samples/sec	 accuracy=32.212500	 loss=3.104029	 lr=0.086714
Epoch[072] Batch [1799]/[3759]	Speed: 116.813848 samples/sec	 accuracy=32.234375	 loss=3.103744	 lr=0.086705
Epoch[072] Batch [1849]/[3759]	Speed: 117.256269 samples/sec	 accuracy=32.206926	 loss=3.104846	 lr=0.086697
Epoch[072] Batch [1899]/[3759]	Speed: 117.025831 samples/sec	 accuracy=32.221217	 loss=3.104316	 lr=0.086688
Epoch[072] Batch [1949]/[3759]	Speed: 117.467798 samples/sec	 accuracy=32.189103	 loss=3.105375	 lr=0.086679
Epoch[072] Batch [1999]/[3759]	Speed: 116.886219 samples/sec	 accuracy=32.196094	 loss=3.104918	 lr=0.086670
Epoch[072] Batch [2049]/[3759]	Speed: 117.704793 samples/sec	 accuracy=32.194360	 loss=3.105127	 lr=0.086661
Epoch[072] Batch [2099]/[3759]	Speed: 117.112154 samples/sec	 accuracy=32.206845	 loss=3.104556	 lr=0.086653
Epoch[072] Batch [2149]/[3759]	Speed: 117.300249 samples/sec	 accuracy=32.207122	 loss=3.103579	 lr=0.086644
Epoch[072] Batch [2199]/[3759]	Speed: 117.434243 samples/sec	 accuracy=32.187500	 loss=3.104360	 lr=0.086635
Epoch[072] Batch [2249]/[3759]	Speed: 118.029821 samples/sec	 accuracy=32.199306	 loss=3.104179	 lr=0.086626
Epoch[072] Batch [2299]/[3759]	Speed: 118.132205 samples/sec	 accuracy=32.207201	 loss=3.104075	 lr=0.086618
Epoch[072] Batch [2349]/[3759]	Speed: 116.724331 samples/sec	 accuracy=32.205452	 loss=3.104913	 lr=0.086609
Epoch[072] Batch [2399]/[3759]	Speed: 116.863459 samples/sec	 accuracy=32.210938	 loss=3.104035	 lr=0.086600
Epoch[072] Batch [2449]/[3759]	Speed: 116.885957 samples/sec	 accuracy=32.177296	 loss=3.104598	 lr=0.086591
Epoch[072] Batch [2499]/[3759]	Speed: 117.453774 samples/sec	 accuracy=32.158125	 loss=3.105836	 lr=0.086582
Epoch[072] Batch [2549]/[3759]	Speed: 116.957179 samples/sec	 accuracy=32.177083	 loss=3.104610	 lr=0.086574
Epoch[072] Batch [2599]/[3759]	Speed: 118.263145 samples/sec	 accuracy=32.161058	 loss=3.105873	 lr=0.086565
Epoch[072] Batch [2649]/[3759]	Speed: 117.854644 samples/sec	 accuracy=32.137972	 loss=3.107245	 lr=0.086556
Epoch[072] Batch [2699]/[3759]	Speed: 116.399778 samples/sec	 accuracy=32.133681	 loss=3.107166	 lr=0.086547
Epoch[072] Batch [2749]/[3759]	Speed: 117.120857 samples/sec	 accuracy=32.134091	 loss=3.106767	 lr=0.086538
Epoch[072] Batch [2799]/[3759]	Speed: 118.026506 samples/sec	 accuracy=32.124442	 loss=3.107409	 lr=0.086530
Epoch[072] Batch [2849]/[3759]	Speed: 117.570013 samples/sec	 accuracy=32.127741	 loss=3.107406	 lr=0.086521
Epoch[072] Batch [2899]/[3759]	Speed: 117.358455 samples/sec	 accuracy=32.137931	 loss=3.107974	 lr=0.086512
Epoch[072] Batch [2949]/[3759]	Speed: 117.625457 samples/sec	 accuracy=32.104873	 loss=3.109197	 lr=0.086503
Epoch[072] Batch [2999]/[3759]	Speed: 117.769996 samples/sec	 accuracy=32.094792	 loss=3.110404	 lr=0.086494
Epoch[072] Batch [3049]/[3759]	Speed: 117.630322 samples/sec	 accuracy=32.098873	 loss=3.109013	 lr=0.086486
Epoch[072] Batch [3099]/[3759]	Speed: 116.705240 samples/sec	 accuracy=32.108367	 loss=3.108346	 lr=0.086477
Epoch[072] Batch [3149]/[3759]	Speed: 117.033932 samples/sec	 accuracy=32.102183	 loss=3.107943	 lr=0.086468
Epoch[072] Batch [3199]/[3759]	Speed: 118.029333 samples/sec	 accuracy=32.095215	 loss=3.108444	 lr=0.086459
Epoch[072] Batch [3249]/[3759]	Speed: 116.825708 samples/sec	 accuracy=32.099038	 loss=3.108416	 lr=0.086450
Epoch[072] Batch [3299]/[3759]	Speed: 117.775198 samples/sec	 accuracy=32.068655	 loss=3.109692	 lr=0.086441
Epoch[072] Batch [3349]/[3759]	Speed: 117.694364 samples/sec	 accuracy=32.054104	 loss=3.110706	 lr=0.086433
Epoch[072] Batch [3399]/[3759]	Speed: 116.740862 samples/sec	 accuracy=32.062500	 loss=3.110622	 lr=0.086424
Epoch[072] Batch [3449]/[3759]	Speed: 117.054304 samples/sec	 accuracy=32.043025	 loss=3.111814	 lr=0.086415
Epoch[072] Batch [3499]/[3759]	Speed: 116.842317 samples/sec	 accuracy=32.009375	 loss=3.113131	 lr=0.086406
Epoch[072] Batch [3549]/[3759]	Speed: 117.967095 samples/sec	 accuracy=31.985475	 loss=3.113881	 lr=0.086397
Epoch[072] Batch [3599]/[3759]	Speed: 118.196793 samples/sec	 accuracy=31.975694	 loss=3.114348	 lr=0.086388
Epoch[072] Batch [3649]/[3759]	Speed: 116.755229 samples/sec	 accuracy=31.984161	 loss=3.114424	 lr=0.086380
Epoch[072] Batch [3699]/[3759]	Speed: 117.550623 samples/sec	 accuracy=31.984797	 loss=3.113950	 lr=0.086371
Epoch[072] Batch [3749]/[3759]	Speed: 120.294901 samples/sec	 accuracy=31.977083	 loss=3.114848	 lr=0.086362
Batch [0049]/[0303]: acc-top1=34.468750 acc-top5=59.781250
Batch [0099]/[0303]: acc-top1=32.265625 acc-top5=58.812500
Batch [0149]/[0303]: acc-top1=33.520833 acc-top5=60.479167
Batch [0199]/[0303]: acc-top1=35.640625 acc-top5=62.210938
Batch [0249]/[0303]: acc-top1=34.875000 acc-top5=61.862500
Batch [0299]/[0303]: acc-top1=34.515625 acc-top5=61.875000
[Epoch 072] training: accuracy=31.976590	 loss=3.114682
[Epoch 072] speed: 116 samples/sec	time cost: 2178.958024
[Epoch 072] validation: acc-top1=34.596741 acc-top5=61.865718 loss=3.008119
Epoch[073] Batch [0049]/[3760]	Speed: 86.743973 samples/sec	 accuracy=32.218750	 loss=3.128454	 lr=0.086351
Epoch[073] Batch [0099]/[3760]	Speed: 117.977227 samples/sec	 accuracy=32.343750	 loss=3.110283	 lr=0.086343
Epoch[073] Batch [0149]/[3760]	Speed: 117.105545 samples/sec	 accuracy=32.489583	 loss=3.093299	 lr=0.086334
Epoch[073] Batch [0199]/[3760]	Speed: 118.163401 samples/sec	 accuracy=32.664062	 loss=3.076672	 lr=0.086325
Epoch[073] Batch [0249]/[3760]	Speed: 116.988906 samples/sec	 accuracy=32.650000	 loss=3.071780	 lr=0.086316
Epoch[073] Batch [0299]/[3760]	Speed: 117.307406 samples/sec	 accuracy=32.432292	 loss=3.074569	 lr=0.086307
Epoch[073] Batch [0349]/[3760]	Speed: 117.511674 samples/sec	 accuracy=32.339286	 loss=3.082691	 lr=0.086298
Epoch[073] Batch [0399]/[3760]	Speed: 117.143571 samples/sec	 accuracy=32.359375	 loss=3.083291	 lr=0.086289
Epoch[073] Batch [0449]/[3760]	Speed: 117.232789 samples/sec	 accuracy=32.232639	 loss=3.093208	 lr=0.086281
Epoch[073] Batch [0499]/[3760]	Speed: 117.539351 samples/sec	 accuracy=32.343750	 loss=3.093253	 lr=0.086272
Epoch[073] Batch [0549]/[3760]	Speed: 117.541767 samples/sec	 accuracy=32.315341	 loss=3.092832	 lr=0.086263
Epoch[073] Batch [0599]/[3760]	Speed: 117.337492 samples/sec	 accuracy=32.307292	 loss=3.096614	 lr=0.086254
Epoch[073] Batch [0649]/[3760]	Speed: 117.568503 samples/sec	 accuracy=32.343750	 loss=3.093102	 lr=0.086245
Epoch[073] Batch [0699]/[3760]	Speed: 117.753524 samples/sec	 accuracy=32.301339	 loss=3.094460	 lr=0.086236
Epoch[073] Batch [0749]/[3760]	Speed: 117.863118 samples/sec	 accuracy=32.243750	 loss=3.093627	 lr=0.086227
Epoch[073] Batch [0799]/[3760]	Speed: 117.433104 samples/sec	 accuracy=32.193359	 loss=3.097459	 lr=0.086218
Epoch[073] Batch [0849]/[3760]	Speed: 117.660569 samples/sec	 accuracy=32.213235	 loss=3.097751	 lr=0.086209
Epoch[073] Batch [0899]/[3760]	Speed: 117.500691 samples/sec	 accuracy=32.234375	 loss=3.095517	 lr=0.086201
Epoch[073] Batch [0949]/[3760]	Speed: 116.885734 samples/sec	 accuracy=32.271382	 loss=3.094477	 lr=0.086192
Epoch[073] Batch [0999]/[3760]	Speed: 118.204703 samples/sec	 accuracy=32.251562	 loss=3.098343	 lr=0.086183
Epoch[073] Batch [1049]/[3760]	Speed: 117.571230 samples/sec	 accuracy=32.250000	 loss=3.099798	 lr=0.086174
Epoch[073] Batch [1099]/[3760]	Speed: 117.942822 samples/sec	 accuracy=32.204545	 loss=3.099052	 lr=0.086165
Epoch[073] Batch [1149]/[3760]	Speed: 117.731752 samples/sec	 accuracy=32.236413	 loss=3.099472	 lr=0.086156
Epoch[073] Batch [1199]/[3760]	Speed: 118.182033 samples/sec	 accuracy=32.205729	 loss=3.101146	 lr=0.086147
Epoch[073] Batch [1249]/[3760]	Speed: 118.421014 samples/sec	 accuracy=32.206250	 loss=3.101857	 lr=0.086138
Epoch[073] Batch [1299]/[3760]	Speed: 118.198072 samples/sec	 accuracy=32.168269	 loss=3.102257	 lr=0.086129
Epoch[073] Batch [1349]/[3760]	Speed: 117.756633 samples/sec	 accuracy=32.158565	 loss=3.103767	 lr=0.086120
Epoch[073] Batch [1399]/[3760]	Speed: 117.849698 samples/sec	 accuracy=32.121652	 loss=3.104801	 lr=0.086111
Epoch[073] Batch [1449]/[3760]	Speed: 117.380789 samples/sec	 accuracy=32.143319	 loss=3.103320	 lr=0.086103
Epoch[073] Batch [1499]/[3760]	Speed: 117.444667 samples/sec	 accuracy=32.133333	 loss=3.102655	 lr=0.086094
Epoch[073] Batch [1549]/[3760]	Speed: 116.913073 samples/sec	 accuracy=32.129032	 loss=3.104712	 lr=0.086085
Epoch[073] Batch [1599]/[3760]	Speed: 117.784196 samples/sec	 accuracy=32.120117	 loss=3.106257	 lr=0.086076
Epoch[073] Batch [1649]/[3760]	Speed: 116.989083 samples/sec	 accuracy=32.119318	 loss=3.104871	 lr=0.086067
Epoch[073] Batch [1699]/[3760]	Speed: 117.616165 samples/sec	 accuracy=32.131434	 loss=3.104127	 lr=0.086058
Epoch[073] Batch [1749]/[3760]	Speed: 117.103248 samples/sec	 accuracy=32.105357	 loss=3.106192	 lr=0.086049
Epoch[073] Batch [1799]/[3760]	Speed: 118.275414 samples/sec	 accuracy=32.114583	 loss=3.106345	 lr=0.086040
Epoch[073] Batch [1849]/[3760]	Speed: 117.638327 samples/sec	 accuracy=32.126689	 loss=3.105380	 lr=0.086031
Epoch[073] Batch [1899]/[3760]	Speed: 117.657008 samples/sec	 accuracy=32.100329	 loss=3.106513	 lr=0.086022
Epoch[073] Batch [1949]/[3760]	Speed: 118.153477 samples/sec	 accuracy=32.056090	 loss=3.108264	 lr=0.086013
Epoch[073] Batch [1999]/[3760]	Speed: 117.776066 samples/sec	 accuracy=32.028906	 loss=3.109185	 lr=0.086004
Epoch[073] Batch [2049]/[3760]	Speed: 117.311196 samples/sec	 accuracy=32.006860	 loss=3.110778	 lr=0.085995
Epoch[073] Batch [2099]/[3760]	Speed: 117.119028 samples/sec	 accuracy=31.959821	 loss=3.113384	 lr=0.085986
Epoch[073] Batch [2149]/[3760]	Speed: 116.781390 samples/sec	 accuracy=31.936047	 loss=3.114520	 lr=0.085977
Epoch[073] Batch [2199]/[3760]	Speed: 117.388902 samples/sec	 accuracy=31.948153	 loss=3.114410	 lr=0.085968
Epoch[073] Batch [2249]/[3760]	Speed: 117.568168 samples/sec	 accuracy=31.949306	 loss=3.114190	 lr=0.085959
Epoch[073] Batch [2299]/[3760]	Speed: 117.339663 samples/sec	 accuracy=31.921875	 loss=3.115523	 lr=0.085951
Epoch[073] Batch [2349]/[3760]	Speed: 116.735457 samples/sec	 accuracy=31.924202	 loss=3.115357	 lr=0.085942
Epoch[073] Batch [2399]/[3760]	Speed: 117.405874 samples/sec	 accuracy=31.899089	 loss=3.116570	 lr=0.085933
Epoch[073] Batch [2449]/[3760]	Speed: 116.010157 samples/sec	 accuracy=31.913903	 loss=3.115602	 lr=0.085924
Epoch[073] Batch [2499]/[3760]	Speed: 117.008195 samples/sec	 accuracy=31.899375	 loss=3.116406	 lr=0.085915
Epoch[073] Batch [2549]/[3760]	Speed: 117.019430 samples/sec	 accuracy=31.887868	 loss=3.117421	 lr=0.085906
Epoch[073] Batch [2599]/[3760]	Speed: 116.408271 samples/sec	 accuracy=31.900841	 loss=3.116622	 lr=0.085897
Epoch[073] Batch [2649]/[3760]	Speed: 117.088700 samples/sec	 accuracy=31.914505	 loss=3.116036	 lr=0.085888
Epoch[073] Batch [2699]/[3760]	Speed: 117.624693 samples/sec	 accuracy=31.917824	 loss=3.115952	 lr=0.085879
Epoch[073] Batch [2749]/[3760]	Speed: 117.057397 samples/sec	 accuracy=31.915341	 loss=3.116387	 lr=0.085870
Epoch[073] Batch [2799]/[3760]	Speed: 117.217923 samples/sec	 accuracy=31.904576	 loss=3.117246	 lr=0.085861
Epoch[073] Batch [2849]/[3760]	Speed: 117.895427 samples/sec	 accuracy=31.882675	 loss=3.118582	 lr=0.085852
Epoch[073] Batch [2899]/[3760]	Speed: 117.751430 samples/sec	 accuracy=31.863147	 loss=3.119214	 lr=0.085843
Epoch[073] Batch [2949]/[3760]	Speed: 117.326850 samples/sec	 accuracy=31.862288	 loss=3.121081	 lr=0.085834
Epoch[073] Batch [2999]/[3760]	Speed: 116.788644 samples/sec	 accuracy=31.878125	 loss=3.120370	 lr=0.085825
Epoch[073] Batch [3049]/[3760]	Speed: 117.599520 samples/sec	 accuracy=31.887807	 loss=3.120333	 lr=0.085816
Epoch[073] Batch [3099]/[3760]	Speed: 117.850782 samples/sec	 accuracy=31.879032	 loss=3.120649	 lr=0.085807
Epoch[073] Batch [3149]/[3760]	Speed: 117.309483 samples/sec	 accuracy=31.883929	 loss=3.120933	 lr=0.085798
Epoch[073] Batch [3199]/[3760]	Speed: 117.558519 samples/sec	 accuracy=31.877441	 loss=3.120552	 lr=0.085789
Epoch[073] Batch [3249]/[3760]	Speed: 117.070532 samples/sec	 accuracy=31.881250	 loss=3.120621	 lr=0.085780
Epoch[073] Batch [3299]/[3760]	Speed: 117.695557 samples/sec	 accuracy=31.869792	 loss=3.121694	 lr=0.085771
Epoch[073] Batch [3349]/[3760]	Speed: 116.927020 samples/sec	 accuracy=31.870336	 loss=3.121517	 lr=0.085762
Epoch[073] Batch [3399]/[3760]	Speed: 117.409793 samples/sec	 accuracy=31.869026	 loss=3.121434	 lr=0.085753
Epoch[073] Batch [3449]/[3760]	Speed: 116.575046 samples/sec	 accuracy=31.855525	 loss=3.122011	 lr=0.085744
Epoch[073] Batch [3499]/[3760]	Speed: 118.154997 samples/sec	 accuracy=31.843304	 loss=3.122421	 lr=0.085735
Epoch[073] Batch [3549]/[3760]	Speed: 118.163217 samples/sec	 accuracy=31.849472	 loss=3.122427	 lr=0.085726
Epoch[073] Batch [3599]/[3760]	Speed: 117.525008 samples/sec	 accuracy=31.832465	 loss=3.122850	 lr=0.085717
Epoch[073] Batch [3649]/[3760]	Speed: 117.931452 samples/sec	 accuracy=31.851884	 loss=3.121913	 lr=0.085708
Epoch[073] Batch [3699]/[3760]	Speed: 117.024383 samples/sec	 accuracy=31.856419	 loss=3.122261	 lr=0.085699
Epoch[073] Batch [3749]/[3760]	Speed: 120.564827 samples/sec	 accuracy=31.859167	 loss=3.122277	 lr=0.085690
Batch [0049]/[0303]: acc-top1=34.843750 acc-top5=62.031250
Batch [0099]/[0303]: acc-top1=33.515625 acc-top5=59.750000
Batch [0149]/[0303]: acc-top1=35.239583 acc-top5=61.312500
Batch [0199]/[0303]: acc-top1=37.304688 acc-top5=63.000000
Batch [0249]/[0303]: acc-top1=36.093750 acc-top5=62.275000
Batch [0299]/[0303]: acc-top1=34.921875 acc-top5=61.718750
[Epoch 073] training: accuracy=31.856715	 loss=3.122218
[Epoch 073] speed: 116 samples/sec	time cost: 2177.036623
[Epoch 073] validation: acc-top1=34.828795 acc-top5=61.716172 loss=2.987804
Epoch[074] Batch [0049]/[3760]	Speed: 87.040607 samples/sec	 accuracy=32.343750	 loss=3.033843	 lr=0.085679
Epoch[074] Batch [0099]/[3760]	Speed: 117.595881 samples/sec	 accuracy=32.218750	 loss=3.068466	 lr=0.085670
Epoch[074] Batch [0149]/[3760]	Speed: 118.126649 samples/sec	 accuracy=32.031250	 loss=3.087101	 lr=0.085661
Epoch[074] Batch [0199]/[3760]	Speed: 118.164457 samples/sec	 accuracy=31.703125	 loss=3.093028	 lr=0.085652
Epoch[074] Batch [0249]/[3760]	Speed: 117.201846 samples/sec	 accuracy=31.843750	 loss=3.103065	 lr=0.085643
Epoch[074] Batch [0299]/[3760]	Speed: 116.989945 samples/sec	 accuracy=31.984375	 loss=3.095705	 lr=0.085634
Epoch[074] Batch [0349]/[3760]	Speed: 118.091501 samples/sec	 accuracy=31.955357	 loss=3.102991	 lr=0.085624
Epoch[074] Batch [0399]/[3760]	Speed: 116.563883 samples/sec	 accuracy=32.089844	 loss=3.095748	 lr=0.085615
Epoch[074] Batch [0449]/[3760]	Speed: 117.317578 samples/sec	 accuracy=32.225694	 loss=3.087681	 lr=0.085606
Epoch[074] Batch [0499]/[3760]	Speed: 117.716025 samples/sec	 accuracy=32.362500	 loss=3.086360	 lr=0.085597
Epoch[074] Batch [0549]/[3760]	Speed: 117.734162 samples/sec	 accuracy=32.468750	 loss=3.081151	 lr=0.085588
Epoch[074] Batch [0599]/[3760]	Speed: 117.481264 samples/sec	 accuracy=32.403646	 loss=3.084970	 lr=0.085579
Epoch[074] Batch [0649]/[3760]	Speed: 116.710502 samples/sec	 accuracy=32.502404	 loss=3.080744	 lr=0.085570
Epoch[074] Batch [0699]/[3760]	Speed: 117.380924 samples/sec	 accuracy=32.537946	 loss=3.080624	 lr=0.085561
Epoch[074] Batch [0749]/[3760]	Speed: 117.681228 samples/sec	 accuracy=32.445833	 loss=3.084546	 lr=0.085552
Epoch[074] Batch [0799]/[3760]	Speed: 117.878042 samples/sec	 accuracy=32.412109	 loss=3.089581	 lr=0.085543
Epoch[074] Batch [0849]/[3760]	Speed: 117.750338 samples/sec	 accuracy=32.397059	 loss=3.092082	 lr=0.085534
Epoch[074] Batch [0899]/[3760]	Speed: 116.956420 samples/sec	 accuracy=32.359375	 loss=3.093428	 lr=0.085525
Epoch[074] Batch [0949]/[3760]	Speed: 117.555430 samples/sec	 accuracy=32.419408	 loss=3.092171	 lr=0.085516
Epoch[074] Batch [0999]/[3760]	Speed: 117.731380 samples/sec	 accuracy=32.464063	 loss=3.089595	 lr=0.085507
Epoch[074] Batch [1049]/[3760]	Speed: 117.900873 samples/sec	 accuracy=32.441964	 loss=3.089043	 lr=0.085498
Epoch[074] Batch [1099]/[3760]	Speed: 117.849388 samples/sec	 accuracy=32.375000	 loss=3.089520	 lr=0.085488
Epoch[074] Batch [1149]/[3760]	Speed: 117.807459 samples/sec	 accuracy=32.423913	 loss=3.089396	 lr=0.085479
Epoch[074] Batch [1199]/[3760]	Speed: 117.358497 samples/sec	 accuracy=32.430990	 loss=3.091527	 lr=0.085470
Epoch[074] Batch [1249]/[3760]	Speed: 117.407176 samples/sec	 accuracy=32.407500	 loss=3.092509	 lr=0.085461
Epoch[074] Batch [1299]/[3760]	Speed: 117.110074 samples/sec	 accuracy=32.381010	 loss=3.094111	 lr=0.085452
Epoch[074] Batch [1349]/[3760]	Speed: 117.441151 samples/sec	 accuracy=32.425926	 loss=3.093311	 lr=0.085443
Epoch[074] Batch [1399]/[3760]	Speed: 117.294120 samples/sec	 accuracy=32.389509	 loss=3.096382	 lr=0.085434
Epoch[074] Batch [1449]/[3760]	Speed: 118.192796 samples/sec	 accuracy=32.364224	 loss=3.095999	 lr=0.085425
Epoch[074] Batch [1499]/[3760]	Speed: 116.995286 samples/sec	 accuracy=32.352083	 loss=3.096616	 lr=0.085416
Epoch[074] Batch [1549]/[3760]	Speed: 117.295875 samples/sec	 accuracy=32.341734	 loss=3.096857	 lr=0.085407
Epoch[074] Batch [1599]/[3760]	Speed: 117.657340 samples/sec	 accuracy=32.340820	 loss=3.097593	 lr=0.085397
Epoch[074] Batch [1649]/[3760]	Speed: 117.954406 samples/sec	 accuracy=32.312500	 loss=3.098145	 lr=0.085388
Epoch[074] Batch [1699]/[3760]	Speed: 117.990694 samples/sec	 accuracy=32.297794	 loss=3.099705	 lr=0.085379
Epoch[074] Batch [1749]/[3760]	Speed: 117.729710 samples/sec	 accuracy=32.351786	 loss=3.098343	 lr=0.085370
Epoch[074] Batch [1799]/[3760]	Speed: 117.500640 samples/sec	 accuracy=32.377604	 loss=3.097017	 lr=0.085361
Epoch[074] Batch [1849]/[3760]	Speed: 118.215064 samples/sec	 accuracy=32.374155	 loss=3.096858	 lr=0.085352
Epoch[074] Batch [1899]/[3760]	Speed: 118.050642 samples/sec	 accuracy=32.356908	 loss=3.098399	 lr=0.085343
Epoch[074] Batch [1949]/[3760]	Speed: 117.590997 samples/sec	 accuracy=32.310897	 loss=3.100385	 lr=0.085334
Epoch[074] Batch [1999]/[3760]	Speed: 117.494462 samples/sec	 accuracy=32.319531	 loss=3.099256	 lr=0.085325
Epoch[074] Batch [2049]/[3760]	Speed: 117.453181 samples/sec	 accuracy=32.316311	 loss=3.099337	 lr=0.085315
Epoch[074] Batch [2099]/[3760]	Speed: 117.100536 samples/sec	 accuracy=32.298363	 loss=3.099019	 lr=0.085306
Epoch[074] Batch [2149]/[3760]	Speed: 118.391274 samples/sec	 accuracy=32.292151	 loss=3.101166	 lr=0.085297
Epoch[074] Batch [2199]/[3760]	Speed: 117.952260 samples/sec	 accuracy=32.317472	 loss=3.100319	 lr=0.085288
Epoch[074] Batch [2249]/[3760]	Speed: 117.584174 samples/sec	 accuracy=32.302778	 loss=3.100812	 lr=0.085279
Epoch[074] Batch [2299]/[3760]	Speed: 118.112476 samples/sec	 accuracy=32.311821	 loss=3.100271	 lr=0.085270
Epoch[074] Batch [2349]/[3760]	Speed: 117.369136 samples/sec	 accuracy=32.298537	 loss=3.099547	 lr=0.085261
Epoch[074] Batch [2399]/[3760]	Speed: 118.405154 samples/sec	 accuracy=32.293620	 loss=3.099376	 lr=0.085251
Epoch[074] Batch [2449]/[3760]	Speed: 117.297984 samples/sec	 accuracy=32.260204	 loss=3.100091	 lr=0.085242
Epoch[074] Batch [2499]/[3760]	Speed: 117.905143 samples/sec	 accuracy=32.273750	 loss=3.099746	 lr=0.085233
Epoch[074] Batch [2549]/[3760]	Speed: 118.062204 samples/sec	 accuracy=32.286152	 loss=3.099051	 lr=0.085224
Epoch[074] Batch [2599]/[3760]	Speed: 117.975955 samples/sec	 accuracy=32.260216	 loss=3.099455	 lr=0.085215
Epoch[074] Batch [2649]/[3760]	Speed: 116.602390 samples/sec	 accuracy=32.264741	 loss=3.100363	 lr=0.085206
Epoch[074] Batch [2699]/[3760]	Speed: 117.398539 samples/sec	 accuracy=32.251157	 loss=3.100514	 lr=0.085197
Epoch[074] Batch [2749]/[3760]	Speed: 117.329590 samples/sec	 accuracy=32.277273	 loss=3.099342	 lr=0.085187
Epoch[074] Batch [2799]/[3760]	Speed: 117.557340 samples/sec	 accuracy=32.258929	 loss=3.099653	 lr=0.085178
Epoch[074] Batch [2849]/[3760]	Speed: 117.024826 samples/sec	 accuracy=32.241228	 loss=3.099797	 lr=0.085169
Epoch[074] Batch [2899]/[3760]	Speed: 116.854713 samples/sec	 accuracy=32.231681	 loss=3.101074	 lr=0.085160
Epoch[074] Batch [2949]/[3760]	Speed: 117.368063 samples/sec	 accuracy=32.236758	 loss=3.101370	 lr=0.085151
Epoch[074] Batch [2999]/[3760]	Speed: 116.360880 samples/sec	 accuracy=32.244792	 loss=3.101318	 lr=0.085142
Epoch[074] Batch [3049]/[3760]	Speed: 117.820148 samples/sec	 accuracy=32.235656	 loss=3.102077	 lr=0.085132
Epoch[074] Batch [3099]/[3760]	Speed: 117.106881 samples/sec	 accuracy=32.225806	 loss=3.103143	 lr=0.085123
Epoch[074] Batch [3149]/[3760]	Speed: 118.447248 samples/sec	 accuracy=32.209821	 loss=3.104006	 lr=0.085114
Epoch[074] Batch [3199]/[3760]	Speed: 117.497097 samples/sec	 accuracy=32.212402	 loss=3.104148	 lr=0.085105
Epoch[074] Batch [3249]/[3760]	Speed: 117.779401 samples/sec	 accuracy=32.199038	 loss=3.105235	 lr=0.085096
Epoch[074] Batch [3299]/[3760]	Speed: 117.025733 samples/sec	 accuracy=32.215909	 loss=3.104396	 lr=0.085086
Epoch[074] Batch [3349]/[3760]	Speed: 117.545461 samples/sec	 accuracy=32.218284	 loss=3.104815	 lr=0.085077
Epoch[074] Batch [3399]/[3760]	Speed: 117.050183 samples/sec	 accuracy=32.209559	 loss=3.106112	 lr=0.085068
Epoch[074] Batch [3449]/[3760]	Speed: 116.979902 samples/sec	 accuracy=32.223732	 loss=3.105394	 lr=0.085059
Epoch[074] Batch [3499]/[3760]	Speed: 117.566545 samples/sec	 accuracy=32.228125	 loss=3.105577	 lr=0.085050
Epoch[074] Batch [3549]/[3760]	Speed: 116.747605 samples/sec	 accuracy=32.220070	 loss=3.106486	 lr=0.085040
Epoch[074] Batch [3599]/[3760]	Speed: 117.794474 samples/sec	 accuracy=32.220486	 loss=3.106704	 lr=0.085031
Epoch[074] Batch [3649]/[3760]	Speed: 117.905039 samples/sec	 accuracy=32.235017	 loss=3.105563	 lr=0.085022
Epoch[074] Batch [3699]/[3760]	Speed: 118.041798 samples/sec	 accuracy=32.222128	 loss=3.106191	 lr=0.085013
Epoch[074] Batch [3749]/[3760]	Speed: 120.642004 samples/sec	 accuracy=32.223750	 loss=3.106035	 lr=0.085004
Batch [0049]/[0303]: acc-top1=32.781250 acc-top5=58.593750
Batch [0099]/[0303]: acc-top1=29.921875 acc-top5=55.859375
Batch [0149]/[0303]: acc-top1=31.635417 acc-top5=57.479167
Batch [0199]/[0303]: acc-top1=34.039062 acc-top5=59.757812
Batch [0249]/[0303]: acc-top1=33.650000 acc-top5=59.900000
Batch [0299]/[0303]: acc-top1=32.932292 acc-top5=59.703125
[Epoch 074] training: accuracy=32.222407	 loss=3.106022
[Epoch 074] speed: 117 samples/sec	time cost: 2175.100307
[Epoch 074] validation: acc-top1=32.900165 acc-top5=59.653465 loss=3.167469
Epoch[075] Batch [0049]/[3759]	Speed: 88.222246 samples/sec	 accuracy=32.000000	 loss=3.127266	 lr=0.084993
Epoch[075] Batch [0099]/[3759]	Speed: 117.412663 samples/sec	 accuracy=31.500000	 loss=3.120029	 lr=0.084983
Epoch[075] Batch [0149]/[3759]	Speed: 117.367342 samples/sec	 accuracy=31.937500	 loss=3.095644	 lr=0.084974
Epoch[075] Batch [0199]/[3759]	Speed: 117.669561 samples/sec	 accuracy=31.976562	 loss=3.095635	 lr=0.084965
Epoch[075] Batch [0249]/[3759]	Speed: 117.591742 samples/sec	 accuracy=32.031250	 loss=3.091927	 lr=0.084956
Epoch[075] Batch [0299]/[3759]	Speed: 116.677350 samples/sec	 accuracy=32.104167	 loss=3.096894	 lr=0.084946
Epoch[075] Batch [0349]/[3759]	Speed: 117.302159 samples/sec	 accuracy=32.120536	 loss=3.093223	 lr=0.084937
Epoch[075] Batch [0399]/[3759]	Speed: 118.521221 samples/sec	 accuracy=32.109375	 loss=3.095661	 lr=0.084928
Epoch[075] Batch [0449]/[3759]	Speed: 116.968233 samples/sec	 accuracy=32.083333	 loss=3.096532	 lr=0.084919
Epoch[075] Batch [0499]/[3759]	Speed: 117.772651 samples/sec	 accuracy=32.193750	 loss=3.091749	 lr=0.084910
Epoch[075] Batch [0549]/[3759]	Speed: 117.654954 samples/sec	 accuracy=32.230114	 loss=3.090228	 lr=0.084900
Epoch[075] Batch [0599]/[3759]	Speed: 117.769657 samples/sec	 accuracy=32.309896	 loss=3.089429	 lr=0.084891
Epoch[075] Batch [0649]/[3759]	Speed: 117.555523 samples/sec	 accuracy=32.362981	 loss=3.087121	 lr=0.084882
Epoch[075] Batch [0699]/[3759]	Speed: 117.087339 samples/sec	 accuracy=32.383929	 loss=3.087145	 lr=0.084873
Epoch[075] Batch [0749]/[3759]	Speed: 117.855005 samples/sec	 accuracy=32.250000	 loss=3.091896	 lr=0.084863
Epoch[075] Batch [0799]/[3759]	Speed: 117.805237 samples/sec	 accuracy=32.197266	 loss=3.094200	 lr=0.084854
Epoch[075] Batch [0849]/[3759]	Speed: 117.345499 samples/sec	 accuracy=32.231618	 loss=3.094373	 lr=0.084845
Epoch[075] Batch [0899]/[3759]	Speed: 117.540258 samples/sec	 accuracy=32.206597	 loss=3.094584	 lr=0.084836
Epoch[075] Batch [0949]/[3759]	Speed: 118.022669 samples/sec	 accuracy=32.226974	 loss=3.090925	 lr=0.084826
Epoch[075] Batch [0999]/[3759]	Speed: 117.208409 samples/sec	 accuracy=32.262500	 loss=3.090365	 lr=0.084817
Epoch[075] Batch [1049]/[3759]	Speed: 117.649713 samples/sec	 accuracy=32.254464	 loss=3.091004	 lr=0.084808
Epoch[075] Batch [1099]/[3759]	Speed: 117.598130 samples/sec	 accuracy=32.228693	 loss=3.092357	 lr=0.084799
Epoch[075] Batch [1149]/[3759]	Speed: 118.134354 samples/sec	 accuracy=32.180707	 loss=3.096037	 lr=0.084789
Epoch[075] Batch [1199]/[3759]	Speed: 118.151271 samples/sec	 accuracy=32.187500	 loss=3.094016	 lr=0.084780
Epoch[075] Batch [1249]/[3759]	Speed: 117.377974 samples/sec	 accuracy=32.170000	 loss=3.094451	 lr=0.084771
Epoch[075] Batch [1299]/[3759]	Speed: 117.602008 samples/sec	 accuracy=32.201923	 loss=3.095686	 lr=0.084762
Epoch[075] Batch [1349]/[3759]	Speed: 117.189108 samples/sec	 accuracy=32.193287	 loss=3.095230	 lr=0.084752
Epoch[075] Batch [1399]/[3759]	Speed: 117.364273 samples/sec	 accuracy=32.203125	 loss=3.094833	 lr=0.084743
Epoch[075] Batch [1449]/[3759]	Speed: 116.894664 samples/sec	 accuracy=32.181034	 loss=3.096508	 lr=0.084734
Epoch[075] Batch [1499]/[3759]	Speed: 117.094021 samples/sec	 accuracy=32.208333	 loss=3.095580	 lr=0.084724
Epoch[075] Batch [1549]/[3759]	Speed: 117.115851 samples/sec	 accuracy=32.208669	 loss=3.095673	 lr=0.084715
Epoch[075] Batch [1599]/[3759]	Speed: 116.910842 samples/sec	 accuracy=32.209961	 loss=3.097142	 lr=0.084706
Epoch[075] Batch [1649]/[3759]	Speed: 116.839321 samples/sec	 accuracy=32.207386	 loss=3.097435	 lr=0.084697
Epoch[075] Batch [1699]/[3759]	Speed: 116.710283 samples/sec	 accuracy=32.246324	 loss=3.096286	 lr=0.084687
Epoch[075] Batch [1749]/[3759]	Speed: 117.919391 samples/sec	 accuracy=32.250893	 loss=3.095934	 lr=0.084678
Epoch[075] Batch [1799]/[3759]	Speed: 117.571714 samples/sec	 accuracy=32.234375	 loss=3.096570	 lr=0.084669
Epoch[075] Batch [1849]/[3759]	Speed: 117.673056 samples/sec	 accuracy=32.205236	 loss=3.097550	 lr=0.084659
Epoch[075] Batch [1899]/[3759]	Speed: 117.199711 samples/sec	 accuracy=32.171053	 loss=3.099893	 lr=0.084650
Epoch[075] Batch [1949]/[3759]	Speed: 118.034290 samples/sec	 accuracy=32.201923	 loss=3.098155	 lr=0.084641
Epoch[075] Batch [1999]/[3759]	Speed: 117.490614 samples/sec	 accuracy=32.240625	 loss=3.096742	 lr=0.084632
Epoch[075] Batch [2049]/[3759]	Speed: 117.179487 samples/sec	 accuracy=32.262957	 loss=3.096259	 lr=0.084622
Epoch[075] Batch [2099]/[3759]	Speed: 117.652158 samples/sec	 accuracy=32.239583	 loss=3.097215	 lr=0.084613
Epoch[075] Batch [2149]/[3759]	Speed: 117.618773 samples/sec	 accuracy=32.253634	 loss=3.096781	 lr=0.084604
Epoch[075] Batch [2199]/[3759]	Speed: 117.060122 samples/sec	 accuracy=32.232955	 loss=3.097104	 lr=0.084594
Epoch[075] Batch [2249]/[3759]	Speed: 117.596226 samples/sec	 accuracy=32.220139	 loss=3.097713	 lr=0.084585
Epoch[075] Batch [2299]/[3759]	Speed: 117.524198 samples/sec	 accuracy=32.224185	 loss=3.098236	 lr=0.084576
Epoch[075] Batch [2349]/[3759]	Speed: 117.348425 samples/sec	 accuracy=32.255319	 loss=3.097433	 lr=0.084566
Epoch[075] Batch [2399]/[3759]	Speed: 116.813660 samples/sec	 accuracy=32.255208	 loss=3.097540	 lr=0.084557
Epoch[075] Batch [2449]/[3759]	Speed: 117.750846 samples/sec	 accuracy=32.200255	 loss=3.099417	 lr=0.084548
Epoch[075] Batch [2499]/[3759]	Speed: 117.532343 samples/sec	 accuracy=32.226250	 loss=3.098638	 lr=0.084538
Epoch[075] Batch [2549]/[3759]	Speed: 117.666325 samples/sec	 accuracy=32.232230	 loss=3.099634	 lr=0.084529
Epoch[075] Batch [2599]/[3759]	Speed: 117.990167 samples/sec	 accuracy=32.246394	 loss=3.098546	 lr=0.084520
Epoch[075] Batch [2649]/[3759]	Speed: 116.602390 samples/sec	 accuracy=32.229953	 loss=3.099647	 lr=0.084510
Epoch[075] Batch [2699]/[3759]	Speed: 117.415588 samples/sec	 accuracy=32.239583	 loss=3.099623	 lr=0.084501
Epoch[075] Batch [2749]/[3759]	Speed: 117.394604 samples/sec	 accuracy=32.218750	 loss=3.100044	 lr=0.084492
Epoch[075] Batch [2799]/[3759]	Speed: 117.169496 samples/sec	 accuracy=32.221540	 loss=3.100378	 lr=0.084482
Epoch[075] Batch [2849]/[3759]	Speed: 116.705898 samples/sec	 accuracy=32.256031	 loss=3.099520	 lr=0.084473
Epoch[075] Batch [2899]/[3759]	Speed: 118.012402 samples/sec	 accuracy=32.260237	 loss=3.099085	 lr=0.084464
Epoch[075] Batch [2949]/[3759]	Speed: 117.887251 samples/sec	 accuracy=32.248411	 loss=3.099809	 lr=0.084454
Epoch[075] Batch [2999]/[3759]	Speed: 117.865486 samples/sec	 accuracy=32.246875	 loss=3.099946	 lr=0.084445
Epoch[075] Batch [3049]/[3759]	Speed: 117.146674 samples/sec	 accuracy=32.237193	 loss=3.100682	 lr=0.084436
Epoch[075] Batch [3099]/[3759]	Speed: 116.825207 samples/sec	 accuracy=32.240927	 loss=3.100487	 lr=0.084426
Epoch[075] Batch [3149]/[3759]	Speed: 117.637183 samples/sec	 accuracy=32.224206	 loss=3.101402	 lr=0.084417
Epoch[075] Batch [3199]/[3759]	Speed: 117.293346 samples/sec	 accuracy=32.223145	 loss=3.101600	 lr=0.084408
Epoch[075] Batch [3249]/[3759]	Speed: 117.362172 samples/sec	 accuracy=32.211058	 loss=3.102202	 lr=0.084398
Epoch[075] Batch [3299]/[3759]	Speed: 117.879166 samples/sec	 accuracy=32.193655	 loss=3.102826	 lr=0.084389
Epoch[075] Batch [3349]/[3759]	Speed: 117.037075 samples/sec	 accuracy=32.173974	 loss=3.103983	 lr=0.084380
Epoch[075] Batch [3399]/[3759]	Speed: 117.680626 samples/sec	 accuracy=32.167279	 loss=3.103864	 lr=0.084370
Epoch[075] Batch [3449]/[3759]	Speed: 117.577589 samples/sec	 accuracy=32.148098	 loss=3.104530	 lr=0.084361
Epoch[075] Batch [3499]/[3759]	Speed: 117.277605 samples/sec	 accuracy=32.151786	 loss=3.105066	 lr=0.084351
Epoch[075] Batch [3549]/[3759]	Speed: 117.825978 samples/sec	 accuracy=32.149208	 loss=3.105255	 lr=0.084342
Epoch[075] Batch [3599]/[3759]	Speed: 117.860538 samples/sec	 accuracy=32.127170	 loss=3.105490	 lr=0.084333
Epoch[075] Batch [3649]/[3759]	Speed: 117.492591 samples/sec	 accuracy=32.141695	 loss=3.104683	 lr=0.084323
Epoch[075] Batch [3699]/[3759]	Speed: 117.311439 samples/sec	 accuracy=32.129223	 loss=3.105102	 lr=0.084314
Epoch[075] Batch [3749]/[3759]	Speed: 120.450840 samples/sec	 accuracy=32.122917	 loss=3.105792	 lr=0.084305
Batch [0049]/[0303]: acc-top1=32.312500 acc-top5=58.375000
Batch [0099]/[0303]: acc-top1=29.281250 acc-top5=55.453125
Batch [0149]/[0303]: acc-top1=31.250000 acc-top5=58.114583
Batch [0199]/[0303]: acc-top1=33.335938 acc-top5=60.351562
Batch [0249]/[0303]: acc-top1=32.412500 acc-top5=60.043750
Batch [0299]/[0303]: acc-top1=32.041667 acc-top5=59.979167
[Epoch 075] training: accuracy=32.125399	 loss=3.105561
[Epoch 075] speed: 116 samples/sec	time cost: 2176.203942
[Epoch 075] validation: acc-top1=32.131807 acc-top5=60.009282 loss=3.165461
Epoch[076] Batch [0049]/[3760]	Speed: 87.578758 samples/sec	 accuracy=35.593750	 loss=2.950844	 lr=0.084293
Epoch[076] Batch [0099]/[3760]	Speed: 117.284901 samples/sec	 accuracy=34.546875	 loss=2.981633	 lr=0.084284
Epoch[076] Batch [0149]/[3760]	Speed: 117.829993 samples/sec	 accuracy=34.156250	 loss=3.007000	 lr=0.084275
Epoch[076] Batch [0199]/[3760]	Speed: 117.388337 samples/sec	 accuracy=33.273438	 loss=3.027141	 lr=0.084265
Epoch[076] Batch [0249]/[3760]	Speed: 118.321123 samples/sec	 accuracy=33.337500	 loss=3.025984	 lr=0.084256
Epoch[076] Batch [0299]/[3760]	Speed: 117.365319 samples/sec	 accuracy=33.166667	 loss=3.042506	 lr=0.084246
Epoch[076] Batch [0349]/[3760]	Speed: 118.082352 samples/sec	 accuracy=33.107143	 loss=3.046004	 lr=0.084237
Epoch[076] Batch [0399]/[3760]	Speed: 117.129926 samples/sec	 accuracy=32.992188	 loss=3.051390	 lr=0.084228
Epoch[076] Batch [0449]/[3760]	Speed: 117.678255 samples/sec	 accuracy=32.920139	 loss=3.051202	 lr=0.084218
Epoch[076] Batch [0499]/[3760]	Speed: 117.403761 samples/sec	 accuracy=32.956250	 loss=3.055085	 lr=0.084209
Epoch[076] Batch [0549]/[3760]	Speed: 117.641065 samples/sec	 accuracy=33.039773	 loss=3.053049	 lr=0.084199
Epoch[076] Batch [0599]/[3760]	Speed: 118.052288 samples/sec	 accuracy=33.007812	 loss=3.056475	 lr=0.084190
Epoch[076] Batch [0649]/[3760]	Speed: 117.229099 samples/sec	 accuracy=32.966346	 loss=3.060779	 lr=0.084181
Epoch[076] Batch [0699]/[3760]	Speed: 117.411139 samples/sec	 accuracy=32.964286	 loss=3.063088	 lr=0.084171
Epoch[076] Batch [0749]/[3760]	Speed: 116.910342 samples/sec	 accuracy=32.987500	 loss=3.064710	 lr=0.084162
Epoch[076] Batch [0799]/[3760]	Speed: 117.660344 samples/sec	 accuracy=32.996094	 loss=3.060637	 lr=0.084152
Epoch[076] Batch [0849]/[3760]	Speed: 117.841650 samples/sec	 accuracy=32.948529	 loss=3.063652	 lr=0.084143
Epoch[076] Batch [0899]/[3760]	Speed: 118.169399 samples/sec	 accuracy=32.885417	 loss=3.063310	 lr=0.084134
Epoch[076] Batch [0949]/[3760]	Speed: 117.428569 samples/sec	 accuracy=32.952303	 loss=3.060220	 lr=0.084124
Epoch[076] Batch [0999]/[3760]	Speed: 118.895508 samples/sec	 accuracy=32.895313	 loss=3.064360	 lr=0.084115
Epoch[076] Batch [1049]/[3760]	Speed: 117.302028 samples/sec	 accuracy=32.866071	 loss=3.067688	 lr=0.084105
Epoch[076] Batch [1099]/[3760]	Speed: 118.099989 samples/sec	 accuracy=32.805398	 loss=3.070706	 lr=0.084096
Epoch[076] Batch [1149]/[3760]	Speed: 117.285201 samples/sec	 accuracy=32.739130	 loss=3.072973	 lr=0.084086
Epoch[076] Batch [1199]/[3760]	Speed: 117.242570 samples/sec	 accuracy=32.756510	 loss=3.073071	 lr=0.084077
Epoch[076] Batch [1249]/[3760]	Speed: 117.837727 samples/sec	 accuracy=32.726250	 loss=3.074155	 lr=0.084068
Epoch[076] Batch [1299]/[3760]	Speed: 117.616915 samples/sec	 accuracy=32.683894	 loss=3.075436	 lr=0.084058
Epoch[076] Batch [1349]/[3760]	Speed: 116.974294 samples/sec	 accuracy=32.699074	 loss=3.076187	 lr=0.084049
Epoch[076] Batch [1399]/[3760]	Speed: 117.107063 samples/sec	 accuracy=32.690848	 loss=3.076322	 lr=0.084039
Epoch[076] Batch [1449]/[3760]	Speed: 118.038502 samples/sec	 accuracy=32.695043	 loss=3.077396	 lr=0.084030
Epoch[076] Batch [1499]/[3760]	Speed: 117.984728 samples/sec	 accuracy=32.694792	 loss=3.079117	 lr=0.084020
Epoch[076] Batch [1549]/[3760]	Speed: 117.685690 samples/sec	 accuracy=32.647177	 loss=3.079898	 lr=0.084011
Epoch[076] Batch [1599]/[3760]	Speed: 118.881422 samples/sec	 accuracy=32.666016	 loss=3.081315	 lr=0.084001
Epoch[076] Batch [1649]/[3760]	Speed: 117.662791 samples/sec	 accuracy=32.661932	 loss=3.081201	 lr=0.083992
Epoch[076] Batch [1699]/[3760]	Speed: 116.874892 samples/sec	 accuracy=32.696691	 loss=3.080573	 lr=0.083982
Epoch[076] Batch [1749]/[3760]	Speed: 117.838758 samples/sec	 accuracy=32.673214	 loss=3.080791	 lr=0.083973
Epoch[076] Batch [1799]/[3760]	Speed: 117.362489 samples/sec	 accuracy=32.698785	 loss=3.080076	 lr=0.083964
Epoch[076] Batch [1849]/[3760]	Speed: 117.565690 samples/sec	 accuracy=32.692568	 loss=3.080061	 lr=0.083954
Epoch[076] Batch [1899]/[3760]	Speed: 118.294262 samples/sec	 accuracy=32.670230	 loss=3.080999	 lr=0.083945
Epoch[076] Batch [1949]/[3760]	Speed: 117.985982 samples/sec	 accuracy=32.637019	 loss=3.081912	 lr=0.083935
Epoch[076] Batch [1999]/[3760]	Speed: 117.957272 samples/sec	 accuracy=32.653906	 loss=3.081194	 lr=0.083926
Epoch[076] Batch [2049]/[3760]	Speed: 116.552834 samples/sec	 accuracy=32.639482	 loss=3.082387	 lr=0.083916
Epoch[076] Batch [2099]/[3760]	Speed: 116.991041 samples/sec	 accuracy=32.649554	 loss=3.080877	 lr=0.083907
Epoch[076] Batch [2149]/[3760]	Speed: 117.499155 samples/sec	 accuracy=32.633721	 loss=3.082345	 lr=0.083897
Epoch[076] Batch [2199]/[3760]	Speed: 117.570561 samples/sec	 accuracy=32.610085	 loss=3.083084	 lr=0.083888
Epoch[076] Batch [2249]/[3760]	Speed: 117.584213 samples/sec	 accuracy=32.611111	 loss=3.083317	 lr=0.083878
Epoch[076] Batch [2299]/[3760]	Speed: 118.088298 samples/sec	 accuracy=32.625679	 loss=3.083049	 lr=0.083869
Epoch[076] Batch [2349]/[3760]	Speed: 117.619535 samples/sec	 accuracy=32.613697	 loss=3.084804	 lr=0.083859
Epoch[076] Batch [2399]/[3760]	Speed: 117.061269 samples/sec	 accuracy=32.580729	 loss=3.085696	 lr=0.083850
Epoch[076] Batch [2449]/[3760]	Speed: 117.112966 samples/sec	 accuracy=32.579082	 loss=3.086532	 lr=0.083840
Epoch[076] Batch [2499]/[3760]	Speed: 117.378324 samples/sec	 accuracy=32.585625	 loss=3.086069	 lr=0.083831
Epoch[076] Batch [2549]/[3760]	Speed: 117.201172 samples/sec	 accuracy=32.599265	 loss=3.085468	 lr=0.083821
Epoch[076] Batch [2599]/[3760]	Speed: 117.835881 samples/sec	 accuracy=32.567308	 loss=3.087134	 lr=0.083812
Epoch[076] Batch [2649]/[3760]	Speed: 117.319247 samples/sec	 accuracy=32.566627	 loss=3.088402	 lr=0.083802
Epoch[076] Batch [2699]/[3760]	Speed: 117.926536 samples/sec	 accuracy=32.549769	 loss=3.088518	 lr=0.083793
Epoch[076] Batch [2749]/[3760]	Speed: 118.088808 samples/sec	 accuracy=32.536932	 loss=3.089156	 lr=0.083783
Epoch[076] Batch [2799]/[3760]	Speed: 116.632981 samples/sec	 accuracy=32.537946	 loss=3.089027	 lr=0.083774
Epoch[076] Batch [2849]/[3760]	Speed: 118.143474 samples/sec	 accuracy=32.532346	 loss=3.088755	 lr=0.083764
Epoch[076] Batch [2899]/[3760]	Speed: 117.810583 samples/sec	 accuracy=32.518319	 loss=3.088987	 lr=0.083755
Epoch[076] Batch [2949]/[3760]	Speed: 117.876329 samples/sec	 accuracy=32.501059	 loss=3.089598	 lr=0.083745
Epoch[076] Batch [2999]/[3760]	Speed: 117.441309 samples/sec	 accuracy=32.504688	 loss=3.089600	 lr=0.083736
Epoch[076] Batch [3049]/[3760]	Speed: 117.365510 samples/sec	 accuracy=32.514344	 loss=3.089768	 lr=0.083726
Epoch[076] Batch [3099]/[3760]	Speed: 118.164157 samples/sec	 accuracy=32.496472	 loss=3.090525	 lr=0.083717
Epoch[076] Batch [3149]/[3760]	Speed: 117.688891 samples/sec	 accuracy=32.478175	 loss=3.091205	 lr=0.083707
Epoch[076] Batch [3199]/[3760]	Speed: 117.829803 samples/sec	 accuracy=32.493652	 loss=3.090519	 lr=0.083698
Epoch[076] Batch [3249]/[3760]	Speed: 117.473031 samples/sec	 accuracy=32.469712	 loss=3.092093	 lr=0.083688
Epoch[076] Batch [3299]/[3760]	Speed: 117.931160 samples/sec	 accuracy=32.464015	 loss=3.092021	 lr=0.083679
Epoch[076] Batch [3349]/[3760]	Speed: 117.088893 samples/sec	 accuracy=32.464552	 loss=3.092415	 lr=0.083669
Epoch[076] Batch [3399]/[3760]	Speed: 118.265565 samples/sec	 accuracy=32.461857	 loss=3.092826	 lr=0.083660
Epoch[076] Batch [3449]/[3760]	Speed: 117.770481 samples/sec	 accuracy=32.469656	 loss=3.091856	 lr=0.083650
Epoch[076] Batch [3499]/[3760]	Speed: 117.468637 samples/sec	 accuracy=32.484375	 loss=3.091572	 lr=0.083640
Epoch[076] Batch [3549]/[3760]	Speed: 117.397429 samples/sec	 accuracy=32.490317	 loss=3.091115	 lr=0.083631
Epoch[076] Batch [3599]/[3760]	Speed: 118.134141 samples/sec	 accuracy=32.459635	 loss=3.092249	 lr=0.083621
Epoch[076] Batch [3649]/[3760]	Speed: 118.256542 samples/sec	 accuracy=32.468750	 loss=3.092468	 lr=0.083612
Epoch[076] Batch [3699]/[3760]	Speed: 117.908422 samples/sec	 accuracy=32.465794	 loss=3.092367	 lr=0.083602
Epoch[076] Batch [3749]/[3760]	Speed: 120.779864 samples/sec	 accuracy=32.465000	 loss=3.092559	 lr=0.083593
Batch [0049]/[0303]: acc-top1=33.406250 acc-top5=57.562500
Batch [0099]/[0303]: acc-top1=31.484375 acc-top5=56.265625
Batch [0149]/[0303]: acc-top1=33.208333 acc-top5=58.166667
Batch [0199]/[0303]: acc-top1=35.289062 acc-top5=60.421875
Batch [0249]/[0303]: acc-top1=34.450000 acc-top5=60.531250
Batch [0299]/[0303]: acc-top1=33.833333 acc-top5=60.447917
[Epoch 076] training: accuracy=32.470911	 loss=3.092402
[Epoch 076] speed: 117 samples/sec	time cost: 2174.380882
[Epoch 076] validation: acc-top1=33.797442 acc-top5=60.411510 loss=3.141108
Epoch[077] Batch [0049]/[3760]	Speed: 88.926310 samples/sec	 accuracy=31.500000	 loss=3.068134	 lr=0.083581
Epoch[077] Batch [0099]/[3760]	Speed: 117.731873 samples/sec	 accuracy=32.203125	 loss=3.070275	 lr=0.083572
Epoch[077] Batch [0149]/[3760]	Speed: 117.574584 samples/sec	 accuracy=32.395833	 loss=3.070129	 lr=0.083562
Epoch[077] Batch [0199]/[3760]	Speed: 117.463630 samples/sec	 accuracy=32.468750	 loss=3.070263	 lr=0.083553
Epoch[077] Batch [0249]/[3760]	Speed: 116.917608 samples/sec	 accuracy=32.468750	 loss=3.057297	 lr=0.083543
Epoch[077] Batch [0299]/[3760]	Speed: 117.171920 samples/sec	 accuracy=32.244792	 loss=3.063439	 lr=0.083533
Epoch[077] Batch [0349]/[3760]	Speed: 117.546809 samples/sec	 accuracy=32.120536	 loss=3.069738	 lr=0.083524
Epoch[077] Batch [0399]/[3760]	Speed: 116.839715 samples/sec	 accuracy=32.414062	 loss=3.064081	 lr=0.083514
Epoch[077] Batch [0449]/[3760]	Speed: 116.961209 samples/sec	 accuracy=32.479167	 loss=3.063391	 lr=0.083505
Epoch[077] Batch [0499]/[3760]	Speed: 117.619454 samples/sec	 accuracy=32.440625	 loss=3.066144	 lr=0.083495
Epoch[077] Batch [0549]/[3760]	Speed: 116.980078 samples/sec	 accuracy=32.505682	 loss=3.067828	 lr=0.083486
Epoch[077] Batch [0599]/[3760]	Speed: 117.753021 samples/sec	 accuracy=32.583333	 loss=3.068543	 lr=0.083476
Epoch[077] Batch [0649]/[3760]	Speed: 116.977458 samples/sec	 accuracy=32.586538	 loss=3.066412	 lr=0.083466
Epoch[077] Batch [0699]/[3760]	Speed: 117.879205 samples/sec	 accuracy=32.562500	 loss=3.071621	 lr=0.083457
Epoch[077] Batch [0749]/[3760]	Speed: 117.322951 samples/sec	 accuracy=32.604167	 loss=3.071077	 lr=0.083447
Epoch[077] Batch [0799]/[3760]	Speed: 117.943952 samples/sec	 accuracy=32.587891	 loss=3.072057	 lr=0.083438
Epoch[077] Batch [0849]/[3760]	Speed: 117.155492 samples/sec	 accuracy=32.612132	 loss=3.072121	 lr=0.083428
Epoch[077] Batch [0899]/[3760]	Speed: 117.309215 samples/sec	 accuracy=32.595486	 loss=3.073315	 lr=0.083418
Epoch[077] Batch [0949]/[3760]	Speed: 117.172592 samples/sec	 accuracy=32.672697	 loss=3.071970	 lr=0.083409
Epoch[077] Batch [0999]/[3760]	Speed: 117.403749 samples/sec	 accuracy=32.625000	 loss=3.076899	 lr=0.083399
Epoch[077] Batch [1049]/[3760]	Speed: 117.480820 samples/sec	 accuracy=32.623512	 loss=3.078392	 lr=0.083390
Epoch[077] Batch [1099]/[3760]	Speed: 117.202797 samples/sec	 accuracy=32.603693	 loss=3.078256	 lr=0.083380
Epoch[077] Batch [1149]/[3760]	Speed: 118.059803 samples/sec	 accuracy=32.614130	 loss=3.078120	 lr=0.083370
Epoch[077] Batch [1199]/[3760]	Speed: 116.579872 samples/sec	 accuracy=32.606771	 loss=3.078781	 lr=0.083361
Epoch[077] Batch [1249]/[3760]	Speed: 117.034076 samples/sec	 accuracy=32.653750	 loss=3.076909	 lr=0.083351
Epoch[077] Batch [1299]/[3760]	Speed: 117.625161 samples/sec	 accuracy=32.631010	 loss=3.078183	 lr=0.083342
Epoch[077] Batch [1349]/[3760]	Speed: 117.597456 samples/sec	 accuracy=32.645833	 loss=3.078565	 lr=0.083332
Epoch[077] Batch [1399]/[3760]	Speed: 117.043963 samples/sec	 accuracy=32.639509	 loss=3.079687	 lr=0.083322
Epoch[077] Batch [1449]/[3760]	Speed: 116.956640 samples/sec	 accuracy=32.593750	 loss=3.083140	 lr=0.083313
Epoch[077] Batch [1499]/[3760]	Speed: 117.399120 samples/sec	 accuracy=32.575000	 loss=3.084903	 lr=0.083303
Epoch[077] Batch [1549]/[3760]	Speed: 117.569079 samples/sec	 accuracy=32.601815	 loss=3.083590	 lr=0.083294
Epoch[077] Batch [1599]/[3760]	Speed: 117.469910 samples/sec	 accuracy=32.581055	 loss=3.084364	 lr=0.083284
Epoch[077] Batch [1649]/[3760]	Speed: 117.975905 samples/sec	 accuracy=32.609848	 loss=3.082615	 lr=0.083274
Epoch[077] Batch [1699]/[3760]	Speed: 117.483096 samples/sec	 accuracy=32.574449	 loss=3.084686	 lr=0.083265
Epoch[077] Batch [1749]/[3760]	Speed: 117.319188 samples/sec	 accuracy=32.559821	 loss=3.085448	 lr=0.083255
Epoch[077] Batch [1799]/[3760]	Speed: 116.350703 samples/sec	 accuracy=32.577257	 loss=3.084459	 lr=0.083245
Epoch[077] Batch [1849]/[3760]	Speed: 118.495125 samples/sec	 accuracy=32.586993	 loss=3.084213	 lr=0.083236
Epoch[077] Batch [1899]/[3760]	Speed: 117.015717 samples/sec	 accuracy=32.554276	 loss=3.084968	 lr=0.083226
Epoch[077] Batch [1949]/[3760]	Speed: 117.136265 samples/sec	 accuracy=32.562500	 loss=3.084309	 lr=0.083217
Epoch[077] Batch [1999]/[3760]	Speed: 118.026879 samples/sec	 accuracy=32.567188	 loss=3.084071	 lr=0.083207
Epoch[077] Batch [2049]/[3760]	Speed: 117.303734 samples/sec	 accuracy=32.554878	 loss=3.083589	 lr=0.083197
Epoch[077] Batch [2099]/[3760]	Speed: 117.408965 samples/sec	 accuracy=32.540923	 loss=3.084052	 lr=0.083188
Epoch[077] Batch [2149]/[3760]	Speed: 118.516626 samples/sec	 accuracy=32.529797	 loss=3.084329	 lr=0.083178
Epoch[077] Batch [2199]/[3760]	Speed: 117.356346 samples/sec	 accuracy=32.540483	 loss=3.083993	 lr=0.083168
Epoch[077] Batch [2249]/[3760]	Speed: 117.346714 samples/sec	 accuracy=32.561806	 loss=3.083435	 lr=0.083159
Epoch[077] Batch [2299]/[3760]	Speed: 117.832792 samples/sec	 accuracy=32.557745	 loss=3.083873	 lr=0.083149
Epoch[077] Batch [2349]/[3760]	Speed: 116.333548 samples/sec	 accuracy=32.571144	 loss=3.083487	 lr=0.083139
Epoch[077] Batch [2399]/[3760]	Speed: 117.792497 samples/sec	 accuracy=32.575521	 loss=3.084466	 lr=0.083130
Epoch[077] Batch [2449]/[3760]	Speed: 117.782066 samples/sec	 accuracy=32.607143	 loss=3.082997	 lr=0.083120
Epoch[077] Batch [2499]/[3760]	Speed: 117.539656 samples/sec	 accuracy=32.578750	 loss=3.084582	 lr=0.083110
Epoch[077] Batch [2549]/[3760]	Speed: 117.953298 samples/sec	 accuracy=32.596814	 loss=3.084013	 lr=0.083101
Epoch[077] Batch [2599]/[3760]	Speed: 117.644415 samples/sec	 accuracy=32.591346	 loss=3.085192	 lr=0.083091
Epoch[077] Batch [2649]/[3760]	Speed: 117.071110 samples/sec	 accuracy=32.587854	 loss=3.085888	 lr=0.083081
Epoch[077] Batch [2699]/[3760]	Speed: 117.316208 samples/sec	 accuracy=32.601852	 loss=3.085849	 lr=0.083072
Epoch[077] Batch [2749]/[3760]	Speed: 117.223541 samples/sec	 accuracy=32.588636	 loss=3.086607	 lr=0.083062
Epoch[077] Batch [2799]/[3760]	Speed: 117.137878 samples/sec	 accuracy=32.576451	 loss=3.086838	 lr=0.083052
Epoch[077] Batch [2849]/[3760]	Speed: 118.113634 samples/sec	 accuracy=32.578399	 loss=3.087037	 lr=0.083043
Epoch[077] Batch [2899]/[3760]	Speed: 116.684685 samples/sec	 accuracy=32.551185	 loss=3.087917	 lr=0.083033
Epoch[077] Batch [2949]/[3760]	Speed: 117.852994 samples/sec	 accuracy=32.572034	 loss=3.087905	 lr=0.083023
Epoch[077] Batch [2999]/[3760]	Speed: 117.696216 samples/sec	 accuracy=32.560937	 loss=3.088317	 lr=0.083014
Epoch[077] Batch [3049]/[3760]	Speed: 117.225100 samples/sec	 accuracy=32.572746	 loss=3.087711	 lr=0.083004
Epoch[077] Batch [3099]/[3760]	Speed: 117.868528 samples/sec	 accuracy=32.556452	 loss=3.088422	 lr=0.082994
Epoch[077] Batch [3149]/[3760]	Speed: 117.984105 samples/sec	 accuracy=32.553075	 loss=3.088929	 lr=0.082985
Epoch[077] Batch [3199]/[3760]	Speed: 116.259971 samples/sec	 accuracy=32.550293	 loss=3.090200	 lr=0.082975
Epoch[077] Batch [3249]/[3760]	Speed: 117.483279 samples/sec	 accuracy=32.550000	 loss=3.090149	 lr=0.082965
Epoch[077] Batch [3299]/[3760]	Speed: 117.290955 samples/sec	 accuracy=32.537405	 loss=3.090529	 lr=0.082955
Epoch[077] Batch [3349]/[3760]	Speed: 118.163774 samples/sec	 accuracy=32.524720	 loss=3.091030	 lr=0.082946
Epoch[077] Batch [3399]/[3760]	Speed: 117.740751 samples/sec	 accuracy=32.537684	 loss=3.090458	 lr=0.082936
Epoch[077] Batch [3449]/[3760]	Speed: 117.055607 samples/sec	 accuracy=32.522192	 loss=3.091109	 lr=0.082926
Epoch[077] Batch [3499]/[3760]	Speed: 117.963789 samples/sec	 accuracy=32.524107	 loss=3.091433	 lr=0.082917
Epoch[077] Batch [3549]/[3760]	Speed: 116.656009 samples/sec	 accuracy=32.532130	 loss=3.091435	 lr=0.082907
Epoch[077] Batch [3599]/[3760]	Speed: 117.227290 samples/sec	 accuracy=32.534722	 loss=3.091544	 lr=0.082897
Epoch[077] Batch [3649]/[3760]	Speed: 117.504698 samples/sec	 accuracy=32.535103	 loss=3.091738	 lr=0.082887
Epoch[077] Batch [3699]/[3760]	Speed: 117.466724 samples/sec	 accuracy=32.544341	 loss=3.091020	 lr=0.082878
Epoch[077] Batch [3749]/[3760]	Speed: 120.459900 samples/sec	 accuracy=32.505000	 loss=3.092592	 lr=0.082868
Batch [0049]/[0303]: acc-top1=36.125000 acc-top5=61.343750
Batch [0099]/[0303]: acc-top1=32.000000 acc-top5=58.093750
Batch [0149]/[0303]: acc-top1=32.593750 acc-top5=58.750000
Batch [0199]/[0303]: acc-top1=34.484375 acc-top5=60.265625
Batch [0249]/[0303]: acc-top1=33.868750 acc-top5=60.156250
Batch [0299]/[0303]: acc-top1=33.135417 acc-top5=59.890625
[Epoch 077] training: accuracy=32.492520	 loss=3.092985
[Epoch 077] speed: 116 samples/sec	time cost: 2176.432032
[Epoch 077] validation: acc-top1=33.054868 acc-top5=59.880363 loss=3.111995
Epoch[078] Batch [0049]/[3759]	Speed: 86.396707 samples/sec	 accuracy=33.187500	 loss=3.067136	 lr=0.082856
Epoch[078] Batch [0099]/[3759]	Speed: 116.475167 samples/sec	 accuracy=34.062500	 loss=3.021253	 lr=0.082847
Epoch[078] Batch [0149]/[3759]	Speed: 117.045770 samples/sec	 accuracy=33.822917	 loss=3.053927	 lr=0.082837
Epoch[078] Batch [0199]/[3759]	Speed: 116.857684 samples/sec	 accuracy=33.320312	 loss=3.075083	 lr=0.082827
Epoch[078] Batch [0249]/[3759]	Speed: 116.864493 samples/sec	 accuracy=33.418750	 loss=3.063409	 lr=0.082817
Epoch[078] Batch [0299]/[3759]	Speed: 116.884196 samples/sec	 accuracy=33.255208	 loss=3.061322	 lr=0.082808
Epoch[078] Batch [0349]/[3759]	Speed: 117.522956 samples/sec	 accuracy=33.205357	 loss=3.061952	 lr=0.082798
Epoch[078] Batch [0399]/[3759]	Speed: 117.016766 samples/sec	 accuracy=32.871094	 loss=3.078736	 lr=0.082788
Epoch[078] Batch [0449]/[3759]	Speed: 117.208790 samples/sec	 accuracy=32.934028	 loss=3.077966	 lr=0.082779
Epoch[078] Batch [0499]/[3759]	Speed: 117.418655 samples/sec	 accuracy=33.015625	 loss=3.071390	 lr=0.082769
Epoch[078] Batch [0549]/[3759]	Speed: 117.888812 samples/sec	 accuracy=33.028409	 loss=3.073965	 lr=0.082759
Epoch[078] Batch [0599]/[3759]	Speed: 117.274921 samples/sec	 accuracy=32.963542	 loss=3.075466	 lr=0.082749
Epoch[078] Batch [0649]/[3759]	Speed: 117.865260 samples/sec	 accuracy=32.975962	 loss=3.076028	 lr=0.082740
Epoch[078] Batch [0699]/[3759]	Speed: 117.047949 samples/sec	 accuracy=32.933036	 loss=3.075776	 lr=0.082730
Epoch[078] Batch [0749]/[3759]	Speed: 117.218245 samples/sec	 accuracy=32.883333	 loss=3.077414	 lr=0.082720
Epoch[078] Batch [0799]/[3759]	Speed: 116.387381 samples/sec	 accuracy=32.896484	 loss=3.075559	 lr=0.082710
Epoch[078] Batch [0849]/[3759]	Speed: 117.419407 samples/sec	 accuracy=32.922794	 loss=3.076133	 lr=0.082701
Epoch[078] Batch [0899]/[3759]	Speed: 117.433272 samples/sec	 accuracy=32.949653	 loss=3.074502	 lr=0.082691
Epoch[078] Batch [0949]/[3759]	Speed: 117.655896 samples/sec	 accuracy=32.962171	 loss=3.071810	 lr=0.082681
Epoch[078] Batch [0999]/[3759]	Speed: 117.113369 samples/sec	 accuracy=33.001562	 loss=3.071459	 lr=0.082671
Epoch[078] Batch [1049]/[3759]	Speed: 117.532460 samples/sec	 accuracy=33.017857	 loss=3.069785	 lr=0.082662
Epoch[078] Batch [1099]/[3759]	Speed: 118.073649 samples/sec	 accuracy=33.021307	 loss=3.069158	 lr=0.082652
Epoch[078] Batch [1149]/[3759]	Speed: 117.476785 samples/sec	 accuracy=33.020380	 loss=3.066787	 lr=0.082642
Epoch[078] Batch [1199]/[3759]	Speed: 117.413659 samples/sec	 accuracy=32.984375	 loss=3.068226	 lr=0.082632
Epoch[078] Batch [1249]/[3759]	Speed: 117.056878 samples/sec	 accuracy=32.986250	 loss=3.068267	 lr=0.082622
Epoch[078] Batch [1299]/[3759]	Speed: 117.195095 samples/sec	 accuracy=32.973558	 loss=3.067961	 lr=0.082613
Epoch[078] Batch [1349]/[3759]	Speed: 117.314154 samples/sec	 accuracy=32.938657	 loss=3.070352	 lr=0.082603
Epoch[078] Batch [1399]/[3759]	Speed: 117.438308 samples/sec	 accuracy=32.943080	 loss=3.068245	 lr=0.082593
Epoch[078] Batch [1449]/[3759]	Speed: 117.557835 samples/sec	 accuracy=32.982759	 loss=3.068174	 lr=0.082583
Epoch[078] Batch [1499]/[3759]	Speed: 116.454477 samples/sec	 accuracy=32.991667	 loss=3.066426	 lr=0.082574
Epoch[078] Batch [1549]/[3759]	Speed: 117.247671 samples/sec	 accuracy=32.935484	 loss=3.069153	 lr=0.082564
Epoch[078] Batch [1599]/[3759]	Speed: 117.711848 samples/sec	 accuracy=32.914062	 loss=3.070242	 lr=0.082554
Epoch[078] Batch [1649]/[3759]	Speed: 116.756170 samples/sec	 accuracy=32.914773	 loss=3.070237	 lr=0.082544
Epoch[078] Batch [1699]/[3759]	Speed: 117.960818 samples/sec	 accuracy=32.923713	 loss=3.070464	 lr=0.082534
Epoch[078] Batch [1749]/[3759]	Speed: 117.385572 samples/sec	 accuracy=32.902679	 loss=3.069968	 lr=0.082525
Epoch[078] Batch [1799]/[3759]	Speed: 117.532391 samples/sec	 accuracy=32.874132	 loss=3.071733	 lr=0.082515
Epoch[078] Batch [1849]/[3759]	Speed: 117.797952 samples/sec	 accuracy=32.876689	 loss=3.071453	 lr=0.082505
Epoch[078] Batch [1899]/[3759]	Speed: 116.564200 samples/sec	 accuracy=32.893914	 loss=3.071374	 lr=0.082495
Epoch[078] Batch [1949]/[3759]	Speed: 117.310025 samples/sec	 accuracy=32.877404	 loss=3.072048	 lr=0.082485
Epoch[078] Batch [1999]/[3759]	Speed: 117.727507 samples/sec	 accuracy=32.849219	 loss=3.072113	 lr=0.082476
Epoch[078] Batch [2049]/[3759]	Speed: 117.769777 samples/sec	 accuracy=32.827744	 loss=3.072370	 lr=0.082466
Epoch[078] Batch [2099]/[3759]	Speed: 116.460718 samples/sec	 accuracy=32.833333	 loss=3.072332	 lr=0.082456
Epoch[078] Batch [2149]/[3759]	Speed: 117.271324 samples/sec	 accuracy=32.860465	 loss=3.072018	 lr=0.082446
Epoch[078] Batch [2199]/[3759]	Speed: 117.343298 samples/sec	 accuracy=32.851562	 loss=3.072625	 lr=0.082436
Epoch[078] Batch [2249]/[3759]	Speed: 116.796062 samples/sec	 accuracy=32.843750	 loss=3.072945	 lr=0.082427
Epoch[078] Batch [2299]/[3759]	Speed: 117.622433 samples/sec	 accuracy=32.814538	 loss=3.074498	 lr=0.082417
Epoch[078] Batch [2349]/[3759]	Speed: 117.897333 samples/sec	 accuracy=32.773936	 loss=3.076102	 lr=0.082407
Epoch[078] Batch [2399]/[3759]	Speed: 117.274933 samples/sec	 accuracy=32.772786	 loss=3.076468	 lr=0.082397
Epoch[078] Batch [2449]/[3759]	Speed: 117.383874 samples/sec	 accuracy=32.792092	 loss=3.076050	 lr=0.082387
Epoch[078] Batch [2499]/[3759]	Speed: 117.137401 samples/sec	 accuracy=32.768750	 loss=3.077123	 lr=0.082377
Epoch[078] Batch [2549]/[3759]	Speed: 117.513092 samples/sec	 accuracy=32.757353	 loss=3.077095	 lr=0.082368
Epoch[078] Batch [2599]/[3759]	Speed: 117.058170 samples/sec	 accuracy=32.740385	 loss=3.078195	 lr=0.082358
Epoch[078] Batch [2649]/[3759]	Speed: 117.244223 samples/sec	 accuracy=32.737618	 loss=3.078594	 lr=0.082348
Epoch[078] Batch [2699]/[3759]	Speed: 117.736493 samples/sec	 accuracy=32.740162	 loss=3.077383	 lr=0.082338
Epoch[078] Batch [2749]/[3759]	Speed: 117.152177 samples/sec	 accuracy=32.712500	 loss=3.078118	 lr=0.082328
Epoch[078] Batch [2799]/[3759]	Speed: 117.882975 samples/sec	 accuracy=32.704799	 loss=3.078450	 lr=0.082318
Epoch[078] Batch [2849]/[3759]	Speed: 117.900917 samples/sec	 accuracy=32.733004	 loss=3.077633	 lr=0.082309
Epoch[078] Batch [2899]/[3759]	Speed: 117.354302 samples/sec	 accuracy=32.713901	 loss=3.077474	 lr=0.082299
Epoch[078] Batch [2949]/[3759]	Speed: 116.738115 samples/sec	 accuracy=32.698623	 loss=3.077477	 lr=0.082289
Epoch[078] Batch [2999]/[3759]	Speed: 117.140603 samples/sec	 accuracy=32.700521	 loss=3.077333	 lr=0.082279
Epoch[078] Batch [3049]/[3759]	Speed: 117.456204 samples/sec	 accuracy=32.692111	 loss=3.077897	 lr=0.082269
Epoch[078] Batch [3099]/[3759]	Speed: 118.051069 samples/sec	 accuracy=32.702621	 loss=3.078232	 lr=0.082259
Epoch[078] Batch [3149]/[3759]	Speed: 117.422975 samples/sec	 accuracy=32.708333	 loss=3.078525	 lr=0.082249
Epoch[078] Batch [3199]/[3759]	Speed: 117.349589 samples/sec	 accuracy=32.701660	 loss=3.078753	 lr=0.082240
Epoch[078] Batch [3249]/[3759]	Speed: 117.245061 samples/sec	 accuracy=32.714423	 loss=3.078605	 lr=0.082230
Epoch[078] Batch [3299]/[3759]	Speed: 117.791643 samples/sec	 accuracy=32.725852	 loss=3.077861	 lr=0.082220
Epoch[078] Batch [3349]/[3759]	Speed: 117.578959 samples/sec	 accuracy=32.730410	 loss=3.078066	 lr=0.082210
Epoch[078] Batch [3399]/[3759]	Speed: 118.106024 samples/sec	 accuracy=32.718290	 loss=3.078109	 lr=0.082200
Epoch[078] Batch [3449]/[3759]	Speed: 117.043738 samples/sec	 accuracy=32.702446	 loss=3.078585	 lr=0.082190
Epoch[078] Batch [3499]/[3759]	Speed: 117.752206 samples/sec	 accuracy=32.698661	 loss=3.078543	 lr=0.082180
Epoch[078] Batch [3549]/[3759]	Speed: 117.221331 samples/sec	 accuracy=32.709947	 loss=3.078107	 lr=0.082171
Epoch[078] Batch [3599]/[3759]	Speed: 117.222285 samples/sec	 accuracy=32.720920	 loss=3.077798	 lr=0.082161
Epoch[078] Batch [3649]/[3759]	Speed: 116.221849 samples/sec	 accuracy=32.726027	 loss=3.077236	 lr=0.082151
Epoch[078] Batch [3699]/[3759]	Speed: 117.525503 samples/sec	 accuracy=32.712838	 loss=3.078120	 lr=0.082141
Epoch[078] Batch [3749]/[3759]	Speed: 120.555576 samples/sec	 accuracy=32.700000	 loss=3.078499	 lr=0.082131
Batch [0049]/[0303]: acc-top1=34.375000 acc-top5=60.312500
Batch [0099]/[0303]: acc-top1=30.828125 acc-top5=56.500000
Batch [0149]/[0303]: acc-top1=32.406250 acc-top5=57.770833
Batch [0199]/[0303]: acc-top1=35.273438 acc-top5=60.500000
Batch [0249]/[0303]: acc-top1=34.437500 acc-top5=60.193750
Batch [0299]/[0303]: acc-top1=33.989583 acc-top5=60.348958
[Epoch 078] training: accuracy=32.700685	 loss=3.078437
[Epoch 078] speed: 116 samples/sec	time cost: 2179.376678
[Epoch 078] validation: acc-top1=33.931518 acc-top5=60.292904 loss=3.070862
Epoch[079] Batch [0049]/[3760]	Speed: 86.690677 samples/sec	 accuracy=34.656250	 loss=2.993677	 lr=0.082119
Epoch[079] Batch [0099]/[3760]	Speed: 117.464484 samples/sec	 accuracy=34.125000	 loss=2.996643	 lr=0.082109
Epoch[079] Batch [0149]/[3760]	Speed: 117.231657 samples/sec	 accuracy=34.114583	 loss=3.013979	 lr=0.082100
Epoch[079] Batch [0199]/[3760]	Speed: 116.806590 samples/sec	 accuracy=33.875000	 loss=3.019818	 lr=0.082090
Epoch[079] Batch [0249]/[3760]	Speed: 117.633944 samples/sec	 accuracy=33.393750	 loss=3.040712	 lr=0.082080
Epoch[079] Batch [0299]/[3760]	Speed: 116.491172 samples/sec	 accuracy=33.578125	 loss=3.032349	 lr=0.082070
Epoch[079] Batch [0349]/[3760]	Speed: 117.216367 samples/sec	 accuracy=33.709821	 loss=3.032430	 lr=0.082060
Epoch[079] Batch [0399]/[3760]	Speed: 117.704690 samples/sec	 accuracy=33.738281	 loss=3.032921	 lr=0.082050
Epoch[079] Batch [0449]/[3760]	Speed: 117.540302 samples/sec	 accuracy=33.798611	 loss=3.028617	 lr=0.082040
Epoch[079] Batch [0499]/[3760]	Speed: 116.592725 samples/sec	 accuracy=33.768750	 loss=3.033337	 lr=0.082030
Epoch[079] Batch [0549]/[3760]	Speed: 117.402292 samples/sec	 accuracy=33.678977	 loss=3.037658	 lr=0.082020
Epoch[079] Batch [0599]/[3760]	Speed: 117.333179 samples/sec	 accuracy=33.677083	 loss=3.036456	 lr=0.082011
Epoch[079] Batch [0649]/[3760]	Speed: 117.695765 samples/sec	 accuracy=33.680288	 loss=3.040021	 lr=0.082001
Epoch[079] Batch [0699]/[3760]	Speed: 117.908752 samples/sec	 accuracy=33.654018	 loss=3.039283	 lr=0.081991
Epoch[079] Batch [0749]/[3760]	Speed: 117.694677 samples/sec	 accuracy=33.554167	 loss=3.043052	 lr=0.081981
Epoch[079] Batch [0799]/[3760]	Speed: 117.768603 samples/sec	 accuracy=33.619141	 loss=3.042614	 lr=0.081971
Epoch[079] Batch [0849]/[3760]	Speed: 116.252080 samples/sec	 accuracy=33.568015	 loss=3.046011	 lr=0.081961
Epoch[079] Batch [0899]/[3760]	Speed: 117.742216 samples/sec	 accuracy=33.579861	 loss=3.047256	 lr=0.081951
Epoch[079] Batch [0949]/[3760]	Speed: 117.295813 samples/sec	 accuracy=33.500000	 loss=3.045979	 lr=0.081941
Epoch[079] Batch [0999]/[3760]	Speed: 117.916437 samples/sec	 accuracy=33.421875	 loss=3.048022	 lr=0.081931
Epoch[079] Batch [1049]/[3760]	Speed: 117.256366 samples/sec	 accuracy=33.383929	 loss=3.047856	 lr=0.081921
Epoch[079] Batch [1099]/[3760]	Speed: 117.833703 samples/sec	 accuracy=33.350852	 loss=3.049077	 lr=0.081911
Epoch[079] Batch [1149]/[3760]	Speed: 117.597445 samples/sec	 accuracy=33.316576	 loss=3.051190	 lr=0.081901
Epoch[079] Batch [1199]/[3760]	Speed: 116.964081 samples/sec	 accuracy=33.234375	 loss=3.055003	 lr=0.081891
Epoch[079] Batch [1249]/[3760]	Speed: 117.827635 samples/sec	 accuracy=33.210000	 loss=3.055693	 lr=0.081882
Epoch[079] Batch [1299]/[3760]	Speed: 117.365752 samples/sec	 accuracy=33.207933	 loss=3.054222	 lr=0.081872
Epoch[079] Batch [1349]/[3760]	Speed: 116.728649 samples/sec	 accuracy=33.159722	 loss=3.056364	 lr=0.081862
Epoch[079] Batch [1399]/[3760]	Speed: 116.662203 samples/sec	 accuracy=33.135045	 loss=3.058350	 lr=0.081852
Epoch[079] Batch [1449]/[3760]	Speed: 118.191292 samples/sec	 accuracy=33.145474	 loss=3.056967	 lr=0.081842
Epoch[079] Batch [1499]/[3760]	Speed: 118.312589 samples/sec	 accuracy=33.135417	 loss=3.057597	 lr=0.081832
Epoch[079] Batch [1549]/[3760]	Speed: 117.508135 samples/sec	 accuracy=33.101815	 loss=3.057561	 lr=0.081822
Epoch[079] Batch [1599]/[3760]	Speed: 116.791192 samples/sec	 accuracy=33.087891	 loss=3.056648	 lr=0.081812
Epoch[079] Batch [1649]/[3760]	Speed: 117.282421 samples/sec	 accuracy=33.118371	 loss=3.055882	 lr=0.081802
Epoch[079] Batch [1699]/[3760]	Speed: 117.411845 samples/sec	 accuracy=33.092831	 loss=3.057311	 lr=0.081792
Epoch[079] Batch [1749]/[3760]	Speed: 117.655599 samples/sec	 accuracy=33.066964	 loss=3.058877	 lr=0.081782
Epoch[079] Batch [1799]/[3760]	Speed: 117.278668 samples/sec	 accuracy=33.034722	 loss=3.060372	 lr=0.081772
Epoch[079] Batch [1849]/[3760]	Speed: 117.961076 samples/sec	 accuracy=33.020270	 loss=3.062872	 lr=0.081762
Epoch[079] Batch [1899]/[3760]	Speed: 117.130266 samples/sec	 accuracy=32.981086	 loss=3.065016	 lr=0.081752
Epoch[079] Batch [1949]/[3760]	Speed: 117.059846 samples/sec	 accuracy=32.969551	 loss=3.065960	 lr=0.081742
Epoch[079] Batch [1999]/[3760]	Speed: 118.449047 samples/sec	 accuracy=32.942187	 loss=3.067851	 lr=0.081732
Epoch[079] Batch [2049]/[3760]	Speed: 116.702512 samples/sec	 accuracy=32.903201	 loss=3.068151	 lr=0.081722
Epoch[079] Batch [2099]/[3760]	Speed: 117.074208 samples/sec	 accuracy=32.894345	 loss=3.069152	 lr=0.081712
Epoch[079] Batch [2149]/[3760]	Speed: 116.979058 samples/sec	 accuracy=32.911337	 loss=3.068547	 lr=0.081702
Epoch[079] Batch [2199]/[3760]	Speed: 117.552629 samples/sec	 accuracy=32.933949	 loss=3.068256	 lr=0.081692
Epoch[079] Batch [2249]/[3760]	Speed: 117.484634 samples/sec	 accuracy=32.925000	 loss=3.068885	 lr=0.081682
Epoch[079] Batch [2299]/[3760]	Speed: 117.279850 samples/sec	 accuracy=32.938859	 loss=3.069131	 lr=0.081672
Epoch[079] Batch [2349]/[3760]	Speed: 118.058245 samples/sec	 accuracy=32.913564	 loss=3.071099	 lr=0.081662
Epoch[079] Batch [2399]/[3760]	Speed: 117.025761 samples/sec	 accuracy=32.904948	 loss=3.072235	 lr=0.081652
Epoch[079] Batch [2449]/[3760]	Speed: 117.584805 samples/sec	 accuracy=32.909439	 loss=3.072167	 lr=0.081642
Epoch[079] Batch [2499]/[3760]	Speed: 117.345269 samples/sec	 accuracy=32.853125	 loss=3.074179	 lr=0.081632
Epoch[079] Batch [2549]/[3760]	Speed: 117.297976 samples/sec	 accuracy=32.866422	 loss=3.073250	 lr=0.081622
Epoch[079] Batch [2599]/[3760]	Speed: 117.673319 samples/sec	 accuracy=32.858173	 loss=3.072530	 lr=0.081612
Epoch[079] Batch [2649]/[3760]	Speed: 117.592739 samples/sec	 accuracy=32.831958	 loss=3.074415	 lr=0.081603
Epoch[079] Batch [2699]/[3760]	Speed: 118.099305 samples/sec	 accuracy=32.812500	 loss=3.075235	 lr=0.081593
Epoch[079] Batch [2749]/[3760]	Speed: 117.004768 samples/sec	 accuracy=32.812500	 loss=3.075777	 lr=0.081583
Epoch[079] Batch [2799]/[3760]	Speed: 116.558247 samples/sec	 accuracy=32.818638	 loss=3.075849	 lr=0.081573
Epoch[079] Batch [2849]/[3760]	Speed: 117.336975 samples/sec	 accuracy=32.825110	 loss=3.075307	 lr=0.081563
Epoch[079] Batch [2899]/[3760]	Speed: 117.444825 samples/sec	 accuracy=32.819504	 loss=3.076199	 lr=0.081553
Epoch[079] Batch [2949]/[3760]	Speed: 117.512964 samples/sec	 accuracy=32.825212	 loss=3.075673	 lr=0.081542
Epoch[079] Batch [2999]/[3760]	Speed: 116.798609 samples/sec	 accuracy=32.810938	 loss=3.076353	 lr=0.081532
Epoch[079] Batch [3049]/[3760]	Speed: 117.535458 samples/sec	 accuracy=32.793545	 loss=3.077140	 lr=0.081522
Epoch[079] Batch [3099]/[3760]	Speed: 117.731630 samples/sec	 accuracy=32.784274	 loss=3.077182	 lr=0.081512
Epoch[079] Batch [3149]/[3760]	Speed: 117.836992 samples/sec	 accuracy=32.765873	 loss=3.077689	 lr=0.081502
Epoch[079] Batch [3199]/[3760]	Speed: 116.946657 samples/sec	 accuracy=32.768066	 loss=3.077943	 lr=0.081492
Epoch[079] Batch [3249]/[3760]	Speed: 117.483608 samples/sec	 accuracy=32.761058	 loss=3.077509	 lr=0.081482
Epoch[079] Batch [3299]/[3760]	Speed: 117.681010 samples/sec	 accuracy=32.762311	 loss=3.077513	 lr=0.081472
Epoch[079] Batch [3349]/[3760]	Speed: 117.195106 samples/sec	 accuracy=32.768190	 loss=3.077516	 lr=0.081462
Epoch[079] Batch [3399]/[3760]	Speed: 116.879782 samples/sec	 accuracy=32.783548	 loss=3.077371	 lr=0.081452
Epoch[079] Batch [3449]/[3760]	Speed: 117.611239 samples/sec	 accuracy=32.788496	 loss=3.077194	 lr=0.081442
Epoch[079] Batch [3499]/[3760]	Speed: 117.212084 samples/sec	 accuracy=32.787054	 loss=3.077194	 lr=0.081432
Epoch[079] Batch [3549]/[3760]	Speed: 117.224140 samples/sec	 accuracy=32.760563	 loss=3.077317	 lr=0.081422
Epoch[079] Batch [3599]/[3760]	Speed: 118.277620 samples/sec	 accuracy=32.757378	 loss=3.077384	 lr=0.081412
Epoch[079] Batch [3649]/[3760]	Speed: 117.303413 samples/sec	 accuracy=32.753425	 loss=3.077764	 lr=0.081402
Epoch[079] Batch [3699]/[3760]	Speed: 117.740497 samples/sec	 accuracy=32.747044	 loss=3.077712	 lr=0.081392
Epoch[079] Batch [3749]/[3760]	Speed: 120.477364 samples/sec	 accuracy=32.734167	 loss=3.077741	 lr=0.081382
Batch [0049]/[0303]: acc-top1=34.281250 acc-top5=59.562500
Batch [0099]/[0303]: acc-top1=30.671875 acc-top5=56.500000
Batch [0149]/[0303]: acc-top1=31.875000 acc-top5=57.520833
Batch [0199]/[0303]: acc-top1=34.476562 acc-top5=59.859375
Batch [0249]/[0303]: acc-top1=33.893750 acc-top5=59.887500
Batch [0299]/[0303]: acc-top1=33.708333 acc-top5=59.911458
[Epoch 079] training: accuracy=32.732713	 loss=3.077738
[Epoch 079] speed: 116 samples/sec	time cost: 2177.728354
[Epoch 079] validation: acc-top1=33.616955 acc-top5=59.864893 loss=3.144302
Epoch[080] Batch [0049]/[3760]	Speed: 88.130295 samples/sec	 accuracy=33.468750	 loss=3.032818	 lr=0.081370
Epoch[080] Batch [0099]/[3760]	Speed: 117.976207 samples/sec	 accuracy=34.187500	 loss=2.996118	 lr=0.081360
Epoch[080] Batch [0149]/[3760]	Speed: 118.013964 samples/sec	 accuracy=34.000000	 loss=3.001140	 lr=0.081350
Epoch[080] Batch [0199]/[3760]	Speed: 117.159057 samples/sec	 accuracy=33.656250	 loss=3.021903	 lr=0.081340
Epoch[080] Batch [0249]/[3760]	Speed: 117.937766 samples/sec	 accuracy=33.287500	 loss=3.038829	 lr=0.081330
Epoch[080] Batch [0299]/[3760]	Speed: 117.215171 samples/sec	 accuracy=33.317708	 loss=3.036219	 lr=0.081320
Epoch[080] Batch [0349]/[3760]	Speed: 118.802598 samples/sec	 accuracy=33.308036	 loss=3.040804	 lr=0.081310
Epoch[080] Batch [0399]/[3760]	Speed: 117.544022 samples/sec	 accuracy=33.335938	 loss=3.041637	 lr=0.081300
Epoch[080] Batch [0449]/[3760]	Speed: 117.495894 samples/sec	 accuracy=33.267361	 loss=3.044106	 lr=0.081290
Epoch[080] Batch [0499]/[3760]	Speed: 117.372089 samples/sec	 accuracy=33.046875	 loss=3.050819	 lr=0.081280
Epoch[080] Batch [0549]/[3760]	Speed: 117.468202 samples/sec	 accuracy=33.173295	 loss=3.047933	 lr=0.081270
Epoch[080] Batch [0599]/[3760]	Speed: 118.197615 samples/sec	 accuracy=33.138021	 loss=3.047152	 lr=0.081259
Epoch[080] Batch [0649]/[3760]	Speed: 118.240590 samples/sec	 accuracy=33.144231	 loss=3.045300	 lr=0.081249
Epoch[080] Batch [0699]/[3760]	Speed: 117.324050 samples/sec	 accuracy=32.984375	 loss=3.052642	 lr=0.081239
Epoch[080] Batch [0749]/[3760]	Speed: 117.073562 samples/sec	 accuracy=32.918750	 loss=3.053472	 lr=0.081229
Epoch[080] Batch [0799]/[3760]	Speed: 117.936813 samples/sec	 accuracy=32.957031	 loss=3.053496	 lr=0.081219
Epoch[080] Batch [0849]/[3760]	Speed: 117.753548 samples/sec	 accuracy=32.954044	 loss=3.054433	 lr=0.081209
Epoch[080] Batch [0899]/[3760]	Speed: 117.798156 samples/sec	 accuracy=32.946181	 loss=3.056463	 lr=0.081199
Epoch[080] Batch [0949]/[3760]	Speed: 117.438045 samples/sec	 accuracy=33.013158	 loss=3.054268	 lr=0.081189
Epoch[080] Batch [0999]/[3760]	Speed: 117.499286 samples/sec	 accuracy=32.975000	 loss=3.054537	 lr=0.081179
Epoch[080] Batch [1049]/[3760]	Speed: 117.615675 samples/sec	 accuracy=32.977679	 loss=3.056440	 lr=0.081169
Epoch[080] Batch [1099]/[3760]	Speed: 116.963301 samples/sec	 accuracy=33.026989	 loss=3.057430	 lr=0.081159
Epoch[080] Batch [1149]/[3760]	Speed: 117.733038 samples/sec	 accuracy=33.040761	 loss=3.057320	 lr=0.081149
Epoch[080] Batch [1199]/[3760]	Speed: 116.789236 samples/sec	 accuracy=33.049479	 loss=3.056339	 lr=0.081139
Epoch[080] Batch [1249]/[3760]	Speed: 117.702503 samples/sec	 accuracy=33.002500	 loss=3.058001	 lr=0.081128
Epoch[080] Batch [1299]/[3760]	Speed: 117.415432 samples/sec	 accuracy=32.978365	 loss=3.059411	 lr=0.081118
Epoch[080] Batch [1349]/[3760]	Speed: 117.434635 samples/sec	 accuracy=32.964120	 loss=3.058968	 lr=0.081108
Epoch[080] Batch [1399]/[3760]	Speed: 117.575627 samples/sec	 accuracy=32.982143	 loss=3.058952	 lr=0.081098
Epoch[080] Batch [1449]/[3760]	Speed: 117.702156 samples/sec	 accuracy=32.978448	 loss=3.058146	 lr=0.081088
Epoch[080] Batch [1499]/[3760]	Speed: 117.396907 samples/sec	 accuracy=32.916667	 loss=3.059693	 lr=0.081078
Epoch[080] Batch [1549]/[3760]	Speed: 116.832771 samples/sec	 accuracy=32.911290	 loss=3.060946	 lr=0.081068
Epoch[080] Batch [1599]/[3760]	Speed: 117.314593 samples/sec	 accuracy=32.925781	 loss=3.060388	 lr=0.081058
Epoch[080] Batch [1649]/[3760]	Speed: 116.959689 samples/sec	 accuracy=32.946023	 loss=3.059296	 lr=0.081048
Epoch[080] Batch [1699]/[3760]	Speed: 116.807230 samples/sec	 accuracy=32.942096	 loss=3.059805	 lr=0.081038
Epoch[080] Batch [1749]/[3760]	Speed: 117.273556 samples/sec	 accuracy=32.936607	 loss=3.059310	 lr=0.081027
Epoch[080] Batch [1799]/[3760]	Speed: 118.441688 samples/sec	 accuracy=32.928819	 loss=3.058865	 lr=0.081017
Epoch[080] Batch [1849]/[3760]	Speed: 117.013178 samples/sec	 accuracy=32.908784	 loss=3.059216	 lr=0.081007
Epoch[080] Batch [1899]/[3760]	Speed: 118.195503 samples/sec	 accuracy=32.909539	 loss=3.060079	 lr=0.080997
Epoch[080] Batch [1949]/[3760]	Speed: 117.414767 samples/sec	 accuracy=32.890224	 loss=3.060919	 lr=0.080987
Epoch[080] Batch [1999]/[3760]	Speed: 117.339542 samples/sec	 accuracy=32.803906	 loss=3.065191	 lr=0.080977
Epoch[080] Batch [2049]/[3760]	Speed: 117.402306 samples/sec	 accuracy=32.838415	 loss=3.065719	 lr=0.080967
Epoch[080] Batch [2099]/[3760]	Speed: 117.365159 samples/sec	 accuracy=32.857887	 loss=3.065549	 lr=0.080957
Epoch[080] Batch [2149]/[3760]	Speed: 117.821300 samples/sec	 accuracy=32.877907	 loss=3.064828	 lr=0.080946
Epoch[080] Batch [2199]/[3760]	Speed: 117.354779 samples/sec	 accuracy=32.870739	 loss=3.065295	 lr=0.080936
Epoch[080] Batch [2249]/[3760]	Speed: 118.175423 samples/sec	 accuracy=32.860417	 loss=3.066610	 lr=0.080926
Epoch[080] Batch [2299]/[3760]	Speed: 118.095868 samples/sec	 accuracy=32.865489	 loss=3.066339	 lr=0.080916
Epoch[080] Batch [2349]/[3760]	Speed: 117.708287 samples/sec	 accuracy=32.857048	 loss=3.067473	 lr=0.080906
Epoch[080] Batch [2399]/[3760]	Speed: 118.326753 samples/sec	 accuracy=32.847005	 loss=3.067919	 lr=0.080896
Epoch[080] Batch [2449]/[3760]	Speed: 117.441808 samples/sec	 accuracy=32.852041	 loss=3.068123	 lr=0.080886
Epoch[080] Batch [2499]/[3760]	Speed: 117.261357 samples/sec	 accuracy=32.849375	 loss=3.068326	 lr=0.080875
Epoch[080] Batch [2549]/[3760]	Speed: 116.891470 samples/sec	 accuracy=32.831495	 loss=3.069886	 lr=0.080865
Epoch[080] Batch [2599]/[3760]	Speed: 117.333467 samples/sec	 accuracy=32.825721	 loss=3.070110	 lr=0.080855
Epoch[080] Batch [2649]/[3760]	Speed: 117.367819 samples/sec	 accuracy=32.837854	 loss=3.069218	 lr=0.080845
Epoch[080] Batch [2699]/[3760]	Speed: 117.395558 samples/sec	 accuracy=32.817130	 loss=3.069860	 lr=0.080835
Epoch[080] Batch [2749]/[3760]	Speed: 117.547362 samples/sec	 accuracy=32.831250	 loss=3.069488	 lr=0.080825
Epoch[080] Batch [2799]/[3760]	Speed: 117.335913 samples/sec	 accuracy=32.808036	 loss=3.070222	 lr=0.080815
Epoch[080] Batch [2849]/[3760]	Speed: 117.502832 samples/sec	 accuracy=32.810307	 loss=3.070008	 lr=0.080804
Epoch[080] Batch [2899]/[3760]	Speed: 117.672823 samples/sec	 accuracy=32.808190	 loss=3.070440	 lr=0.080794
Epoch[080] Batch [2949]/[3760]	Speed: 117.281442 samples/sec	 accuracy=32.800318	 loss=3.071173	 lr=0.080784
Epoch[080] Batch [2999]/[3760]	Speed: 117.424743 samples/sec	 accuracy=32.802604	 loss=3.071166	 lr=0.080774
Epoch[080] Batch [3049]/[3760]	Speed: 117.324344 samples/sec	 accuracy=32.762295	 loss=3.072765	 lr=0.080764
Epoch[080] Batch [3099]/[3760]	Speed: 117.609633 samples/sec	 accuracy=32.753528	 loss=3.072978	 lr=0.080754
Epoch[080] Batch [3149]/[3760]	Speed: 117.935401 samples/sec	 accuracy=32.739087	 loss=3.073417	 lr=0.080743
Epoch[080] Batch [3199]/[3760]	Speed: 117.921188 samples/sec	 accuracy=32.750488	 loss=3.072867	 lr=0.080733
Epoch[080] Batch [3249]/[3760]	Speed: 117.329872 samples/sec	 accuracy=32.748558	 loss=3.073400	 lr=0.080723
Epoch[080] Batch [3299]/[3760]	Speed: 117.120582 samples/sec	 accuracy=32.739583	 loss=3.073585	 lr=0.080713
Epoch[080] Batch [3349]/[3760]	Speed: 116.678312 samples/sec	 accuracy=32.722015	 loss=3.074973	 lr=0.080703
Epoch[080] Batch [3399]/[3760]	Speed: 116.430633 samples/sec	 accuracy=32.739430	 loss=3.074207	 lr=0.080693
Epoch[080] Batch [3449]/[3760]	Speed: 116.821356 samples/sec	 accuracy=32.753623	 loss=3.073946	 lr=0.080682
Epoch[080] Batch [3499]/[3760]	Speed: 117.440643 samples/sec	 accuracy=32.750446	 loss=3.074775	 lr=0.080672
Epoch[080] Batch [3549]/[3760]	Speed: 117.788086 samples/sec	 accuracy=32.733275	 loss=3.075776	 lr=0.080662
Epoch[080] Batch [3599]/[3760]	Speed: 117.687613 samples/sec	 accuracy=32.740017	 loss=3.075610	 lr=0.080652
Epoch[080] Batch [3649]/[3760]	Speed: 117.365667 samples/sec	 accuracy=32.770120	 loss=3.074889	 lr=0.080642
Epoch[080] Batch [3699]/[3760]	Speed: 116.695068 samples/sec	 accuracy=32.779561	 loss=3.074139	 lr=0.080631
Epoch[080] Batch [3749]/[3760]	Speed: 120.790200 samples/sec	 accuracy=32.779167	 loss=3.073776	 lr=0.080621
Batch [0049]/[0303]: acc-top1=32.593750 acc-top5=58.843750
Batch [0099]/[0303]: acc-top1=29.531250 acc-top5=56.718750
Batch [0149]/[0303]: acc-top1=32.052083 acc-top5=58.041667
Batch [0199]/[0303]: acc-top1=35.328125 acc-top5=60.765625
Batch [0249]/[0303]: acc-top1=34.181250 acc-top5=60.487500
Batch [0299]/[0303]: acc-top1=33.187500 acc-top5=60.213542
[Epoch 080] training: accuracy=32.778424	 loss=3.073921
[Epoch 080] speed: 117 samples/sec	time cost: 2175.733698
[Epoch 080] validation: acc-top1=33.188944 acc-top5=60.210396 loss=3.181545
Epoch[081] Batch [0049]/[3759]	Speed: 85.430793 samples/sec	 accuracy=32.968750	 loss=3.003276	 lr=0.080609
Epoch[081] Batch [0099]/[3759]	Speed: 116.884476 samples/sec	 accuracy=33.703125	 loss=2.993372	 lr=0.080599
Epoch[081] Batch [0149]/[3759]	Speed: 117.099313 samples/sec	 accuracy=33.781250	 loss=3.006394	 lr=0.080589
Epoch[081] Batch [0199]/[3759]	Speed: 117.033554 samples/sec	 accuracy=33.625000	 loss=3.024892	 lr=0.080578
Epoch[081] Batch [0249]/[3759]	Speed: 117.282585 samples/sec	 accuracy=33.281250	 loss=3.042096	 lr=0.080568
Epoch[081] Batch [0299]/[3759]	Speed: 116.555113 samples/sec	 accuracy=33.375000	 loss=3.041096	 lr=0.080558
Epoch[081] Batch [0349]/[3759]	Speed: 116.972688 samples/sec	 accuracy=33.424107	 loss=3.040321	 lr=0.080548
Epoch[081] Batch [0399]/[3759]	Speed: 117.239680 samples/sec	 accuracy=33.273438	 loss=3.047368	 lr=0.080538
Epoch[081] Batch [0449]/[3759]	Speed: 117.159641 samples/sec	 accuracy=33.135417	 loss=3.050848	 lr=0.080527
Epoch[081] Batch [0499]/[3759]	Speed: 117.036175 samples/sec	 accuracy=33.190625	 loss=3.048589	 lr=0.080517
Epoch[081] Batch [0549]/[3759]	Speed: 117.266996 samples/sec	 accuracy=33.224432	 loss=3.046375	 lr=0.080507
Epoch[081] Batch [0599]/[3759]	Speed: 117.404668 samples/sec	 accuracy=33.213542	 loss=3.050255	 lr=0.080497
Epoch[081] Batch [0649]/[3759]	Speed: 117.522576 samples/sec	 accuracy=33.218750	 loss=3.050641	 lr=0.080486
Epoch[081] Batch [0699]/[3759]	Speed: 117.442325 samples/sec	 accuracy=33.254464	 loss=3.050926	 lr=0.080476
Epoch[081] Batch [0749]/[3759]	Speed: 117.184530 samples/sec	 accuracy=33.164583	 loss=3.051112	 lr=0.080466
Epoch[081] Batch [0799]/[3759]	Speed: 117.209645 samples/sec	 accuracy=33.078125	 loss=3.055607	 lr=0.080456
Epoch[081] Batch [0849]/[3759]	Speed: 116.963143 samples/sec	 accuracy=33.097426	 loss=3.054592	 lr=0.080446
Epoch[081] Batch [0899]/[3759]	Speed: 117.722652 samples/sec	 accuracy=33.161458	 loss=3.053834	 lr=0.080435
Epoch[081] Batch [0949]/[3759]	Speed: 117.633881 samples/sec	 accuracy=33.177632	 loss=3.053211	 lr=0.080425
Epoch[081] Batch [0999]/[3759]	Speed: 117.627242 samples/sec	 accuracy=33.160937	 loss=3.054202	 lr=0.080415
Epoch[081] Batch [1049]/[3759]	Speed: 116.190213 samples/sec	 accuracy=33.101190	 loss=3.056469	 lr=0.080405
Epoch[081] Batch [1099]/[3759]	Speed: 117.873574 samples/sec	 accuracy=33.073864	 loss=3.057853	 lr=0.080394
Epoch[081] Batch [1149]/[3759]	Speed: 117.640770 samples/sec	 accuracy=33.065217	 loss=3.057825	 lr=0.080384
Epoch[081] Batch [1199]/[3759]	Speed: 117.660511 samples/sec	 accuracy=33.075521	 loss=3.057467	 lr=0.080374
Epoch[081] Batch [1249]/[3759]	Speed: 116.956638 samples/sec	 accuracy=33.085000	 loss=3.057489	 lr=0.080364
Epoch[081] Batch [1299]/[3759]	Speed: 116.991174 samples/sec	 accuracy=33.081731	 loss=3.059503	 lr=0.080353
Epoch[081] Batch [1349]/[3759]	Speed: 118.317270 samples/sec	 accuracy=33.143519	 loss=3.057588	 lr=0.080343
Epoch[081] Batch [1399]/[3759]	Speed: 117.023679 samples/sec	 accuracy=33.137277	 loss=3.056607	 lr=0.080333
Epoch[081] Batch [1449]/[3759]	Speed: 117.733800 samples/sec	 accuracy=33.151940	 loss=3.055387	 lr=0.080323
Epoch[081] Batch [1499]/[3759]	Speed: 116.860640 samples/sec	 accuracy=33.178125	 loss=3.055753	 lr=0.080312
Epoch[081] Batch [1549]/[3759]	Speed: 117.931147 samples/sec	 accuracy=33.169355	 loss=3.054819	 lr=0.080302
Epoch[081] Batch [1599]/[3759]	Speed: 117.329115 samples/sec	 accuracy=33.165039	 loss=3.054320	 lr=0.080292
Epoch[081] Batch [1649]/[3759]	Speed: 117.170645 samples/sec	 accuracy=33.127841	 loss=3.056637	 lr=0.080282
Epoch[081] Batch [1699]/[3759]	Speed: 117.594604 samples/sec	 accuracy=33.124081	 loss=3.057207	 lr=0.080271
Epoch[081] Batch [1749]/[3759]	Speed: 117.067400 samples/sec	 accuracy=33.115179	 loss=3.058235	 lr=0.080261
Epoch[081] Batch [1799]/[3759]	Speed: 116.285896 samples/sec	 accuracy=33.147569	 loss=3.057461	 lr=0.080251
Epoch[081] Batch [1849]/[3759]	Speed: 117.275597 samples/sec	 accuracy=33.168919	 loss=3.057423	 lr=0.080241
Epoch[081] Batch [1899]/[3759]	Speed: 117.537659 samples/sec	 accuracy=33.180921	 loss=3.057488	 lr=0.080230
Epoch[081] Batch [1949]/[3759]	Speed: 117.728661 samples/sec	 accuracy=33.168269	 loss=3.057529	 lr=0.080220
Epoch[081] Batch [1999]/[3759]	Speed: 117.171281 samples/sec	 accuracy=33.146094	 loss=3.058292	 lr=0.080210
Epoch[081] Batch [2049]/[3759]	Speed: 117.089013 samples/sec	 accuracy=33.095274	 loss=3.060584	 lr=0.080199
Epoch[081] Batch [2099]/[3759]	Speed: 117.581042 samples/sec	 accuracy=33.123512	 loss=3.059136	 lr=0.080189
Epoch[081] Batch [2149]/[3759]	Speed: 118.194403 samples/sec	 accuracy=33.147529	 loss=3.058419	 lr=0.080179
Epoch[081] Batch [2199]/[3759]	Speed: 116.661778 samples/sec	 accuracy=33.132102	 loss=3.058485	 lr=0.080169
Epoch[081] Batch [2249]/[3759]	Speed: 117.305270 samples/sec	 accuracy=33.127778	 loss=3.059059	 lr=0.080158
Epoch[081] Batch [2299]/[3759]	Speed: 116.837596 samples/sec	 accuracy=33.107337	 loss=3.059483	 lr=0.080148
Epoch[081] Batch [2349]/[3759]	Speed: 116.960225 samples/sec	 accuracy=33.059840	 loss=3.060803	 lr=0.080138
Epoch[081] Batch [2399]/[3759]	Speed: 117.207857 samples/sec	 accuracy=33.057943	 loss=3.061394	 lr=0.080127
Epoch[081] Batch [2449]/[3759]	Speed: 116.863133 samples/sec	 accuracy=33.049745	 loss=3.061399	 lr=0.080117
Epoch[081] Batch [2499]/[3759]	Speed: 116.926835 samples/sec	 accuracy=33.057500	 loss=3.061213	 lr=0.080107
Epoch[081] Batch [2549]/[3759]	Speed: 117.042905 samples/sec	 accuracy=33.040441	 loss=3.062682	 lr=0.080097
Epoch[081] Batch [2599]/[3759]	Speed: 117.258179 samples/sec	 accuracy=33.039062	 loss=3.062818	 lr=0.080086
Epoch[081] Batch [2649]/[3759]	Speed: 116.718230 samples/sec	 accuracy=33.048349	 loss=3.062668	 lr=0.080076
Epoch[081] Batch [2699]/[3759]	Speed: 117.268746 samples/sec	 accuracy=33.041088	 loss=3.062194	 lr=0.080066
Epoch[081] Batch [2749]/[3759]	Speed: 117.182138 samples/sec	 accuracy=33.029545	 loss=3.062192	 lr=0.080055
Epoch[081] Batch [2799]/[3759]	Speed: 117.124822 samples/sec	 accuracy=33.010603	 loss=3.063096	 lr=0.080045
Epoch[081] Batch [2849]/[3759]	Speed: 118.231504 samples/sec	 accuracy=33.042763	 loss=3.061527	 lr=0.080035
Epoch[081] Batch [2899]/[3759]	Speed: 118.238101 samples/sec	 accuracy=33.038793	 loss=3.062072	 lr=0.080024
Epoch[081] Batch [2949]/[3759]	Speed: 117.843775 samples/sec	 accuracy=33.021186	 loss=3.061992	 lr=0.080014
Epoch[081] Batch [2999]/[3759]	Speed: 116.959304 samples/sec	 accuracy=33.013542	 loss=3.063079	 lr=0.080004
Epoch[081] Batch [3049]/[3759]	Speed: 117.117829 samples/sec	 accuracy=33.002561	 loss=3.063654	 lr=0.079993
Epoch[081] Batch [3099]/[3759]	Speed: 116.908966 samples/sec	 accuracy=33.005040	 loss=3.063345	 lr=0.079983
Epoch[081] Batch [3149]/[3759]	Speed: 117.229160 samples/sec	 accuracy=32.996528	 loss=3.063298	 lr=0.079973
Epoch[081] Batch [3199]/[3759]	Speed: 118.059585 samples/sec	 accuracy=33.012695	 loss=3.062824	 lr=0.079962
Epoch[081] Batch [3249]/[3759]	Speed: 117.397860 samples/sec	 accuracy=33.006731	 loss=3.063579	 lr=0.079952
Epoch[081] Batch [3299]/[3759]	Speed: 118.249028 samples/sec	 accuracy=32.995265	 loss=3.063661	 lr=0.079942
Epoch[081] Batch [3349]/[3759]	Speed: 117.791674 samples/sec	 accuracy=32.975280	 loss=3.064768	 lr=0.079931
Epoch[081] Batch [3399]/[3759]	Speed: 117.674472 samples/sec	 accuracy=32.959099	 loss=3.065400	 lr=0.079921
Epoch[081] Batch [3449]/[3759]	Speed: 117.522974 samples/sec	 accuracy=32.965127	 loss=3.065633	 lr=0.079911
Epoch[081] Batch [3499]/[3759]	Speed: 118.062530 samples/sec	 accuracy=32.966518	 loss=3.065708	 lr=0.079900
Epoch[081] Batch [3549]/[3759]	Speed: 117.589225 samples/sec	 accuracy=32.943662	 loss=3.066537	 lr=0.079890
Epoch[081] Batch [3599]/[3759]	Speed: 117.100126 samples/sec	 accuracy=32.931424	 loss=3.067402	 lr=0.079880
Epoch[081] Batch [3649]/[3759]	Speed: 117.137734 samples/sec	 accuracy=32.939212	 loss=3.067676	 lr=0.079869
Epoch[081] Batch [3699]/[3759]	Speed: 117.396951 samples/sec	 accuracy=32.945524	 loss=3.066918	 lr=0.079859
Epoch[081] Batch [3749]/[3759]	Speed: 120.535158 samples/sec	 accuracy=32.955417	 loss=3.066303	 lr=0.079849
Batch [0049]/[0303]: acc-top1=33.218750 acc-top5=58.562500
Batch [0099]/[0303]: acc-top1=29.765625 acc-top5=55.515625
Batch [0149]/[0303]: acc-top1=31.343750 acc-top5=56.781250
Batch [0199]/[0303]: acc-top1=33.406250 acc-top5=59.210938
Batch [0249]/[0303]: acc-top1=32.656250 acc-top5=58.837500
Batch [0299]/[0303]: acc-top1=32.869792 acc-top5=59.322917
[Epoch 081] training: accuracy=32.955074	 loss=3.066444
[Epoch 081] speed: 116 samples/sec	time cost: 2178.467669
[Epoch 081] validation: acc-top1=32.853754 acc-top5=59.313119 loss=3.194040
Epoch[082] Batch [0049]/[3760]	Speed: 87.481642 samples/sec	 accuracy=33.187500	 loss=3.014494	 lr=0.079837
Epoch[082] Batch [0099]/[3760]	Speed: 116.704607 samples/sec	 accuracy=33.859375	 loss=3.006474	 lr=0.079826
Epoch[082] Batch [0149]/[3760]	Speed: 117.672943 samples/sec	 accuracy=33.510417	 loss=3.017799	 lr=0.079816
Epoch[082] Batch [0199]/[3760]	Speed: 115.978136 samples/sec	 accuracy=33.265625	 loss=3.028986	 lr=0.079805
Epoch[082] Batch [0249]/[3760]	Speed: 117.546326 samples/sec	 accuracy=33.125000	 loss=3.035980	 lr=0.079795
Epoch[082] Batch [0299]/[3760]	Speed: 117.370672 samples/sec	 accuracy=33.427083	 loss=3.028682	 lr=0.079785
Epoch[082] Batch [0349]/[3760]	Speed: 118.921036 samples/sec	 accuracy=33.361607	 loss=3.035165	 lr=0.079774
Epoch[082] Batch [0399]/[3760]	Speed: 117.666973 samples/sec	 accuracy=33.261719	 loss=3.041745	 lr=0.079764
Epoch[082] Batch [0449]/[3760]	Speed: 117.245152 samples/sec	 accuracy=33.090278	 loss=3.048169	 lr=0.079754
Epoch[082] Batch [0499]/[3760]	Speed: 117.546571 samples/sec	 accuracy=33.078125	 loss=3.050159	 lr=0.079743
Epoch[082] Batch [0549]/[3760]	Speed: 116.996934 samples/sec	 accuracy=33.028409	 loss=3.052546	 lr=0.079733
Epoch[082] Batch [0599]/[3760]	Speed: 118.158264 samples/sec	 accuracy=33.062500	 loss=3.051376	 lr=0.079723
Epoch[082] Batch [0649]/[3760]	Speed: 117.057783 samples/sec	 accuracy=33.158654	 loss=3.047171	 lr=0.079712
Epoch[082] Batch [0699]/[3760]	Speed: 117.634947 samples/sec	 accuracy=33.084821	 loss=3.048716	 lr=0.079702
Epoch[082] Batch [0749]/[3760]	Speed: 116.756866 samples/sec	 accuracy=33.114583	 loss=3.043120	 lr=0.079691
Epoch[082] Batch [0799]/[3760]	Speed: 117.444845 samples/sec	 accuracy=33.140625	 loss=3.042553	 lr=0.079681
Epoch[082] Batch [0849]/[3760]	Speed: 117.014882 samples/sec	 accuracy=33.205882	 loss=3.041577	 lr=0.079671
Epoch[082] Batch [0899]/[3760]	Speed: 117.339766 samples/sec	 accuracy=33.140625	 loss=3.043186	 lr=0.079660
Epoch[082] Batch [0949]/[3760]	Speed: 117.092605 samples/sec	 accuracy=33.164474	 loss=3.044366	 lr=0.079650
Epoch[082] Batch [0999]/[3760]	Speed: 116.660840 samples/sec	 accuracy=33.175000	 loss=3.044428	 lr=0.079640
Epoch[082] Batch [1049]/[3760]	Speed: 116.846459 samples/sec	 accuracy=33.178571	 loss=3.044991	 lr=0.079629
Epoch[082] Batch [1099]/[3760]	Speed: 116.616549 samples/sec	 accuracy=33.122159	 loss=3.044783	 lr=0.079619
Epoch[082] Batch [1149]/[3760]	Speed: 116.621106 samples/sec	 accuracy=33.089674	 loss=3.048365	 lr=0.079608
Epoch[082] Batch [1199]/[3760]	Speed: 116.948631 samples/sec	 accuracy=33.106771	 loss=3.047047	 lr=0.079598
Epoch[082] Batch [1249]/[3760]	Speed: 117.853421 samples/sec	 accuracy=33.091250	 loss=3.047563	 lr=0.079588
Epoch[082] Batch [1299]/[3760]	Speed: 117.162152 samples/sec	 accuracy=33.120192	 loss=3.048064	 lr=0.079577
Epoch[082] Batch [1349]/[3760]	Speed: 117.949504 samples/sec	 accuracy=33.144676	 loss=3.048679	 lr=0.079567
Epoch[082] Batch [1399]/[3760]	Speed: 117.539303 samples/sec	 accuracy=33.143973	 loss=3.049618	 lr=0.079556
Epoch[082] Batch [1449]/[3760]	Speed: 117.780938 samples/sec	 accuracy=33.184267	 loss=3.048546	 lr=0.079546
Epoch[082] Batch [1499]/[3760]	Speed: 117.058909 samples/sec	 accuracy=33.118750	 loss=3.051230	 lr=0.079536
Epoch[082] Batch [1549]/[3760]	Speed: 117.739444 samples/sec	 accuracy=33.100806	 loss=3.053797	 lr=0.079525
Epoch[082] Batch [1599]/[3760]	Speed: 117.262337 samples/sec	 accuracy=33.044922	 loss=3.056649	 lr=0.079515
Epoch[082] Batch [1649]/[3760]	Speed: 117.480647 samples/sec	 accuracy=33.038826	 loss=3.055960	 lr=0.079504
Epoch[082] Batch [1699]/[3760]	Speed: 118.041778 samples/sec	 accuracy=33.007353	 loss=3.056349	 lr=0.079494
Epoch[082] Batch [1749]/[3760]	Speed: 117.526215 samples/sec	 accuracy=33.032143	 loss=3.055004	 lr=0.079484
Epoch[082] Batch [1799]/[3760]	Speed: 117.994782 samples/sec	 accuracy=33.030382	 loss=3.055298	 lr=0.079473
Epoch[082] Batch [1849]/[3760]	Speed: 116.844660 samples/sec	 accuracy=33.053209	 loss=3.055481	 lr=0.079463
Epoch[082] Batch [1899]/[3760]	Speed: 117.569041 samples/sec	 accuracy=33.078125	 loss=3.053834	 lr=0.079452
Epoch[082] Batch [1949]/[3760]	Speed: 117.867635 samples/sec	 accuracy=33.062500	 loss=3.055566	 lr=0.079442
Epoch[082] Batch [1999]/[3760]	Speed: 117.619528 samples/sec	 accuracy=33.078125	 loss=3.055836	 lr=0.079431
Epoch[082] Batch [2049]/[3760]	Speed: 117.171179 samples/sec	 accuracy=33.087652	 loss=3.055086	 lr=0.079421
Epoch[082] Batch [2099]/[3760]	Speed: 118.208826 samples/sec	 accuracy=33.073661	 loss=3.053736	 lr=0.079411
Epoch[082] Batch [2149]/[3760]	Speed: 117.369608 samples/sec	 accuracy=33.078488	 loss=3.053149	 lr=0.079400
Epoch[082] Batch [2199]/[3760]	Speed: 117.004330 samples/sec	 accuracy=33.065341	 loss=3.054832	 lr=0.079390
Epoch[082] Batch [2249]/[3760]	Speed: 117.073404 samples/sec	 accuracy=33.047917	 loss=3.055387	 lr=0.079379
Epoch[082] Batch [2299]/[3760]	Speed: 117.918366 samples/sec	 accuracy=33.046875	 loss=3.055365	 lr=0.079369
Epoch[082] Batch [2349]/[3760]	Speed: 117.521714 samples/sec	 accuracy=33.054521	 loss=3.055076	 lr=0.079358
Epoch[082] Batch [2399]/[3760]	Speed: 118.039210 samples/sec	 accuracy=33.037760	 loss=3.055584	 lr=0.079348
Epoch[082] Batch [2449]/[3760]	Speed: 117.315824 samples/sec	 accuracy=33.044643	 loss=3.056121	 lr=0.079338
Epoch[082] Batch [2499]/[3760]	Speed: 117.102877 samples/sec	 accuracy=33.016875	 loss=3.057021	 lr=0.079327
Epoch[082] Batch [2549]/[3760]	Speed: 116.877848 samples/sec	 accuracy=33.017157	 loss=3.057963	 lr=0.079317
Epoch[082] Batch [2599]/[3760]	Speed: 118.043953 samples/sec	 accuracy=33.040865	 loss=3.057472	 lr=0.079306
Epoch[082] Batch [2649]/[3760]	Speed: 117.387488 samples/sec	 accuracy=33.047170	 loss=3.057754	 lr=0.079296
Epoch[082] Batch [2699]/[3760]	Speed: 118.899901 samples/sec	 accuracy=33.048611	 loss=3.057914	 lr=0.079285
Epoch[082] Batch [2749]/[3760]	Speed: 118.149104 samples/sec	 accuracy=33.052841	 loss=3.056975	 lr=0.079275
Epoch[082] Batch [2799]/[3760]	Speed: 117.416166 samples/sec	 accuracy=33.047433	 loss=3.056658	 lr=0.079264
Epoch[082] Batch [2849]/[3760]	Speed: 117.604238 samples/sec	 accuracy=33.070724	 loss=3.055945	 lr=0.079254
Epoch[082] Batch [2899]/[3760]	Speed: 117.702039 samples/sec	 accuracy=33.068427	 loss=3.056293	 lr=0.079243
Epoch[082] Batch [2949]/[3760]	Speed: 117.050456 samples/sec	 accuracy=33.091102	 loss=3.055362	 lr=0.079233
Epoch[082] Batch [2999]/[3760]	Speed: 117.647084 samples/sec	 accuracy=33.088021	 loss=3.055340	 lr=0.079223
Epoch[082] Batch [3049]/[3760]	Speed: 117.628494 samples/sec	 accuracy=33.069672	 loss=3.056418	 lr=0.079212
Epoch[082] Batch [3099]/[3760]	Speed: 117.302898 samples/sec	 accuracy=33.068548	 loss=3.056196	 lr=0.079202
Epoch[082] Batch [3149]/[3760]	Speed: 117.324873 samples/sec	 accuracy=33.071925	 loss=3.055477	 lr=0.079191
Epoch[082] Batch [3199]/[3760]	Speed: 117.383391 samples/sec	 accuracy=33.064941	 loss=3.056474	 lr=0.079181
Epoch[082] Batch [3249]/[3760]	Speed: 116.838573 samples/sec	 accuracy=33.068269	 loss=3.056032	 lr=0.079170
Epoch[082] Batch [3299]/[3760]	Speed: 116.998605 samples/sec	 accuracy=33.061080	 loss=3.056333	 lr=0.079160
Epoch[082] Batch [3349]/[3760]	Speed: 118.192260 samples/sec	 accuracy=33.068097	 loss=3.056689	 lr=0.079149
Epoch[082] Batch [3399]/[3760]	Speed: 117.787052 samples/sec	 accuracy=33.060202	 loss=3.057003	 lr=0.079139
Epoch[082] Batch [3449]/[3760]	Speed: 117.697944 samples/sec	 accuracy=33.080163	 loss=3.056465	 lr=0.079128
Epoch[082] Batch [3499]/[3760]	Speed: 118.057782 samples/sec	 accuracy=33.070089	 loss=3.057031	 lr=0.079118
Epoch[082] Batch [3549]/[3760]	Speed: 117.149661 samples/sec	 accuracy=33.038732	 loss=3.057553	 lr=0.079107
Epoch[082] Batch [3599]/[3760]	Speed: 117.386540 samples/sec	 accuracy=33.054688	 loss=3.056871	 lr=0.079097
Epoch[082] Batch [3649]/[3760]	Speed: 118.296755 samples/sec	 accuracy=33.071490	 loss=3.056767	 lr=0.079086
Epoch[082] Batch [3699]/[3760]	Speed: 118.555087 samples/sec	 accuracy=33.065456	 loss=3.056738	 lr=0.079076
Epoch[082] Batch [3749]/[3760]	Speed: 120.097114 samples/sec	 accuracy=33.070000	 loss=3.056519	 lr=0.079065
Batch [0049]/[0303]: acc-top1=33.593750 acc-top5=59.125000
Batch [0099]/[0303]: acc-top1=31.921875 acc-top5=58.359375
Batch [0149]/[0303]: acc-top1=33.687500 acc-top5=59.791667
Batch [0199]/[0303]: acc-top1=36.101562 acc-top5=61.859375
Batch [0249]/[0303]: acc-top1=34.950000 acc-top5=61.575000
Batch [0299]/[0303]: acc-top1=34.458333 acc-top5=61.421875
[Epoch 082] training: accuracy=33.077211	 loss=3.056504
[Epoch 082] speed: 116 samples/sec	time cost: 2175.890413
[Epoch 082] validation: acc-top1=34.493606 acc-top5=61.458333 loss=3.054896
Epoch[083] Batch [0049]/[3760]	Speed: 88.194249 samples/sec	 accuracy=33.718750	 loss=3.052630	 lr=0.079053
Epoch[083] Batch [0099]/[3760]	Speed: 117.348599 samples/sec	 accuracy=33.546875	 loss=3.018403	 lr=0.079042
Epoch[083] Batch [0149]/[3760]	Speed: 117.801064 samples/sec	 accuracy=33.489583	 loss=3.032055	 lr=0.079032
Epoch[083] Batch [0199]/[3760]	Speed: 117.570625 samples/sec	 accuracy=33.632812	 loss=3.029573	 lr=0.079021
Epoch[083] Batch [0249]/[3760]	Speed: 117.137842 samples/sec	 accuracy=33.606250	 loss=3.032086	 lr=0.079011
Epoch[083] Batch [0299]/[3760]	Speed: 117.052321 samples/sec	 accuracy=33.848958	 loss=3.023054	 lr=0.079000
Epoch[083] Batch [0349]/[3760]	Speed: 117.855170 samples/sec	 accuracy=33.839286	 loss=3.028193	 lr=0.078990
Epoch[083] Batch [0399]/[3760]	Speed: 117.043648 samples/sec	 accuracy=33.714844	 loss=3.029620	 lr=0.078979
Epoch[083] Batch [0449]/[3760]	Speed: 117.242462 samples/sec	 accuracy=33.642361	 loss=3.030650	 lr=0.078969
Epoch[083] Batch [0499]/[3760]	Speed: 118.058814 samples/sec	 accuracy=33.553125	 loss=3.031984	 lr=0.078958
Epoch[083] Batch [0549]/[3760]	Speed: 117.316748 samples/sec	 accuracy=33.593750	 loss=3.036321	 lr=0.078948
Epoch[083] Batch [0599]/[3760]	Speed: 117.658723 samples/sec	 accuracy=33.505208	 loss=3.038225	 lr=0.078937
Epoch[083] Batch [0649]/[3760]	Speed: 116.985287 samples/sec	 accuracy=33.468750	 loss=3.041699	 lr=0.078927
Epoch[083] Batch [0699]/[3760]	Speed: 117.274099 samples/sec	 accuracy=33.500000	 loss=3.040644	 lr=0.078916
Epoch[083] Batch [0749]/[3760]	Speed: 116.227234 samples/sec	 accuracy=33.547917	 loss=3.039324	 lr=0.078906
Epoch[083] Batch [0799]/[3760]	Speed: 117.257747 samples/sec	 accuracy=33.609375	 loss=3.036271	 lr=0.078895
Epoch[083] Batch [0849]/[3760]	Speed: 117.350605 samples/sec	 accuracy=33.586397	 loss=3.039158	 lr=0.078885
Epoch[083] Batch [0899]/[3760]	Speed: 117.504785 samples/sec	 accuracy=33.616319	 loss=3.039368	 lr=0.078874
Epoch[083] Batch [0949]/[3760]	Speed: 116.979429 samples/sec	 accuracy=33.616776	 loss=3.039908	 lr=0.078863
Epoch[083] Batch [0999]/[3760]	Speed: 115.978428 samples/sec	 accuracy=33.539062	 loss=3.041857	 lr=0.078853
Epoch[083] Batch [1049]/[3760]	Speed: 118.276224 samples/sec	 accuracy=33.497024	 loss=3.042540	 lr=0.078842
Epoch[083] Batch [1099]/[3760]	Speed: 117.876357 samples/sec	 accuracy=33.509943	 loss=3.041247	 lr=0.078832
Epoch[083] Batch [1149]/[3760]	Speed: 117.374699 samples/sec	 accuracy=33.527174	 loss=3.041673	 lr=0.078821
Epoch[083] Batch [1199]/[3760]	Speed: 117.467504 samples/sec	 accuracy=33.557292	 loss=3.041737	 lr=0.078811
Epoch[083] Batch [1249]/[3760]	Speed: 117.883412 samples/sec	 accuracy=33.503750	 loss=3.040736	 lr=0.078800
Epoch[083] Batch [1299]/[3760]	Speed: 117.498593 samples/sec	 accuracy=33.510817	 loss=3.042593	 lr=0.078790
Epoch[083] Batch [1349]/[3760]	Speed: 116.825629 samples/sec	 accuracy=33.516204	 loss=3.044771	 lr=0.078779
Epoch[083] Batch [1399]/[3760]	Speed: 117.761522 samples/sec	 accuracy=33.506696	 loss=3.046125	 lr=0.078769
Epoch[083] Batch [1449]/[3760]	Speed: 117.271594 samples/sec	 accuracy=33.495690	 loss=3.047104	 lr=0.078758
Epoch[083] Batch [1499]/[3760]	Speed: 117.282892 samples/sec	 accuracy=33.435417	 loss=3.048849	 lr=0.078747
Epoch[083] Batch [1549]/[3760]	Speed: 117.683033 samples/sec	 accuracy=33.361895	 loss=3.050701	 lr=0.078737
Epoch[083] Batch [1599]/[3760]	Speed: 118.395346 samples/sec	 accuracy=33.425781	 loss=3.047974	 lr=0.078726
Epoch[083] Batch [1649]/[3760]	Speed: 117.269667 samples/sec	 accuracy=33.453598	 loss=3.047727	 lr=0.078716
Epoch[083] Batch [1699]/[3760]	Speed: 117.643107 samples/sec	 accuracy=33.452206	 loss=3.048053	 lr=0.078705
Epoch[083] Batch [1749]/[3760]	Speed: 116.830187 samples/sec	 accuracy=33.470536	 loss=3.049258	 lr=0.078695
Epoch[083] Batch [1799]/[3760]	Speed: 117.574101 samples/sec	 accuracy=33.480903	 loss=3.048744	 lr=0.078684
Epoch[083] Batch [1849]/[3760]	Speed: 117.356820 samples/sec	 accuracy=33.456926	 loss=3.048741	 lr=0.078674
Epoch[083] Batch [1899]/[3760]	Speed: 118.054351 samples/sec	 accuracy=33.451480	 loss=3.049123	 lr=0.078663
Epoch[083] Batch [1949]/[3760]	Speed: 117.545425 samples/sec	 accuracy=33.400641	 loss=3.051861	 lr=0.078652
Epoch[083] Batch [1999]/[3760]	Speed: 117.541080 samples/sec	 accuracy=33.414062	 loss=3.051507	 lr=0.078642
Epoch[083] Batch [2049]/[3760]	Speed: 117.642377 samples/sec	 accuracy=33.365854	 loss=3.053441	 lr=0.078631
Epoch[083] Batch [2099]/[3760]	Speed: 117.487719 samples/sec	 accuracy=33.343006	 loss=3.054049	 lr=0.078621
Epoch[083] Batch [2149]/[3760]	Speed: 117.015114 samples/sec	 accuracy=33.361919	 loss=3.054344	 lr=0.078610
Epoch[083] Batch [2199]/[3760]	Speed: 117.191695 samples/sec	 accuracy=33.321733	 loss=3.056732	 lr=0.078600
Epoch[083] Batch [2249]/[3760]	Speed: 117.302739 samples/sec	 accuracy=33.300694	 loss=3.057122	 lr=0.078589
Epoch[083] Batch [2299]/[3760]	Speed: 118.694050 samples/sec	 accuracy=33.297554	 loss=3.057051	 lr=0.078578
Epoch[083] Batch [2349]/[3760]	Speed: 118.076987 samples/sec	 accuracy=33.287899	 loss=3.057102	 lr=0.078568
Epoch[083] Batch [2399]/[3760]	Speed: 117.310238 samples/sec	 accuracy=33.285156	 loss=3.057463	 lr=0.078557
Epoch[083] Batch [2449]/[3760]	Speed: 117.205461 samples/sec	 accuracy=33.261480	 loss=3.058833	 lr=0.078547
Epoch[083] Batch [2499]/[3760]	Speed: 116.645334 samples/sec	 accuracy=33.250000	 loss=3.059455	 lr=0.078536
Epoch[083] Batch [2549]/[3760]	Speed: 117.432634 samples/sec	 accuracy=33.251838	 loss=3.058037	 lr=0.078525
Epoch[083] Batch [2599]/[3760]	Speed: 116.683725 samples/sec	 accuracy=33.243389	 loss=3.058599	 lr=0.078515
Epoch[083] Batch [2649]/[3760]	Speed: 117.670384 samples/sec	 accuracy=33.230542	 loss=3.058228	 lr=0.078504
Epoch[083] Batch [2699]/[3760]	Speed: 117.809949 samples/sec	 accuracy=33.231481	 loss=3.058241	 lr=0.078494
Epoch[083] Batch [2749]/[3760]	Speed: 117.918558 samples/sec	 accuracy=33.251705	 loss=3.056824	 lr=0.078483
Epoch[083] Batch [2799]/[3760]	Speed: 117.291239 samples/sec	 accuracy=33.271205	 loss=3.056504	 lr=0.078472
Epoch[083] Batch [2849]/[3760]	Speed: 117.109698 samples/sec	 accuracy=33.249452	 loss=3.057069	 lr=0.078462
Epoch[083] Batch [2899]/[3760]	Speed: 117.303701 samples/sec	 accuracy=33.247845	 loss=3.057141	 lr=0.078451
Epoch[083] Batch [2949]/[3760]	Speed: 118.064088 samples/sec	 accuracy=33.256356	 loss=3.056860	 lr=0.078441
Epoch[083] Batch [2999]/[3760]	Speed: 117.903736 samples/sec	 accuracy=33.259896	 loss=3.057368	 lr=0.078430
Epoch[083] Batch [3049]/[3760]	Speed: 118.210392 samples/sec	 accuracy=33.273566	 loss=3.056665	 lr=0.078419
Epoch[083] Batch [3099]/[3760]	Speed: 117.699572 samples/sec	 accuracy=33.268145	 loss=3.056426	 lr=0.078409
Epoch[083] Batch [3149]/[3760]	Speed: 118.214911 samples/sec	 accuracy=33.269841	 loss=3.055740	 lr=0.078398
Epoch[083] Batch [3199]/[3760]	Speed: 117.882641 samples/sec	 accuracy=33.275879	 loss=3.055487	 lr=0.078388
Epoch[083] Batch [3249]/[3760]	Speed: 117.067848 samples/sec	 accuracy=33.251442	 loss=3.055504	 lr=0.078377
Epoch[083] Batch [3299]/[3760]	Speed: 117.817594 samples/sec	 accuracy=33.235795	 loss=3.055639	 lr=0.078366
Epoch[083] Batch [3349]/[3760]	Speed: 117.598056 samples/sec	 accuracy=33.237407	 loss=3.055220	 lr=0.078356
Epoch[083] Batch [3399]/[3760]	Speed: 117.550557 samples/sec	 accuracy=33.223805	 loss=3.055416	 lr=0.078345
Epoch[083] Batch [3449]/[3760]	Speed: 117.229205 samples/sec	 accuracy=33.221467	 loss=3.055740	 lr=0.078334
Epoch[083] Batch [3499]/[3760]	Speed: 117.645785 samples/sec	 accuracy=33.232589	 loss=3.055539	 lr=0.078324
Epoch[083] Batch [3549]/[3760]	Speed: 117.747980 samples/sec	 accuracy=33.233715	 loss=3.055462	 lr=0.078313
Epoch[083] Batch [3599]/[3760]	Speed: 117.874172 samples/sec	 accuracy=33.251302	 loss=3.054958	 lr=0.078303
Epoch[083] Batch [3649]/[3760]	Speed: 117.437378 samples/sec	 accuracy=33.261986	 loss=3.054788	 lr=0.078292
Epoch[083] Batch [3699]/[3760]	Speed: 118.660291 samples/sec	 accuracy=33.267736	 loss=3.054387	 lr=0.078281
Epoch[083] Batch [3749]/[3760]	Speed: 120.754489 samples/sec	 accuracy=33.258333	 loss=3.054840	 lr=0.078271
Batch [0049]/[0303]: acc-top1=31.406250 acc-top5=54.468750
Batch [0099]/[0303]: acc-top1=29.250000 acc-top5=53.234375
Batch [0149]/[0303]: acc-top1=29.822917 acc-top5=54.552083
Batch [0199]/[0303]: acc-top1=32.210938 acc-top5=56.773438
Batch [0249]/[0303]: acc-top1=30.993750 acc-top5=55.925000
Batch [0299]/[0303]: acc-top1=30.723958 acc-top5=56.119792
[Epoch 083] training: accuracy=33.250083	 loss=3.055384
[Epoch 083] speed: 117 samples/sec	time cost: 2176.077933
[Epoch 083] validation: acc-top1=30.744637 acc-top5=56.146865 loss=3.667589
Epoch[084] Batch [0049]/[3759]	Speed: 87.981405 samples/sec	 accuracy=32.687500	 loss=3.023841	 lr=0.078258
Epoch[084] Batch [0099]/[3759]	Speed: 117.554726 samples/sec	 accuracy=32.687500	 loss=3.039896	 lr=0.078247
Epoch[084] Batch [0149]/[3759]	Speed: 117.477304 samples/sec	 accuracy=33.447917	 loss=3.014438	 lr=0.078237
Epoch[084] Batch [0199]/[3759]	Speed: 117.009389 samples/sec	 accuracy=33.328125	 loss=3.019720	 lr=0.078226
Epoch[084] Batch [0249]/[3759]	Speed: 117.228898 samples/sec	 accuracy=33.362500	 loss=3.019449	 lr=0.078215
Epoch[084] Batch [0299]/[3759]	Speed: 117.418458 samples/sec	 accuracy=33.536458	 loss=3.018579	 lr=0.078205
Epoch[084] Batch [0349]/[3759]	Speed: 117.179843 samples/sec	 accuracy=33.504464	 loss=3.023627	 lr=0.078194
Epoch[084] Batch [0399]/[3759]	Speed: 117.010856 samples/sec	 accuracy=33.363281	 loss=3.027934	 lr=0.078183
Epoch[084] Batch [0449]/[3759]	Speed: 118.008198 samples/sec	 accuracy=33.506944	 loss=3.024099	 lr=0.078173
Epoch[084] Batch [0499]/[3759]	Speed: 117.412354 samples/sec	 accuracy=33.431250	 loss=3.026320	 lr=0.078162
Epoch[084] Batch [0549]/[3759]	Speed: 117.291093 samples/sec	 accuracy=33.480114	 loss=3.025327	 lr=0.078151
Epoch[084] Batch [0599]/[3759]	Speed: 117.356439 samples/sec	 accuracy=33.447917	 loss=3.025187	 lr=0.078141
Epoch[084] Batch [0649]/[3759]	Speed: 117.904956 samples/sec	 accuracy=33.437500	 loss=3.030150	 lr=0.078130
Epoch[084] Batch [0699]/[3759]	Speed: 117.493958 samples/sec	 accuracy=33.437500	 loss=3.032086	 lr=0.078119
Epoch[084] Batch [0749]/[3759]	Speed: 116.731550 samples/sec	 accuracy=33.429167	 loss=3.030233	 lr=0.078109
Epoch[084] Batch [0799]/[3759]	Speed: 117.498339 samples/sec	 accuracy=33.431641	 loss=3.030694	 lr=0.078098
Epoch[084] Batch [0849]/[3759]	Speed: 117.259817 samples/sec	 accuracy=33.437500	 loss=3.031055	 lr=0.078087
Epoch[084] Batch [0899]/[3759]	Speed: 118.043003 samples/sec	 accuracy=33.449653	 loss=3.030089	 lr=0.078077
Epoch[084] Batch [0949]/[3759]	Speed: 117.329723 samples/sec	 accuracy=33.412829	 loss=3.033528	 lr=0.078066
Epoch[084] Batch [0999]/[3759]	Speed: 117.005436 samples/sec	 accuracy=33.409375	 loss=3.035502	 lr=0.078055
Epoch[084] Batch [1049]/[3759]	Speed: 117.944623 samples/sec	 accuracy=33.300595	 loss=3.041382	 lr=0.078045
Epoch[084] Batch [1099]/[3759]	Speed: 117.173659 samples/sec	 accuracy=33.265625	 loss=3.043754	 lr=0.078034
Epoch[084] Batch [1149]/[3759]	Speed: 117.512349 samples/sec	 accuracy=33.277174	 loss=3.044328	 lr=0.078023
Epoch[084] Batch [1199]/[3759]	Speed: 117.624956 samples/sec	 accuracy=33.277344	 loss=3.042458	 lr=0.078013
Epoch[084] Batch [1249]/[3759]	Speed: 117.573148 samples/sec	 accuracy=33.291250	 loss=3.041085	 lr=0.078002
Epoch[084] Batch [1299]/[3759]	Speed: 117.218114 samples/sec	 accuracy=33.323317	 loss=3.040354	 lr=0.077991
Epoch[084] Batch [1349]/[3759]	Speed: 118.012401 samples/sec	 accuracy=33.304398	 loss=3.041056	 lr=0.077981
Epoch[084] Batch [1399]/[3759]	Speed: 118.343508 samples/sec	 accuracy=33.339286	 loss=3.040591	 lr=0.077970
Epoch[084] Batch [1449]/[3759]	Speed: 117.333810 samples/sec	 accuracy=33.321121	 loss=3.042687	 lr=0.077959
Epoch[084] Batch [1499]/[3759]	Speed: 116.948724 samples/sec	 accuracy=33.303125	 loss=3.043447	 lr=0.077949
Epoch[084] Batch [1549]/[3759]	Speed: 117.762244 samples/sec	 accuracy=33.303427	 loss=3.044871	 lr=0.077938
Epoch[084] Batch [1599]/[3759]	Speed: 118.151138 samples/sec	 accuracy=33.266602	 loss=3.045914	 lr=0.077927
Epoch[084] Batch [1649]/[3759]	Speed: 117.452472 samples/sec	 accuracy=33.269886	 loss=3.046168	 lr=0.077916
Epoch[084] Batch [1699]/[3759]	Speed: 116.339896 samples/sec	 accuracy=33.283088	 loss=3.044072	 lr=0.077906
Epoch[084] Batch [1749]/[3759]	Speed: 117.285312 samples/sec	 accuracy=33.298214	 loss=3.044733	 lr=0.077895
Epoch[084] Batch [1799]/[3759]	Speed: 117.973566 samples/sec	 accuracy=33.294271	 loss=3.044834	 lr=0.077884
Epoch[084] Batch [1849]/[3759]	Speed: 116.756844 samples/sec	 accuracy=33.337838	 loss=3.042865	 lr=0.077874
Epoch[084] Batch [1899]/[3759]	Speed: 117.222779 samples/sec	 accuracy=33.309211	 loss=3.043668	 lr=0.077863
Epoch[084] Batch [1949]/[3759]	Speed: 117.262619 samples/sec	 accuracy=33.295673	 loss=3.043646	 lr=0.077852
Epoch[084] Batch [1999]/[3759]	Speed: 117.585282 samples/sec	 accuracy=33.283594	 loss=3.044674	 lr=0.077842
Epoch[084] Batch [2049]/[3759]	Speed: 117.937566 samples/sec	 accuracy=33.303354	 loss=3.044093	 lr=0.077831
Epoch[084] Batch [2099]/[3759]	Speed: 117.361130 samples/sec	 accuracy=33.290923	 loss=3.044443	 lr=0.077820
Epoch[084] Batch [2149]/[3759]	Speed: 118.597550 samples/sec	 accuracy=33.296512	 loss=3.043259	 lr=0.077809
Epoch[084] Batch [2199]/[3759]	Speed: 117.233488 samples/sec	 accuracy=33.289773	 loss=3.044543	 lr=0.077799
Epoch[084] Batch [2249]/[3759]	Speed: 117.568186 samples/sec	 accuracy=33.291667	 loss=3.044086	 lr=0.077788
Epoch[084] Batch [2299]/[3759]	Speed: 117.662892 samples/sec	 accuracy=33.273777	 loss=3.043739	 lr=0.077777
Epoch[084] Batch [2349]/[3759]	Speed: 117.154307 samples/sec	 accuracy=33.248670	 loss=3.044901	 lr=0.077767
Epoch[084] Batch [2399]/[3759]	Speed: 117.344271 samples/sec	 accuracy=33.256510	 loss=3.043978	 lr=0.077756
Epoch[084] Batch [2449]/[3759]	Speed: 117.109819 samples/sec	 accuracy=33.225128	 loss=3.045229	 lr=0.077745
Epoch[084] Batch [2499]/[3759]	Speed: 116.865770 samples/sec	 accuracy=33.247500	 loss=3.044092	 lr=0.077734
Epoch[084] Batch [2549]/[3759]	Speed: 117.268175 samples/sec	 accuracy=33.273284	 loss=3.043973	 lr=0.077724
Epoch[084] Batch [2599]/[3759]	Speed: 118.273830 samples/sec	 accuracy=33.274038	 loss=3.043849	 lr=0.077713
Epoch[084] Batch [2649]/[3759]	Speed: 117.456826 samples/sec	 accuracy=33.253538	 loss=3.045050	 lr=0.077702
Epoch[084] Batch [2699]/[3759]	Speed: 117.713420 samples/sec	 accuracy=33.254051	 loss=3.045451	 lr=0.077691
Epoch[084] Batch [2749]/[3759]	Speed: 116.390805 samples/sec	 accuracy=33.260795	 loss=3.044606	 lr=0.077681
Epoch[084] Batch [2799]/[3759]	Speed: 118.326138 samples/sec	 accuracy=33.271763	 loss=3.043902	 lr=0.077670
Epoch[084] Batch [2849]/[3759]	Speed: 117.601004 samples/sec	 accuracy=33.270833	 loss=3.043816	 lr=0.077659
Epoch[084] Batch [2899]/[3759]	Speed: 116.894878 samples/sec	 accuracy=33.267780	 loss=3.043813	 lr=0.077648
Epoch[084] Batch [2949]/[3759]	Speed: 118.140598 samples/sec	 accuracy=33.278602	 loss=3.044392	 lr=0.077638
Epoch[084] Batch [2999]/[3759]	Speed: 117.404628 samples/sec	 accuracy=33.267708	 loss=3.044663	 lr=0.077627
Epoch[084] Batch [3049]/[3759]	Speed: 117.754480 samples/sec	 accuracy=33.252561	 loss=3.044680	 lr=0.077616
Epoch[084] Batch [3099]/[3759]	Speed: 117.806644 samples/sec	 accuracy=33.240927	 loss=3.045121	 lr=0.077605
Epoch[084] Batch [3149]/[3759]	Speed: 117.712430 samples/sec	 accuracy=33.277778	 loss=3.044729	 lr=0.077595
Epoch[084] Batch [3199]/[3759]	Speed: 117.662813 samples/sec	 accuracy=33.270996	 loss=3.045471	 lr=0.077584
Epoch[084] Batch [3249]/[3759]	Speed: 117.073290 samples/sec	 accuracy=33.273558	 loss=3.045643	 lr=0.077573
Epoch[084] Batch [3299]/[3759]	Speed: 117.177331 samples/sec	 accuracy=33.271307	 loss=3.046150	 lr=0.077562
Epoch[084] Batch [3349]/[3759]	Speed: 117.616376 samples/sec	 accuracy=33.256530	 loss=3.046626	 lr=0.077552
Epoch[084] Batch [3399]/[3759]	Speed: 117.538378 samples/sec	 accuracy=33.256434	 loss=3.046299	 lr=0.077541
Epoch[084] Batch [3449]/[3759]	Speed: 117.383314 samples/sec	 accuracy=33.240489	 loss=3.046821	 lr=0.077530
Epoch[084] Batch [3499]/[3759]	Speed: 118.933114 samples/sec	 accuracy=33.218750	 loss=3.047564	 lr=0.077519
Epoch[084] Batch [3549]/[3759]	Speed: 117.498936 samples/sec	 accuracy=33.224472	 loss=3.046765	 lr=0.077509
Epoch[084] Batch [3599]/[3759]	Speed: 118.022688 samples/sec	 accuracy=33.221354	 loss=3.046904	 lr=0.077498
Epoch[084] Batch [3649]/[3759]	Speed: 117.703891 samples/sec	 accuracy=33.208048	 loss=3.047615	 lr=0.077487
Epoch[084] Batch [3699]/[3759]	Speed: 117.491641 samples/sec	 accuracy=33.213682	 loss=3.047195	 lr=0.077476
Epoch[084] Batch [3749]/[3759]	Speed: 120.508626 samples/sec	 accuracy=33.194167	 loss=3.047894	 lr=0.077465
Batch [0049]/[0303]: acc-top1=32.375000 acc-top5=59.000000
Batch [0099]/[0303]: acc-top1=31.171875 acc-top5=58.843750
Batch [0149]/[0303]: acc-top1=32.718750 acc-top5=59.583333
Batch [0199]/[0303]: acc-top1=35.679688 acc-top5=61.664063
Batch [0249]/[0303]: acc-top1=35.581250 acc-top5=62.037500
Batch [0299]/[0303]: acc-top1=35.041667 acc-top5=61.807292
[Epoch 084] training: accuracy=33.194916	 loss=3.048067
[Epoch 084] speed: 116 samples/sec	time cost: 2174.228709
[Epoch 084] validation: acc-top1=34.973185 acc-top5=61.721328 loss=3.039829
Epoch[085] Batch [0049]/[3760]	Speed: 87.509220 samples/sec	 accuracy=35.406250	 loss=2.944578	 lr=0.077453
Epoch[085] Batch [0099]/[3760]	Speed: 118.011770 samples/sec	 accuracy=34.484375	 loss=2.978831	 lr=0.077442
Epoch[085] Batch [0149]/[3760]	Speed: 117.459965 samples/sec	 accuracy=34.802083	 loss=2.948843	 lr=0.077431
Epoch[085] Batch [0199]/[3760]	Speed: 117.588189 samples/sec	 accuracy=34.382812	 loss=2.979713	 lr=0.077420
Epoch[085] Batch [0249]/[3760]	Speed: 117.353197 samples/sec	 accuracy=34.068750	 loss=2.994519	 lr=0.077410
Epoch[085] Batch [0299]/[3760]	Speed: 118.187291 samples/sec	 accuracy=34.057292	 loss=2.990169	 lr=0.077399
Epoch[085] Batch [0349]/[3760]	Speed: 116.089546 samples/sec	 accuracy=34.111607	 loss=2.992152	 lr=0.077388
Epoch[085] Batch [0399]/[3760]	Speed: 117.282779 samples/sec	 accuracy=34.023438	 loss=2.998512	 lr=0.077377
Epoch[085] Batch [0449]/[3760]	Speed: 118.264528 samples/sec	 accuracy=34.253472	 loss=2.991261	 lr=0.077366
Epoch[085] Batch [0499]/[3760]	Speed: 117.254399 samples/sec	 accuracy=34.340625	 loss=2.985255	 lr=0.077356
Epoch[085] Batch [0549]/[3760]	Speed: 117.750938 samples/sec	 accuracy=34.255682	 loss=2.992054	 lr=0.077345
Epoch[085] Batch [0599]/[3760]	Speed: 117.459105 samples/sec	 accuracy=34.299479	 loss=2.990933	 lr=0.077334
Epoch[085] Batch [0649]/[3760]	Speed: 117.351414 samples/sec	 accuracy=34.228365	 loss=2.996177	 lr=0.077323
Epoch[085] Batch [0699]/[3760]	Speed: 117.233454 samples/sec	 accuracy=34.220982	 loss=2.998440	 lr=0.077312
Epoch[085] Batch [0749]/[3760]	Speed: 117.159710 samples/sec	 accuracy=34.227083	 loss=2.998042	 lr=0.077302
Epoch[085] Batch [0799]/[3760]	Speed: 117.197834 samples/sec	 accuracy=34.255859	 loss=2.996430	 lr=0.077291
Epoch[085] Batch [0849]/[3760]	Speed: 116.781273 samples/sec	 accuracy=34.148897	 loss=3.001540	 lr=0.077280
Epoch[085] Batch [0899]/[3760]	Speed: 117.059625 samples/sec	 accuracy=34.128472	 loss=3.001172	 lr=0.077269
Epoch[085] Batch [0949]/[3760]	Speed: 117.489096 samples/sec	 accuracy=34.123355	 loss=3.002062	 lr=0.077258
Epoch[085] Batch [0999]/[3760]	Speed: 116.860616 samples/sec	 accuracy=34.104688	 loss=3.003307	 lr=0.077248
Epoch[085] Batch [1049]/[3760]	Speed: 117.312089 samples/sec	 accuracy=34.080357	 loss=3.003517	 lr=0.077237
Epoch[085] Batch [1099]/[3760]	Speed: 117.294545 samples/sec	 accuracy=34.045455	 loss=3.003302	 lr=0.077226
Epoch[085] Batch [1149]/[3760]	Speed: 117.475671 samples/sec	 accuracy=34.057065	 loss=3.001390	 lr=0.077215
Epoch[085] Batch [1199]/[3760]	Speed: 118.071722 samples/sec	 accuracy=34.007812	 loss=3.003664	 lr=0.077204
Epoch[085] Batch [1249]/[3760]	Speed: 117.722263 samples/sec	 accuracy=34.010000	 loss=3.003461	 lr=0.077194
Epoch[085] Batch [1299]/[3760]	Speed: 117.076261 samples/sec	 accuracy=33.963942	 loss=3.003857	 lr=0.077183
Epoch[085] Batch [1349]/[3760]	Speed: 117.460599 samples/sec	 accuracy=33.965278	 loss=3.005266	 lr=0.077172
Epoch[085] Batch [1399]/[3760]	Speed: 117.655445 samples/sec	 accuracy=33.936384	 loss=3.007455	 lr=0.077161
Epoch[085] Batch [1449]/[3760]	Speed: 117.335614 samples/sec	 accuracy=33.901940	 loss=3.007898	 lr=0.077150
Epoch[085] Batch [1499]/[3760]	Speed: 117.318989 samples/sec	 accuracy=33.900000	 loss=3.008575	 lr=0.077139
Epoch[085] Batch [1549]/[3760]	Speed: 117.373908 samples/sec	 accuracy=33.820565	 loss=3.012213	 lr=0.077129
Epoch[085] Batch [1599]/[3760]	Speed: 118.044730 samples/sec	 accuracy=33.812500	 loss=3.013260	 lr=0.077118
Epoch[085] Batch [1649]/[3760]	Speed: 117.706843 samples/sec	 accuracy=33.805871	 loss=3.015447	 lr=0.077107
Epoch[085] Batch [1699]/[3760]	Speed: 117.281486 samples/sec	 accuracy=33.828125	 loss=3.016315	 lr=0.077096
Epoch[085] Batch [1749]/[3760]	Speed: 117.354959 samples/sec	 accuracy=33.855357	 loss=3.016014	 lr=0.077085
Epoch[085] Batch [1799]/[3760]	Speed: 117.260794 samples/sec	 accuracy=33.850694	 loss=3.015778	 lr=0.077074
Epoch[085] Batch [1849]/[3760]	Speed: 117.670591 samples/sec	 accuracy=33.843750	 loss=3.015808	 lr=0.077064
Epoch[085] Batch [1899]/[3760]	Speed: 116.874738 samples/sec	 accuracy=33.853618	 loss=3.015894	 lr=0.077053
Epoch[085] Batch [1949]/[3760]	Speed: 117.531679 samples/sec	 accuracy=33.815705	 loss=3.018109	 lr=0.077042
Epoch[085] Batch [1999]/[3760]	Speed: 116.705119 samples/sec	 accuracy=33.789062	 loss=3.019131	 lr=0.077031
Epoch[085] Batch [2049]/[3760]	Speed: 117.636384 samples/sec	 accuracy=33.803354	 loss=3.018965	 lr=0.077020
Epoch[085] Batch [2099]/[3760]	Speed: 117.701591 samples/sec	 accuracy=33.787946	 loss=3.020653	 lr=0.077009
Epoch[085] Batch [2149]/[3760]	Speed: 117.512374 samples/sec	 accuracy=33.787791	 loss=3.019601	 lr=0.076998
Epoch[085] Batch [2199]/[3760]	Speed: 117.431784 samples/sec	 accuracy=33.789773	 loss=3.019007	 lr=0.076988
Epoch[085] Batch [2249]/[3760]	Speed: 118.176645 samples/sec	 accuracy=33.753472	 loss=3.020153	 lr=0.076977
Epoch[085] Batch [2299]/[3760]	Speed: 117.785283 samples/sec	 accuracy=33.790761	 loss=3.018872	 lr=0.076966
Epoch[085] Batch [2349]/[3760]	Speed: 116.974409 samples/sec	 accuracy=33.769947	 loss=3.020401	 lr=0.076955
Epoch[085] Batch [2399]/[3760]	Speed: 117.485316 samples/sec	 accuracy=33.750651	 loss=3.020368	 lr=0.076944
Epoch[085] Batch [2449]/[3760]	Speed: 117.553388 samples/sec	 accuracy=33.744898	 loss=3.020463	 lr=0.076933
Epoch[085] Batch [2499]/[3760]	Speed: 117.640949 samples/sec	 accuracy=33.748125	 loss=3.020673	 lr=0.076922
Epoch[085] Batch [2549]/[3760]	Speed: 117.105775 samples/sec	 accuracy=33.726716	 loss=3.021404	 lr=0.076912
Epoch[085] Batch [2599]/[3760]	Speed: 118.360518 samples/sec	 accuracy=33.716346	 loss=3.022167	 lr=0.076901
Epoch[085] Batch [2649]/[3760]	Speed: 116.723325 samples/sec	 accuracy=33.727005	 loss=3.021773	 lr=0.076890
Epoch[085] Batch [2699]/[3760]	Speed: 117.744260 samples/sec	 accuracy=33.733796	 loss=3.021593	 lr=0.076879
Epoch[085] Batch [2749]/[3760]	Speed: 116.627050 samples/sec	 accuracy=33.739773	 loss=3.022156	 lr=0.076868
Epoch[085] Batch [2799]/[3760]	Speed: 116.775623 samples/sec	 accuracy=33.756138	 loss=3.022035	 lr=0.076857
Epoch[085] Batch [2849]/[3760]	Speed: 117.835833 samples/sec	 accuracy=33.747259	 loss=3.022189	 lr=0.076846
Epoch[085] Batch [2899]/[3760]	Speed: 117.099996 samples/sec	 accuracy=33.750539	 loss=3.022385	 lr=0.076835
Epoch[085] Batch [2949]/[3760]	Speed: 117.205910 samples/sec	 accuracy=33.726695	 loss=3.023019	 lr=0.076824
Epoch[085] Batch [2999]/[3760]	Speed: 117.475335 samples/sec	 accuracy=33.723958	 loss=3.023517	 lr=0.076814
Epoch[085] Batch [3049]/[3760]	Speed: 118.036273 samples/sec	 accuracy=33.731557	 loss=3.023518	 lr=0.076803
Epoch[085] Batch [3099]/[3760]	Speed: 117.204463 samples/sec	 accuracy=33.714214	 loss=3.024596	 lr=0.076792
Epoch[085] Batch [3149]/[3760]	Speed: 118.178348 samples/sec	 accuracy=33.688492	 loss=3.025915	 lr=0.076781
Epoch[085] Batch [3199]/[3760]	Speed: 117.970146 samples/sec	 accuracy=33.676758	 loss=3.027101	 lr=0.076770
Epoch[085] Batch [3249]/[3760]	Speed: 116.842766 samples/sec	 accuracy=33.666827	 loss=3.028883	 lr=0.076759
Epoch[085] Batch [3299]/[3760]	Speed: 117.482082 samples/sec	 accuracy=33.651515	 loss=3.029017	 lr=0.076748
Epoch[085] Batch [3349]/[3760]	Speed: 117.297449 samples/sec	 accuracy=33.636194	 loss=3.030242	 lr=0.076737
Epoch[085] Batch [3399]/[3760]	Speed: 116.460666 samples/sec	 accuracy=33.631893	 loss=3.030129	 lr=0.076726
Epoch[085] Batch [3449]/[3760]	Speed: 117.889464 samples/sec	 accuracy=33.626812	 loss=3.030515	 lr=0.076716
Epoch[085] Batch [3499]/[3760]	Speed: 117.299215 samples/sec	 accuracy=33.625893	 loss=3.030333	 lr=0.076705
Epoch[085] Batch [3549]/[3760]	Speed: 117.238147 samples/sec	 accuracy=33.635563	 loss=3.030207	 lr=0.076694
Epoch[085] Batch [3599]/[3760]	Speed: 117.469313 samples/sec	 accuracy=33.646701	 loss=3.030035	 lr=0.076683
Epoch[085] Batch [3649]/[3760]	Speed: 118.316462 samples/sec	 accuracy=33.639555	 loss=3.030749	 lr=0.076672
Epoch[085] Batch [3699]/[3760]	Speed: 117.664758 samples/sec	 accuracy=33.630912	 loss=3.030890	 lr=0.076661
Epoch[085] Batch [3749]/[3760]	Speed: 120.822972 samples/sec	 accuracy=33.613750	 loss=3.031300	 lr=0.076650
Batch [0049]/[0303]: acc-top1=32.593750 acc-top5=59.468750
Batch [0099]/[0303]: acc-top1=29.625000 acc-top5=56.062500
Batch [0149]/[0303]: acc-top1=30.583333 acc-top5=57.427083
Batch [0199]/[0303]: acc-top1=33.554688 acc-top5=59.703125
Batch [0249]/[0303]: acc-top1=32.918750 acc-top5=59.418750
Batch [0299]/[0303]: acc-top1=32.708333 acc-top5=59.656250
[Epoch 085] training: accuracy=33.610372	 loss=3.031668
[Epoch 085] speed: 116 samples/sec	time cost: 2175.065951
[Epoch 085] validation: acc-top1=32.647483 acc-top5=59.622525 loss=3.152413
Epoch[086] Batch [0049]/[3760]	Speed: 86.582152 samples/sec	 accuracy=33.375000	 loss=3.046019	 lr=0.076637
Epoch[086] Batch [0099]/[3760]	Speed: 117.603447 samples/sec	 accuracy=33.531250	 loss=3.042112	 lr=0.076626
Epoch[086] Batch [0149]/[3760]	Speed: 117.186233 samples/sec	 accuracy=33.468750	 loss=3.032300	 lr=0.076615
Epoch[086] Batch [0199]/[3760]	Speed: 117.225097 samples/sec	 accuracy=33.835938	 loss=3.014717	 lr=0.076604
Epoch[086] Batch [0249]/[3760]	Speed: 117.288759 samples/sec	 accuracy=33.787500	 loss=3.011526	 lr=0.076593
Epoch[086] Batch [0299]/[3760]	Speed: 117.800561 samples/sec	 accuracy=33.911458	 loss=3.009635	 lr=0.076582
Epoch[086] Batch [0349]/[3760]	Speed: 117.403980 samples/sec	 accuracy=33.656250	 loss=3.018402	 lr=0.076571
Epoch[086] Batch [0399]/[3760]	Speed: 116.468468 samples/sec	 accuracy=33.535156	 loss=3.015066	 lr=0.076561
Epoch[086] Batch [0449]/[3760]	Speed: 117.584547 samples/sec	 accuracy=33.635417	 loss=3.007084	 lr=0.076550
Epoch[086] Batch [0499]/[3760]	Speed: 117.459053 samples/sec	 accuracy=33.596875	 loss=3.010415	 lr=0.076539
Epoch[086] Batch [0549]/[3760]	Speed: 118.144207 samples/sec	 accuracy=33.650568	 loss=3.011948	 lr=0.076528
Epoch[086] Batch [0599]/[3760]	Speed: 117.550042 samples/sec	 accuracy=33.648438	 loss=3.011517	 lr=0.076517
Epoch[086] Batch [0649]/[3760]	Speed: 117.228768 samples/sec	 accuracy=33.670673	 loss=3.011409	 lr=0.076506
Epoch[086] Batch [0699]/[3760]	Speed: 116.695797 samples/sec	 accuracy=33.642857	 loss=3.011947	 lr=0.076495
Epoch[086] Batch [0749]/[3760]	Speed: 117.423328 samples/sec	 accuracy=33.529167	 loss=3.015531	 lr=0.076484
Epoch[086] Batch [0799]/[3760]	Speed: 117.655559 samples/sec	 accuracy=33.527344	 loss=3.015854	 lr=0.076473
Epoch[086] Batch [0849]/[3760]	Speed: 117.185887 samples/sec	 accuracy=33.571691	 loss=3.013134	 lr=0.076462
Epoch[086] Batch [0899]/[3760]	Speed: 117.384004 samples/sec	 accuracy=33.600694	 loss=3.011786	 lr=0.076451
Epoch[086] Batch [0949]/[3760]	Speed: 117.220423 samples/sec	 accuracy=33.521382	 loss=3.014493	 lr=0.076440
Epoch[086] Batch [0999]/[3760]	Speed: 116.889071 samples/sec	 accuracy=33.593750	 loss=3.014351	 lr=0.076429
Epoch[086] Batch [1049]/[3760]	Speed: 117.025150 samples/sec	 accuracy=33.572917	 loss=3.016516	 lr=0.076418
Epoch[086] Batch [1099]/[3760]	Speed: 117.883884 samples/sec	 accuracy=33.609375	 loss=3.016110	 lr=0.076407
Epoch[086] Batch [1149]/[3760]	Speed: 117.637563 samples/sec	 accuracy=33.638587	 loss=3.017174	 lr=0.076396
Epoch[086] Batch [1199]/[3760]	Speed: 116.761087 samples/sec	 accuracy=33.653646	 loss=3.016488	 lr=0.076386
Epoch[086] Batch [1249]/[3760]	Speed: 117.458446 samples/sec	 accuracy=33.556250	 loss=3.020246	 lr=0.076375
Epoch[086] Batch [1299]/[3760]	Speed: 117.145774 samples/sec	 accuracy=33.528846	 loss=3.022437	 lr=0.076364
Epoch[086] Batch [1349]/[3760]	Speed: 117.332051 samples/sec	 accuracy=33.532407	 loss=3.025308	 lr=0.076353
Epoch[086] Batch [1399]/[3760]	Speed: 117.896063 samples/sec	 accuracy=33.515625	 loss=3.027908	 lr=0.076342
Epoch[086] Batch [1449]/[3760]	Speed: 116.930673 samples/sec	 accuracy=33.524784	 loss=3.028031	 lr=0.076331
Epoch[086] Batch [1499]/[3760]	Speed: 117.636482 samples/sec	 accuracy=33.590625	 loss=3.027391	 lr=0.076320
Epoch[086] Batch [1549]/[3760]	Speed: 117.415529 samples/sec	 accuracy=33.597782	 loss=3.027446	 lr=0.076309
Epoch[086] Batch [1599]/[3760]	Speed: 117.126083 samples/sec	 accuracy=33.650391	 loss=3.026218	 lr=0.076298
Epoch[086] Batch [1649]/[3760]	Speed: 117.352238 samples/sec	 accuracy=33.648674	 loss=3.027509	 lr=0.076287
Epoch[086] Batch [1699]/[3760]	Speed: 118.008485 samples/sec	 accuracy=33.635110	 loss=3.028650	 lr=0.076276
Epoch[086] Batch [1749]/[3760]	Speed: 117.202028 samples/sec	 accuracy=33.618750	 loss=3.028721	 lr=0.076265
Epoch[086] Batch [1799]/[3760]	Speed: 116.811293 samples/sec	 accuracy=33.628472	 loss=3.028777	 lr=0.076254
Epoch[086] Batch [1849]/[3760]	Speed: 116.869363 samples/sec	 accuracy=33.643581	 loss=3.028345	 lr=0.076243
Epoch[086] Batch [1899]/[3760]	Speed: 117.229627 samples/sec	 accuracy=33.688322	 loss=3.027382	 lr=0.076232
Epoch[086] Batch [1949]/[3760]	Speed: 116.861632 samples/sec	 accuracy=33.716346	 loss=3.027245	 lr=0.076221
Epoch[086] Batch [1999]/[3760]	Speed: 117.406945 samples/sec	 accuracy=33.700781	 loss=3.027882	 lr=0.076210
Epoch[086] Batch [2049]/[3760]	Speed: 118.046628 samples/sec	 accuracy=33.705793	 loss=3.028150	 lr=0.076199
Epoch[086] Batch [2099]/[3760]	Speed: 116.690053 samples/sec	 accuracy=33.709077	 loss=3.026687	 lr=0.076188
Epoch[086] Batch [2149]/[3760]	Speed: 117.679822 samples/sec	 accuracy=33.697674	 loss=3.026825	 lr=0.076177
Epoch[086] Batch [2199]/[3760]	Speed: 117.285424 samples/sec	 accuracy=33.659091	 loss=3.028155	 lr=0.076166
Epoch[086] Batch [2249]/[3760]	Speed: 117.691991 samples/sec	 accuracy=33.674306	 loss=3.026097	 lr=0.076155
Epoch[086] Batch [2299]/[3760]	Speed: 116.600988 samples/sec	 accuracy=33.673913	 loss=3.026048	 lr=0.076144
Epoch[086] Batch [2349]/[3760]	Speed: 117.346107 samples/sec	 accuracy=33.660904	 loss=3.026042	 lr=0.076133
Epoch[086] Batch [2399]/[3760]	Speed: 117.939246 samples/sec	 accuracy=33.652995	 loss=3.026174	 lr=0.076122
Epoch[086] Batch [2449]/[3760]	Speed: 117.483810 samples/sec	 accuracy=33.675383	 loss=3.024892	 lr=0.076111
Epoch[086] Batch [2499]/[3760]	Speed: 117.337833 samples/sec	 accuracy=33.653750	 loss=3.025455	 lr=0.076100
Epoch[086] Batch [2549]/[3760]	Speed: 117.447317 samples/sec	 accuracy=33.649510	 loss=3.025277	 lr=0.076089
Epoch[086] Batch [2599]/[3760]	Speed: 117.606128 samples/sec	 accuracy=33.620192	 loss=3.025005	 lr=0.076078
Epoch[086] Batch [2649]/[3760]	Speed: 117.088359 samples/sec	 accuracy=33.588443	 loss=3.025814	 lr=0.076067
Epoch[086] Batch [2699]/[3760]	Speed: 117.250488 samples/sec	 accuracy=33.577546	 loss=3.026457	 lr=0.076056
Epoch[086] Batch [2749]/[3760]	Speed: 117.187124 samples/sec	 accuracy=33.567045	 loss=3.025924	 lr=0.076045
Epoch[086] Batch [2799]/[3760]	Speed: 117.343255 samples/sec	 accuracy=33.565290	 loss=3.026845	 lr=0.076034
Epoch[086] Batch [2849]/[3760]	Speed: 117.572327 samples/sec	 accuracy=33.540570	 loss=3.027261	 lr=0.076023
Epoch[086] Batch [2899]/[3760]	Speed: 116.712943 samples/sec	 accuracy=33.536638	 loss=3.027219	 lr=0.076012
Epoch[086] Batch [2949]/[3760]	Speed: 117.745214 samples/sec	 accuracy=33.551907	 loss=3.027026	 lr=0.076001
Epoch[086] Batch [2999]/[3760]	Speed: 117.050770 samples/sec	 accuracy=33.538021	 loss=3.027846	 lr=0.075990
Epoch[086] Batch [3049]/[3760]	Speed: 117.174275 samples/sec	 accuracy=33.563012	 loss=3.027008	 lr=0.075979
Epoch[086] Batch [3099]/[3760]	Speed: 117.142780 samples/sec	 accuracy=33.555948	 loss=3.027650	 lr=0.075968
Epoch[086] Batch [3149]/[3760]	Speed: 117.539838 samples/sec	 accuracy=33.546627	 loss=3.028355	 lr=0.075957
Epoch[086] Batch [3199]/[3760]	Speed: 118.100322 samples/sec	 accuracy=33.560547	 loss=3.027287	 lr=0.075946
Epoch[086] Batch [3249]/[3760]	Speed: 117.530899 samples/sec	 accuracy=33.559615	 loss=3.027580	 lr=0.075935
Epoch[086] Batch [3299]/[3760]	Speed: 117.167357 samples/sec	 accuracy=33.542614	 loss=3.028292	 lr=0.075924
Epoch[086] Batch [3349]/[3760]	Speed: 116.624369 samples/sec	 accuracy=33.563433	 loss=3.027540	 lr=0.075913
Epoch[086] Batch [3399]/[3760]	Speed: 116.893376 samples/sec	 accuracy=33.550092	 loss=3.027925	 lr=0.075902
Epoch[086] Batch [3449]/[3760]	Speed: 117.326903 samples/sec	 accuracy=33.542120	 loss=3.028629	 lr=0.075891
Epoch[086] Batch [3499]/[3760]	Speed: 116.303560 samples/sec	 accuracy=33.531250	 loss=3.028687	 lr=0.075880
Epoch[086] Batch [3549]/[3760]	Speed: 117.018612 samples/sec	 accuracy=33.529489	 loss=3.029279	 lr=0.075869
Epoch[086] Batch [3599]/[3760]	Speed: 117.138405 samples/sec	 accuracy=33.523003	 loss=3.029668	 lr=0.075858
Epoch[086] Batch [3649]/[3760]	Speed: 117.358711 samples/sec	 accuracy=33.509418	 loss=3.030569	 lr=0.075847
Epoch[086] Batch [3699]/[3760]	Speed: 117.001828 samples/sec	 accuracy=33.500422	 loss=3.031481	 lr=0.075836
Epoch[086] Batch [3749]/[3760]	Speed: 120.253230 samples/sec	 accuracy=33.504583	 loss=3.032315	 lr=0.075825
Batch [0049]/[0303]: acc-top1=34.562500 acc-top5=61.343750
Batch [0099]/[0303]: acc-top1=32.328125 acc-top5=59.468750
Batch [0149]/[0303]: acc-top1=34.062500 acc-top5=60.739583
Batch [0199]/[0303]: acc-top1=36.812500 acc-top5=62.976562
Batch [0249]/[0303]: acc-top1=35.450000 acc-top5=62.225000
Batch [0299]/[0303]: acc-top1=35.260417 acc-top5=62.276042
[Epoch 086] training: accuracy=33.508976	 loss=3.032366
[Epoch 086] speed: 116 samples/sec	time cost: 2178.691269
[Epoch 086] validation: acc-top1=35.189769 acc-top5=62.262789 loss=3.022544
Epoch[087] Batch [0049]/[3759]	Speed: 86.470082 samples/sec	 accuracy=33.750000	 loss=3.001150	 lr=0.075811
Epoch[087] Batch [0099]/[3759]	Speed: 117.523532 samples/sec	 accuracy=34.015625	 loss=3.028166	 lr=0.075800
Epoch[087] Batch [0149]/[3759]	Speed: 117.179375 samples/sec	 accuracy=33.739583	 loss=3.011649	 lr=0.075789
Epoch[087] Batch [0199]/[3759]	Speed: 117.419994 samples/sec	 accuracy=34.070312	 loss=3.003245	 lr=0.075778
Epoch[087] Batch [0249]/[3759]	Speed: 117.055770 samples/sec	 accuracy=34.118750	 loss=3.001084	 lr=0.075767
Epoch[087] Batch [0299]/[3759]	Speed: 117.314654 samples/sec	 accuracy=34.265625	 loss=3.002106	 lr=0.075756
Epoch[087] Batch [0349]/[3759]	Speed: 116.465249 samples/sec	 accuracy=34.218750	 loss=2.998660	 lr=0.075745
Epoch[087] Batch [0399]/[3759]	Speed: 117.218235 samples/sec	 accuracy=34.269531	 loss=2.997958	 lr=0.075734
Epoch[087] Batch [0449]/[3759]	Speed: 117.367584 samples/sec	 accuracy=34.284722	 loss=2.996606	 lr=0.075723
Epoch[087] Batch [0499]/[3759]	Speed: 117.515635 samples/sec	 accuracy=34.156250	 loss=3.000143	 lr=0.075712
Epoch[087] Batch [0549]/[3759]	Speed: 116.164494 samples/sec	 accuracy=34.150568	 loss=3.004355	 lr=0.075701
Epoch[087] Batch [0599]/[3759]	Speed: 116.857881 samples/sec	 accuracy=34.148438	 loss=3.003054	 lr=0.075690
Epoch[087] Batch [0649]/[3759]	Speed: 117.521754 samples/sec	 accuracy=34.158654	 loss=3.004198	 lr=0.075679
Epoch[087] Batch [0699]/[3759]	Speed: 117.099045 samples/sec	 accuracy=34.082589	 loss=3.006980	 lr=0.075668
Epoch[087] Batch [0749]/[3759]	Speed: 116.729244 samples/sec	 accuracy=34.041667	 loss=3.008615	 lr=0.075656
Epoch[087] Batch [0799]/[3759]	Speed: 118.182323 samples/sec	 accuracy=34.001953	 loss=3.005656	 lr=0.075645
Epoch[087] Batch [0849]/[3759]	Speed: 117.124669 samples/sec	 accuracy=34.022059	 loss=3.004111	 lr=0.075634
Epoch[087] Batch [0899]/[3759]	Speed: 117.448311 samples/sec	 accuracy=34.019097	 loss=3.002788	 lr=0.075623
Epoch[087] Batch [0949]/[3759]	Speed: 117.380349 samples/sec	 accuracy=34.008224	 loss=3.004809	 lr=0.075612
Epoch[087] Batch [0999]/[3759]	Speed: 118.183008 samples/sec	 accuracy=34.050000	 loss=3.004154	 lr=0.075601
Epoch[087] Batch [1049]/[3759]	Speed: 117.492048 samples/sec	 accuracy=33.988095	 loss=3.006215	 lr=0.075590
Epoch[087] Batch [1099]/[3759]	Speed: 117.031146 samples/sec	 accuracy=33.930398	 loss=3.008236	 lr=0.075579
Epoch[087] Batch [1149]/[3759]	Speed: 116.652559 samples/sec	 accuracy=33.967391	 loss=3.007366	 lr=0.075568
Epoch[087] Batch [1199]/[3759]	Speed: 117.815157 samples/sec	 accuracy=33.928385	 loss=3.007243	 lr=0.075557
Epoch[087] Batch [1249]/[3759]	Speed: 117.811609 samples/sec	 accuracy=33.906250	 loss=3.007430	 lr=0.075546
Epoch[087] Batch [1299]/[3759]	Speed: 116.696311 samples/sec	 accuracy=33.913462	 loss=3.007808	 lr=0.075535
Epoch[087] Batch [1349]/[3759]	Speed: 116.340750 samples/sec	 accuracy=33.903935	 loss=3.008649	 lr=0.075523
Epoch[087] Batch [1399]/[3759]	Speed: 116.790618 samples/sec	 accuracy=33.850446	 loss=3.011945	 lr=0.075512
Epoch[087] Batch [1449]/[3759]	Speed: 117.219692 samples/sec	 accuracy=33.822198	 loss=3.012319	 lr=0.075501
Epoch[087] Batch [1499]/[3759]	Speed: 117.452059 samples/sec	 accuracy=33.779167	 loss=3.014197	 lr=0.075490
Epoch[087] Batch [1549]/[3759]	Speed: 116.962053 samples/sec	 accuracy=33.763105	 loss=3.015482	 lr=0.075479
Epoch[087] Batch [1599]/[3759]	Speed: 116.613062 samples/sec	 accuracy=33.726562	 loss=3.016130	 lr=0.075468
Epoch[087] Batch [1649]/[3759]	Speed: 117.630752 samples/sec	 accuracy=33.740530	 loss=3.017268	 lr=0.075457
Epoch[087] Batch [1699]/[3759]	Speed: 117.711099 samples/sec	 accuracy=33.765625	 loss=3.016961	 lr=0.075446
Epoch[087] Batch [1749]/[3759]	Speed: 117.104952 samples/sec	 accuracy=33.770536	 loss=3.016525	 lr=0.075435
Epoch[087] Batch [1799]/[3759]	Speed: 116.958289 samples/sec	 accuracy=33.776910	 loss=3.017233	 lr=0.075424
Epoch[087] Batch [1849]/[3759]	Speed: 117.745026 samples/sec	 accuracy=33.821791	 loss=3.015091	 lr=0.075412
Epoch[087] Batch [1899]/[3759]	Speed: 117.484692 samples/sec	 accuracy=33.828947	 loss=3.014410	 lr=0.075401
Epoch[087] Batch [1949]/[3759]	Speed: 116.883927 samples/sec	 accuracy=33.848558	 loss=3.013804	 lr=0.075390
Epoch[087] Batch [1999]/[3759]	Speed: 116.977877 samples/sec	 accuracy=33.860937	 loss=3.013850	 lr=0.075379
Epoch[087] Batch [2049]/[3759]	Speed: 118.575344 samples/sec	 accuracy=33.832317	 loss=3.014308	 lr=0.075368
Epoch[087] Batch [2099]/[3759]	Speed: 117.063060 samples/sec	 accuracy=33.805060	 loss=3.015196	 lr=0.075357
Epoch[087] Batch [2149]/[3759]	Speed: 116.801280 samples/sec	 accuracy=33.819767	 loss=3.015006	 lr=0.075346
Epoch[087] Batch [2199]/[3759]	Speed: 117.815716 samples/sec	 accuracy=33.808949	 loss=3.016188	 lr=0.075335
Epoch[087] Batch [2249]/[3759]	Speed: 117.012577 samples/sec	 accuracy=33.786111	 loss=3.017666	 lr=0.075324
Epoch[087] Batch [2299]/[3759]	Speed: 117.897821 samples/sec	 accuracy=33.775136	 loss=3.018358	 lr=0.075312
Epoch[087] Batch [2349]/[3759]	Speed: 117.741321 samples/sec	 accuracy=33.796543	 loss=3.017448	 lr=0.075301
Epoch[087] Batch [2399]/[3759]	Speed: 117.120191 samples/sec	 accuracy=33.805339	 loss=3.016707	 lr=0.075290
Epoch[087] Batch [2449]/[3759]	Speed: 117.420805 samples/sec	 accuracy=33.813138	 loss=3.016871	 lr=0.075279
Epoch[087] Batch [2499]/[3759]	Speed: 117.574182 samples/sec	 accuracy=33.784375	 loss=3.018089	 lr=0.075268
Epoch[087] Batch [2549]/[3759]	Speed: 117.530481 samples/sec	 accuracy=33.787377	 loss=3.016857	 lr=0.075257
Epoch[087] Batch [2599]/[3759]	Speed: 116.623469 samples/sec	 accuracy=33.754808	 loss=3.018666	 lr=0.075246
Epoch[087] Batch [2649]/[3759]	Speed: 118.311256 samples/sec	 accuracy=33.725236	 loss=3.019764	 lr=0.075235
Epoch[087] Batch [2699]/[3759]	Speed: 117.388143 samples/sec	 accuracy=33.704861	 loss=3.020712	 lr=0.075223
Epoch[087] Batch [2749]/[3759]	Speed: 117.101084 samples/sec	 accuracy=33.710227	 loss=3.019945	 lr=0.075212
Epoch[087] Batch [2799]/[3759]	Speed: 116.516901 samples/sec	 accuracy=33.708147	 loss=3.019619	 lr=0.075201
Epoch[087] Batch [2849]/[3759]	Speed: 117.580818 samples/sec	 accuracy=33.706689	 loss=3.020092	 lr=0.075190
Epoch[087] Batch [2899]/[3759]	Speed: 117.809143 samples/sec	 accuracy=33.698276	 loss=3.020687	 lr=0.075179
Epoch[087] Batch [2949]/[3759]	Speed: 116.315030 samples/sec	 accuracy=33.701801	 loss=3.020672	 lr=0.075168
Epoch[087] Batch [2999]/[3759]	Speed: 117.648219 samples/sec	 accuracy=33.709896	 loss=3.021635	 lr=0.075157
Epoch[087] Batch [3049]/[3759]	Speed: 116.865594 samples/sec	 accuracy=33.727971	 loss=3.020410	 lr=0.075145
Epoch[087] Batch [3099]/[3759]	Speed: 117.854410 samples/sec	 accuracy=33.742944	 loss=3.020262	 lr=0.075134
Epoch[087] Batch [3149]/[3759]	Speed: 117.404364 samples/sec	 accuracy=33.741071	 loss=3.019721	 lr=0.075123
Epoch[087] Batch [3199]/[3759]	Speed: 117.374857 samples/sec	 accuracy=33.741699	 loss=3.020096	 lr=0.075112
Epoch[087] Batch [3249]/[3759]	Speed: 117.024130 samples/sec	 accuracy=33.752404	 loss=3.019623	 lr=0.075101
Epoch[087] Batch [3299]/[3759]	Speed: 117.914579 samples/sec	 accuracy=33.736742	 loss=3.019919	 lr=0.075090
Epoch[087] Batch [3349]/[3759]	Speed: 117.588930 samples/sec	 accuracy=33.736940	 loss=3.019787	 lr=0.075079
Epoch[087] Batch [3399]/[3759]	Speed: 116.431398 samples/sec	 accuracy=33.723346	 loss=3.020332	 lr=0.075067
Epoch[087] Batch [3449]/[3759]	Speed: 118.012160 samples/sec	 accuracy=33.718750	 loss=3.020571	 lr=0.075056
Epoch[087] Batch [3499]/[3759]	Speed: 117.462361 samples/sec	 accuracy=33.735714	 loss=3.019877	 lr=0.075045
Epoch[087] Batch [3549]/[3759]	Speed: 117.879644 samples/sec	 accuracy=33.736796	 loss=3.020117	 lr=0.075034
Epoch[087] Batch [3599]/[3759]	Speed: 118.814418 samples/sec	 accuracy=33.746094	 loss=3.019882	 lr=0.075023
Epoch[087] Batch [3649]/[3759]	Speed: 116.788521 samples/sec	 accuracy=33.739726	 loss=3.019750	 lr=0.075012
Epoch[087] Batch [3699]/[3759]	Speed: 117.830079 samples/sec	 accuracy=33.743666	 loss=3.019159	 lr=0.075000
Epoch[087] Batch [3749]/[3759]	Speed: 119.460331 samples/sec	 accuracy=33.765833	 loss=3.018096	 lr=0.074989
Batch [0049]/[0303]: acc-top1=29.031250 acc-top5=56.531250
Batch [0099]/[0303]: acc-top1=27.890625 acc-top5=55.312500
Batch [0149]/[0303]: acc-top1=29.802083 acc-top5=57.187500
Batch [0199]/[0303]: acc-top1=33.453125 acc-top5=59.734375
Batch [0249]/[0303]: acc-top1=32.975000 acc-top5=59.306250
Batch [0299]/[0303]: acc-top1=32.411458 acc-top5=59.322917
[Epoch 087] training: accuracy=33.763135	 loss=3.018301
[Epoch 087] speed: 116 samples/sec	time cost: 2179.201158
[Epoch 087] validation: acc-top1=32.513408 acc-top5=59.375000 loss=3.187811
Epoch[088] Batch [0049]/[3760]	Speed: 87.698170 samples/sec	 accuracy=33.750000	 loss=3.001086	 lr=0.074976
Epoch[088] Batch [0099]/[3760]	Speed: 118.009099 samples/sec	 accuracy=33.968750	 loss=2.982076	 lr=0.074965
Epoch[088] Batch [0149]/[3760]	Speed: 116.877808 samples/sec	 accuracy=33.947917	 loss=2.988145	 lr=0.074954
Epoch[088] Batch [0199]/[3760]	Speed: 117.237142 samples/sec	 accuracy=34.375000	 loss=2.991556	 lr=0.074942
Epoch[088] Batch [0249]/[3760]	Speed: 117.781105 samples/sec	 accuracy=34.425000	 loss=2.985296	 lr=0.074931
Epoch[088] Batch [0299]/[3760]	Speed: 117.214982 samples/sec	 accuracy=34.468750	 loss=2.984322	 lr=0.074920
Epoch[088] Batch [0349]/[3760]	Speed: 117.372535 samples/sec	 accuracy=34.272321	 loss=2.982450	 lr=0.074909
Epoch[088] Batch [0399]/[3760]	Speed: 117.117538 samples/sec	 accuracy=34.191406	 loss=2.985612	 lr=0.074898
Epoch[088] Batch [0449]/[3760]	Speed: 116.429502 samples/sec	 accuracy=34.218750	 loss=2.982251	 lr=0.074887
Epoch[088] Batch [0499]/[3760]	Speed: 117.073774 samples/sec	 accuracy=34.121875	 loss=2.986256	 lr=0.074875
Epoch[088] Batch [0549]/[3760]	Speed: 117.677688 samples/sec	 accuracy=34.056818	 loss=2.989360	 lr=0.074864
Epoch[088] Batch [0599]/[3760]	Speed: 117.409798 samples/sec	 accuracy=34.005208	 loss=2.991632	 lr=0.074853
Epoch[088] Batch [0649]/[3760]	Speed: 117.373460 samples/sec	 accuracy=34.045673	 loss=2.993015	 lr=0.074842
Epoch[088] Batch [0699]/[3760]	Speed: 117.455010 samples/sec	 accuracy=34.071429	 loss=2.993564	 lr=0.074831
Epoch[088] Batch [0749]/[3760]	Speed: 117.200891 samples/sec	 accuracy=34.170833	 loss=2.993342	 lr=0.074819
Epoch[088] Batch [0799]/[3760]	Speed: 116.657559 samples/sec	 accuracy=34.167969	 loss=2.993290	 lr=0.074808
Epoch[088] Batch [0849]/[3760]	Speed: 117.645160 samples/sec	 accuracy=34.204044	 loss=2.993043	 lr=0.074797
Epoch[088] Batch [0899]/[3760]	Speed: 116.729204 samples/sec	 accuracy=34.185764	 loss=2.994995	 lr=0.074786
Epoch[088] Batch [0949]/[3760]	Speed: 117.068976 samples/sec	 accuracy=34.259868	 loss=2.993428	 lr=0.074775
Epoch[088] Batch [0999]/[3760]	Speed: 117.304790 samples/sec	 accuracy=34.223437	 loss=2.992903	 lr=0.074763
Epoch[088] Batch [1049]/[3760]	Speed: 117.379554 samples/sec	 accuracy=34.194940	 loss=2.996316	 lr=0.074752
Epoch[088] Batch [1099]/[3760]	Speed: 117.408507 samples/sec	 accuracy=34.129261	 loss=2.998902	 lr=0.074741
Epoch[088] Batch [1149]/[3760]	Speed: 116.844887 samples/sec	 accuracy=34.130435	 loss=2.997980	 lr=0.074730
Epoch[088] Batch [1199]/[3760]	Speed: 117.668989 samples/sec	 accuracy=34.125000	 loss=2.998427	 lr=0.074719
Epoch[088] Batch [1249]/[3760]	Speed: 118.132507 samples/sec	 accuracy=34.192500	 loss=2.997583	 lr=0.074707
Epoch[088] Batch [1299]/[3760]	Speed: 117.418135 samples/sec	 accuracy=34.158654	 loss=2.999849	 lr=0.074696
Epoch[088] Batch [1349]/[3760]	Speed: 117.184743 samples/sec	 accuracy=34.111111	 loss=3.001806	 lr=0.074685
Epoch[088] Batch [1399]/[3760]	Speed: 117.044646 samples/sec	 accuracy=34.117188	 loss=3.002085	 lr=0.074674
Epoch[088] Batch [1449]/[3760]	Speed: 117.537624 samples/sec	 accuracy=34.151940	 loss=3.002497	 lr=0.074663
Epoch[088] Batch [1499]/[3760]	Speed: 117.795958 samples/sec	 accuracy=34.147917	 loss=3.002227	 lr=0.074651
Epoch[088] Batch [1549]/[3760]	Speed: 118.209849 samples/sec	 accuracy=34.087702	 loss=3.004465	 lr=0.074640
Epoch[088] Batch [1599]/[3760]	Speed: 116.801793 samples/sec	 accuracy=34.006836	 loss=3.007781	 lr=0.074629
Epoch[088] Batch [1649]/[3760]	Speed: 116.998288 samples/sec	 accuracy=33.979167	 loss=3.009278	 lr=0.074618
Epoch[088] Batch [1699]/[3760]	Speed: 117.303934 samples/sec	 accuracy=33.976103	 loss=3.009727	 lr=0.074606
Epoch[088] Batch [1749]/[3760]	Speed: 117.434831 samples/sec	 accuracy=33.950893	 loss=3.009911	 lr=0.074595
Epoch[088] Batch [1799]/[3760]	Speed: 117.561687 samples/sec	 accuracy=33.967014	 loss=3.009716	 lr=0.074584
Epoch[088] Batch [1849]/[3760]	Speed: 117.443912 samples/sec	 accuracy=33.918919	 loss=3.010571	 lr=0.074573
Epoch[088] Batch [1899]/[3760]	Speed: 117.328210 samples/sec	 accuracy=33.919408	 loss=3.010492	 lr=0.074561
Epoch[088] Batch [1949]/[3760]	Speed: 117.235249 samples/sec	 accuracy=33.930288	 loss=3.009535	 lr=0.074550
Epoch[088] Batch [1999]/[3760]	Speed: 116.781627 samples/sec	 accuracy=33.945312	 loss=3.008697	 lr=0.074539
Epoch[088] Batch [2049]/[3760]	Speed: 117.648829 samples/sec	 accuracy=33.903963	 loss=3.010452	 lr=0.074528
Epoch[088] Batch [2099]/[3760]	Speed: 118.230329 samples/sec	 accuracy=33.901786	 loss=3.011128	 lr=0.074517
Epoch[088] Batch [2149]/[3760]	Speed: 117.409857 samples/sec	 accuracy=33.950581	 loss=3.009974	 lr=0.074505
Epoch[088] Batch [2199]/[3760]	Speed: 118.093724 samples/sec	 accuracy=33.928977	 loss=3.010034	 lr=0.074494
Epoch[088] Batch [2249]/[3760]	Speed: 117.709363 samples/sec	 accuracy=33.928472	 loss=3.009901	 lr=0.074483
Epoch[088] Batch [2299]/[3760]	Speed: 118.252218 samples/sec	 accuracy=33.922554	 loss=3.010424	 lr=0.074472
Epoch[088] Batch [2349]/[3760]	Speed: 117.759590 samples/sec	 accuracy=33.954787	 loss=3.009515	 lr=0.074460
Epoch[088] Batch [2399]/[3760]	Speed: 116.314083 samples/sec	 accuracy=33.947266	 loss=3.009408	 lr=0.074449
Epoch[088] Batch [2449]/[3760]	Speed: 117.742584 samples/sec	 accuracy=33.980867	 loss=3.009171	 lr=0.074438
Epoch[088] Batch [2499]/[3760]	Speed: 117.749162 samples/sec	 accuracy=34.012500	 loss=3.007111	 lr=0.074427
Epoch[088] Batch [2549]/[3760]	Speed: 117.657457 samples/sec	 accuracy=34.011029	 loss=3.007875	 lr=0.074415
Epoch[088] Batch [2599]/[3760]	Speed: 118.010100 samples/sec	 accuracy=34.021034	 loss=3.007983	 lr=0.074404
Epoch[088] Batch [2649]/[3760]	Speed: 117.266646 samples/sec	 accuracy=34.022406	 loss=3.008492	 lr=0.074393
Epoch[088] Batch [2699]/[3760]	Speed: 117.610564 samples/sec	 accuracy=34.051505	 loss=3.008163	 lr=0.074382
Epoch[088] Batch [2749]/[3760]	Speed: 117.726695 samples/sec	 accuracy=34.041477	 loss=3.008150	 lr=0.074370
Epoch[088] Batch [2799]/[3760]	Speed: 116.799940 samples/sec	 accuracy=34.055804	 loss=3.007727	 lr=0.074359
Epoch[088] Batch [2849]/[3760]	Speed: 117.634637 samples/sec	 accuracy=34.040570	 loss=3.007856	 lr=0.074348
Epoch[088] Batch [2899]/[3760]	Speed: 117.540064 samples/sec	 accuracy=34.032866	 loss=3.007527	 lr=0.074336
Epoch[088] Batch [2949]/[3760]	Speed: 117.356897 samples/sec	 accuracy=34.028072	 loss=3.007768	 lr=0.074325
Epoch[088] Batch [2999]/[3760]	Speed: 117.051134 samples/sec	 accuracy=34.033854	 loss=3.006220	 lr=0.074314
Epoch[088] Batch [3049]/[3760]	Speed: 116.841001 samples/sec	 accuracy=34.022029	 loss=3.007194	 lr=0.074303
Epoch[088] Batch [3099]/[3760]	Speed: 116.954118 samples/sec	 accuracy=34.013609	 loss=3.007251	 lr=0.074291
Epoch[088] Batch [3149]/[3760]	Speed: 118.192561 samples/sec	 accuracy=34.003968	 loss=3.007622	 lr=0.074280
Epoch[088] Batch [3199]/[3760]	Speed: 116.530742 samples/sec	 accuracy=33.992676	 loss=3.007552	 lr=0.074269
Epoch[088] Batch [3249]/[3760]	Speed: 116.677067 samples/sec	 accuracy=33.978846	 loss=3.008393	 lr=0.074258
Epoch[088] Batch [3299]/[3760]	Speed: 117.679128 samples/sec	 accuracy=33.968750	 loss=3.008397	 lr=0.074246
Epoch[088] Batch [3349]/[3760]	Speed: 117.670501 samples/sec	 accuracy=33.971082	 loss=3.009041	 lr=0.074235
Epoch[088] Batch [3399]/[3760]	Speed: 117.673628 samples/sec	 accuracy=33.951746	 loss=3.009623	 lr=0.074224
Epoch[088] Batch [3449]/[3760]	Speed: 117.214489 samples/sec	 accuracy=33.943841	 loss=3.009405	 lr=0.074212
Epoch[088] Batch [3499]/[3760]	Speed: 117.813398 samples/sec	 accuracy=33.931250	 loss=3.009730	 lr=0.074201
Epoch[088] Batch [3549]/[3760]	Speed: 117.378758 samples/sec	 accuracy=33.949384	 loss=3.009577	 lr=0.074190
Epoch[088] Batch [3599]/[3760]	Speed: 117.200065 samples/sec	 accuracy=33.952691	 loss=3.009703	 lr=0.074179
Epoch[088] Batch [3649]/[3760]	Speed: 117.610961 samples/sec	 accuracy=33.943921	 loss=3.010046	 lr=0.074167
Epoch[088] Batch [3699]/[3760]	Speed: 117.580901 samples/sec	 accuracy=33.961993	 loss=3.009787	 lr=0.074156
Epoch[088] Batch [3749]/[3760]	Speed: 121.077615 samples/sec	 accuracy=33.973333	 loss=3.009998	 lr=0.074145
Batch [0049]/[0303]: acc-top1=35.500000 acc-top5=61.031250
Batch [0099]/[0303]: acc-top1=32.562500 acc-top5=57.609375
Batch [0149]/[0303]: acc-top1=33.635417 acc-top5=59.000000
Batch [0199]/[0303]: acc-top1=36.617188 acc-top5=61.585938
Batch [0249]/[0303]: acc-top1=36.050000 acc-top5=61.768750
Batch [0299]/[0303]: acc-top1=35.567708 acc-top5=61.791667
[Epoch 088] training: accuracy=33.973570	 loss=3.010304
[Epoch 088] speed: 116 samples/sec	time cost: 2176.777462
[Epoch 088] validation: acc-top1=35.612624 acc-top5=61.917285 loss=2.961470
Epoch[089] Batch [0049]/[3760]	Speed: 87.086582 samples/sec	 accuracy=34.875000	 loss=2.941512	 lr=0.074131
Epoch[089] Batch [0099]/[3760]	Speed: 117.487987 samples/sec	 accuracy=34.421875	 loss=2.979387	 lr=0.074120
Epoch[089] Batch [0149]/[3760]	Speed: 117.449305 samples/sec	 accuracy=34.541667	 loss=2.979028	 lr=0.074109
Epoch[089] Batch [0199]/[3760]	Speed: 116.974801 samples/sec	 accuracy=34.492188	 loss=2.987405	 lr=0.074097
Epoch[089] Batch [0249]/[3760]	Speed: 117.240831 samples/sec	 accuracy=34.768750	 loss=2.972638	 lr=0.074086
Epoch[089] Batch [0299]/[3760]	Speed: 117.779020 samples/sec	 accuracy=34.760417	 loss=2.973340	 lr=0.074075
Epoch[089] Batch [0349]/[3760]	Speed: 117.970418 samples/sec	 accuracy=34.495536	 loss=2.983161	 lr=0.074063
Epoch[089] Batch [0399]/[3760]	Speed: 118.257311 samples/sec	 accuracy=34.382812	 loss=2.996656	 lr=0.074052
Epoch[089] Batch [0449]/[3760]	Speed: 117.429012 samples/sec	 accuracy=34.496528	 loss=2.989957	 lr=0.074041
Epoch[089] Batch [0499]/[3760]	Speed: 117.781750 samples/sec	 accuracy=34.437500	 loss=2.984227	 lr=0.074029
Epoch[089] Batch [0549]/[3760]	Speed: 116.841684 samples/sec	 accuracy=34.511364	 loss=2.982870	 lr=0.074018
Epoch[089] Batch [0599]/[3760]	Speed: 117.066552 samples/sec	 accuracy=34.390625	 loss=2.988051	 lr=0.074007
Epoch[089] Batch [0649]/[3760]	Speed: 117.413835 samples/sec	 accuracy=34.495192	 loss=2.986542	 lr=0.073995
Epoch[089] Batch [0699]/[3760]	Speed: 117.162334 samples/sec	 accuracy=34.370536	 loss=2.990334	 lr=0.073984
Epoch[089] Batch [0749]/[3760]	Speed: 116.758545 samples/sec	 accuracy=34.252083	 loss=2.995477	 lr=0.073973
Epoch[089] Batch [0799]/[3760]	Speed: 116.774886 samples/sec	 accuracy=34.167969	 loss=3.000626	 lr=0.073962
Epoch[089] Batch [0849]/[3760]	Speed: 117.691589 samples/sec	 accuracy=34.233456	 loss=2.999882	 lr=0.073950
Epoch[089] Batch [0899]/[3760]	Speed: 117.433528 samples/sec	 accuracy=34.208333	 loss=3.000797	 lr=0.073939
Epoch[089] Batch [0949]/[3760]	Speed: 117.759669 samples/sec	 accuracy=34.289474	 loss=2.996945	 lr=0.073928
Epoch[089] Batch [0999]/[3760]	Speed: 115.977383 samples/sec	 accuracy=34.310938	 loss=2.996949	 lr=0.073916
Epoch[089] Batch [1049]/[3760]	Speed: 117.803749 samples/sec	 accuracy=34.245536	 loss=2.999319	 lr=0.073905
Epoch[089] Batch [1099]/[3760]	Speed: 116.837507 samples/sec	 accuracy=34.241477	 loss=2.998973	 lr=0.073894
Epoch[089] Batch [1149]/[3760]	Speed: 117.786138 samples/sec	 accuracy=34.258152	 loss=2.997346	 lr=0.073882
Epoch[089] Batch [1199]/[3760]	Speed: 117.656456 samples/sec	 accuracy=34.199219	 loss=3.000251	 lr=0.073871
Epoch[089] Batch [1249]/[3760]	Speed: 117.254015 samples/sec	 accuracy=34.208750	 loss=2.997785	 lr=0.073860
Epoch[089] Batch [1299]/[3760]	Speed: 117.412736 samples/sec	 accuracy=34.180288	 loss=2.999176	 lr=0.073848
Epoch[089] Batch [1349]/[3760]	Speed: 117.082990 samples/sec	 accuracy=34.151620	 loss=3.000428	 lr=0.073837
Epoch[089] Batch [1399]/[3760]	Speed: 117.426979 samples/sec	 accuracy=34.140625	 loss=2.999555	 lr=0.073826
Epoch[089] Batch [1449]/[3760]	Speed: 117.760599 samples/sec	 accuracy=34.135776	 loss=2.999784	 lr=0.073814
Epoch[089] Batch [1499]/[3760]	Speed: 116.981663 samples/sec	 accuracy=34.158333	 loss=2.999238	 lr=0.073803
Epoch[089] Batch [1549]/[3760]	Speed: 117.578985 samples/sec	 accuracy=34.138105	 loss=3.001130	 lr=0.073792
Epoch[089] Batch [1599]/[3760]	Speed: 117.416419 samples/sec	 accuracy=34.153320	 loss=3.000641	 lr=0.073780
Epoch[089] Batch [1649]/[3760]	Speed: 117.702545 samples/sec	 accuracy=34.169508	 loss=2.999813	 lr=0.073769
Epoch[089] Batch [1699]/[3760]	Speed: 117.610667 samples/sec	 accuracy=34.126838	 loss=3.001863	 lr=0.073758
Epoch[089] Batch [1749]/[3760]	Speed: 117.815960 samples/sec	 accuracy=34.111607	 loss=3.002382	 lr=0.073746
Epoch[089] Batch [1799]/[3760]	Speed: 116.329153 samples/sec	 accuracy=34.134549	 loss=3.000904	 lr=0.073735
Epoch[089] Batch [1849]/[3760]	Speed: 117.504541 samples/sec	 accuracy=34.141892	 loss=3.001433	 lr=0.073723
Epoch[089] Batch [1899]/[3760]	Speed: 118.446002 samples/sec	 accuracy=34.141447	 loss=3.002019	 lr=0.073712
Epoch[089] Batch [1949]/[3760]	Speed: 116.594314 samples/sec	 accuracy=34.147436	 loss=3.000775	 lr=0.073701
Epoch[089] Batch [1999]/[3760]	Speed: 117.966302 samples/sec	 accuracy=34.154688	 loss=3.000081	 lr=0.073689
Epoch[089] Batch [2049]/[3760]	Speed: 117.540256 samples/sec	 accuracy=34.114329	 loss=3.001996	 lr=0.073678
Epoch[089] Batch [2099]/[3760]	Speed: 117.382443 samples/sec	 accuracy=34.094494	 loss=3.001960	 lr=0.073667
Epoch[089] Batch [2149]/[3760]	Speed: 117.103530 samples/sec	 accuracy=34.092297	 loss=3.002588	 lr=0.073655
Epoch[089] Batch [2199]/[3760]	Speed: 117.431006 samples/sec	 accuracy=34.076705	 loss=3.002884	 lr=0.073644
Epoch[089] Batch [2249]/[3760]	Speed: 117.232304 samples/sec	 accuracy=34.063194	 loss=3.003342	 lr=0.073633
Epoch[089] Batch [2299]/[3760]	Speed: 117.234900 samples/sec	 accuracy=34.079484	 loss=3.002265	 lr=0.073621
Epoch[089] Batch [2349]/[3760]	Speed: 117.772720 samples/sec	 accuracy=34.069149	 loss=3.002526	 lr=0.073610
Epoch[089] Batch [2399]/[3760]	Speed: 117.477527 samples/sec	 accuracy=34.070312	 loss=3.002710	 lr=0.073598
Epoch[089] Batch [2449]/[3760]	Speed: 117.103300 samples/sec	 accuracy=34.040816	 loss=3.003958	 lr=0.073587
Epoch[089] Batch [2499]/[3760]	Speed: 118.042869 samples/sec	 accuracy=34.036250	 loss=3.004717	 lr=0.073576
Epoch[089] Batch [2549]/[3760]	Speed: 117.270106 samples/sec	 accuracy=34.033088	 loss=3.005466	 lr=0.073564
Epoch[089] Batch [2599]/[3760]	Speed: 117.601546 samples/sec	 accuracy=34.013221	 loss=3.006309	 lr=0.073553
Epoch[089] Batch [2649]/[3760]	Speed: 117.840972 samples/sec	 accuracy=34.022995	 loss=3.005679	 lr=0.073542
Epoch[089] Batch [2699]/[3760]	Speed: 117.535209 samples/sec	 accuracy=34.045139	 loss=3.004397	 lr=0.073530
Epoch[089] Batch [2749]/[3760]	Speed: 117.265814 samples/sec	 accuracy=34.057386	 loss=3.005039	 lr=0.073519
Epoch[089] Batch [2799]/[3760]	Speed: 118.267244 samples/sec	 accuracy=34.069196	 loss=3.004500	 lr=0.073507
Epoch[089] Batch [2849]/[3760]	Speed: 117.218631 samples/sec	 accuracy=34.066338	 loss=3.004222	 lr=0.073496
Epoch[089] Batch [2899]/[3760]	Speed: 117.387492 samples/sec	 accuracy=34.072737	 loss=3.003259	 lr=0.073485
Epoch[089] Batch [2949]/[3760]	Speed: 117.390234 samples/sec	 accuracy=34.079449	 loss=3.003322	 lr=0.073473
Epoch[089] Batch [2999]/[3760]	Speed: 117.547342 samples/sec	 accuracy=34.072396	 loss=3.003374	 lr=0.073462
Epoch[089] Batch [3049]/[3760]	Speed: 117.404258 samples/sec	 accuracy=34.089139	 loss=3.003900	 lr=0.073451
Epoch[089] Batch [3099]/[3760]	Speed: 117.364772 samples/sec	 accuracy=34.059476	 loss=3.005224	 lr=0.073439
Epoch[089] Batch [3149]/[3760]	Speed: 117.941574 samples/sec	 accuracy=34.045635	 loss=3.005756	 lr=0.073428
Epoch[089] Batch [3199]/[3760]	Speed: 117.213986 samples/sec	 accuracy=34.027344	 loss=3.006329	 lr=0.073416
Epoch[089] Batch [3249]/[3760]	Speed: 117.160231 samples/sec	 accuracy=34.045673	 loss=3.005887	 lr=0.073405
Epoch[089] Batch [3299]/[3760]	Speed: 116.532603 samples/sec	 accuracy=34.052083	 loss=3.006267	 lr=0.073394
Epoch[089] Batch [3349]/[3760]	Speed: 118.288457 samples/sec	 accuracy=34.057836	 loss=3.005711	 lr=0.073382
Epoch[089] Batch [3399]/[3760]	Speed: 117.369907 samples/sec	 accuracy=34.065717	 loss=3.005886	 lr=0.073371
Epoch[089] Batch [3449]/[3760]	Speed: 117.766857 samples/sec	 accuracy=34.076087	 loss=3.004995	 lr=0.073359
Epoch[089] Batch [3499]/[3760]	Speed: 118.489845 samples/sec	 accuracy=34.089732	 loss=3.004430	 lr=0.073348
Epoch[089] Batch [3549]/[3760]	Speed: 117.432267 samples/sec	 accuracy=34.072623	 loss=3.005139	 lr=0.073337
Epoch[089] Batch [3599]/[3760]	Speed: 116.959847 samples/sec	 accuracy=34.062934	 loss=3.006155	 lr=0.073325
Epoch[089] Batch [3649]/[3760]	Speed: 117.177234 samples/sec	 accuracy=34.057363	 loss=3.006502	 lr=0.073314
Epoch[089] Batch [3699]/[3760]	Speed: 117.680808 samples/sec	 accuracy=34.078547	 loss=3.005815	 lr=0.073302
Epoch[089] Batch [3749]/[3760]	Speed: 120.578662 samples/sec	 accuracy=34.085000	 loss=3.005152	 lr=0.073291
Batch [0049]/[0303]: acc-top1=34.218750 acc-top5=59.218750
Batch [0099]/[0303]: acc-top1=32.171875 acc-top5=57.921875
Batch [0149]/[0303]: acc-top1=33.708333 acc-top5=59.270833
Batch [0199]/[0303]: acc-top1=36.585938 acc-top5=61.937500
Batch [0249]/[0303]: acc-top1=35.237500 acc-top5=60.925000
Batch [0299]/[0303]: acc-top1=34.708333 acc-top5=60.817708
[Epoch 089] training: accuracy=34.079953	 loss=3.005410
[Epoch 089] speed: 116 samples/sec	time cost: 2176.926054
[Epoch 089] validation: acc-top1=34.637995 acc-top5=60.751856 loss=3.177131
Epoch[090] Batch [0049]/[3759]	Speed: 86.967581 samples/sec	 accuracy=35.843750	 loss=2.899209	 lr=0.073277
Epoch[090] Batch [0099]/[3759]	Speed: 118.339271 samples/sec	 accuracy=35.656250	 loss=2.945034	 lr=0.073266
Epoch[090] Batch [0149]/[3759]	Speed: 117.547775 samples/sec	 accuracy=35.302083	 loss=2.959105	 lr=0.073254
Epoch[090] Batch [0199]/[3759]	Speed: 116.487222 samples/sec	 accuracy=35.195312	 loss=2.960948	 lr=0.073243
Epoch[090] Batch [0249]/[3759]	Speed: 117.077414 samples/sec	 accuracy=35.200000	 loss=2.962535	 lr=0.073232
Epoch[090] Batch [0299]/[3759]	Speed: 117.130985 samples/sec	 accuracy=35.114583	 loss=2.965298	 lr=0.073220
Epoch[090] Batch [0349]/[3759]	Speed: 117.336658 samples/sec	 accuracy=34.950893	 loss=2.968729	 lr=0.073209
Epoch[090] Batch [0399]/[3759]	Speed: 118.020152 samples/sec	 accuracy=34.726562	 loss=2.981881	 lr=0.073197
Epoch[090] Batch [0449]/[3759]	Speed: 116.663036 samples/sec	 accuracy=34.795139	 loss=2.977357	 lr=0.073186
Epoch[090] Batch [0499]/[3759]	Speed: 118.263316 samples/sec	 accuracy=34.690625	 loss=2.975260	 lr=0.073174
Epoch[090] Batch [0549]/[3759]	Speed: 117.266623 samples/sec	 accuracy=34.630682	 loss=2.981358	 lr=0.073163
Epoch[090] Batch [0599]/[3759]	Speed: 117.259981 samples/sec	 accuracy=34.557292	 loss=2.987838	 lr=0.073152
Epoch[090] Batch [0649]/[3759]	Speed: 117.565889 samples/sec	 accuracy=34.545673	 loss=2.990252	 lr=0.073140
Epoch[090] Batch [0699]/[3759]	Speed: 117.765959 samples/sec	 accuracy=34.495536	 loss=2.986649	 lr=0.073129
Epoch[090] Batch [0749]/[3759]	Speed: 117.071179 samples/sec	 accuracy=34.512500	 loss=2.989621	 lr=0.073117
Epoch[090] Batch [0799]/[3759]	Speed: 117.204734 samples/sec	 accuracy=34.455078	 loss=2.991048	 lr=0.073106
Epoch[090] Batch [0849]/[3759]	Speed: 117.104882 samples/sec	 accuracy=34.380515	 loss=2.994882	 lr=0.073094
Epoch[090] Batch [0899]/[3759]	Speed: 117.521550 samples/sec	 accuracy=34.390625	 loss=2.996348	 lr=0.073083
Epoch[090] Batch [0949]/[3759]	Speed: 116.665745 samples/sec	 accuracy=34.353618	 loss=2.999038	 lr=0.073072
Epoch[090] Batch [0999]/[3759]	Speed: 117.031672 samples/sec	 accuracy=34.268750	 loss=2.998995	 lr=0.073060
Epoch[090] Batch [1049]/[3759]	Speed: 117.251936 samples/sec	 accuracy=34.206845	 loss=3.003118	 lr=0.073049
Epoch[090] Batch [1099]/[3759]	Speed: 117.400808 samples/sec	 accuracy=34.188920	 loss=3.002773	 lr=0.073037
Epoch[090] Batch [1149]/[3759]	Speed: 117.504422 samples/sec	 accuracy=34.164402	 loss=3.001672	 lr=0.073026
Epoch[090] Batch [1199]/[3759]	Speed: 117.326211 samples/sec	 accuracy=34.199219	 loss=3.000379	 lr=0.073014
Epoch[090] Batch [1249]/[3759]	Speed: 117.418063 samples/sec	 accuracy=34.215000	 loss=3.000106	 lr=0.073003
Epoch[090] Batch [1299]/[3759]	Speed: 117.064650 samples/sec	 accuracy=34.230769	 loss=2.998352	 lr=0.072991
Epoch[090] Batch [1349]/[3759]	Speed: 117.541948 samples/sec	 accuracy=34.260417	 loss=2.997957	 lr=0.072980
Epoch[090] Batch [1399]/[3759]	Speed: 116.987203 samples/sec	 accuracy=34.257812	 loss=2.997314	 lr=0.072968
Epoch[090] Batch [1449]/[3759]	Speed: 117.694776 samples/sec	 accuracy=34.272629	 loss=2.996128	 lr=0.072957
Epoch[090] Batch [1499]/[3759]	Speed: 117.129674 samples/sec	 accuracy=34.281250	 loss=2.994541	 lr=0.072946
Epoch[090] Batch [1549]/[3759]	Speed: 116.833856 samples/sec	 accuracy=34.299395	 loss=2.994006	 lr=0.072934
Epoch[090] Batch [1599]/[3759]	Speed: 118.005148 samples/sec	 accuracy=34.307617	 loss=2.994793	 lr=0.072923
Epoch[090] Batch [1649]/[3759]	Speed: 116.854392 samples/sec	 accuracy=34.294508	 loss=2.995262	 lr=0.072911
Epoch[090] Batch [1699]/[3759]	Speed: 117.990230 samples/sec	 accuracy=34.266544	 loss=2.996861	 lr=0.072900
Epoch[090] Batch [1749]/[3759]	Speed: 118.150990 samples/sec	 accuracy=34.208036	 loss=2.999636	 lr=0.072888
Epoch[090] Batch [1799]/[3759]	Speed: 117.952176 samples/sec	 accuracy=34.255208	 loss=2.997339	 lr=0.072877
Epoch[090] Batch [1849]/[3759]	Speed: 118.319470 samples/sec	 accuracy=34.255068	 loss=2.997694	 lr=0.072865
Epoch[090] Batch [1899]/[3759]	Speed: 117.201844 samples/sec	 accuracy=34.234375	 loss=2.997645	 lr=0.072854
Epoch[090] Batch [1949]/[3759]	Speed: 118.066394 samples/sec	 accuracy=34.225160	 loss=2.997792	 lr=0.072842
Epoch[090] Batch [1999]/[3759]	Speed: 117.573134 samples/sec	 accuracy=34.235156	 loss=2.996720	 lr=0.072831
Epoch[090] Batch [2049]/[3759]	Speed: 117.311307 samples/sec	 accuracy=34.220274	 loss=2.997167	 lr=0.072819
Epoch[090] Batch [2099]/[3759]	Speed: 116.670550 samples/sec	 accuracy=34.220238	 loss=2.997807	 lr=0.072808
Epoch[090] Batch [2149]/[3759]	Speed: 116.898933 samples/sec	 accuracy=34.224564	 loss=2.997098	 lr=0.072796
Epoch[090] Batch [2199]/[3759]	Speed: 116.875985 samples/sec	 accuracy=34.222301	 loss=2.997529	 lr=0.072785
Epoch[090] Batch [2249]/[3759]	Speed: 117.641221 samples/sec	 accuracy=34.204861	 loss=2.998811	 lr=0.072773
Epoch[090] Batch [2299]/[3759]	Speed: 117.097502 samples/sec	 accuracy=34.198370	 loss=2.999343	 lr=0.072762
Epoch[090] Batch [2349]/[3759]	Speed: 117.355608 samples/sec	 accuracy=34.196144	 loss=3.000302	 lr=0.072751
Epoch[090] Batch [2399]/[3759]	Speed: 117.839210 samples/sec	 accuracy=34.220703	 loss=2.998732	 lr=0.072739
Epoch[090] Batch [2449]/[3759]	Speed: 117.670232 samples/sec	 accuracy=34.181122	 loss=3.000424	 lr=0.072728
Epoch[090] Batch [2499]/[3759]	Speed: 117.589102 samples/sec	 accuracy=34.183125	 loss=3.000815	 lr=0.072716
Epoch[090] Batch [2549]/[3759]	Speed: 117.876136 samples/sec	 accuracy=34.189338	 loss=3.001298	 lr=0.072705
Epoch[090] Batch [2599]/[3759]	Speed: 117.750212 samples/sec	 accuracy=34.190505	 loss=3.001090	 lr=0.072693
Epoch[090] Batch [2649]/[3759]	Speed: 118.691078 samples/sec	 accuracy=34.199882	 loss=3.000144	 lr=0.072682
Epoch[090] Batch [2699]/[3759]	Speed: 117.445035 samples/sec	 accuracy=34.227431	 loss=2.999340	 lr=0.072670
Epoch[090] Batch [2749]/[3759]	Speed: 117.997804 samples/sec	 accuracy=34.217614	 loss=3.000253	 lr=0.072659
Epoch[090] Batch [2799]/[3759]	Speed: 117.688957 samples/sec	 accuracy=34.233259	 loss=2.999606	 lr=0.072647
Epoch[090] Batch [2849]/[3759]	Speed: 117.961769 samples/sec	 accuracy=34.238487	 loss=2.999710	 lr=0.072636
Epoch[090] Batch [2899]/[3759]	Speed: 117.981574 samples/sec	 accuracy=34.228987	 loss=2.999836	 lr=0.072624
Epoch[090] Batch [2949]/[3759]	Speed: 116.764365 samples/sec	 accuracy=34.245233	 loss=2.999669	 lr=0.072613
Epoch[090] Batch [2999]/[3759]	Speed: 117.618386 samples/sec	 accuracy=34.236979	 loss=3.000221	 lr=0.072601
Epoch[090] Batch [3049]/[3759]	Speed: 117.594542 samples/sec	 accuracy=34.232582	 loss=3.000466	 lr=0.072590
Epoch[090] Batch [3099]/[3759]	Speed: 117.626874 samples/sec	 accuracy=34.225806	 loss=3.000073	 lr=0.072578
Epoch[090] Batch [3149]/[3759]	Speed: 117.790598 samples/sec	 accuracy=34.240575	 loss=2.999292	 lr=0.072567
Epoch[090] Batch [3199]/[3759]	Speed: 117.578729 samples/sec	 accuracy=34.242188	 loss=2.999436	 lr=0.072555
Epoch[090] Batch [3249]/[3759]	Speed: 116.973711 samples/sec	 accuracy=34.258654	 loss=2.998142	 lr=0.072544
Epoch[090] Batch [3299]/[3759]	Speed: 117.296719 samples/sec	 accuracy=34.258996	 loss=2.997783	 lr=0.072532
Epoch[090] Batch [3349]/[3759]	Speed: 117.317925 samples/sec	 accuracy=34.256996	 loss=2.997800	 lr=0.072521
Epoch[090] Batch [3399]/[3759]	Speed: 117.460171 samples/sec	 accuracy=34.242647	 loss=2.999047	 lr=0.072509
Epoch[090] Batch [3449]/[3759]	Speed: 117.856344 samples/sec	 accuracy=34.225091	 loss=3.000134	 lr=0.072497
Epoch[090] Batch [3499]/[3759]	Speed: 117.148778 samples/sec	 accuracy=34.224107	 loss=2.999850	 lr=0.072486
Epoch[090] Batch [3549]/[3759]	Speed: 117.191483 samples/sec	 accuracy=34.224912	 loss=2.999400	 lr=0.072474
Epoch[090] Batch [3599]/[3759]	Speed: 117.088275 samples/sec	 accuracy=34.214410	 loss=2.999980	 lr=0.072463
Epoch[090] Batch [3649]/[3759]	Speed: 117.194296 samples/sec	 accuracy=34.203339	 loss=2.999483	 lr=0.072451
Epoch[090] Batch [3699]/[3759]	Speed: 117.701114 samples/sec	 accuracy=34.192990	 loss=2.999748	 lr=0.072440
Epoch[090] Batch [3749]/[3759]	Speed: 118.883891 samples/sec	 accuracy=34.203333	 loss=2.998980	 lr=0.072428
Batch [0049]/[0303]: acc-top1=35.500000 acc-top5=61.187500
Batch [0099]/[0303]: acc-top1=33.734375 acc-top5=59.453125
Batch [0149]/[0303]: acc-top1=34.958333 acc-top5=60.656250
Batch [0199]/[0303]: acc-top1=37.531250 acc-top5=62.437500
Batch [0249]/[0303]: acc-top1=37.337500 acc-top5=62.606250
Batch [0299]/[0303]: acc-top1=36.442708 acc-top5=62.671875
[Epoch 090] training: accuracy=34.208732	 loss=2.998932
[Epoch 090] speed: 116 samples/sec	time cost: 2175.937352
[Epoch 090] validation: acc-top1=36.473804 acc-top5=62.711427 loss=2.942223
Epoch[091] Batch [0049]/[3760]	Speed: 87.170333 samples/sec	 accuracy=35.812500	 loss=2.896171	 lr=0.072415
Epoch[091] Batch [0099]/[3760]	Speed: 117.431775 samples/sec	 accuracy=34.828125	 loss=2.956996	 lr=0.072403
Epoch[091] Batch [0149]/[3760]	Speed: 116.696423 samples/sec	 accuracy=34.885417	 loss=2.953642	 lr=0.072392
Epoch[091] Batch [0199]/[3760]	Speed: 117.372469 samples/sec	 accuracy=34.750000	 loss=2.961935	 lr=0.072380
Epoch[091] Batch [0249]/[3760]	Speed: 117.595281 samples/sec	 accuracy=34.893750	 loss=2.956719	 lr=0.072369
Epoch[091] Batch [0299]/[3760]	Speed: 117.949847 samples/sec	 accuracy=34.885417	 loss=2.967560	 lr=0.072357
Epoch[091] Batch [0349]/[3760]	Speed: 116.999421 samples/sec	 accuracy=34.857143	 loss=2.965017	 lr=0.072346
Epoch[091] Batch [0399]/[3760]	Speed: 117.945842 samples/sec	 accuracy=34.730469	 loss=2.977143	 lr=0.072334
Epoch[091] Batch [0449]/[3760]	Speed: 117.059163 samples/sec	 accuracy=34.687500	 loss=2.976024	 lr=0.072322
Epoch[091] Batch [0499]/[3760]	Speed: 117.670381 samples/sec	 accuracy=34.684375	 loss=2.977267	 lr=0.072311
Epoch[091] Batch [0549]/[3760]	Speed: 118.241628 samples/sec	 accuracy=34.494318	 loss=2.988141	 lr=0.072299
Epoch[091] Batch [0599]/[3760]	Speed: 117.301457 samples/sec	 accuracy=34.578125	 loss=2.986805	 lr=0.072288
Epoch[091] Batch [0649]/[3760]	Speed: 116.826018 samples/sec	 accuracy=34.569712	 loss=2.987429	 lr=0.072276
Epoch[091] Batch [0699]/[3760]	Speed: 118.028722 samples/sec	 accuracy=34.595982	 loss=2.988349	 lr=0.072265
Epoch[091] Batch [0749]/[3760]	Speed: 116.689059 samples/sec	 accuracy=34.620833	 loss=2.986691	 lr=0.072253
Epoch[091] Batch [0799]/[3760]	Speed: 117.922767 samples/sec	 accuracy=34.724609	 loss=2.983356	 lr=0.072242
Epoch[091] Batch [0849]/[3760]	Speed: 116.966557 samples/sec	 accuracy=34.726103	 loss=2.983193	 lr=0.072230
Epoch[091] Batch [0899]/[3760]	Speed: 117.665924 samples/sec	 accuracy=34.678819	 loss=2.984218	 lr=0.072219
Epoch[091] Batch [0949]/[3760]	Speed: 117.683909 samples/sec	 accuracy=34.610197	 loss=2.987895	 lr=0.072207
Epoch[091] Batch [0999]/[3760]	Speed: 117.569074 samples/sec	 accuracy=34.556250	 loss=2.992130	 lr=0.072195
Epoch[091] Batch [1049]/[3760]	Speed: 118.072917 samples/sec	 accuracy=34.552083	 loss=2.991701	 lr=0.072184
Epoch[091] Batch [1099]/[3760]	Speed: 118.033920 samples/sec	 accuracy=34.454545	 loss=2.994329	 lr=0.072172
Epoch[091] Batch [1149]/[3760]	Speed: 118.470685 samples/sec	 accuracy=34.447011	 loss=2.994972	 lr=0.072161
Epoch[091] Batch [1199]/[3760]	Speed: 118.159629 samples/sec	 accuracy=34.419271	 loss=2.997982	 lr=0.072149
Epoch[091] Batch [1249]/[3760]	Speed: 117.048936 samples/sec	 accuracy=34.426250	 loss=2.997714	 lr=0.072138
Epoch[091] Batch [1299]/[3760]	Speed: 117.732725 samples/sec	 accuracy=34.451923	 loss=2.995692	 lr=0.072126
Epoch[091] Batch [1349]/[3760]	Speed: 117.995818 samples/sec	 accuracy=34.420139	 loss=2.997065	 lr=0.072114
Epoch[091] Batch [1399]/[3760]	Speed: 117.512479 samples/sec	 accuracy=34.397321	 loss=2.996982	 lr=0.072103
Epoch[091] Batch [1449]/[3760]	Speed: 117.404204 samples/sec	 accuracy=34.335129	 loss=2.999800	 lr=0.072091
Epoch[091] Batch [1499]/[3760]	Speed: 117.154104 samples/sec	 accuracy=34.287500	 loss=3.001285	 lr=0.072080
Epoch[091] Batch [1549]/[3760]	Speed: 117.558583 samples/sec	 accuracy=34.317540	 loss=2.999798	 lr=0.072068
Epoch[091] Batch [1599]/[3760]	Speed: 118.468640 samples/sec	 accuracy=34.334961	 loss=2.996835	 lr=0.072057
Epoch[091] Batch [1649]/[3760]	Speed: 115.503472 samples/sec	 accuracy=34.341856	 loss=2.996440	 lr=0.072045
Epoch[091] Batch [1699]/[3760]	Speed: 117.698608 samples/sec	 accuracy=34.326287	 loss=2.996515	 lr=0.072033
Epoch[091] Batch [1749]/[3760]	Speed: 116.922698 samples/sec	 accuracy=34.318750	 loss=2.996302	 lr=0.072022
Epoch[091] Batch [1799]/[3760]	Speed: 117.873866 samples/sec	 accuracy=34.314236	 loss=2.996251	 lr=0.072010
Epoch[091] Batch [1849]/[3760]	Speed: 117.160102 samples/sec	 accuracy=34.358108	 loss=2.995096	 lr=0.071999
Epoch[091] Batch [1899]/[3760]	Speed: 117.865591 samples/sec	 accuracy=34.339638	 loss=2.995364	 lr=0.071987
Epoch[091] Batch [1949]/[3760]	Speed: 117.524795 samples/sec	 accuracy=34.353365	 loss=2.995951	 lr=0.071976
Epoch[091] Batch [1999]/[3760]	Speed: 116.166044 samples/sec	 accuracy=34.310938	 loss=2.996230	 lr=0.071964
Epoch[091] Batch [2049]/[3760]	Speed: 117.321396 samples/sec	 accuracy=34.329268	 loss=2.994189	 lr=0.071952
Epoch[091] Batch [2099]/[3760]	Speed: 117.470217 samples/sec	 accuracy=34.319940	 loss=2.994402	 lr=0.071941
Epoch[091] Batch [2149]/[3760]	Speed: 117.414573 samples/sec	 accuracy=34.305959	 loss=2.994362	 lr=0.071929
Epoch[091] Batch [2199]/[3760]	Speed: 117.541107 samples/sec	 accuracy=34.316761	 loss=2.994202	 lr=0.071918
Epoch[091] Batch [2249]/[3760]	Speed: 117.684166 samples/sec	 accuracy=34.320833	 loss=2.993928	 lr=0.071906
Epoch[091] Batch [2299]/[3760]	Speed: 117.840068 samples/sec	 accuracy=34.328125	 loss=2.993848	 lr=0.071894
Epoch[091] Batch [2349]/[3760]	Speed: 118.429598 samples/sec	 accuracy=34.332447	 loss=2.994737	 lr=0.071883
Epoch[091] Batch [2399]/[3760]	Speed: 117.032057 samples/sec	 accuracy=34.324870	 loss=2.994843	 lr=0.071871
Epoch[091] Batch [2449]/[3760]	Speed: 117.368452 samples/sec	 accuracy=34.328444	 loss=2.993729	 lr=0.071860
Epoch[091] Batch [2499]/[3760]	Speed: 116.876656 samples/sec	 accuracy=34.314375	 loss=2.993850	 lr=0.071848
Epoch[091] Batch [2549]/[3760]	Speed: 117.082349 samples/sec	 accuracy=34.280025	 loss=2.994060	 lr=0.071836
Epoch[091] Batch [2599]/[3760]	Speed: 117.699042 samples/sec	 accuracy=34.300481	 loss=2.994047	 lr=0.071825
Epoch[091] Batch [2649]/[3760]	Speed: 117.815983 samples/sec	 accuracy=34.310731	 loss=2.993483	 lr=0.071813
Epoch[091] Batch [2699]/[3760]	Speed: 118.331712 samples/sec	 accuracy=34.304977	 loss=2.994109	 lr=0.071802
Epoch[091] Batch [2749]/[3760]	Speed: 117.469651 samples/sec	 accuracy=34.286932	 loss=2.994871	 lr=0.071790
Epoch[091] Batch [2799]/[3760]	Speed: 117.367902 samples/sec	 accuracy=34.294643	 loss=2.994167	 lr=0.071778
Epoch[091] Batch [2849]/[3760]	Speed: 117.677417 samples/sec	 accuracy=34.293311	 loss=2.994302	 lr=0.071767
Epoch[091] Batch [2899]/[3760]	Speed: 116.979560 samples/sec	 accuracy=34.291487	 loss=2.993913	 lr=0.071755
Epoch[091] Batch [2949]/[3760]	Speed: 117.915460 samples/sec	 accuracy=34.313030	 loss=2.993427	 lr=0.071744
Epoch[091] Batch [2999]/[3760]	Speed: 117.786106 samples/sec	 accuracy=34.310417	 loss=2.993224	 lr=0.071732
Epoch[091] Batch [3049]/[3760]	Speed: 117.311843 samples/sec	 accuracy=34.335041	 loss=2.992112	 lr=0.071720
Epoch[091] Batch [3099]/[3760]	Speed: 118.045346 samples/sec	 accuracy=34.346774	 loss=2.992127	 lr=0.071709
Epoch[091] Batch [3149]/[3760]	Speed: 117.321848 samples/sec	 accuracy=34.361111	 loss=2.992028	 lr=0.071697
Epoch[091] Batch [3199]/[3760]	Speed: 117.737528 samples/sec	 accuracy=34.349121	 loss=2.992762	 lr=0.071686
Epoch[091] Batch [3249]/[3760]	Speed: 116.851829 samples/sec	 accuracy=34.336538	 loss=2.993552	 lr=0.071674
Epoch[091] Batch [3299]/[3760]	Speed: 118.216572 samples/sec	 accuracy=34.340909	 loss=2.993519	 lr=0.071662
Epoch[091] Batch [3349]/[3760]	Speed: 115.947680 samples/sec	 accuracy=34.335821	 loss=2.993517	 lr=0.071651
Epoch[091] Batch [3399]/[3760]	Speed: 118.282754 samples/sec	 accuracy=34.323989	 loss=2.994132	 lr=0.071639
Epoch[091] Batch [3449]/[3760]	Speed: 117.954661 samples/sec	 accuracy=34.321105	 loss=2.994516	 lr=0.071627
Epoch[091] Batch [3499]/[3760]	Speed: 117.610400 samples/sec	 accuracy=34.326786	 loss=2.994118	 lr=0.071616
Epoch[091] Batch [3549]/[3760]	Speed: 117.119340 samples/sec	 accuracy=34.329665	 loss=2.994523	 lr=0.071604
Epoch[091] Batch [3599]/[3760]	Speed: 116.706105 samples/sec	 accuracy=34.338976	 loss=2.994206	 lr=0.071592
Epoch[091] Batch [3649]/[3760]	Speed: 118.404405 samples/sec	 accuracy=34.333904	 loss=2.994430	 lr=0.071581
Epoch[091] Batch [3699]/[3760]	Speed: 116.907749 samples/sec	 accuracy=34.338682	 loss=2.994229	 lr=0.071569
Epoch[091] Batch [3749]/[3760]	Speed: 121.072452 samples/sec	 accuracy=34.321250	 loss=2.995435	 lr=0.071558
Batch [0049]/[0303]: acc-top1=33.718750 acc-top5=61.000000
Batch [0099]/[0303]: acc-top1=33.000000 acc-top5=60.406250
Batch [0149]/[0303]: acc-top1=34.687500 acc-top5=61.604167
Batch [0199]/[0303]: acc-top1=37.046875 acc-top5=63.734375
Batch [0249]/[0303]: acc-top1=35.881250 acc-top5=63.281250
Batch [0299]/[0303]: acc-top1=35.557292 acc-top5=63.546875
[Epoch 091] training: accuracy=34.325133	 loss=2.995573
[Epoch 091] speed: 116 samples/sec	time cost: 2174.806381
[Epoch 091] validation: acc-top1=35.586840 acc-top5=63.588078 loss=2.960348
Epoch[092] Batch [0049]/[3760]	Speed: 86.259494 samples/sec	 accuracy=35.781250	 loss=2.884045	 lr=0.071544
Epoch[092] Batch [0099]/[3760]	Speed: 117.646376 samples/sec	 accuracy=35.406250	 loss=2.928338	 lr=0.071532
Epoch[092] Batch [0149]/[3760]	Speed: 117.227601 samples/sec	 accuracy=35.281250	 loss=2.935195	 lr=0.071520
Epoch[092] Batch [0199]/[3760]	Speed: 116.879261 samples/sec	 accuracy=35.476562	 loss=2.936468	 lr=0.071509
Epoch[092] Batch [0249]/[3760]	Speed: 117.601052 samples/sec	 accuracy=35.362500	 loss=2.941887	 lr=0.071497
Epoch[092] Batch [0299]/[3760]	Speed: 117.278960 samples/sec	 accuracy=35.343750	 loss=2.939941	 lr=0.071485
Epoch[092] Batch [0349]/[3760]	Speed: 117.679617 samples/sec	 accuracy=35.044643	 loss=2.956298	 lr=0.071474
Epoch[092] Batch [0399]/[3760]	Speed: 116.619138 samples/sec	 accuracy=35.257812	 loss=2.954523	 lr=0.071462
Epoch[092] Batch [0449]/[3760]	Speed: 117.290648 samples/sec	 accuracy=35.250000	 loss=2.952042	 lr=0.071450
Epoch[092] Batch [0499]/[3760]	Speed: 117.987896 samples/sec	 accuracy=35.178125	 loss=2.954487	 lr=0.071439
Epoch[092] Batch [0549]/[3760]	Speed: 117.717079 samples/sec	 accuracy=35.045455	 loss=2.959815	 lr=0.071427
Epoch[092] Batch [0599]/[3760]	Speed: 117.199258 samples/sec	 accuracy=34.947917	 loss=2.961652	 lr=0.071416
Epoch[092] Batch [0649]/[3760]	Speed: 117.025308 samples/sec	 accuracy=34.764423	 loss=2.967695	 lr=0.071404
Epoch[092] Batch [0699]/[3760]	Speed: 117.059458 samples/sec	 accuracy=34.696429	 loss=2.970048	 lr=0.071392
Epoch[092] Batch [0749]/[3760]	Speed: 117.101736 samples/sec	 accuracy=34.622917	 loss=2.974319	 lr=0.071381
Epoch[092] Batch [0799]/[3760]	Speed: 117.304359 samples/sec	 accuracy=34.591797	 loss=2.973864	 lr=0.071369
Epoch[092] Batch [0849]/[3760]	Speed: 117.049074 samples/sec	 accuracy=34.680147	 loss=2.973278	 lr=0.071357
Epoch[092] Batch [0899]/[3760]	Speed: 117.582544 samples/sec	 accuracy=34.692708	 loss=2.972073	 lr=0.071346
Epoch[092] Batch [0949]/[3760]	Speed: 117.479992 samples/sec	 accuracy=34.726974	 loss=2.973510	 lr=0.071334
Epoch[092] Batch [0999]/[3760]	Speed: 117.568351 samples/sec	 accuracy=34.804688	 loss=2.971983	 lr=0.071322
Epoch[092] Batch [1049]/[3760]	Speed: 117.537557 samples/sec	 accuracy=34.678571	 loss=2.976473	 lr=0.071311
Epoch[092] Batch [1099]/[3760]	Speed: 118.000577 samples/sec	 accuracy=34.700284	 loss=2.974383	 lr=0.071299
Epoch[092] Batch [1149]/[3760]	Speed: 117.006942 samples/sec	 accuracy=34.766304	 loss=2.970697	 lr=0.071287
Epoch[092] Batch [1199]/[3760]	Speed: 117.102132 samples/sec	 accuracy=34.792969	 loss=2.970562	 lr=0.071276
Epoch[092] Batch [1249]/[3760]	Speed: 117.325201 samples/sec	 accuracy=34.746250	 loss=2.971875	 lr=0.071264
Epoch[092] Batch [1299]/[3760]	Speed: 117.698299 samples/sec	 accuracy=34.783654	 loss=2.970179	 lr=0.071252
Epoch[092] Batch [1349]/[3760]	Speed: 117.484703 samples/sec	 accuracy=34.783565	 loss=2.970391	 lr=0.071241
Epoch[092] Batch [1399]/[3760]	Speed: 117.227981 samples/sec	 accuracy=34.804688	 loss=2.969166	 lr=0.071229
Epoch[092] Batch [1449]/[3760]	Speed: 117.466983 samples/sec	 accuracy=34.766164	 loss=2.969892	 lr=0.071217
Epoch[092] Batch [1499]/[3760]	Speed: 117.556639 samples/sec	 accuracy=34.744792	 loss=2.970642	 lr=0.071205
Epoch[092] Batch [1549]/[3760]	Speed: 117.575281 samples/sec	 accuracy=34.752016	 loss=2.970518	 lr=0.071194
Epoch[092] Batch [1599]/[3760]	Speed: 117.534856 samples/sec	 accuracy=34.801758	 loss=2.968223	 lr=0.071182
Epoch[092] Batch [1649]/[3760]	Speed: 117.216089 samples/sec	 accuracy=34.773674	 loss=2.969234	 lr=0.071170
Epoch[092] Batch [1699]/[3760]	Speed: 117.386452 samples/sec	 accuracy=34.784926	 loss=2.968168	 lr=0.071159
Epoch[092] Batch [1749]/[3760]	Speed: 117.496207 samples/sec	 accuracy=34.811607	 loss=2.969164	 lr=0.071147
Epoch[092] Batch [1799]/[3760]	Speed: 117.636357 samples/sec	 accuracy=34.769097	 loss=2.970166	 lr=0.071135
Epoch[092] Batch [1849]/[3760]	Speed: 117.330232 samples/sec	 accuracy=34.736486	 loss=2.969853	 lr=0.071124
Epoch[092] Batch [1899]/[3760]	Speed: 117.238077 samples/sec	 accuracy=34.749178	 loss=2.969611	 lr=0.071112
Epoch[092] Batch [1949]/[3760]	Speed: 117.556449 samples/sec	 accuracy=34.751603	 loss=2.970322	 lr=0.071100
Epoch[092] Batch [1999]/[3760]	Speed: 117.707671 samples/sec	 accuracy=34.748438	 loss=2.970779	 lr=0.071089
Epoch[092] Batch [2049]/[3760]	Speed: 117.743846 samples/sec	 accuracy=34.721799	 loss=2.972286	 lr=0.071077
Epoch[092] Batch [2099]/[3760]	Speed: 116.517611 samples/sec	 accuracy=34.707589	 loss=2.971938	 lr=0.071065
Epoch[092] Batch [2149]/[3760]	Speed: 118.763783 samples/sec	 accuracy=34.651163	 loss=2.973745	 lr=0.071054
Epoch[092] Batch [2199]/[3760]	Speed: 117.368259 samples/sec	 accuracy=34.634943	 loss=2.972680	 lr=0.071042
Epoch[092] Batch [2249]/[3760]	Speed: 118.332241 samples/sec	 accuracy=34.622222	 loss=2.973427	 lr=0.071030
Epoch[092] Batch [2299]/[3760]	Speed: 117.792331 samples/sec	 accuracy=34.632473	 loss=2.973313	 lr=0.071018
Epoch[092] Batch [2349]/[3760]	Speed: 117.252514 samples/sec	 accuracy=34.601729	 loss=2.974121	 lr=0.071007
Epoch[092] Batch [2399]/[3760]	Speed: 117.557796 samples/sec	 accuracy=34.592448	 loss=2.975449	 lr=0.070995
Epoch[092] Batch [2449]/[3760]	Speed: 117.150269 samples/sec	 accuracy=34.568240	 loss=2.976463	 lr=0.070983
Epoch[092] Batch [2499]/[3760]	Speed: 118.042672 samples/sec	 accuracy=34.553750	 loss=2.977014	 lr=0.070972
Epoch[092] Batch [2549]/[3760]	Speed: 117.529737 samples/sec	 accuracy=34.571691	 loss=2.976606	 lr=0.070960
Epoch[092] Batch [2599]/[3760]	Speed: 117.193337 samples/sec	 accuracy=34.570312	 loss=2.976612	 lr=0.070948
Epoch[092] Batch [2649]/[3760]	Speed: 117.467193 samples/sec	 accuracy=34.578420	 loss=2.976203	 lr=0.070936
Epoch[092] Batch [2699]/[3760]	Speed: 117.973548 samples/sec	 accuracy=34.582176	 loss=2.975726	 lr=0.070925
Epoch[092] Batch [2749]/[3760]	Speed: 117.860473 samples/sec	 accuracy=34.568182	 loss=2.976103	 lr=0.070913
Epoch[092] Batch [2799]/[3760]	Speed: 117.120913 samples/sec	 accuracy=34.582031	 loss=2.975212	 lr=0.070901
Epoch[092] Batch [2849]/[3760]	Speed: 117.094641 samples/sec	 accuracy=34.566338	 loss=2.975296	 lr=0.070890
Epoch[092] Batch [2899]/[3760]	Speed: 117.622223 samples/sec	 accuracy=34.567349	 loss=2.976111	 lr=0.070878
Epoch[092] Batch [2949]/[3760]	Speed: 117.008051 samples/sec	 accuracy=34.563030	 loss=2.975931	 lr=0.070866
Epoch[092] Batch [2999]/[3760]	Speed: 116.960136 samples/sec	 accuracy=34.534896	 loss=2.976702	 lr=0.070854
Epoch[092] Batch [3049]/[3760]	Speed: 117.350694 samples/sec	 accuracy=34.525615	 loss=2.976882	 lr=0.070843
Epoch[092] Batch [3099]/[3760]	Speed: 116.532822 samples/sec	 accuracy=34.535282	 loss=2.977394	 lr=0.070831
Epoch[092] Batch [3149]/[3760]	Speed: 117.604155 samples/sec	 accuracy=34.526786	 loss=2.977718	 lr=0.070819
Epoch[092] Batch [3199]/[3760]	Speed: 117.764929 samples/sec	 accuracy=34.503906	 loss=2.978739	 lr=0.070808
Epoch[092] Batch [3249]/[3760]	Speed: 117.642954 samples/sec	 accuracy=34.484615	 loss=2.979888	 lr=0.070796
Epoch[092] Batch [3299]/[3760]	Speed: 117.378967 samples/sec	 accuracy=34.481534	 loss=2.980109	 lr=0.070784
Epoch[092] Batch [3349]/[3760]	Speed: 117.821836 samples/sec	 accuracy=34.485075	 loss=2.979845	 lr=0.070772
Epoch[092] Batch [3399]/[3760]	Speed: 117.688674 samples/sec	 accuracy=34.511489	 loss=2.979551	 lr=0.070761
Epoch[092] Batch [3449]/[3760]	Speed: 116.808775 samples/sec	 accuracy=34.509058	 loss=2.979716	 lr=0.070749
Epoch[092] Batch [3499]/[3760]	Speed: 116.947524 samples/sec	 accuracy=34.498214	 loss=2.980204	 lr=0.070737
Epoch[092] Batch [3549]/[3760]	Speed: 116.805396 samples/sec	 accuracy=34.481514	 loss=2.981016	 lr=0.070725
Epoch[092] Batch [3599]/[3760]	Speed: 117.607792 samples/sec	 accuracy=34.481337	 loss=2.980873	 lr=0.070714
Epoch[092] Batch [3649]/[3760]	Speed: 116.946646 samples/sec	 accuracy=34.472603	 loss=2.980772	 lr=0.070702
Epoch[092] Batch [3699]/[3760]	Speed: 117.105950 samples/sec	 accuracy=34.467061	 loss=2.980388	 lr=0.070690
Epoch[092] Batch [3749]/[3760]	Speed: 119.535012 samples/sec	 accuracy=34.483750	 loss=2.979345	 lr=0.070678
Batch [0049]/[0303]: acc-top1=33.125000 acc-top5=59.781250
Batch [0099]/[0303]: acc-top1=31.390625 acc-top5=57.921875
Batch [0149]/[0303]: acc-top1=33.260417 acc-top5=59.156250
Batch [0199]/[0303]: acc-top1=35.468750 acc-top5=61.031250
Batch [0249]/[0303]: acc-top1=35.018750 acc-top5=60.987500
Batch [0299]/[0303]: acc-top1=34.656250 acc-top5=61.208333
[Epoch 092] training: accuracy=34.489279	 loss=2.979494
[Epoch 092] speed: 116 samples/sec	time cost: 2176.979843
[Epoch 092] validation: acc-top1=34.586427 acc-top5=61.081889 loss=3.163747
Epoch[093] Batch [0049]/[3759]	Speed: 86.928614 samples/sec	 accuracy=35.750000	 loss=2.923022	 lr=0.070664
Epoch[093] Batch [0099]/[3759]	Speed: 117.319462 samples/sec	 accuracy=35.500000	 loss=2.958252	 lr=0.070653
Epoch[093] Batch [0149]/[3759]	Speed: 117.032326 samples/sec	 accuracy=35.604167	 loss=2.952536	 lr=0.070641
Epoch[093] Batch [0199]/[3759]	Speed: 117.299038 samples/sec	 accuracy=35.492188	 loss=2.947496	 lr=0.070629
Epoch[093] Batch [0249]/[3759]	Speed: 116.901410 samples/sec	 accuracy=35.300000	 loss=2.951909	 lr=0.070617
Epoch[093] Batch [0299]/[3759]	Speed: 116.947166 samples/sec	 accuracy=35.286458	 loss=2.959960	 lr=0.070606
Epoch[093] Batch [0349]/[3759]	Speed: 117.406421 samples/sec	 accuracy=35.308036	 loss=2.966178	 lr=0.070594
Epoch[093] Batch [0399]/[3759]	Speed: 117.344605 samples/sec	 accuracy=35.187500	 loss=2.972361	 lr=0.070582
Epoch[093] Batch [0449]/[3759]	Speed: 116.550093 samples/sec	 accuracy=35.204861	 loss=2.968326	 lr=0.070570
Epoch[093] Batch [0499]/[3759]	Speed: 117.776732 samples/sec	 accuracy=35.165625	 loss=2.962447	 lr=0.070559
Epoch[093] Batch [0549]/[3759]	Speed: 117.024488 samples/sec	 accuracy=35.136364	 loss=2.960076	 lr=0.070547
Epoch[093] Batch [0599]/[3759]	Speed: 117.494952 samples/sec	 accuracy=35.125000	 loss=2.956226	 lr=0.070535
Epoch[093] Batch [0649]/[3759]	Speed: 116.702886 samples/sec	 accuracy=35.141827	 loss=2.953897	 lr=0.070523
Epoch[093] Batch [0699]/[3759]	Speed: 117.865759 samples/sec	 accuracy=35.107143	 loss=2.953364	 lr=0.070512
Epoch[093] Batch [0749]/[3759]	Speed: 117.119503 samples/sec	 accuracy=35.089583	 loss=2.955021	 lr=0.070500
Epoch[093] Batch [0799]/[3759]	Speed: 117.929659 samples/sec	 accuracy=34.939453	 loss=2.961497	 lr=0.070488
Epoch[093] Batch [0849]/[3759]	Speed: 116.097636 samples/sec	 accuracy=34.915441	 loss=2.963306	 lr=0.070476
Epoch[093] Batch [0899]/[3759]	Speed: 117.574916 samples/sec	 accuracy=34.963542	 loss=2.959744	 lr=0.070465
Epoch[093] Batch [0949]/[3759]	Speed: 116.977312 samples/sec	 accuracy=34.980263	 loss=2.960371	 lr=0.070453
Epoch[093] Batch [0999]/[3759]	Speed: 117.304085 samples/sec	 accuracy=34.854687	 loss=2.963202	 lr=0.070441
Epoch[093] Batch [1049]/[3759]	Speed: 117.193851 samples/sec	 accuracy=34.872024	 loss=2.960556	 lr=0.070429
Epoch[093] Batch [1099]/[3759]	Speed: 117.116547 samples/sec	 accuracy=34.835227	 loss=2.962352	 lr=0.070417
Epoch[093] Batch [1149]/[3759]	Speed: 117.280832 samples/sec	 accuracy=34.853261	 loss=2.961878	 lr=0.070406
Epoch[093] Batch [1199]/[3759]	Speed: 117.615544 samples/sec	 accuracy=34.791667	 loss=2.963403	 lr=0.070394
Epoch[093] Batch [1249]/[3759]	Speed: 117.176182 samples/sec	 accuracy=34.801250	 loss=2.965372	 lr=0.070382
Epoch[093] Batch [1299]/[3759]	Speed: 117.271962 samples/sec	 accuracy=34.782452	 loss=2.966369	 lr=0.070370
Epoch[093] Batch [1349]/[3759]	Speed: 117.578321 samples/sec	 accuracy=34.765046	 loss=2.967214	 lr=0.070359
Epoch[093] Batch [1399]/[3759]	Speed: 117.336575 samples/sec	 accuracy=34.787946	 loss=2.965784	 lr=0.070347
Epoch[093] Batch [1449]/[3759]	Speed: 117.797790 samples/sec	 accuracy=34.793103	 loss=2.966073	 lr=0.070335
Epoch[093] Batch [1499]/[3759]	Speed: 117.583173 samples/sec	 accuracy=34.751042	 loss=2.968475	 lr=0.070323
Epoch[093] Batch [1549]/[3759]	Speed: 117.049293 samples/sec	 accuracy=34.731855	 loss=2.968776	 lr=0.070311
Epoch[093] Batch [1599]/[3759]	Speed: 118.295259 samples/sec	 accuracy=34.739258	 loss=2.968395	 lr=0.070300
Epoch[093] Batch [1649]/[3759]	Speed: 117.473088 samples/sec	 accuracy=34.729167	 loss=2.968017	 lr=0.070288
Epoch[093] Batch [1699]/[3759]	Speed: 117.038932 samples/sec	 accuracy=34.746324	 loss=2.967004	 lr=0.070276
Epoch[093] Batch [1749]/[3759]	Speed: 117.411458 samples/sec	 accuracy=34.700893	 loss=2.968853	 lr=0.070264
Epoch[093] Batch [1799]/[3759]	Speed: 117.149989 samples/sec	 accuracy=34.714410	 loss=2.968624	 lr=0.070252
Epoch[093] Batch [1849]/[3759]	Speed: 117.957323 samples/sec	 accuracy=34.673986	 loss=2.969537	 lr=0.070241
Epoch[093] Batch [1899]/[3759]	Speed: 117.349197 samples/sec	 accuracy=34.716283	 loss=2.968311	 lr=0.070229
Epoch[093] Batch [1949]/[3759]	Speed: 117.420804 samples/sec	 accuracy=34.685897	 loss=2.969156	 lr=0.070217
Epoch[093] Batch [1999]/[3759]	Speed: 117.444443 samples/sec	 accuracy=34.699219	 loss=2.969525	 lr=0.070205
Epoch[093] Batch [2049]/[3759]	Speed: 117.149027 samples/sec	 accuracy=34.698171	 loss=2.970287	 lr=0.070194
Epoch[093] Batch [2099]/[3759]	Speed: 116.669395 samples/sec	 accuracy=34.695685	 loss=2.971125	 lr=0.070182
Epoch[093] Batch [2149]/[3759]	Speed: 117.291216 samples/sec	 accuracy=34.699128	 loss=2.971034	 lr=0.070170
Epoch[093] Batch [2199]/[3759]	Speed: 117.355942 samples/sec	 accuracy=34.670455	 loss=2.972388	 lr=0.070158
Epoch[093] Batch [2249]/[3759]	Speed: 117.613983 samples/sec	 accuracy=34.685417	 loss=2.972986	 lr=0.070146
Epoch[093] Batch [2299]/[3759]	Speed: 117.144391 samples/sec	 accuracy=34.699049	 loss=2.971980	 lr=0.070135
Epoch[093] Batch [2349]/[3759]	Speed: 117.069778 samples/sec	 accuracy=34.684840	 loss=2.971872	 lr=0.070123
Epoch[093] Batch [2399]/[3759]	Speed: 118.212725 samples/sec	 accuracy=34.690755	 loss=2.971033	 lr=0.070111
Epoch[093] Batch [2449]/[3759]	Speed: 116.979919 samples/sec	 accuracy=34.713010	 loss=2.970091	 lr=0.070099
Epoch[093] Batch [2499]/[3759]	Speed: 118.157619 samples/sec	 accuracy=34.723125	 loss=2.969644	 lr=0.070087
Epoch[093] Batch [2549]/[3759]	Speed: 116.247414 samples/sec	 accuracy=34.726103	 loss=2.969709	 lr=0.070075
Epoch[093] Batch [2599]/[3759]	Speed: 117.037382 samples/sec	 accuracy=34.699519	 loss=2.971189	 lr=0.070064
Epoch[093] Batch [2649]/[3759]	Speed: 117.977528 samples/sec	 accuracy=34.702241	 loss=2.970878	 lr=0.070052
Epoch[093] Batch [2699]/[3759]	Speed: 118.238343 samples/sec	 accuracy=34.712384	 loss=2.970078	 lr=0.070040
Epoch[093] Batch [2749]/[3759]	Speed: 117.371475 samples/sec	 accuracy=34.714773	 loss=2.970192	 lr=0.070028
Epoch[093] Batch [2799]/[3759]	Speed: 117.205200 samples/sec	 accuracy=34.738281	 loss=2.969390	 lr=0.070016
Epoch[093] Batch [2849]/[3759]	Speed: 117.578860 samples/sec	 accuracy=34.740680	 loss=2.970179	 lr=0.070005
Epoch[093] Batch [2899]/[3759]	Speed: 117.036637 samples/sec	 accuracy=34.728987	 loss=2.970831	 lr=0.069993
Epoch[093] Batch [2949]/[3759]	Speed: 117.632514 samples/sec	 accuracy=34.740996	 loss=2.970204	 lr=0.069981
Epoch[093] Batch [2999]/[3759]	Speed: 117.051090 samples/sec	 accuracy=34.718229	 loss=2.970511	 lr=0.069969
Epoch[093] Batch [3049]/[3759]	Speed: 117.747997 samples/sec	 accuracy=34.742828	 loss=2.969833	 lr=0.069957
Epoch[093] Batch [3099]/[3759]	Speed: 116.803645 samples/sec	 accuracy=34.753528	 loss=2.969099	 lr=0.069945
Epoch[093] Batch [3149]/[3759]	Speed: 117.594836 samples/sec	 accuracy=34.728671	 loss=2.969941	 lr=0.069934
Epoch[093] Batch [3199]/[3759]	Speed: 117.499915 samples/sec	 accuracy=34.721680	 loss=2.970474	 lr=0.069922
Epoch[093] Batch [3249]/[3759]	Speed: 117.545228 samples/sec	 accuracy=34.705288	 loss=2.970735	 lr=0.069910
Epoch[093] Batch [3299]/[3759]	Speed: 118.024814 samples/sec	 accuracy=34.703125	 loss=2.970670	 lr=0.069898
Epoch[093] Batch [3349]/[3759]	Speed: 117.494255 samples/sec	 accuracy=34.694963	 loss=2.971127	 lr=0.069886
Epoch[093] Batch [3399]/[3759]	Speed: 116.536718 samples/sec	 accuracy=34.695772	 loss=2.971269	 lr=0.069874
Epoch[093] Batch [3449]/[3759]	Speed: 117.443265 samples/sec	 accuracy=34.704710	 loss=2.970801	 lr=0.069863
Epoch[093] Batch [3499]/[3759]	Speed: 117.242368 samples/sec	 accuracy=34.726339	 loss=2.970005	 lr=0.069851
Epoch[093] Batch [3549]/[3759]	Speed: 117.054687 samples/sec	 accuracy=34.695423	 loss=2.971259	 lr=0.069839
Epoch[093] Batch [3599]/[3759]	Speed: 116.914006 samples/sec	 accuracy=34.677083	 loss=2.972058	 lr=0.069827
Epoch[093] Batch [3649]/[3759]	Speed: 117.902636 samples/sec	 accuracy=34.673801	 loss=2.972352	 lr=0.069815
Epoch[093] Batch [3699]/[3759]	Speed: 117.440240 samples/sec	 accuracy=34.645693	 loss=2.973054	 lr=0.069803
Epoch[093] Batch [3749]/[3759]	Speed: 120.575061 samples/sec	 accuracy=34.649167	 loss=2.972669	 lr=0.069792
Batch [0049]/[0303]: acc-top1=35.656250 acc-top5=63.593750
Batch [0099]/[0303]: acc-top1=34.281250 acc-top5=60.796875
Batch [0149]/[0303]: acc-top1=35.843750 acc-top5=61.760417
Batch [0199]/[0303]: acc-top1=38.085938 acc-top5=63.851562
Batch [0249]/[0303]: acc-top1=37.343750 acc-top5=63.881250
Batch [0299]/[0303]: acc-top1=36.578125 acc-top5=63.619792
[Epoch 093] training: accuracy=34.649757	 loss=2.972589
[Epoch 093] speed: 116 samples/sec	time cost: 2177.058865
[Epoch 093] validation: acc-top1=36.509901 acc-top5=63.613861 loss=2.921809
Epoch[094] Batch [0049]/[3760]	Speed: 87.902987 samples/sec	 accuracy=34.531250	 loss=2.902887	 lr=0.069778
Epoch[094] Batch [0099]/[3760]	Speed: 117.479955 samples/sec	 accuracy=35.468750	 loss=2.912804	 lr=0.069766
Epoch[094] Batch [0149]/[3760]	Speed: 117.225307 samples/sec	 accuracy=35.375000	 loss=2.929137	 lr=0.069754
Epoch[094] Batch [0199]/[3760]	Speed: 116.831390 samples/sec	 accuracy=35.000000	 loss=2.944141	 lr=0.069742
Epoch[094] Batch [0249]/[3760]	Speed: 117.356961 samples/sec	 accuracy=35.231250	 loss=2.932299	 lr=0.069730
Epoch[094] Batch [0299]/[3760]	Speed: 117.081153 samples/sec	 accuracy=35.322917	 loss=2.929178	 lr=0.069718
Epoch[094] Batch [0349]/[3760]	Speed: 117.903236 samples/sec	 accuracy=35.379464	 loss=2.926552	 lr=0.069707
Epoch[094] Batch [0399]/[3760]	Speed: 117.283104 samples/sec	 accuracy=35.210938	 loss=2.937342	 lr=0.069695
Epoch[094] Batch [0449]/[3760]	Speed: 117.537557 samples/sec	 accuracy=35.135417	 loss=2.940779	 lr=0.069683
Epoch[094] Batch [0499]/[3760]	Speed: 116.081308 samples/sec	 accuracy=35.134375	 loss=2.944625	 lr=0.069671
Epoch[094] Batch [0549]/[3760]	Speed: 117.831412 samples/sec	 accuracy=35.105114	 loss=2.945752	 lr=0.069659
Epoch[094] Batch [0599]/[3760]	Speed: 117.010739 samples/sec	 accuracy=35.091146	 loss=2.945764	 lr=0.069647
Epoch[094] Batch [0649]/[3760]	Speed: 117.675920 samples/sec	 accuracy=35.000000	 loss=2.949658	 lr=0.069635
Epoch[094] Batch [0699]/[3760]	Speed: 118.561017 samples/sec	 accuracy=34.986607	 loss=2.949114	 lr=0.069624
Epoch[094] Batch [0749]/[3760]	Speed: 117.781061 samples/sec	 accuracy=35.070833	 loss=2.947487	 lr=0.069612
Epoch[094] Batch [0799]/[3760]	Speed: 118.111843 samples/sec	 accuracy=35.011719	 loss=2.949389	 lr=0.069600
Epoch[094] Batch [0849]/[3760]	Speed: 117.309048 samples/sec	 accuracy=35.064338	 loss=2.947977	 lr=0.069588
Epoch[094] Batch [0899]/[3760]	Speed: 117.512322 samples/sec	 accuracy=35.057292	 loss=2.950486	 lr=0.069576
Epoch[094] Batch [0949]/[3760]	Speed: 117.131463 samples/sec	 accuracy=35.052632	 loss=2.949131	 lr=0.069564
Epoch[094] Batch [0999]/[3760]	Speed: 117.595679 samples/sec	 accuracy=35.043750	 loss=2.948706	 lr=0.069552
Epoch[094] Batch [1049]/[3760]	Speed: 116.745546 samples/sec	 accuracy=35.072917	 loss=2.946697	 lr=0.069540
Epoch[094] Batch [1099]/[3760]	Speed: 117.381031 samples/sec	 accuracy=35.125000	 loss=2.947873	 lr=0.069529
Epoch[094] Batch [1149]/[3760]	Speed: 117.698568 samples/sec	 accuracy=35.072011	 loss=2.951532	 lr=0.069517
Epoch[094] Batch [1199]/[3760]	Speed: 117.811166 samples/sec	 accuracy=35.104167	 loss=2.951929	 lr=0.069505
Epoch[094] Batch [1249]/[3760]	Speed: 118.183674 samples/sec	 accuracy=35.093750	 loss=2.953439	 lr=0.069493
Epoch[094] Batch [1299]/[3760]	Speed: 117.200785 samples/sec	 accuracy=35.100962	 loss=2.952430	 lr=0.069481
Epoch[094] Batch [1349]/[3760]	Speed: 116.923987 samples/sec	 accuracy=35.092593	 loss=2.952777	 lr=0.069469
Epoch[094] Batch [1399]/[3760]	Speed: 118.076847 samples/sec	 accuracy=35.024554	 loss=2.955476	 lr=0.069457
Epoch[094] Batch [1449]/[3760]	Speed: 117.502065 samples/sec	 accuracy=35.011853	 loss=2.956328	 lr=0.069445
Epoch[094] Batch [1499]/[3760]	Speed: 117.739187 samples/sec	 accuracy=35.082292	 loss=2.954076	 lr=0.069434
Epoch[094] Batch [1549]/[3760]	Speed: 117.417212 samples/sec	 accuracy=35.116935	 loss=2.951971	 lr=0.069422
Epoch[094] Batch [1599]/[3760]	Speed: 117.634919 samples/sec	 accuracy=35.047852	 loss=2.954924	 lr=0.069410
Epoch[094] Batch [1649]/[3760]	Speed: 116.661642 samples/sec	 accuracy=35.015152	 loss=2.955194	 lr=0.069398
Epoch[094] Batch [1699]/[3760]	Speed: 117.288606 samples/sec	 accuracy=35.030331	 loss=2.954770	 lr=0.069386
Epoch[094] Batch [1749]/[3760]	Speed: 116.810779 samples/sec	 accuracy=35.033036	 loss=2.955319	 lr=0.069374
Epoch[094] Batch [1799]/[3760]	Speed: 117.246153 samples/sec	 accuracy=35.028646	 loss=2.955751	 lr=0.069362
Epoch[094] Batch [1849]/[3760]	Speed: 117.360289 samples/sec	 accuracy=35.038851	 loss=2.954468	 lr=0.069350
Epoch[094] Batch [1899]/[3760]	Speed: 117.633520 samples/sec	 accuracy=35.020559	 loss=2.954538	 lr=0.069338
Epoch[094] Batch [1949]/[3760]	Speed: 117.297699 samples/sec	 accuracy=35.000000	 loss=2.956883	 lr=0.069327
Epoch[094] Batch [1999]/[3760]	Speed: 117.669751 samples/sec	 accuracy=34.989063	 loss=2.956525	 lr=0.069315
Epoch[094] Batch [2049]/[3760]	Speed: 117.369211 samples/sec	 accuracy=34.967226	 loss=2.958014	 lr=0.069303
Epoch[094] Batch [2099]/[3760]	Speed: 117.342311 samples/sec	 accuracy=34.950893	 loss=2.958330	 lr=0.069291
Epoch[094] Batch [2149]/[3760]	Speed: 118.281791 samples/sec	 accuracy=34.928052	 loss=2.959293	 lr=0.069279
Epoch[094] Batch [2199]/[3760]	Speed: 115.732136 samples/sec	 accuracy=34.911222	 loss=2.960426	 lr=0.069267
Epoch[094] Batch [2249]/[3760]	Speed: 117.164559 samples/sec	 accuracy=34.901389	 loss=2.960524	 lr=0.069255
Epoch[094] Batch [2299]/[3760]	Speed: 116.191573 samples/sec	 accuracy=34.869565	 loss=2.962013	 lr=0.069243
Epoch[094] Batch [2349]/[3760]	Speed: 116.854473 samples/sec	 accuracy=34.881649	 loss=2.961157	 lr=0.069231
Epoch[094] Batch [2399]/[3760]	Speed: 117.413694 samples/sec	 accuracy=34.863281	 loss=2.961669	 lr=0.069219
Epoch[094] Batch [2449]/[3760]	Speed: 117.288756 samples/sec	 accuracy=34.823342	 loss=2.962912	 lr=0.069208
Epoch[094] Batch [2499]/[3760]	Speed: 117.301370 samples/sec	 accuracy=34.814375	 loss=2.963902	 lr=0.069196
Epoch[094] Batch [2549]/[3760]	Speed: 117.705398 samples/sec	 accuracy=34.827819	 loss=2.963965	 lr=0.069184
Epoch[094] Batch [2599]/[3760]	Speed: 117.256864 samples/sec	 accuracy=34.861779	 loss=2.963709	 lr=0.069172
Epoch[094] Batch [2649]/[3760]	Speed: 116.782753 samples/sec	 accuracy=34.879127	 loss=2.963190	 lr=0.069160
Epoch[094] Batch [2699]/[3760]	Speed: 117.258350 samples/sec	 accuracy=34.891204	 loss=2.962313	 lr=0.069148
Epoch[094] Batch [2749]/[3760]	Speed: 117.149625 samples/sec	 accuracy=34.907955	 loss=2.961851	 lr=0.069136
Epoch[094] Batch [2799]/[3760]	Speed: 117.037215 samples/sec	 accuracy=34.910714	 loss=2.962276	 lr=0.069124
Epoch[094] Batch [2849]/[3760]	Speed: 117.118588 samples/sec	 accuracy=34.943531	 loss=2.961014	 lr=0.069112
Epoch[094] Batch [2899]/[3760]	Speed: 117.253312 samples/sec	 accuracy=34.929957	 loss=2.961862	 lr=0.069100
Epoch[094] Batch [2949]/[3760]	Speed: 118.263009 samples/sec	 accuracy=34.952331	 loss=2.962574	 lr=0.069088
Epoch[094] Batch [2999]/[3760]	Speed: 117.276721 samples/sec	 accuracy=34.943750	 loss=2.963509	 lr=0.069076
Epoch[094] Batch [3049]/[3760]	Speed: 115.747527 samples/sec	 accuracy=34.935451	 loss=2.963811	 lr=0.069065
Epoch[094] Batch [3099]/[3760]	Speed: 118.256192 samples/sec	 accuracy=34.922379	 loss=2.964282	 lr=0.069053
Epoch[094] Batch [3149]/[3760]	Speed: 117.398477 samples/sec	 accuracy=34.926587	 loss=2.963469	 lr=0.069041
Epoch[094] Batch [3199]/[3760]	Speed: 116.761419 samples/sec	 accuracy=34.944336	 loss=2.962772	 lr=0.069029
Epoch[094] Batch [3249]/[3760]	Speed: 117.847162 samples/sec	 accuracy=34.947115	 loss=2.962393	 lr=0.069017
Epoch[094] Batch [3299]/[3760]	Speed: 117.523495 samples/sec	 accuracy=34.952178	 loss=2.962699	 lr=0.069005
Epoch[094] Batch [3349]/[3760]	Speed: 117.142433 samples/sec	 accuracy=34.958022	 loss=2.962502	 lr=0.068993
Epoch[094] Batch [3399]/[3760]	Speed: 117.131322 samples/sec	 accuracy=34.960938	 loss=2.962985	 lr=0.068981
Epoch[094] Batch [3449]/[3760]	Speed: 117.600542 samples/sec	 accuracy=34.948822	 loss=2.962986	 lr=0.068969
Epoch[094] Batch [3499]/[3760]	Speed: 117.273315 samples/sec	 accuracy=34.943750	 loss=2.962995	 lr=0.068957
Epoch[094] Batch [3549]/[3760]	Speed: 117.722783 samples/sec	 accuracy=34.951585	 loss=2.963026	 lr=0.068945
Epoch[094] Batch [3599]/[3760]	Speed: 116.609087 samples/sec	 accuracy=34.932726	 loss=2.963680	 lr=0.068933
Epoch[094] Batch [3649]/[3760]	Speed: 117.489177 samples/sec	 accuracy=34.916524	 loss=2.964208	 lr=0.068921
Epoch[094] Batch [3699]/[3760]	Speed: 118.456450 samples/sec	 accuracy=34.912584	 loss=2.964991	 lr=0.068909
Epoch[094] Batch [3749]/[3760]	Speed: 120.227378 samples/sec	 accuracy=34.894583	 loss=2.965691	 lr=0.068897
Batch [0049]/[0303]: acc-top1=36.875000 acc-top5=62.312500
Batch [0099]/[0303]: acc-top1=33.812500 acc-top5=60.359375
Batch [0149]/[0303]: acc-top1=35.260417 acc-top5=61.447917
Batch [0199]/[0303]: acc-top1=37.593750 acc-top5=63.320312
Batch [0249]/[0303]: acc-top1=37.006250 acc-top5=63.212500
Batch [0299]/[0303]: acc-top1=36.171875 acc-top5=62.984375
[Epoch 094] training: accuracy=34.889877	 loss=2.966103
[Epoch 094] speed: 116 samples/sec	time cost: 2176.632136
[Epoch 094] validation: acc-top1=36.210809 acc-top5=63.046617 loss=2.930863
Epoch[095] Batch [0049]/[3760]	Speed: 87.285028 samples/sec	 accuracy=35.718750	 loss=2.913305	 lr=0.068883
Epoch[095] Batch [0099]/[3760]	Speed: 117.691248 samples/sec	 accuracy=35.781250	 loss=2.897248	 lr=0.068871
Epoch[095] Batch [0149]/[3760]	Speed: 116.782374 samples/sec	 accuracy=35.854167	 loss=2.906789	 lr=0.068859
Epoch[095] Batch [0199]/[3760]	Speed: 117.591443 samples/sec	 accuracy=35.484375	 loss=2.927077	 lr=0.068847
Epoch[095] Batch [0249]/[3760]	Speed: 117.487800 samples/sec	 accuracy=35.412500	 loss=2.925952	 lr=0.068835
Epoch[095] Batch [0299]/[3760]	Speed: 116.911992 samples/sec	 accuracy=35.250000	 loss=2.935177	 lr=0.068823
Epoch[095] Batch [0349]/[3760]	Speed: 117.725911 samples/sec	 accuracy=35.254464	 loss=2.930368	 lr=0.068811
Epoch[095] Batch [0399]/[3760]	Speed: 118.399596 samples/sec	 accuracy=35.402344	 loss=2.931774	 lr=0.068800
Epoch[095] Batch [0449]/[3760]	Speed: 116.766156 samples/sec	 accuracy=35.461806	 loss=2.930801	 lr=0.068788
Epoch[095] Batch [0499]/[3760]	Speed: 117.058971 samples/sec	 accuracy=35.459375	 loss=2.931599	 lr=0.068776
Epoch[095] Batch [0549]/[3760]	Speed: 117.662555 samples/sec	 accuracy=35.406250	 loss=2.931861	 lr=0.068764
Epoch[095] Batch [0599]/[3760]	Speed: 117.484356 samples/sec	 accuracy=35.333333	 loss=2.933206	 lr=0.068752
Epoch[095] Batch [0649]/[3760]	Speed: 117.925096 samples/sec	 accuracy=35.326923	 loss=2.931412	 lr=0.068740
Epoch[095] Batch [0699]/[3760]	Speed: 117.506459 samples/sec	 accuracy=35.245536	 loss=2.935846	 lr=0.068728
Epoch[095] Batch [0749]/[3760]	Speed: 116.923836 samples/sec	 accuracy=35.210417	 loss=2.938773	 lr=0.068716
Epoch[095] Batch [0799]/[3760]	Speed: 118.254927 samples/sec	 accuracy=35.146484	 loss=2.941360	 lr=0.068704
Epoch[095] Batch [0849]/[3760]	Speed: 117.314842 samples/sec	 accuracy=35.161765	 loss=2.940436	 lr=0.068692
Epoch[095] Batch [0899]/[3760]	Speed: 117.071915 samples/sec	 accuracy=35.144097	 loss=2.939151	 lr=0.068680
Epoch[095] Batch [0949]/[3760]	Speed: 117.913746 samples/sec	 accuracy=35.116776	 loss=2.940271	 lr=0.068668
Epoch[095] Batch [0999]/[3760]	Speed: 116.575805 samples/sec	 accuracy=35.068750	 loss=2.941194	 lr=0.068656
Epoch[095] Batch [1049]/[3760]	Speed: 117.892900 samples/sec	 accuracy=35.019345	 loss=2.942416	 lr=0.068644
Epoch[095] Batch [1099]/[3760]	Speed: 116.832118 samples/sec	 accuracy=34.995739	 loss=2.945539	 lr=0.068632
Epoch[095] Batch [1149]/[3760]	Speed: 117.430544 samples/sec	 accuracy=35.025815	 loss=2.946305	 lr=0.068620
Epoch[095] Batch [1199]/[3760]	Speed: 118.128321 samples/sec	 accuracy=35.059896	 loss=2.946084	 lr=0.068608
Epoch[095] Batch [1249]/[3760]	Speed: 117.552946 samples/sec	 accuracy=35.060000	 loss=2.945285	 lr=0.068596
Epoch[095] Batch [1299]/[3760]	Speed: 117.676138 samples/sec	 accuracy=35.072115	 loss=2.946079	 lr=0.068584
Epoch[095] Batch [1349]/[3760]	Speed: 118.091209 samples/sec	 accuracy=35.071759	 loss=2.947171	 lr=0.068572
Epoch[095] Batch [1399]/[3760]	Speed: 117.366835 samples/sec	 accuracy=35.023438	 loss=2.948375	 lr=0.068560
Epoch[095] Batch [1449]/[3760]	Speed: 117.571367 samples/sec	 accuracy=35.021552	 loss=2.948594	 lr=0.068548
Epoch[095] Batch [1499]/[3760]	Speed: 118.315513 samples/sec	 accuracy=34.966667	 loss=2.951536	 lr=0.068536
Epoch[095] Batch [1549]/[3760]	Speed: 118.149199 samples/sec	 accuracy=34.965726	 loss=2.951343	 lr=0.068524
Epoch[095] Batch [1599]/[3760]	Speed: 118.177458 samples/sec	 accuracy=34.986328	 loss=2.951125	 lr=0.068512
Epoch[095] Batch [1649]/[3760]	Speed: 117.270608 samples/sec	 accuracy=34.976326	 loss=2.951906	 lr=0.068500
Epoch[095] Batch [1699]/[3760]	Speed: 117.894630 samples/sec	 accuracy=34.999081	 loss=2.950669	 lr=0.068488
Epoch[095] Batch [1749]/[3760]	Speed: 117.005788 samples/sec	 accuracy=35.023214	 loss=2.949798	 lr=0.068476
Epoch[095] Batch [1799]/[3760]	Speed: 117.744725 samples/sec	 accuracy=35.013021	 loss=2.950326	 lr=0.068464
Epoch[095] Batch [1849]/[3760]	Speed: 117.167477 samples/sec	 accuracy=34.987331	 loss=2.951550	 lr=0.068452
Epoch[095] Batch [1899]/[3760]	Speed: 117.489698 samples/sec	 accuracy=34.962171	 loss=2.953152	 lr=0.068440
Epoch[095] Batch [1949]/[3760]	Speed: 117.403107 samples/sec	 accuracy=34.973558	 loss=2.953697	 lr=0.068428
Epoch[095] Batch [1999]/[3760]	Speed: 116.963669 samples/sec	 accuracy=34.953906	 loss=2.953862	 lr=0.068416
Epoch[095] Batch [2049]/[3760]	Speed: 118.159478 samples/sec	 accuracy=34.935213	 loss=2.954540	 lr=0.068404
Epoch[095] Batch [2099]/[3760]	Speed: 117.340547 samples/sec	 accuracy=34.916667	 loss=2.955902	 lr=0.068392
Epoch[095] Batch [2149]/[3760]	Speed: 117.298516 samples/sec	 accuracy=34.882267	 loss=2.956635	 lr=0.068380
Epoch[095] Batch [2199]/[3760]	Speed: 117.345844 samples/sec	 accuracy=34.882812	 loss=2.956937	 lr=0.068368
Epoch[095] Batch [2249]/[3760]	Speed: 117.201856 samples/sec	 accuracy=34.902083	 loss=2.956044	 lr=0.068356
Epoch[095] Batch [2299]/[3760]	Speed: 117.968631 samples/sec	 accuracy=34.897418	 loss=2.956037	 lr=0.068344
Epoch[095] Batch [2349]/[3760]	Speed: 118.050636 samples/sec	 accuracy=34.876995	 loss=2.957900	 lr=0.068332
Epoch[095] Batch [2399]/[3760]	Speed: 117.321832 samples/sec	 accuracy=34.910156	 loss=2.956940	 lr=0.068320
Epoch[095] Batch [2449]/[3760]	Speed: 118.066063 samples/sec	 accuracy=34.905612	 loss=2.957872	 lr=0.068308
Epoch[095] Batch [2499]/[3760]	Speed: 117.494219 samples/sec	 accuracy=34.905625	 loss=2.959417	 lr=0.068296
Epoch[095] Batch [2549]/[3760]	Speed: 118.004626 samples/sec	 accuracy=34.888480	 loss=2.959631	 lr=0.068284
Epoch[095] Batch [2599]/[3760]	Speed: 118.036348 samples/sec	 accuracy=34.866587	 loss=2.959564	 lr=0.068272
Epoch[095] Batch [2649]/[3760]	Speed: 116.829784 samples/sec	 accuracy=34.866156	 loss=2.959446	 lr=0.068260
Epoch[095] Batch [2699]/[3760]	Speed: 116.196640 samples/sec	 accuracy=34.877894	 loss=2.959251	 lr=0.068248
Epoch[095] Batch [2749]/[3760]	Speed: 117.594967 samples/sec	 accuracy=34.870455	 loss=2.960291	 lr=0.068236
Epoch[095] Batch [2799]/[3760]	Speed: 117.142229 samples/sec	 accuracy=34.864397	 loss=2.960388	 lr=0.068224
Epoch[095] Batch [2849]/[3760]	Speed: 117.362600 samples/sec	 accuracy=34.859101	 loss=2.959896	 lr=0.068212
Epoch[095] Batch [2899]/[3760]	Speed: 117.170081 samples/sec	 accuracy=34.860453	 loss=2.959799	 lr=0.068200
Epoch[095] Batch [2949]/[3760]	Speed: 117.764818 samples/sec	 accuracy=34.855932	 loss=2.960204	 lr=0.068188
Epoch[095] Batch [2999]/[3760]	Speed: 117.474461 samples/sec	 accuracy=34.844271	 loss=2.960214	 lr=0.068176
Epoch[095] Batch [3049]/[3760]	Speed: 117.883360 samples/sec	 accuracy=34.826844	 loss=2.960632	 lr=0.068164
Epoch[095] Batch [3099]/[3760]	Speed: 118.310494 samples/sec	 accuracy=34.835685	 loss=2.960432	 lr=0.068152
Epoch[095] Batch [3149]/[3760]	Speed: 116.804855 samples/sec	 accuracy=34.822917	 loss=2.960985	 lr=0.068140
Epoch[095] Batch [3199]/[3760]	Speed: 117.444389 samples/sec	 accuracy=34.824219	 loss=2.960758	 lr=0.068128
Epoch[095] Batch [3249]/[3760]	Speed: 117.025729 samples/sec	 accuracy=34.841827	 loss=2.960896	 lr=0.068116
Epoch[095] Batch [3299]/[3760]	Speed: 118.177369 samples/sec	 accuracy=34.839489	 loss=2.960904	 lr=0.068104
Epoch[095] Batch [3349]/[3760]	Speed: 117.558785 samples/sec	 accuracy=34.854944	 loss=2.960904	 lr=0.068092
Epoch[095] Batch [3399]/[3760]	Speed: 117.150974 samples/sec	 accuracy=34.851103	 loss=2.961713	 lr=0.068080
Epoch[095] Batch [3449]/[3760]	Speed: 117.525619 samples/sec	 accuracy=34.865942	 loss=2.961423	 lr=0.068068
Epoch[095] Batch [3499]/[3760]	Speed: 117.442841 samples/sec	 accuracy=34.879018	 loss=2.960821	 lr=0.068056
Epoch[095] Batch [3549]/[3760]	Speed: 117.529629 samples/sec	 accuracy=34.885563	 loss=2.960607	 lr=0.068044
Epoch[095] Batch [3599]/[3760]	Speed: 117.379532 samples/sec	 accuracy=34.888021	 loss=2.960255	 lr=0.068032
Epoch[095] Batch [3649]/[3760]	Speed: 117.308482 samples/sec	 accuracy=34.885702	 loss=2.960101	 lr=0.068020
Epoch[095] Batch [3699]/[3760]	Speed: 117.134638 samples/sec	 accuracy=34.887247	 loss=2.960056	 lr=0.068008
Epoch[095] Batch [3749]/[3760]	Speed: 120.202133 samples/sec	 accuracy=34.890000	 loss=2.960468	 lr=0.067996
Batch [0049]/[0303]: acc-top1=35.468750 acc-top5=61.468750
Batch [0099]/[0303]: acc-top1=31.953125 acc-top5=58.281250
Batch [0149]/[0303]: acc-top1=33.406250 acc-top5=59.718750
Batch [0199]/[0303]: acc-top1=35.507812 acc-top5=61.148438
Batch [0249]/[0303]: acc-top1=34.762500 acc-top5=61.275000
Batch [0299]/[0303]: acc-top1=34.536458 acc-top5=61.541667
[Epoch 095] training: accuracy=34.889461	 loss=2.960623
[Epoch 095] speed: 116 samples/sec	time cost: 2175.069653
[Epoch 095] validation: acc-top1=34.576114 acc-top5=61.545998 loss=3.185577
Epoch[096] Batch [0049]/[3759]	Speed: 88.468764 samples/sec	 accuracy=35.468750	 loss=2.979218	 lr=0.067982
Epoch[096] Batch [0099]/[3759]	Speed: 117.721974 samples/sec	 accuracy=35.937500	 loss=2.933112	 lr=0.067970
Epoch[096] Batch [0149]/[3759]	Speed: 117.492440 samples/sec	 accuracy=35.958333	 loss=2.921243	 lr=0.067958
Epoch[096] Batch [0199]/[3759]	Speed: 117.519044 samples/sec	 accuracy=35.468750	 loss=2.943812	 lr=0.067945
Epoch[096] Batch [0249]/[3759]	Speed: 117.571806 samples/sec	 accuracy=35.356250	 loss=2.948689	 lr=0.067933
Epoch[096] Batch [0299]/[3759]	Speed: 116.682886 samples/sec	 accuracy=35.380208	 loss=2.949951	 lr=0.067921
Epoch[096] Batch [0349]/[3759]	Speed: 117.698947 samples/sec	 accuracy=35.401786	 loss=2.941578	 lr=0.067909
Epoch[096] Batch [0399]/[3759]	Speed: 116.864507 samples/sec	 accuracy=35.320312	 loss=2.944771	 lr=0.067897
Epoch[096] Batch [0449]/[3759]	Speed: 117.185850 samples/sec	 accuracy=35.527778	 loss=2.939485	 lr=0.067885
Epoch[096] Batch [0499]/[3759]	Speed: 117.286488 samples/sec	 accuracy=35.503125	 loss=2.942112	 lr=0.067873
Epoch[096] Batch [0549]/[3759]	Speed: 117.173456 samples/sec	 accuracy=35.562500	 loss=2.934198	 lr=0.067861
Epoch[096] Batch [0599]/[3759]	Speed: 117.446588 samples/sec	 accuracy=35.468750	 loss=2.936647	 lr=0.067849
Epoch[096] Batch [0649]/[3759]	Speed: 116.540263 samples/sec	 accuracy=35.521635	 loss=2.933502	 lr=0.067837
Epoch[096] Batch [0699]/[3759]	Speed: 116.848566 samples/sec	 accuracy=35.524554	 loss=2.934360	 lr=0.067825
Epoch[096] Batch [0749]/[3759]	Speed: 116.927096 samples/sec	 accuracy=35.470833	 loss=2.934763	 lr=0.067813
Epoch[096] Batch [0799]/[3759]	Speed: 117.349555 samples/sec	 accuracy=35.568359	 loss=2.933022	 lr=0.067801
Epoch[096] Batch [0849]/[3759]	Speed: 117.184139 samples/sec	 accuracy=35.505515	 loss=2.933481	 lr=0.067789
Epoch[096] Batch [0899]/[3759]	Speed: 117.247154 samples/sec	 accuracy=35.552083	 loss=2.930086	 lr=0.067777
Epoch[096] Batch [0949]/[3759]	Speed: 117.642902 samples/sec	 accuracy=35.549342	 loss=2.930406	 lr=0.067765
Epoch[096] Batch [0999]/[3759]	Speed: 117.532347 samples/sec	 accuracy=35.634375	 loss=2.928184	 lr=0.067753
Epoch[096] Batch [1049]/[3759]	Speed: 116.688550 samples/sec	 accuracy=35.626488	 loss=2.927734	 lr=0.067741
Epoch[096] Batch [1099]/[3759]	Speed: 116.866718 samples/sec	 accuracy=35.700284	 loss=2.926117	 lr=0.067729
Epoch[096] Batch [1149]/[3759]	Speed: 116.566973 samples/sec	 accuracy=35.701087	 loss=2.928902	 lr=0.067717
Epoch[096] Batch [1199]/[3759]	Speed: 116.574638 samples/sec	 accuracy=35.657552	 loss=2.930661	 lr=0.067704
Epoch[096] Batch [1249]/[3759]	Speed: 117.024821 samples/sec	 accuracy=35.625000	 loss=2.931867	 lr=0.067692
Epoch[096] Batch [1299]/[3759]	Speed: 117.418412 samples/sec	 accuracy=35.643029	 loss=2.932145	 lr=0.067680
Epoch[096] Batch [1349]/[3759]	Speed: 117.485320 samples/sec	 accuracy=35.633102	 loss=2.932247	 lr=0.067668
Epoch[096] Batch [1399]/[3759]	Speed: 117.442416 samples/sec	 accuracy=35.622768	 loss=2.933465	 lr=0.067656
Epoch[096] Batch [1449]/[3759]	Speed: 117.041829 samples/sec	 accuracy=35.595905	 loss=2.933997	 lr=0.067644
Epoch[096] Batch [1499]/[3759]	Speed: 118.186664 samples/sec	 accuracy=35.604167	 loss=2.934125	 lr=0.067632
Epoch[096] Batch [1549]/[3759]	Speed: 116.828811 samples/sec	 accuracy=35.607863	 loss=2.934371	 lr=0.067620
Epoch[096] Batch [1599]/[3759]	Speed: 116.921111 samples/sec	 accuracy=35.556641	 loss=2.934777	 lr=0.067608
Epoch[096] Batch [1649]/[3759]	Speed: 117.680028 samples/sec	 accuracy=35.546402	 loss=2.934534	 lr=0.067596
Epoch[096] Batch [1699]/[3759]	Speed: 117.569323 samples/sec	 accuracy=35.528493	 loss=2.934925	 lr=0.067584
Epoch[096] Batch [1749]/[3759]	Speed: 117.857021 samples/sec	 accuracy=35.496429	 loss=2.935589	 lr=0.067572
Epoch[096] Batch [1799]/[3759]	Speed: 117.068320 samples/sec	 accuracy=35.480035	 loss=2.936328	 lr=0.067560
Epoch[096] Batch [1849]/[3759]	Speed: 117.761146 samples/sec	 accuracy=35.423142	 loss=2.938934	 lr=0.067548
Epoch[096] Batch [1899]/[3759]	Speed: 117.339096 samples/sec	 accuracy=35.391447	 loss=2.939736	 lr=0.067536
Epoch[096] Batch [1949]/[3759]	Speed: 117.301838 samples/sec	 accuracy=35.398237	 loss=2.940630	 lr=0.067523
Epoch[096] Batch [1999]/[3759]	Speed: 117.303295 samples/sec	 accuracy=35.373438	 loss=2.941031	 lr=0.067511
Epoch[096] Batch [2049]/[3759]	Speed: 116.922358 samples/sec	 accuracy=35.384909	 loss=2.941163	 lr=0.067499
Epoch[096] Batch [2099]/[3759]	Speed: 116.918149 samples/sec	 accuracy=35.358631	 loss=2.942340	 lr=0.067487
Epoch[096] Batch [2149]/[3759]	Speed: 117.429384 samples/sec	 accuracy=35.314680	 loss=2.943210	 lr=0.067475
Epoch[096] Batch [2199]/[3759]	Speed: 117.218761 samples/sec	 accuracy=35.352273	 loss=2.942728	 lr=0.067463
Epoch[096] Batch [2249]/[3759]	Speed: 117.126824 samples/sec	 accuracy=35.333333	 loss=2.944097	 lr=0.067451
Epoch[096] Batch [2299]/[3759]	Speed: 117.046431 samples/sec	 accuracy=35.321332	 loss=2.944270	 lr=0.067439
Epoch[096] Batch [2349]/[3759]	Speed: 117.052061 samples/sec	 accuracy=35.335106	 loss=2.943685	 lr=0.067427
Epoch[096] Batch [2399]/[3759]	Speed: 116.710383 samples/sec	 accuracy=35.333984	 loss=2.944104	 lr=0.067415
Epoch[096] Batch [2449]/[3759]	Speed: 115.941275 samples/sec	 accuracy=35.348214	 loss=2.943185	 lr=0.067403
Epoch[096] Batch [2499]/[3759]	Speed: 117.635220 samples/sec	 accuracy=35.328750	 loss=2.943999	 lr=0.067390
Epoch[096] Batch [2549]/[3759]	Speed: 116.916264 samples/sec	 accuracy=35.300858	 loss=2.945462	 lr=0.067378
Epoch[096] Batch [2599]/[3759]	Speed: 116.974766 samples/sec	 accuracy=35.302885	 loss=2.944897	 lr=0.067366
Epoch[096] Batch [2649]/[3759]	Speed: 118.336185 samples/sec	 accuracy=35.324292	 loss=2.944162	 lr=0.067354
Epoch[096] Batch [2699]/[3759]	Speed: 118.178875 samples/sec	 accuracy=35.337384	 loss=2.944291	 lr=0.067342
Epoch[096] Batch [2749]/[3759]	Speed: 117.094552 samples/sec	 accuracy=35.307955	 loss=2.944722	 lr=0.067330
Epoch[096] Batch [2799]/[3759]	Speed: 117.778919 samples/sec	 accuracy=35.303571	 loss=2.945751	 lr=0.067318
Epoch[096] Batch [2849]/[3759]	Speed: 118.015855 samples/sec	 accuracy=35.279057	 loss=2.946587	 lr=0.067306
Epoch[096] Batch [2899]/[3759]	Speed: 117.929233 samples/sec	 accuracy=35.267241	 loss=2.947082	 lr=0.067294
Epoch[096] Batch [2949]/[3759]	Speed: 117.522913 samples/sec	 accuracy=35.249470	 loss=2.947449	 lr=0.067282
Epoch[096] Batch [2999]/[3759]	Speed: 118.083974 samples/sec	 accuracy=35.227083	 loss=2.948944	 lr=0.067270
Epoch[096] Batch [3049]/[3759]	Speed: 117.479339 samples/sec	 accuracy=35.224898	 loss=2.948660	 lr=0.067257
Epoch[096] Batch [3099]/[3759]	Speed: 117.273913 samples/sec	 accuracy=35.215726	 loss=2.948854	 lr=0.067245
Epoch[096] Batch [3149]/[3759]	Speed: 117.427183 samples/sec	 accuracy=35.186508	 loss=2.949318	 lr=0.067233
Epoch[096] Batch [3199]/[3759]	Speed: 117.608221 samples/sec	 accuracy=35.173828	 loss=2.949475	 lr=0.067221
Epoch[096] Batch [3249]/[3759]	Speed: 116.583647 samples/sec	 accuracy=35.144231	 loss=2.950181	 lr=0.067209
Epoch[096] Batch [3299]/[3759]	Speed: 116.527661 samples/sec	 accuracy=35.152462	 loss=2.950524	 lr=0.067197
Epoch[096] Batch [3349]/[3759]	Speed: 117.274912 samples/sec	 accuracy=35.153451	 loss=2.950771	 lr=0.067185
Epoch[096] Batch [3399]/[3759]	Speed: 117.244214 samples/sec	 accuracy=35.170956	 loss=2.950779	 lr=0.067173
Epoch[096] Batch [3449]/[3759]	Speed: 118.294268 samples/sec	 accuracy=35.158062	 loss=2.951399	 lr=0.067161
Epoch[096] Batch [3499]/[3759]	Speed: 117.674960 samples/sec	 accuracy=35.158036	 loss=2.951751	 lr=0.067148
Epoch[096] Batch [3549]/[3759]	Speed: 116.677229 samples/sec	 accuracy=35.137324	 loss=2.951968	 lr=0.067136
Epoch[096] Batch [3599]/[3759]	Speed: 117.405078 samples/sec	 accuracy=35.138889	 loss=2.951641	 lr=0.067124
Epoch[096] Batch [3649]/[3759]	Speed: 116.434401 samples/sec	 accuracy=35.126712	 loss=2.952100	 lr=0.067112
Epoch[096] Batch [3699]/[3759]	Speed: 118.185708 samples/sec	 accuracy=35.118666	 loss=2.951583	 lr=0.067100
Epoch[096] Batch [3749]/[3759]	Speed: 120.123165 samples/sec	 accuracy=35.132917	 loss=2.951168	 lr=0.067088
Batch [0049]/[0303]: acc-top1=35.000000 acc-top5=60.187500
Batch [0099]/[0303]: acc-top1=33.468750 acc-top5=58.734375
Batch [0149]/[0303]: acc-top1=35.510417 acc-top5=60.354167
Batch [0199]/[0303]: acc-top1=37.281250 acc-top5=62.375000
Batch [0249]/[0303]: acc-top1=36.737500 acc-top5=62.287500
Batch [0299]/[0303]: acc-top1=36.218750 acc-top5=62.114583
[Epoch 096] training: accuracy=35.131102	 loss=2.951404
[Epoch 096] speed: 116 samples/sec	time cost: 2177.927941
[Epoch 096] validation: acc-top1=36.143771 acc-top5=62.144183 loss=2.969986
Epoch[097] Batch [0049]/[3760]	Speed: 89.347652 samples/sec	 accuracy=36.062500	 loss=2.904113	 lr=0.067074
Epoch[097] Batch [0099]/[3760]	Speed: 117.632955 samples/sec	 accuracy=35.375000	 loss=2.928287	 lr=0.067061
Epoch[097] Batch [0149]/[3760]	Speed: 117.118216 samples/sec	 accuracy=35.093750	 loss=2.947251	 lr=0.067049
Epoch[097] Batch [0199]/[3760]	Speed: 117.595012 samples/sec	 accuracy=35.007812	 loss=2.948129	 lr=0.067037
Epoch[097] Batch [0249]/[3760]	Speed: 117.250717 samples/sec	 accuracy=35.231250	 loss=2.931034	 lr=0.067025
Epoch[097] Batch [0299]/[3760]	Speed: 117.048210 samples/sec	 accuracy=35.229167	 loss=2.930539	 lr=0.067013
Epoch[097] Batch [0349]/[3760]	Speed: 117.158798 samples/sec	 accuracy=35.392857	 loss=2.921370	 lr=0.067001
Epoch[097] Batch [0399]/[3760]	Speed: 115.777668 samples/sec	 accuracy=35.277344	 loss=2.929085	 lr=0.066989
Epoch[097] Batch [0449]/[3760]	Speed: 118.078554 samples/sec	 accuracy=35.263889	 loss=2.929282	 lr=0.066977
Epoch[097] Batch [0499]/[3760]	Speed: 116.893532 samples/sec	 accuracy=35.250000	 loss=2.933757	 lr=0.066964
Epoch[097] Batch [0549]/[3760]	Speed: 117.555716 samples/sec	 accuracy=35.176136	 loss=2.937495	 lr=0.066952
Epoch[097] Batch [0599]/[3760]	Speed: 116.882669 samples/sec	 accuracy=35.276042	 loss=2.937542	 lr=0.066940
Epoch[097] Batch [0649]/[3760]	Speed: 116.970655 samples/sec	 accuracy=35.360577	 loss=2.934967	 lr=0.066928
Epoch[097] Batch [0699]/[3760]	Speed: 117.506749 samples/sec	 accuracy=35.272321	 loss=2.938735	 lr=0.066916
Epoch[097] Batch [0749]/[3760]	Speed: 116.789091 samples/sec	 accuracy=35.385417	 loss=2.934374	 lr=0.066904
Epoch[097] Batch [0799]/[3760]	Speed: 117.680845 samples/sec	 accuracy=35.431641	 loss=2.931080	 lr=0.066892
Epoch[097] Batch [0849]/[3760]	Speed: 116.655702 samples/sec	 accuracy=35.457721	 loss=2.930037	 lr=0.066879
Epoch[097] Batch [0899]/[3760]	Speed: 116.582092 samples/sec	 accuracy=35.368056	 loss=2.933154	 lr=0.066867
Epoch[097] Batch [0949]/[3760]	Speed: 117.590975 samples/sec	 accuracy=35.310855	 loss=2.934171	 lr=0.066855
Epoch[097] Batch [0999]/[3760]	Speed: 117.670485 samples/sec	 accuracy=35.368750	 loss=2.931658	 lr=0.066843
Epoch[097] Batch [1049]/[3760]	Speed: 118.160957 samples/sec	 accuracy=35.316964	 loss=2.935839	 lr=0.066831
Epoch[097] Batch [1099]/[3760]	Speed: 117.140937 samples/sec	 accuracy=35.305398	 loss=2.938685	 lr=0.066819
Epoch[097] Batch [1149]/[3760]	Speed: 116.997865 samples/sec	 accuracy=35.288043	 loss=2.939272	 lr=0.066807
Epoch[097] Batch [1199]/[3760]	Speed: 116.639340 samples/sec	 accuracy=35.276042	 loss=2.939129	 lr=0.066794
Epoch[097] Batch [1249]/[3760]	Speed: 117.700162 samples/sec	 accuracy=35.232500	 loss=2.939781	 lr=0.066782
Epoch[097] Batch [1299]/[3760]	Speed: 116.943287 samples/sec	 accuracy=35.289663	 loss=2.937873	 lr=0.066770
Epoch[097] Batch [1349]/[3760]	Speed: 117.791275 samples/sec	 accuracy=35.291667	 loss=2.938223	 lr=0.066758
Epoch[097] Batch [1399]/[3760]	Speed: 117.712378 samples/sec	 accuracy=35.306920	 loss=2.938597	 lr=0.066746
Epoch[097] Batch [1449]/[3760]	Speed: 117.586908 samples/sec	 accuracy=35.298491	 loss=2.938631	 lr=0.066734
Epoch[097] Batch [1499]/[3760]	Speed: 117.770546 samples/sec	 accuracy=35.360417	 loss=2.935435	 lr=0.066722
Epoch[097] Batch [1549]/[3760]	Speed: 116.941058 samples/sec	 accuracy=35.351815	 loss=2.935675	 lr=0.066709
Epoch[097] Batch [1599]/[3760]	Speed: 117.688031 samples/sec	 accuracy=35.320312	 loss=2.936280	 lr=0.066697
Epoch[097] Batch [1649]/[3760]	Speed: 116.651842 samples/sec	 accuracy=35.313447	 loss=2.937310	 lr=0.066685
Epoch[097] Batch [1699]/[3760]	Speed: 117.937439 samples/sec	 accuracy=35.246324	 loss=2.939611	 lr=0.066673
Epoch[097] Batch [1749]/[3760]	Speed: 116.944398 samples/sec	 accuracy=35.252679	 loss=2.940259	 lr=0.066661
Epoch[097] Batch [1799]/[3760]	Speed: 117.788213 samples/sec	 accuracy=35.225694	 loss=2.941271	 lr=0.066649
Epoch[097] Batch [1849]/[3760]	Speed: 117.575649 samples/sec	 accuracy=35.212838	 loss=2.942414	 lr=0.066636
Epoch[097] Batch [1899]/[3760]	Speed: 116.619957 samples/sec	 accuracy=35.230263	 loss=2.941124	 lr=0.066624
Epoch[097] Batch [1949]/[3760]	Speed: 117.359388 samples/sec	 accuracy=35.233974	 loss=2.940573	 lr=0.066612
Epoch[097] Batch [1999]/[3760]	Speed: 116.968962 samples/sec	 accuracy=35.235937	 loss=2.941390	 lr=0.066600
Epoch[097] Batch [2049]/[3760]	Speed: 117.548765 samples/sec	 accuracy=35.237043	 loss=2.940795	 lr=0.066588
Epoch[097] Batch [2099]/[3760]	Speed: 117.364919 samples/sec	 accuracy=35.235119	 loss=2.941355	 lr=0.066576
Epoch[097] Batch [2149]/[3760]	Speed: 116.112540 samples/sec	 accuracy=35.230378	 loss=2.940900	 lr=0.066563
Epoch[097] Batch [2199]/[3760]	Speed: 117.580245 samples/sec	 accuracy=35.243608	 loss=2.941224	 lr=0.066551
Epoch[097] Batch [2249]/[3760]	Speed: 117.556733 samples/sec	 accuracy=35.206944	 loss=2.942470	 lr=0.066539
Epoch[097] Batch [2299]/[3760]	Speed: 117.853686 samples/sec	 accuracy=35.221467	 loss=2.941138	 lr=0.066527
Epoch[097] Batch [2349]/[3760]	Speed: 117.654290 samples/sec	 accuracy=35.240691	 loss=2.940793	 lr=0.066515
Epoch[097] Batch [2399]/[3760]	Speed: 117.373772 samples/sec	 accuracy=35.234375	 loss=2.941038	 lr=0.066503
Epoch[097] Batch [2449]/[3760]	Speed: 116.900239 samples/sec	 accuracy=35.256378	 loss=2.940824	 lr=0.066490
Epoch[097] Batch [2499]/[3760]	Speed: 117.403606 samples/sec	 accuracy=35.219375	 loss=2.942810	 lr=0.066478
Epoch[097] Batch [2549]/[3760]	Speed: 117.679932 samples/sec	 accuracy=35.226716	 loss=2.943627	 lr=0.066466
Epoch[097] Batch [2599]/[3760]	Speed: 117.478605 samples/sec	 accuracy=35.224760	 loss=2.944624	 lr=0.066454
Epoch[097] Batch [2649]/[3760]	Speed: 118.308913 samples/sec	 accuracy=35.208137	 loss=2.945641	 lr=0.066442
Epoch[097] Batch [2699]/[3760]	Speed: 117.849640 samples/sec	 accuracy=35.173611	 loss=2.947049	 lr=0.066429
Epoch[097] Batch [2749]/[3760]	Speed: 118.050635 samples/sec	 accuracy=35.142614	 loss=2.948345	 lr=0.066417
Epoch[097] Batch [2799]/[3760]	Speed: 117.494876 samples/sec	 accuracy=35.146205	 loss=2.948379	 lr=0.066405
Epoch[097] Batch [2849]/[3760]	Speed: 118.174692 samples/sec	 accuracy=35.162829	 loss=2.947514	 lr=0.066393
Epoch[097] Batch [2899]/[3760]	Speed: 118.100612 samples/sec	 accuracy=35.165409	 loss=2.947502	 lr=0.066381
Epoch[097] Batch [2949]/[3760]	Speed: 117.265281 samples/sec	 accuracy=35.171610	 loss=2.947264	 lr=0.066369
Epoch[097] Batch [2999]/[3760]	Speed: 117.065929 samples/sec	 accuracy=35.189583	 loss=2.946244	 lr=0.066356
Epoch[097] Batch [3049]/[3760]	Speed: 117.737154 samples/sec	 accuracy=35.190061	 loss=2.946197	 lr=0.066344
Epoch[097] Batch [3099]/[3760]	Speed: 117.739179 samples/sec	 accuracy=35.199093	 loss=2.945911	 lr=0.066332
Epoch[097] Batch [3149]/[3760]	Speed: 117.763339 samples/sec	 accuracy=35.200397	 loss=2.946084	 lr=0.066320
Epoch[097] Batch [3199]/[3760]	Speed: 117.143491 samples/sec	 accuracy=35.194336	 loss=2.945906	 lr=0.066308
Epoch[097] Batch [3249]/[3760]	Speed: 117.370813 samples/sec	 accuracy=35.205288	 loss=2.944628	 lr=0.066295
Epoch[097] Batch [3299]/[3760]	Speed: 117.591059 samples/sec	 accuracy=35.177557	 loss=2.945441	 lr=0.066283
Epoch[097] Batch [3349]/[3760]	Speed: 117.936648 samples/sec	 accuracy=35.145056	 loss=2.946973	 lr=0.066271
Epoch[097] Batch [3399]/[3760]	Speed: 117.154486 samples/sec	 accuracy=35.154412	 loss=2.947115	 lr=0.066259
Epoch[097] Batch [3449]/[3760]	Speed: 118.002798 samples/sec	 accuracy=35.146739	 loss=2.946959	 lr=0.066247
Epoch[097] Batch [3499]/[3760]	Speed: 117.814513 samples/sec	 accuracy=35.145089	 loss=2.946459	 lr=0.066234
Epoch[097] Batch [3549]/[3760]	Speed: 117.533223 samples/sec	 accuracy=35.145687	 loss=2.946626	 lr=0.066222
Epoch[097] Batch [3599]/[3760]	Speed: 117.581366 samples/sec	 accuracy=35.131944	 loss=2.947101	 lr=0.066210
Epoch[097] Batch [3649]/[3760]	Speed: 117.951054 samples/sec	 accuracy=35.127140	 loss=2.946824	 lr=0.066198
Epoch[097] Batch [3699]/[3760]	Speed: 116.811642 samples/sec	 accuracy=35.111909	 loss=2.947699	 lr=0.066186
Epoch[097] Batch [3749]/[3760]	Speed: 121.155871 samples/sec	 accuracy=35.102500	 loss=2.948083	 lr=0.066173
Batch [0049]/[0303]: acc-top1=36.437500 acc-top5=61.500000
Batch [0099]/[0303]: acc-top1=33.390625 acc-top5=59.843750
Batch [0149]/[0303]: acc-top1=35.343750 acc-top5=61.322917
Batch [0199]/[0303]: acc-top1=38.296875 acc-top5=63.828125
Batch [0249]/[0303]: acc-top1=37.206250 acc-top5=63.543750
Batch [0299]/[0303]: acc-top1=36.458333 acc-top5=63.541667
[Epoch 097] training: accuracy=35.103474	 loss=2.947732
[Epoch 097] speed: 116 samples/sec	time cost: 2174.682642
[Epoch 097] validation: acc-top1=36.478960 acc-top5=63.582921 loss=3.008177
Epoch[098] Batch [0049]/[3759]	Speed: 87.686267 samples/sec	 accuracy=35.906250	 loss=2.886578	 lr=0.066159
Epoch[098] Batch [0099]/[3759]	Speed: 116.437878 samples/sec	 accuracy=36.218750	 loss=2.875015	 lr=0.066147
Epoch[098] Batch [0149]/[3759]	Speed: 117.239451 samples/sec	 accuracy=36.135417	 loss=2.891496	 lr=0.066134
Epoch[098] Batch [0199]/[3759]	Speed: 117.304611 samples/sec	 accuracy=35.921875	 loss=2.906364	 lr=0.066122
Epoch[098] Batch [0249]/[3759]	Speed: 117.792132 samples/sec	 accuracy=36.143750	 loss=2.900766	 lr=0.066110
Epoch[098] Batch [0299]/[3759]	Speed: 116.883792 samples/sec	 accuracy=36.171875	 loss=2.901132	 lr=0.066098
Epoch[098] Batch [0349]/[3759]	Speed: 117.497318 samples/sec	 accuracy=36.044643	 loss=2.902986	 lr=0.066086
Epoch[098] Batch [0399]/[3759]	Speed: 117.929679 samples/sec	 accuracy=36.058594	 loss=2.906089	 lr=0.066073
Epoch[098] Batch [0449]/[3759]	Speed: 117.281982 samples/sec	 accuracy=36.079861	 loss=2.905517	 lr=0.066061
Epoch[098] Batch [0499]/[3759]	Speed: 117.653348 samples/sec	 accuracy=35.959375	 loss=2.907275	 lr=0.066049
Epoch[098] Batch [0549]/[3759]	Speed: 117.325837 samples/sec	 accuracy=35.934659	 loss=2.913446	 lr=0.066037
Epoch[098] Batch [0599]/[3759]	Speed: 117.156343 samples/sec	 accuracy=35.835938	 loss=2.918934	 lr=0.066024
Epoch[098] Batch [0649]/[3759]	Speed: 117.877430 samples/sec	 accuracy=35.939904	 loss=2.912774	 lr=0.066012
Epoch[098] Batch [0699]/[3759]	Speed: 116.788881 samples/sec	 accuracy=36.020089	 loss=2.908615	 lr=0.066000
Epoch[098] Batch [0749]/[3759]	Speed: 117.274340 samples/sec	 accuracy=35.958333	 loss=2.911572	 lr=0.065988
Epoch[098] Batch [0799]/[3759]	Speed: 117.192461 samples/sec	 accuracy=35.802734	 loss=2.917333	 lr=0.065976
Epoch[098] Batch [0849]/[3759]	Speed: 117.312599 samples/sec	 accuracy=35.707721	 loss=2.923905	 lr=0.065963
Epoch[098] Batch [0899]/[3759]	Speed: 117.516581 samples/sec	 accuracy=35.765625	 loss=2.921396	 lr=0.065951
Epoch[098] Batch [0949]/[3759]	Speed: 117.125769 samples/sec	 accuracy=35.736842	 loss=2.923127	 lr=0.065939
Epoch[098] Batch [0999]/[3759]	Speed: 116.306303 samples/sec	 accuracy=35.796875	 loss=2.919601	 lr=0.065927
Epoch[098] Batch [1049]/[3759]	Speed: 117.028716 samples/sec	 accuracy=35.760417	 loss=2.917914	 lr=0.065914
Epoch[098] Batch [1099]/[3759]	Speed: 117.839560 samples/sec	 accuracy=35.718750	 loss=2.919773	 lr=0.065902
Epoch[098] Batch [1149]/[3759]	Speed: 117.213880 samples/sec	 accuracy=35.762228	 loss=2.917364	 lr=0.065890
Epoch[098] Batch [1199]/[3759]	Speed: 117.230799 samples/sec	 accuracy=35.694010	 loss=2.920144	 lr=0.065878
Epoch[098] Batch [1249]/[3759]	Speed: 118.398028 samples/sec	 accuracy=35.625000	 loss=2.921529	 lr=0.065866
Epoch[098] Batch [1299]/[3759]	Speed: 117.599007 samples/sec	 accuracy=35.575721	 loss=2.925535	 lr=0.065853
Epoch[098] Batch [1349]/[3759]	Speed: 117.535610 samples/sec	 accuracy=35.539352	 loss=2.926686	 lr=0.065841
Epoch[098] Batch [1399]/[3759]	Speed: 117.842407 samples/sec	 accuracy=35.520089	 loss=2.927304	 lr=0.065829
Epoch[098] Batch [1449]/[3759]	Speed: 117.602179 samples/sec	 accuracy=35.571121	 loss=2.925393	 lr=0.065817
Epoch[098] Batch [1499]/[3759]	Speed: 117.135951 samples/sec	 accuracy=35.541667	 loss=2.926173	 lr=0.065804
Epoch[098] Batch [1549]/[3759]	Speed: 118.295929 samples/sec	 accuracy=35.565524	 loss=2.925066	 lr=0.065792
Epoch[098] Batch [1599]/[3759]	Speed: 117.247145 samples/sec	 accuracy=35.562500	 loss=2.925037	 lr=0.065780
Epoch[098] Batch [1649]/[3759]	Speed: 118.049735 samples/sec	 accuracy=35.539773	 loss=2.926025	 lr=0.065768
Epoch[098] Batch [1699]/[3759]	Speed: 117.231794 samples/sec	 accuracy=35.494485	 loss=2.927263	 lr=0.065755
Epoch[098] Batch [1749]/[3759]	Speed: 117.045575 samples/sec	 accuracy=35.502679	 loss=2.927666	 lr=0.065743
Epoch[098] Batch [1799]/[3759]	Speed: 117.689453 samples/sec	 accuracy=35.478299	 loss=2.928466	 lr=0.065731
Epoch[098] Batch [1849]/[3759]	Speed: 116.251758 samples/sec	 accuracy=35.467061	 loss=2.930089	 lr=0.065719
Epoch[098] Batch [1899]/[3759]	Speed: 117.701659 samples/sec	 accuracy=35.461349	 loss=2.930530	 lr=0.065706
Epoch[098] Batch [1949]/[3759]	Speed: 117.683947 samples/sec	 accuracy=35.485577	 loss=2.930615	 lr=0.065694
Epoch[098] Batch [1999]/[3759]	Speed: 117.893896 samples/sec	 accuracy=35.478906	 loss=2.929985	 lr=0.065682
Epoch[098] Batch [2049]/[3759]	Speed: 117.749663 samples/sec	 accuracy=35.477896	 loss=2.929647	 lr=0.065670
Epoch[098] Batch [2099]/[3759]	Speed: 117.570315 samples/sec	 accuracy=35.490327	 loss=2.929229	 lr=0.065657
Epoch[098] Batch [2149]/[3759]	Speed: 117.847602 samples/sec	 accuracy=35.497093	 loss=2.929975	 lr=0.065645
Epoch[098] Batch [2199]/[3759]	Speed: 117.989206 samples/sec	 accuracy=35.488636	 loss=2.929838	 lr=0.065633
Epoch[098] Batch [2249]/[3759]	Speed: 118.096208 samples/sec	 accuracy=35.478472	 loss=2.930984	 lr=0.065621
Epoch[098] Batch [2299]/[3759]	Speed: 117.394565 samples/sec	 accuracy=35.457201	 loss=2.931202	 lr=0.065608
Epoch[098] Batch [2349]/[3759]	Speed: 116.718738 samples/sec	 accuracy=35.454787	 loss=2.930550	 lr=0.065596
Epoch[098] Batch [2399]/[3759]	Speed: 117.533232 samples/sec	 accuracy=35.441406	 loss=2.932398	 lr=0.065584
Epoch[098] Batch [2449]/[3759]	Speed: 118.165170 samples/sec	 accuracy=35.435587	 loss=2.932567	 lr=0.065572
Epoch[098] Batch [2499]/[3759]	Speed: 117.423194 samples/sec	 accuracy=35.427500	 loss=2.932124	 lr=0.065559
Epoch[098] Batch [2549]/[3759]	Speed: 118.290369 samples/sec	 accuracy=35.414216	 loss=2.933239	 lr=0.065547
Epoch[098] Batch [2599]/[3759]	Speed: 118.031221 samples/sec	 accuracy=35.430288	 loss=2.932025	 lr=0.065535
Epoch[098] Batch [2649]/[3759]	Speed: 117.861842 samples/sec	 accuracy=35.429835	 loss=2.932375	 lr=0.065523
Epoch[098] Batch [2699]/[3759]	Speed: 118.482151 samples/sec	 accuracy=35.419560	 loss=2.932976	 lr=0.065510
Epoch[098] Batch [2749]/[3759]	Speed: 116.326921 samples/sec	 accuracy=35.428409	 loss=2.932790	 lr=0.065498
Epoch[098] Batch [2799]/[3759]	Speed: 117.646154 samples/sec	 accuracy=35.434710	 loss=2.932522	 lr=0.065486
Epoch[098] Batch [2849]/[3759]	Speed: 118.448172 samples/sec	 accuracy=35.403509	 loss=2.934154	 lr=0.065474
Epoch[098] Batch [2899]/[3759]	Speed: 117.624564 samples/sec	 accuracy=35.394397	 loss=2.935042	 lr=0.065461
Epoch[098] Batch [2949]/[3759]	Speed: 117.823175 samples/sec	 accuracy=35.362818	 loss=2.936271	 lr=0.065449
Epoch[098] Batch [2999]/[3759]	Speed: 117.412977 samples/sec	 accuracy=35.358333	 loss=2.936439	 lr=0.065437
Epoch[098] Batch [3049]/[3759]	Speed: 117.176451 samples/sec	 accuracy=35.345287	 loss=2.936048	 lr=0.065425
Epoch[098] Batch [3099]/[3759]	Speed: 116.975261 samples/sec	 accuracy=35.343750	 loss=2.936313	 lr=0.065412
Epoch[098] Batch [3149]/[3759]	Speed: 117.861892 samples/sec	 accuracy=35.338294	 loss=2.936553	 lr=0.065400
Epoch[098] Batch [3199]/[3759]	Speed: 117.793240 samples/sec	 accuracy=35.340820	 loss=2.936823	 lr=0.065388
Epoch[098] Batch [3249]/[3759]	Speed: 117.184799 samples/sec	 accuracy=35.334135	 loss=2.937151	 lr=0.065375
Epoch[098] Batch [3299]/[3759]	Speed: 117.108137 samples/sec	 accuracy=35.330966	 loss=2.937605	 lr=0.065363
Epoch[098] Batch [3349]/[3759]	Speed: 117.854806 samples/sec	 accuracy=35.321362	 loss=2.937769	 lr=0.065351
Epoch[098] Batch [3399]/[3759]	Speed: 117.792038 samples/sec	 accuracy=35.334559	 loss=2.937641	 lr=0.065339
Epoch[098] Batch [3449]/[3759]	Speed: 117.535013 samples/sec	 accuracy=35.324728	 loss=2.938691	 lr=0.065326
Epoch[098] Batch [3499]/[3759]	Speed: 117.222129 samples/sec	 accuracy=35.321429	 loss=2.939017	 lr=0.065314
Epoch[098] Batch [3549]/[3759]	Speed: 117.917991 samples/sec	 accuracy=35.311620	 loss=2.938824	 lr=0.065302
Epoch[098] Batch [3599]/[3759]	Speed: 116.463679 samples/sec	 accuracy=35.327257	 loss=2.938768	 lr=0.065290
Epoch[098] Batch [3649]/[3759]	Speed: 117.687846 samples/sec	 accuracy=35.339469	 loss=2.938892	 lr=0.065277
Epoch[098] Batch [3699]/[3759]	Speed: 117.644317 samples/sec	 accuracy=35.337416	 loss=2.938490	 lr=0.065265
Epoch[098] Batch [3749]/[3759]	Speed: 120.429240 samples/sec	 accuracy=35.342500	 loss=2.938749	 lr=0.065253
Batch [0049]/[0303]: acc-top1=36.156250 acc-top5=62.562500
Batch [0099]/[0303]: acc-top1=33.046875 acc-top5=59.765625
Batch [0149]/[0303]: acc-top1=34.385417 acc-top5=60.437500
Batch [0199]/[0303]: acc-top1=37.656250 acc-top5=63.187500
Batch [0249]/[0303]: acc-top1=37.381250 acc-top5=63.537500
Batch [0299]/[0303]: acc-top1=36.203125 acc-top5=63.135417
[Epoch 098] training: accuracy=35.346003	 loss=2.938422
[Epoch 098] speed: 116 samples/sec	time cost: 2175.073789
[Epoch 098] validation: acc-top1=36.221122 acc-top5=63.170380 loss=2.961499
Epoch[099] Batch [0049]/[3760]	Speed: 87.620141 samples/sec	 accuracy=34.406250	 loss=2.946379	 lr=0.065238
Epoch[099] Batch [0099]/[3760]	Speed: 116.659411 samples/sec	 accuracy=34.781250	 loss=2.907435	 lr=0.065226
Epoch[099] Batch [0149]/[3760]	Speed: 117.591218 samples/sec	 accuracy=35.364583	 loss=2.908248	 lr=0.065214
Epoch[099] Batch [0199]/[3760]	Speed: 118.057670 samples/sec	 accuracy=35.742188	 loss=2.896421	 lr=0.065201
Epoch[099] Batch [0249]/[3760]	Speed: 117.328170 samples/sec	 accuracy=36.206250	 loss=2.880781	 lr=0.065189
Epoch[099] Batch [0299]/[3760]	Speed: 117.002511 samples/sec	 accuracy=36.078125	 loss=2.890695	 lr=0.065177
Epoch[099] Batch [0349]/[3760]	Speed: 116.941240 samples/sec	 accuracy=35.950893	 loss=2.897628	 lr=0.065164
Epoch[099] Batch [0399]/[3760]	Speed: 117.308056 samples/sec	 accuracy=36.046875	 loss=2.898940	 lr=0.065152
Epoch[099] Batch [0449]/[3760]	Speed: 117.315164 samples/sec	 accuracy=36.125000	 loss=2.894810	 lr=0.065140
Epoch[099] Batch [0499]/[3760]	Speed: 116.932055 samples/sec	 accuracy=36.221875	 loss=2.893525	 lr=0.065128
Epoch[099] Batch [0549]/[3760]	Speed: 116.475356 samples/sec	 accuracy=36.181818	 loss=2.897899	 lr=0.065115
Epoch[099] Batch [0599]/[3760]	Speed: 116.805066 samples/sec	 accuracy=36.250000	 loss=2.895022	 lr=0.065103
Epoch[099] Batch [0649]/[3760]	Speed: 117.443383 samples/sec	 accuracy=36.223558	 loss=2.896030	 lr=0.065091
Epoch[099] Batch [0699]/[3760]	Speed: 116.399805 samples/sec	 accuracy=36.102679	 loss=2.901160	 lr=0.065078
Epoch[099] Batch [0749]/[3760]	Speed: 116.674742 samples/sec	 accuracy=36.133333	 loss=2.900454	 lr=0.065066
Epoch[099] Batch [0799]/[3760]	Speed: 117.813415 samples/sec	 accuracy=36.039062	 loss=2.901872	 lr=0.065054
Epoch[099] Batch [0849]/[3760]	Speed: 116.872195 samples/sec	 accuracy=36.115809	 loss=2.897410	 lr=0.065042
Epoch[099] Batch [0899]/[3760]	Speed: 117.457409 samples/sec	 accuracy=36.090278	 loss=2.898131	 lr=0.065029
Epoch[099] Batch [0949]/[3760]	Speed: 116.830066 samples/sec	 accuracy=36.041118	 loss=2.899208	 lr=0.065017
Epoch[099] Batch [0999]/[3760]	Speed: 116.470664 samples/sec	 accuracy=36.064062	 loss=2.898391	 lr=0.065005
Epoch[099] Batch [1049]/[3760]	Speed: 117.369053 samples/sec	 accuracy=36.099702	 loss=2.897366	 lr=0.064992
Epoch[099] Batch [1099]/[3760]	Speed: 117.686184 samples/sec	 accuracy=36.086648	 loss=2.898598	 lr=0.064980
Epoch[099] Batch [1149]/[3760]	Speed: 117.053611 samples/sec	 accuracy=36.138587	 loss=2.899261	 lr=0.064968
Epoch[099] Batch [1199]/[3760]	Speed: 117.880152 samples/sec	 accuracy=36.115885	 loss=2.901568	 lr=0.064955
Epoch[099] Batch [1249]/[3760]	Speed: 117.653973 samples/sec	 accuracy=36.112500	 loss=2.901837	 lr=0.064943
Epoch[099] Batch [1299]/[3760]	Speed: 116.881269 samples/sec	 accuracy=36.061298	 loss=2.903525	 lr=0.064931
Epoch[099] Batch [1349]/[3760]	Speed: 117.370277 samples/sec	 accuracy=36.062500	 loss=2.904438	 lr=0.064918
Epoch[099] Batch [1399]/[3760]	Speed: 117.281196 samples/sec	 accuracy=36.054688	 loss=2.905103	 lr=0.064906
Epoch[099] Batch [1449]/[3760]	Speed: 117.641770 samples/sec	 accuracy=36.004310	 loss=2.907770	 lr=0.064894
Epoch[099] Batch [1499]/[3760]	Speed: 117.358042 samples/sec	 accuracy=36.001042	 loss=2.908250	 lr=0.064882
Epoch[099] Batch [1549]/[3760]	Speed: 115.997577 samples/sec	 accuracy=36.021169	 loss=2.907754	 lr=0.064869
Epoch[099] Batch [1599]/[3760]	Speed: 118.136744 samples/sec	 accuracy=36.021484	 loss=2.906441	 lr=0.064857
Epoch[099] Batch [1649]/[3760]	Speed: 116.622810 samples/sec	 accuracy=36.003788	 loss=2.907768	 lr=0.064845
Epoch[099] Batch [1699]/[3760]	Speed: 117.645638 samples/sec	 accuracy=36.028493	 loss=2.908030	 lr=0.064832
Epoch[099] Batch [1749]/[3760]	Speed: 117.259742 samples/sec	 accuracy=35.999107	 loss=2.909000	 lr=0.064820
Epoch[099] Batch [1799]/[3760]	Speed: 117.392863 samples/sec	 accuracy=36.005208	 loss=2.908695	 lr=0.064808
Epoch[099] Batch [1849]/[3760]	Speed: 118.160472 samples/sec	 accuracy=36.006757	 loss=2.908012	 lr=0.064795
Epoch[099] Batch [1899]/[3760]	Speed: 117.359126 samples/sec	 accuracy=35.967928	 loss=2.910286	 lr=0.064783
Epoch[099] Batch [1949]/[3760]	Speed: 118.230271 samples/sec	 accuracy=35.951923	 loss=2.911350	 lr=0.064771
Epoch[099] Batch [1999]/[3760]	Speed: 117.153224 samples/sec	 accuracy=35.928125	 loss=2.912655	 lr=0.064758
Epoch[099] Batch [2049]/[3760]	Speed: 117.824536 samples/sec	 accuracy=35.919970	 loss=2.913668	 lr=0.064746
Epoch[099] Batch [2099]/[3760]	Speed: 117.629682 samples/sec	 accuracy=35.895833	 loss=2.914240	 lr=0.064734
Epoch[099] Batch [2149]/[3760]	Speed: 117.915262 samples/sec	 accuracy=35.890262	 loss=2.914022	 lr=0.064721
Epoch[099] Batch [2199]/[3760]	Speed: 117.140595 samples/sec	 accuracy=35.922585	 loss=2.913335	 lr=0.064709
Epoch[099] Batch [2249]/[3760]	Speed: 117.136124 samples/sec	 accuracy=35.912500	 loss=2.914422	 lr=0.064697
Epoch[099] Batch [2299]/[3760]	Speed: 117.197396 samples/sec	 accuracy=35.904891	 loss=2.914392	 lr=0.064684
Epoch[099] Batch [2349]/[3760]	Speed: 117.363791 samples/sec	 accuracy=35.936170	 loss=2.913306	 lr=0.064672
Epoch[099] Batch [2399]/[3760]	Speed: 117.787309 samples/sec	 accuracy=35.915365	 loss=2.912739	 lr=0.064660
Epoch[099] Batch [2449]/[3760]	Speed: 116.468640 samples/sec	 accuracy=35.909439	 loss=2.913768	 lr=0.064647
Epoch[099] Batch [2499]/[3760]	Speed: 117.635294 samples/sec	 accuracy=35.896250	 loss=2.914494	 lr=0.064635
Epoch[099] Batch [2549]/[3760]	Speed: 117.735053 samples/sec	 accuracy=35.895221	 loss=2.914706	 lr=0.064623
Epoch[099] Batch [2599]/[3760]	Speed: 117.433053 samples/sec	 accuracy=35.884014	 loss=2.915691	 lr=0.064610
Epoch[099] Batch [2649]/[3760]	Speed: 117.852929 samples/sec	 accuracy=35.867925	 loss=2.916019	 lr=0.064598
Epoch[099] Batch [2699]/[3760]	Speed: 117.314671 samples/sec	 accuracy=35.879630	 loss=2.915757	 lr=0.064586
Epoch[099] Batch [2749]/[3760]	Speed: 117.406583 samples/sec	 accuracy=35.855114	 loss=2.916486	 lr=0.064573
Epoch[099] Batch [2799]/[3760]	Speed: 117.333194 samples/sec	 accuracy=35.827567	 loss=2.917207	 lr=0.064561
Epoch[099] Batch [2849]/[3760]	Speed: 117.172450 samples/sec	 accuracy=35.829496	 loss=2.917135	 lr=0.064549
Epoch[099] Batch [2899]/[3760]	Speed: 117.528599 samples/sec	 accuracy=35.829203	 loss=2.917272	 lr=0.064536
Epoch[099] Batch [2949]/[3760]	Speed: 116.375916 samples/sec	 accuracy=35.842691	 loss=2.916889	 lr=0.064524
Epoch[099] Batch [2999]/[3760]	Speed: 117.701211 samples/sec	 accuracy=35.877083	 loss=2.915007	 lr=0.064512
Epoch[099] Batch [3049]/[3760]	Speed: 117.620961 samples/sec	 accuracy=35.871414	 loss=2.915937	 lr=0.064499
Epoch[099] Batch [3099]/[3760]	Speed: 118.065716 samples/sec	 accuracy=35.846270	 loss=2.916773	 lr=0.064487
Epoch[099] Batch [3149]/[3760]	Speed: 117.709797 samples/sec	 accuracy=35.836806	 loss=2.916876	 lr=0.064475
Epoch[099] Batch [3199]/[3760]	Speed: 117.713453 samples/sec	 accuracy=35.827148	 loss=2.916945	 lr=0.064462
Epoch[099] Batch [3249]/[3760]	Speed: 117.366638 samples/sec	 accuracy=35.841346	 loss=2.916718	 lr=0.064450
Epoch[099] Batch [3299]/[3760]	Speed: 117.142479 samples/sec	 accuracy=35.855587	 loss=2.916753	 lr=0.064438
Epoch[099] Batch [3349]/[3760]	Speed: 116.011122 samples/sec	 accuracy=35.858209	 loss=2.916960	 lr=0.064425
Epoch[099] Batch [3399]/[3760]	Speed: 117.300904 samples/sec	 accuracy=35.840074	 loss=2.917413	 lr=0.064413
Epoch[099] Batch [3449]/[3760]	Speed: 116.609693 samples/sec	 accuracy=35.828804	 loss=2.917582	 lr=0.064401
Epoch[099] Batch [3499]/[3760]	Speed: 117.661883 samples/sec	 accuracy=35.816964	 loss=2.918323	 lr=0.064388
Epoch[099] Batch [3549]/[3760]	Speed: 116.257880 samples/sec	 accuracy=35.797095	 loss=2.918887	 lr=0.064376
Epoch[099] Batch [3599]/[3760]	Speed: 117.919343 samples/sec	 accuracy=35.803819	 loss=2.918378	 lr=0.064364
Epoch[099] Batch [3649]/[3760]	Speed: 117.376354 samples/sec	 accuracy=35.782962	 loss=2.918873	 lr=0.064351
Epoch[099] Batch [3699]/[3760]	Speed: 116.835372 samples/sec	 accuracy=35.775338	 loss=2.919380	 lr=0.064339
Epoch[099] Batch [3749]/[3760]	Speed: 120.113352 samples/sec	 accuracy=35.776250	 loss=2.919725	 lr=0.064326
Batch [0049]/[0303]: acc-top1=35.312500 acc-top5=60.562500
Batch [0099]/[0303]: acc-top1=33.734375 acc-top5=60.343750
Batch [0149]/[0303]: acc-top1=35.156250 acc-top5=61.385417
Batch [0199]/[0303]: acc-top1=38.171875 acc-top5=63.968750
Batch [0249]/[0303]: acc-top1=37.293750 acc-top5=63.612500
Batch [0299]/[0303]: acc-top1=36.947917 acc-top5=63.203125
[Epoch 099] training: accuracy=35.768368	 loss=2.919993
[Epoch 099] speed: 116 samples/sec	time cost: 2178.599239
[Epoch 099] validation: acc-top1=36.865718 acc-top5=63.144596 loss=2.918059
Epoch[100] Batch [0049]/[3760]	Speed: 87.368177 samples/sec	 accuracy=37.687500	 loss=2.844923	 lr=0.064312
Epoch[100] Batch [0099]/[3760]	Speed: 117.069965 samples/sec	 accuracy=36.921875	 loss=2.882229	 lr=0.064299
Epoch[100] Batch [0149]/[3760]	Speed: 117.237504 samples/sec	 accuracy=37.250000	 loss=2.846677	 lr=0.064287
Epoch[100] Batch [0199]/[3760]	Speed: 117.496243 samples/sec	 accuracy=37.148438	 loss=2.848572	 lr=0.064275
Epoch[100] Batch [0249]/[3760]	Speed: 117.790204 samples/sec	 accuracy=36.856250	 loss=2.860672	 lr=0.064262
Epoch[100] Batch [0299]/[3760]	Speed: 116.957502 samples/sec	 accuracy=36.557292	 loss=2.875466	 lr=0.064250
Epoch[100] Batch [0349]/[3760]	Speed: 118.066178 samples/sec	 accuracy=36.366071	 loss=2.886315	 lr=0.064237
Epoch[100] Batch [0399]/[3760]	Speed: 117.097075 samples/sec	 accuracy=36.257812	 loss=2.889421	 lr=0.064225
Epoch[100] Batch [0449]/[3760]	Speed: 116.479688 samples/sec	 accuracy=36.177083	 loss=2.896327	 lr=0.064213
Epoch[100] Batch [0499]/[3760]	Speed: 117.886175 samples/sec	 accuracy=36.131250	 loss=2.898447	 lr=0.064200
Epoch[100] Batch [0549]/[3760]	Speed: 117.958261 samples/sec	 accuracy=36.130682	 loss=2.900619	 lr=0.064188
Epoch[100] Batch [0599]/[3760]	Speed: 117.458224 samples/sec	 accuracy=36.091146	 loss=2.904833	 lr=0.064176
Epoch[100] Batch [0649]/[3760]	Speed: 118.329743 samples/sec	 accuracy=35.995192	 loss=2.908478	 lr=0.064163
Epoch[100] Batch [0699]/[3760]	Speed: 117.801363 samples/sec	 accuracy=35.970982	 loss=2.910897	 lr=0.064151
Epoch[100] Batch [0749]/[3760]	Speed: 117.691910 samples/sec	 accuracy=35.941667	 loss=2.914821	 lr=0.064139
Epoch[100] Batch [0799]/[3760]	Speed: 117.781772 samples/sec	 accuracy=35.908203	 loss=2.914659	 lr=0.064126
Epoch[100] Batch [0849]/[3760]	Speed: 117.856843 samples/sec	 accuracy=35.860294	 loss=2.916716	 lr=0.064114
Epoch[100] Batch [0899]/[3760]	Speed: 117.913722 samples/sec	 accuracy=35.845486	 loss=2.916267	 lr=0.064101
Epoch[100] Batch [0949]/[3760]	Speed: 118.850556 samples/sec	 accuracy=35.792763	 loss=2.918429	 lr=0.064089
Epoch[100] Batch [0999]/[3760]	Speed: 116.876978 samples/sec	 accuracy=35.790625	 loss=2.918817	 lr=0.064077
Epoch[100] Batch [1049]/[3760]	Speed: 117.217604 samples/sec	 accuracy=35.741071	 loss=2.919480	 lr=0.064064
Epoch[100] Batch [1099]/[3760]	Speed: 117.426916 samples/sec	 accuracy=35.734375	 loss=2.920174	 lr=0.064052
Epoch[100] Batch [1149]/[3760]	Speed: 116.887914 samples/sec	 accuracy=35.756793	 loss=2.918560	 lr=0.064040
Epoch[100] Batch [1199]/[3760]	Speed: 117.556219 samples/sec	 accuracy=35.777344	 loss=2.916818	 lr=0.064027
Epoch[100] Batch [1249]/[3760]	Speed: 117.493597 samples/sec	 accuracy=35.786250	 loss=2.915933	 lr=0.064015
Epoch[100] Batch [1299]/[3760]	Speed: 117.484236 samples/sec	 accuracy=35.764423	 loss=2.918292	 lr=0.064002
Epoch[100] Batch [1349]/[3760]	Speed: 116.345631 samples/sec	 accuracy=35.754630	 loss=2.919343	 lr=0.063990
Epoch[100] Batch [1399]/[3760]	Speed: 117.630809 samples/sec	 accuracy=35.755580	 loss=2.918890	 lr=0.063978
Epoch[100] Batch [1449]/[3760]	Speed: 117.233107 samples/sec	 accuracy=35.725216	 loss=2.920535	 lr=0.063965
Epoch[100] Batch [1499]/[3760]	Speed: 117.770852 samples/sec	 accuracy=35.698958	 loss=2.921967	 lr=0.063953
Epoch[100] Batch [1549]/[3760]	Speed: 117.648483 samples/sec	 accuracy=35.730847	 loss=2.921309	 lr=0.063940
Epoch[100] Batch [1599]/[3760]	Speed: 118.155475 samples/sec	 accuracy=35.733398	 loss=2.919938	 lr=0.063928
Epoch[100] Batch [1649]/[3760]	Speed: 117.352376 samples/sec	 accuracy=35.732008	 loss=2.918823	 lr=0.063916
Epoch[100] Batch [1699]/[3760]	Speed: 117.826240 samples/sec	 accuracy=35.716912	 loss=2.920268	 lr=0.063903
Epoch[100] Batch [1749]/[3760]	Speed: 117.374132 samples/sec	 accuracy=35.721429	 loss=2.919142	 lr=0.063891
Epoch[100] Batch [1799]/[3760]	Speed: 117.200052 samples/sec	 accuracy=35.702257	 loss=2.919372	 lr=0.063879
Epoch[100] Batch [1849]/[3760]	Speed: 117.677610 samples/sec	 accuracy=35.699324	 loss=2.919308	 lr=0.063866
Epoch[100] Batch [1899]/[3760]	Speed: 117.597993 samples/sec	 accuracy=35.701480	 loss=2.919317	 lr=0.063854
Epoch[100] Batch [1949]/[3760]	Speed: 117.727965 samples/sec	 accuracy=35.709135	 loss=2.919032	 lr=0.063841
Epoch[100] Batch [1999]/[3760]	Speed: 117.551172 samples/sec	 accuracy=35.711719	 loss=2.918930	 lr=0.063829
Epoch[100] Batch [2049]/[3760]	Speed: 117.856186 samples/sec	 accuracy=35.728659	 loss=2.918723	 lr=0.063817
Epoch[100] Batch [2099]/[3760]	Speed: 116.659131 samples/sec	 accuracy=35.697173	 loss=2.919990	 lr=0.063804
Epoch[100] Batch [2149]/[3760]	Speed: 117.106595 samples/sec	 accuracy=35.717297	 loss=2.919662	 lr=0.063792
Epoch[100] Batch [2199]/[3760]	Speed: 117.397555 samples/sec	 accuracy=35.722301	 loss=2.920185	 lr=0.063779
Epoch[100] Batch [2249]/[3760]	Speed: 117.193866 samples/sec	 accuracy=35.693750	 loss=2.920666	 lr=0.063767
Epoch[100] Batch [2299]/[3760]	Speed: 117.413169 samples/sec	 accuracy=35.696332	 loss=2.920315	 lr=0.063755
Epoch[100] Batch [2349]/[3760]	Speed: 117.921725 samples/sec	 accuracy=35.706117	 loss=2.921435	 lr=0.063742
Epoch[100] Batch [2399]/[3760]	Speed: 117.505540 samples/sec	 accuracy=35.683594	 loss=2.921851	 lr=0.063730
Epoch[100] Batch [2449]/[3760]	Speed: 117.705394 samples/sec	 accuracy=35.663903	 loss=2.922169	 lr=0.063717
Epoch[100] Batch [2499]/[3760]	Speed: 118.098413 samples/sec	 accuracy=35.674375	 loss=2.921718	 lr=0.063705
Epoch[100] Batch [2549]/[3760]	Speed: 117.467883 samples/sec	 accuracy=35.645221	 loss=2.921684	 lr=0.063693
Epoch[100] Batch [2599]/[3760]	Speed: 117.726789 samples/sec	 accuracy=35.631611	 loss=2.921987	 lr=0.063680
Epoch[100] Batch [2649]/[3760]	Speed: 118.085675 samples/sec	 accuracy=35.631486	 loss=2.922297	 lr=0.063668
Epoch[100] Batch [2699]/[3760]	Speed: 116.401095 samples/sec	 accuracy=35.660301	 loss=2.921658	 lr=0.063655
Epoch[100] Batch [2749]/[3760]	Speed: 117.642013 samples/sec	 accuracy=35.652273	 loss=2.921654	 lr=0.063643
Epoch[100] Batch [2799]/[3760]	Speed: 117.415752 samples/sec	 accuracy=35.647879	 loss=2.922123	 lr=0.063631
Epoch[100] Batch [2849]/[3760]	Speed: 117.594721 samples/sec	 accuracy=35.674890	 loss=2.920657	 lr=0.063618
Epoch[100] Batch [2899]/[3760]	Speed: 117.568067 samples/sec	 accuracy=35.665948	 loss=2.919935	 lr=0.063606
Epoch[100] Batch [2949]/[3760]	Speed: 117.283434 samples/sec	 accuracy=35.685381	 loss=2.919934	 lr=0.063593
Epoch[100] Batch [2999]/[3760]	Speed: 117.862709 samples/sec	 accuracy=35.673437	 loss=2.920293	 lr=0.063581
Epoch[100] Batch [3049]/[3760]	Speed: 117.223410 samples/sec	 accuracy=35.663422	 loss=2.920782	 lr=0.063568
Epoch[100] Batch [3099]/[3760]	Speed: 117.846462 samples/sec	 accuracy=35.665323	 loss=2.919883	 lr=0.063556
Epoch[100] Batch [3149]/[3760]	Speed: 117.938142 samples/sec	 accuracy=35.655258	 loss=2.920103	 lr=0.063544
Epoch[100] Batch [3199]/[3760]	Speed: 117.114777 samples/sec	 accuracy=35.648926	 loss=2.920014	 lr=0.063531
Epoch[100] Batch [3249]/[3760]	Speed: 117.246061 samples/sec	 accuracy=35.646635	 loss=2.920486	 lr=0.063519
Epoch[100] Batch [3299]/[3760]	Speed: 117.485041 samples/sec	 accuracy=35.660985	 loss=2.920030	 lr=0.063506
Epoch[100] Batch [3349]/[3760]	Speed: 118.287556 samples/sec	 accuracy=35.666978	 loss=2.919471	 lr=0.063494
Epoch[100] Batch [3399]/[3760]	Speed: 117.322955 samples/sec	 accuracy=35.676930	 loss=2.919406	 lr=0.063482
Epoch[100] Batch [3449]/[3760]	Speed: 117.994833 samples/sec	 accuracy=35.669384	 loss=2.919605	 lr=0.063469
Epoch[100] Batch [3499]/[3760]	Speed: 117.293951 samples/sec	 accuracy=35.648214	 loss=2.920751	 lr=0.063457
Epoch[100] Batch [3549]/[3760]	Speed: 117.120749 samples/sec	 accuracy=35.652289	 loss=2.920995	 lr=0.063444
Epoch[100] Batch [3599]/[3760]	Speed: 117.397774 samples/sec	 accuracy=35.660590	 loss=2.920259	 lr=0.063432
Epoch[100] Batch [3649]/[3760]	Speed: 117.214516 samples/sec	 accuracy=35.675942	 loss=2.919186	 lr=0.063419
Epoch[100] Batch [3699]/[3760]	Speed: 117.352566 samples/sec	 accuracy=35.656250	 loss=2.919916	 lr=0.063407
Epoch[100] Batch [3749]/[3760]	Speed: 120.486209 samples/sec	 accuracy=35.652917	 loss=2.920129	 lr=0.063395
Batch [0049]/[0303]: acc-top1=36.125000 acc-top5=61.312500
Batch [0099]/[0303]: acc-top1=33.484375 acc-top5=60.406250
Batch [0149]/[0303]: acc-top1=35.614583 acc-top5=61.197917
Batch [0199]/[0303]: acc-top1=38.359375 acc-top5=63.656250
Batch [0249]/[0303]: acc-top1=37.318750 acc-top5=63.131250
Batch [0299]/[0303]: acc-top1=36.671875 acc-top5=63.088542
[Epoch 100] training: accuracy=35.648687	 loss=2.920194
[Epoch 100] speed: 116 samples/sec	time cost: 2174.396685
[Epoch 100] validation: acc-top1=36.576939 acc-top5=63.041460 loss=2.910977
Epoch[101] Batch [0049]/[3759]	Speed: 86.244985 samples/sec	 accuracy=35.625000	 loss=2.911393	 lr=0.063380
Epoch[101] Batch [0099]/[3759]	Speed: 117.498973 samples/sec	 accuracy=35.781250	 loss=2.898349	 lr=0.063367
Epoch[101] Batch [0149]/[3759]	Speed: 117.310848 samples/sec	 accuracy=36.364583	 loss=2.870313	 lr=0.063355
Epoch[101] Batch [0199]/[3759]	Speed: 117.556558 samples/sec	 accuracy=36.492188	 loss=2.868190	 lr=0.063342
Epoch[101] Batch [0249]/[3759]	Speed: 117.627573 samples/sec	 accuracy=36.387500	 loss=2.878020	 lr=0.063330
Epoch[101] Batch [0299]/[3759]	Speed: 116.512207 samples/sec	 accuracy=36.223958	 loss=2.878785	 lr=0.063318
Epoch[101] Batch [0349]/[3759]	Speed: 118.103530 samples/sec	 accuracy=36.236607	 loss=2.883179	 lr=0.063305
Epoch[101] Batch [0399]/[3759]	Speed: 118.140662 samples/sec	 accuracy=36.265625	 loss=2.880892	 lr=0.063293
Epoch[101] Batch [0449]/[3759]	Speed: 117.804464 samples/sec	 accuracy=36.218750	 loss=2.879389	 lr=0.063280
Epoch[101] Batch [0499]/[3759]	Speed: 117.383592 samples/sec	 accuracy=36.193750	 loss=2.883037	 lr=0.063268
Epoch[101] Batch [0549]/[3759]	Speed: 117.493323 samples/sec	 accuracy=35.994318	 loss=2.890566	 lr=0.063255
Epoch[101] Batch [0599]/[3759]	Speed: 117.707008 samples/sec	 accuracy=36.020833	 loss=2.892348	 lr=0.063243
Epoch[101] Batch [0649]/[3759]	Speed: 117.521105 samples/sec	 accuracy=36.117788	 loss=2.888073	 lr=0.063231
Epoch[101] Batch [0699]/[3759]	Speed: 117.517157 samples/sec	 accuracy=35.950893	 loss=2.896964	 lr=0.063218
Epoch[101] Batch [0749]/[3759]	Speed: 116.998208 samples/sec	 accuracy=36.010417	 loss=2.893512	 lr=0.063206
Epoch[101] Batch [0799]/[3759]	Speed: 118.399036 samples/sec	 accuracy=36.042969	 loss=2.892814	 lr=0.063193
Epoch[101] Batch [0849]/[3759]	Speed: 117.091880 samples/sec	 accuracy=36.027574	 loss=2.894096	 lr=0.063181
Epoch[101] Batch [0899]/[3759]	Speed: 117.070997 samples/sec	 accuracy=36.031250	 loss=2.892677	 lr=0.063168
Epoch[101] Batch [0949]/[3759]	Speed: 117.557515 samples/sec	 accuracy=36.000000	 loss=2.895044	 lr=0.063156
Epoch[101] Batch [0999]/[3759]	Speed: 117.422709 samples/sec	 accuracy=36.014063	 loss=2.895748	 lr=0.063143
Epoch[101] Batch [1049]/[3759]	Speed: 117.471182 samples/sec	 accuracy=36.007440	 loss=2.897247	 lr=0.063131
Epoch[101] Batch [1099]/[3759]	Speed: 117.290375 samples/sec	 accuracy=35.978693	 loss=2.899013	 lr=0.063119
Epoch[101] Batch [1149]/[3759]	Speed: 117.889016 samples/sec	 accuracy=35.982337	 loss=2.898871	 lr=0.063106
Epoch[101] Batch [1199]/[3759]	Speed: 116.092204 samples/sec	 accuracy=35.964844	 loss=2.900188	 lr=0.063094
Epoch[101] Batch [1249]/[3759]	Speed: 117.969852 samples/sec	 accuracy=35.926250	 loss=2.901068	 lr=0.063081
Epoch[101] Batch [1299]/[3759]	Speed: 117.248404 samples/sec	 accuracy=35.981971	 loss=2.897971	 lr=0.063069
Epoch[101] Batch [1349]/[3759]	Speed: 117.473469 samples/sec	 accuracy=36.031250	 loss=2.895898	 lr=0.063056
Epoch[101] Batch [1399]/[3759]	Speed: 117.222957 samples/sec	 accuracy=36.046875	 loss=2.895816	 lr=0.063044
Epoch[101] Batch [1449]/[3759]	Speed: 117.095304 samples/sec	 accuracy=36.057112	 loss=2.895391	 lr=0.063031
Epoch[101] Batch [1499]/[3759]	Speed: 117.952937 samples/sec	 accuracy=36.103125	 loss=2.894362	 lr=0.063019
Epoch[101] Batch [1549]/[3759]	Speed: 117.115677 samples/sec	 accuracy=36.093750	 loss=2.895536	 lr=0.063006
Epoch[101] Batch [1599]/[3759]	Speed: 118.164628 samples/sec	 accuracy=36.038086	 loss=2.898215	 lr=0.062994
Epoch[101] Batch [1649]/[3759]	Speed: 117.112219 samples/sec	 accuracy=36.039773	 loss=2.898929	 lr=0.062982
Epoch[101] Batch [1699]/[3759]	Speed: 116.909566 samples/sec	 accuracy=36.035846	 loss=2.899105	 lr=0.062969
Epoch[101] Batch [1749]/[3759]	Speed: 116.257447 samples/sec	 accuracy=36.010714	 loss=2.900653	 lr=0.062957
Epoch[101] Batch [1799]/[3759]	Speed: 117.435182 samples/sec	 accuracy=36.000868	 loss=2.900524	 lr=0.062944
Epoch[101] Batch [1849]/[3759]	Speed: 116.661760 samples/sec	 accuracy=36.000845	 loss=2.901690	 lr=0.062932
Epoch[101] Batch [1899]/[3759]	Speed: 116.853336 samples/sec	 accuracy=35.981908	 loss=2.903601	 lr=0.062919
Epoch[101] Batch [1949]/[3759]	Speed: 117.402894 samples/sec	 accuracy=35.936699	 loss=2.906082	 lr=0.062907
Epoch[101] Batch [1999]/[3759]	Speed: 117.285078 samples/sec	 accuracy=35.952344	 loss=2.904226	 lr=0.062894
Epoch[101] Batch [2049]/[3759]	Speed: 117.996052 samples/sec	 accuracy=35.979421	 loss=2.902877	 lr=0.062882
Epoch[101] Batch [2099]/[3759]	Speed: 115.581104 samples/sec	 accuracy=35.973958	 loss=2.901462	 lr=0.062869
Epoch[101] Batch [2149]/[3759]	Speed: 118.059274 samples/sec	 accuracy=35.957122	 loss=2.901759	 lr=0.062857
Epoch[101] Batch [2199]/[3759]	Speed: 117.035643 samples/sec	 accuracy=35.967330	 loss=2.900633	 lr=0.062845
Epoch[101] Batch [2249]/[3759]	Speed: 117.542995 samples/sec	 accuracy=35.969444	 loss=2.901507	 lr=0.062832
Epoch[101] Batch [2299]/[3759]	Speed: 117.755906 samples/sec	 accuracy=35.993207	 loss=2.899895	 lr=0.062820
Epoch[101] Batch [2349]/[3759]	Speed: 117.410554 samples/sec	 accuracy=36.001330	 loss=2.899347	 lr=0.062807
Epoch[101] Batch [2399]/[3759]	Speed: 117.489847 samples/sec	 accuracy=35.986328	 loss=2.899535	 lr=0.062795
Epoch[101] Batch [2449]/[3759]	Speed: 117.812371 samples/sec	 accuracy=36.008929	 loss=2.898587	 lr=0.062782
Epoch[101] Batch [2499]/[3759]	Speed: 118.350708 samples/sec	 accuracy=36.023750	 loss=2.897805	 lr=0.062770
Epoch[101] Batch [2549]/[3759]	Speed: 117.226659 samples/sec	 accuracy=36.009191	 loss=2.898644	 lr=0.062757
Epoch[101] Batch [2599]/[3759]	Speed: 116.474429 samples/sec	 accuracy=36.003005	 loss=2.899392	 lr=0.062745
Epoch[101] Batch [2649]/[3759]	Speed: 117.479791 samples/sec	 accuracy=35.992335	 loss=2.899318	 lr=0.062732
Epoch[101] Batch [2699]/[3759]	Speed: 117.295660 samples/sec	 accuracy=35.993056	 loss=2.899864	 lr=0.062720
Epoch[101] Batch [2749]/[3759]	Speed: 117.713303 samples/sec	 accuracy=36.018182	 loss=2.899600	 lr=0.062707
Epoch[101] Batch [2799]/[3759]	Speed: 117.087847 samples/sec	 accuracy=36.007812	 loss=2.899418	 lr=0.062695
Epoch[101] Batch [2849]/[3759]	Speed: 117.391571 samples/sec	 accuracy=36.014254	 loss=2.899587	 lr=0.062682
Epoch[101] Batch [2899]/[3759]	Speed: 117.576552 samples/sec	 accuracy=36.020474	 loss=2.900150	 lr=0.062670
Epoch[101] Batch [2949]/[3759]	Speed: 117.831581 samples/sec	 accuracy=36.031780	 loss=2.899762	 lr=0.062657
Epoch[101] Batch [2999]/[3759]	Speed: 115.984423 samples/sec	 accuracy=36.010938	 loss=2.899812	 lr=0.062645
Epoch[101] Batch [3049]/[3759]	Speed: 117.901667 samples/sec	 accuracy=36.013832	 loss=2.899453	 lr=0.062633
Epoch[101] Batch [3099]/[3759]	Speed: 117.798447 samples/sec	 accuracy=36.006048	 loss=2.900258	 lr=0.062620
Epoch[101] Batch [3149]/[3759]	Speed: 117.276387 samples/sec	 accuracy=35.997520	 loss=2.901343	 lr=0.062608
Epoch[101] Batch [3199]/[3759]	Speed: 117.554587 samples/sec	 accuracy=36.012695	 loss=2.900770	 lr=0.062595
Epoch[101] Batch [3249]/[3759]	Speed: 117.889684 samples/sec	 accuracy=36.045192	 loss=2.898959	 lr=0.062583
Epoch[101] Batch [3299]/[3759]	Speed: 117.375085 samples/sec	 accuracy=36.020833	 loss=2.899838	 lr=0.062570
Epoch[101] Batch [3349]/[3759]	Speed: 117.156686 samples/sec	 accuracy=36.012593	 loss=2.899958	 lr=0.062558
Epoch[101] Batch [3399]/[3759]	Speed: 117.670356 samples/sec	 accuracy=35.989430	 loss=2.900914	 lr=0.062545
Epoch[101] Batch [3449]/[3759]	Speed: 117.015337 samples/sec	 accuracy=36.007699	 loss=2.900937	 lr=0.062533
Epoch[101] Batch [3499]/[3759]	Speed: 117.799263 samples/sec	 accuracy=35.992857	 loss=2.901956	 lr=0.062520
Epoch[101] Batch [3549]/[3759]	Speed: 116.656073 samples/sec	 accuracy=35.984155	 loss=2.902254	 lr=0.062508
Epoch[101] Batch [3599]/[3759]	Speed: 116.433658 samples/sec	 accuracy=35.972656	 loss=2.902765	 lr=0.062495
Epoch[101] Batch [3649]/[3759]	Speed: 117.563701 samples/sec	 accuracy=35.974743	 loss=2.903019	 lr=0.062483
Epoch[101] Batch [3699]/[3759]	Speed: 118.053763 samples/sec	 accuracy=35.956503	 loss=2.904032	 lr=0.062470
Epoch[101] Batch [3749]/[3759]	Speed: 120.512189 samples/sec	 accuracy=35.938333	 loss=2.904758	 lr=0.062458
Batch [0049]/[0303]: acc-top1=36.437500 acc-top5=61.843750
Batch [0099]/[0303]: acc-top1=34.265625 acc-top5=61.171875
Batch [0149]/[0303]: acc-top1=36.437500 acc-top5=62.552083
Batch [0199]/[0303]: acc-top1=39.531250 acc-top5=64.890625
Batch [0249]/[0303]: acc-top1=38.137500 acc-top5=64.281250
Batch [0299]/[0303]: acc-top1=37.057292 acc-top5=64.088542
[Epoch 101] training: accuracy=35.940825	 loss=2.904945
[Epoch 101] speed: 116 samples/sec	time cost: 2176.789155
[Epoch 101] validation: acc-top1=37.092616 acc-top5=64.175949 loss=2.964732
Epoch[102] Batch [0049]/[3760]	Speed: 86.848907 samples/sec	 accuracy=34.937500	 loss=2.912548	 lr=0.062443
Epoch[102] Batch [0099]/[3760]	Speed: 117.922783 samples/sec	 accuracy=36.343750	 loss=2.880817	 lr=0.062430
Epoch[102] Batch [0149]/[3760]	Speed: 116.351284 samples/sec	 accuracy=36.229167	 loss=2.896779	 lr=0.062418
Epoch[102] Batch [0199]/[3760]	Speed: 117.362386 samples/sec	 accuracy=36.093750	 loss=2.899402	 lr=0.062405
Epoch[102] Batch [0249]/[3760]	Speed: 116.966599 samples/sec	 accuracy=35.981250	 loss=2.899111	 lr=0.062393
Epoch[102] Batch [0299]/[3760]	Speed: 117.560534 samples/sec	 accuracy=36.109375	 loss=2.894503	 lr=0.062381
Epoch[102] Batch [0349]/[3760]	Speed: 116.917604 samples/sec	 accuracy=36.276786	 loss=2.888823	 lr=0.062368
Epoch[102] Batch [0399]/[3760]	Speed: 117.424172 samples/sec	 accuracy=36.460938	 loss=2.873903	 lr=0.062356
Epoch[102] Batch [0449]/[3760]	Speed: 117.307120 samples/sec	 accuracy=36.493056	 loss=2.875455	 lr=0.062343
Epoch[102] Batch [0499]/[3760]	Speed: 117.473983 samples/sec	 accuracy=36.543750	 loss=2.874377	 lr=0.062331
Epoch[102] Batch [0549]/[3760]	Speed: 117.773549 samples/sec	 accuracy=36.542614	 loss=2.871825	 lr=0.062318
Epoch[102] Batch [0599]/[3760]	Speed: 118.033689 samples/sec	 accuracy=36.507812	 loss=2.872121	 lr=0.062306
Epoch[102] Batch [0649]/[3760]	Speed: 117.196835 samples/sec	 accuracy=36.711538	 loss=2.865134	 lr=0.062293
Epoch[102] Batch [0699]/[3760]	Speed: 117.081257 samples/sec	 accuracy=36.694196	 loss=2.866903	 lr=0.062281
Epoch[102] Batch [0749]/[3760]	Speed: 116.989682 samples/sec	 accuracy=36.679167	 loss=2.867022	 lr=0.062268
Epoch[102] Batch [0799]/[3760]	Speed: 118.015465 samples/sec	 accuracy=36.574219	 loss=2.869693	 lr=0.062256
Epoch[102] Batch [0849]/[3760]	Speed: 117.733806 samples/sec	 accuracy=36.518382	 loss=2.871276	 lr=0.062243
Epoch[102] Batch [0899]/[3760]	Speed: 117.518524 samples/sec	 accuracy=36.451389	 loss=2.873275	 lr=0.062230
Epoch[102] Batch [0949]/[3760]	Speed: 117.492789 samples/sec	 accuracy=36.444079	 loss=2.875334	 lr=0.062218
Epoch[102] Batch [0999]/[3760]	Speed: 117.171992 samples/sec	 accuracy=36.443750	 loss=2.876803	 lr=0.062205
Epoch[102] Batch [1049]/[3760]	Speed: 117.906722 samples/sec	 accuracy=36.388393	 loss=2.880125	 lr=0.062193
Epoch[102] Batch [1099]/[3760]	Speed: 116.929830 samples/sec	 accuracy=36.359375	 loss=2.880474	 lr=0.062180
Epoch[102] Batch [1149]/[3760]	Speed: 117.671400 samples/sec	 accuracy=36.413043	 loss=2.878055	 lr=0.062168
Epoch[102] Batch [1199]/[3760]	Speed: 117.683925 samples/sec	 accuracy=36.442708	 loss=2.878000	 lr=0.062155
Epoch[102] Batch [1249]/[3760]	Speed: 117.655856 samples/sec	 accuracy=36.446250	 loss=2.878647	 lr=0.062143
Epoch[102] Batch [1299]/[3760]	Speed: 117.520099 samples/sec	 accuracy=36.417067	 loss=2.880172	 lr=0.062130
Epoch[102] Batch [1349]/[3760]	Speed: 117.360789 samples/sec	 accuracy=36.434028	 loss=2.878352	 lr=0.062118
Epoch[102] Batch [1399]/[3760]	Speed: 117.307347 samples/sec	 accuracy=36.473214	 loss=2.876293	 lr=0.062105
Epoch[102] Batch [1449]/[3760]	Speed: 118.012039 samples/sec	 accuracy=36.448276	 loss=2.876594	 lr=0.062093
Epoch[102] Batch [1499]/[3760]	Speed: 118.145721 samples/sec	 accuracy=36.475000	 loss=2.874785	 lr=0.062080
Epoch[102] Batch [1549]/[3760]	Speed: 117.013497 samples/sec	 accuracy=36.446573	 loss=2.875356	 lr=0.062068
Epoch[102] Batch [1599]/[3760]	Speed: 118.135199 samples/sec	 accuracy=36.500000	 loss=2.873466	 lr=0.062055
Epoch[102] Batch [1649]/[3760]	Speed: 117.844572 samples/sec	 accuracy=36.473485	 loss=2.875547	 lr=0.062043
Epoch[102] Batch [1699]/[3760]	Speed: 116.634443 samples/sec	 accuracy=36.426471	 loss=2.876392	 lr=0.062030
Epoch[102] Batch [1749]/[3760]	Speed: 117.299190 samples/sec	 accuracy=36.377679	 loss=2.878429	 lr=0.062018
Epoch[102] Batch [1799]/[3760]	Speed: 117.393295 samples/sec	 accuracy=36.387153	 loss=2.877884	 lr=0.062005
Epoch[102] Batch [1849]/[3760]	Speed: 117.726624 samples/sec	 accuracy=36.396959	 loss=2.877077	 lr=0.061993
Epoch[102] Batch [1899]/[3760]	Speed: 118.399560 samples/sec	 accuracy=36.414474	 loss=2.876058	 lr=0.061980
Epoch[102] Batch [1949]/[3760]	Speed: 117.602566 samples/sec	 accuracy=36.381410	 loss=2.878477	 lr=0.061968
Epoch[102] Batch [1999]/[3760]	Speed: 116.783557 samples/sec	 accuracy=36.364844	 loss=2.879761	 lr=0.061955
Epoch[102] Batch [2049]/[3760]	Speed: 117.338825 samples/sec	 accuracy=36.331555	 loss=2.881318	 lr=0.061943
Epoch[102] Batch [2099]/[3760]	Speed: 116.679887 samples/sec	 accuracy=36.359375	 loss=2.879476	 lr=0.061930
Epoch[102] Batch [2149]/[3760]	Speed: 117.228371 samples/sec	 accuracy=36.353924	 loss=2.880996	 lr=0.061918
Epoch[102] Batch [2199]/[3760]	Speed: 117.690705 samples/sec	 accuracy=36.325284	 loss=2.881329	 lr=0.061905
Epoch[102] Batch [2249]/[3760]	Speed: 117.481729 samples/sec	 accuracy=36.277778	 loss=2.882750	 lr=0.061893
Epoch[102] Batch [2299]/[3760]	Speed: 116.684760 samples/sec	 accuracy=36.290082	 loss=2.883193	 lr=0.061880
Epoch[102] Batch [2349]/[3760]	Speed: 117.947212 samples/sec	 accuracy=36.326463	 loss=2.883064	 lr=0.061867
Epoch[102] Batch [2399]/[3760]	Speed: 117.682727 samples/sec	 accuracy=36.330078	 loss=2.883417	 lr=0.061855
Epoch[102] Batch [2449]/[3760]	Speed: 117.191066 samples/sec	 accuracy=36.316964	 loss=2.882960	 lr=0.061842
Epoch[102] Batch [2499]/[3760]	Speed: 117.758785 samples/sec	 accuracy=36.298750	 loss=2.883779	 lr=0.061830
Epoch[102] Batch [2549]/[3760]	Speed: 117.389246 samples/sec	 accuracy=36.307598	 loss=2.883761	 lr=0.061817
Epoch[102] Batch [2599]/[3760]	Speed: 117.326577 samples/sec	 accuracy=36.319712	 loss=2.883422	 lr=0.061805
Epoch[102] Batch [2649]/[3760]	Speed: 116.832964 samples/sec	 accuracy=36.327830	 loss=2.883227	 lr=0.061792
Epoch[102] Batch [2699]/[3760]	Speed: 117.386405 samples/sec	 accuracy=36.296296	 loss=2.885141	 lr=0.061780
Epoch[102] Batch [2749]/[3760]	Speed: 117.480542 samples/sec	 accuracy=36.272727	 loss=2.885338	 lr=0.061767
Epoch[102] Batch [2799]/[3760]	Speed: 117.966041 samples/sec	 accuracy=36.292969	 loss=2.885403	 lr=0.061755
Epoch[102] Batch [2849]/[3760]	Speed: 117.255507 samples/sec	 accuracy=36.268092	 loss=2.885860	 lr=0.061742
Epoch[102] Batch [2899]/[3760]	Speed: 116.493885 samples/sec	 accuracy=36.258621	 loss=2.885954	 lr=0.061730
Epoch[102] Batch [2949]/[3760]	Speed: 117.584255 samples/sec	 accuracy=36.281250	 loss=2.886259	 lr=0.061717
Epoch[102] Batch [2999]/[3760]	Speed: 117.312414 samples/sec	 accuracy=36.263021	 loss=2.887353	 lr=0.061705
Epoch[102] Batch [3049]/[3760]	Speed: 117.419177 samples/sec	 accuracy=36.263320	 loss=2.887954	 lr=0.061692
Epoch[102] Batch [3099]/[3760]	Speed: 117.061340 samples/sec	 accuracy=36.242440	 loss=2.888546	 lr=0.061679
Epoch[102] Batch [3149]/[3760]	Speed: 116.949448 samples/sec	 accuracy=36.250000	 loss=2.887722	 lr=0.061667
Epoch[102] Batch [3199]/[3760]	Speed: 117.182467 samples/sec	 accuracy=36.232422	 loss=2.888619	 lr=0.061654
Epoch[102] Batch [3249]/[3760]	Speed: 116.941515 samples/sec	 accuracy=36.220673	 loss=2.889280	 lr=0.061642
Epoch[102] Batch [3299]/[3760]	Speed: 117.621590 samples/sec	 accuracy=36.205492	 loss=2.890275	 lr=0.061629
Epoch[102] Batch [3349]/[3760]	Speed: 117.454924 samples/sec	 accuracy=36.220616	 loss=2.889649	 lr=0.061617
Epoch[102] Batch [3399]/[3760]	Speed: 117.468193 samples/sec	 accuracy=36.199449	 loss=2.890582	 lr=0.061604
Epoch[102] Batch [3449]/[3760]	Speed: 117.864111 samples/sec	 accuracy=36.187047	 loss=2.890950	 lr=0.061592
Epoch[102] Batch [3499]/[3760]	Speed: 116.971075 samples/sec	 accuracy=36.180357	 loss=2.891910	 lr=0.061579
Epoch[102] Batch [3549]/[3760]	Speed: 117.503288 samples/sec	 accuracy=36.188820	 loss=2.891323	 lr=0.061567
Epoch[102] Batch [3599]/[3760]	Speed: 116.369341 samples/sec	 accuracy=36.194444	 loss=2.890883	 lr=0.061554
Epoch[102] Batch [3649]/[3760]	Speed: 117.712874 samples/sec	 accuracy=36.196062	 loss=2.891775	 lr=0.061541
Epoch[102] Batch [3699]/[3760]	Speed: 117.681067 samples/sec	 accuracy=36.180743	 loss=2.892200	 lr=0.061529
Epoch[102] Batch [3749]/[3760]	Speed: 120.348939 samples/sec	 accuracy=36.147917	 loss=2.892970	 lr=0.061516
Batch [0049]/[0303]: acc-top1=37.968750 acc-top5=64.750000
Batch [0099]/[0303]: acc-top1=33.984375 acc-top5=61.359375
Batch [0149]/[0303]: acc-top1=36.187500 acc-top5=63.020833
Batch [0199]/[0303]: acc-top1=39.031250 acc-top5=65.109375
Batch [0249]/[0303]: acc-top1=38.581250 acc-top5=64.906250
Batch [0299]/[0303]: acc-top1=37.822917 acc-top5=64.869792
[Epoch 102] training: accuracy=36.146941	 loss=2.893135
[Epoch 102] speed: 116 samples/sec	time cost: 2176.461984
[Epoch 102] validation: acc-top1=37.773309 acc-top5=64.841172 loss=2.855371
Epoch[103] Batch [0049]/[3760]	Speed: 85.833244 samples/sec	 accuracy=36.093750	 loss=2.868631	 lr=0.061501
Epoch[103] Batch [0099]/[3760]	Speed: 117.590192 samples/sec	 accuracy=36.656250	 loss=2.851917	 lr=0.061489
Epoch[103] Batch [0149]/[3760]	Speed: 116.887198 samples/sec	 accuracy=36.093750	 loss=2.884849	 lr=0.061476
Epoch[103] Batch [0199]/[3760]	Speed: 117.234937 samples/sec	 accuracy=36.078125	 loss=2.868608	 lr=0.061464
Epoch[103] Batch [0249]/[3760]	Speed: 116.767666 samples/sec	 accuracy=36.131250	 loss=2.872714	 lr=0.061451
Epoch[103] Batch [0299]/[3760]	Speed: 117.459566 samples/sec	 accuracy=36.255208	 loss=2.858392	 lr=0.061439
Epoch[103] Batch [0349]/[3760]	Speed: 117.000930 samples/sec	 accuracy=36.241071	 loss=2.864780	 lr=0.061426
Epoch[103] Batch [0399]/[3760]	Speed: 117.465365 samples/sec	 accuracy=36.125000	 loss=2.870452	 lr=0.061413
Epoch[103] Batch [0449]/[3760]	Speed: 116.546173 samples/sec	 accuracy=36.038194	 loss=2.875689	 lr=0.061401
Epoch[103] Batch [0499]/[3760]	Speed: 116.498631 samples/sec	 accuracy=36.021875	 loss=2.876420	 lr=0.061388
Epoch[103] Batch [0549]/[3760]	Speed: 117.250794 samples/sec	 accuracy=36.045455	 loss=2.877751	 lr=0.061376
Epoch[103] Batch [0599]/[3760]	Speed: 116.911864 samples/sec	 accuracy=36.109375	 loss=2.879398	 lr=0.061363
Epoch[103] Batch [0649]/[3760]	Speed: 117.230750 samples/sec	 accuracy=36.156250	 loss=2.877547	 lr=0.061351
Epoch[103] Batch [0699]/[3760]	Speed: 118.256076 samples/sec	 accuracy=36.058036	 loss=2.879936	 lr=0.061338
Epoch[103] Batch [0749]/[3760]	Speed: 117.839134 samples/sec	 accuracy=36.020833	 loss=2.882432	 lr=0.061326
Epoch[103] Batch [0799]/[3760]	Speed: 117.881262 samples/sec	 accuracy=36.125000	 loss=2.880820	 lr=0.061313
Epoch[103] Batch [0849]/[3760]	Speed: 117.574088 samples/sec	 accuracy=36.185662	 loss=2.878990	 lr=0.061300
Epoch[103] Batch [0899]/[3760]	Speed: 117.410103 samples/sec	 accuracy=36.223958	 loss=2.879932	 lr=0.061288
Epoch[103] Batch [0949]/[3760]	Speed: 117.159036 samples/sec	 accuracy=36.245066	 loss=2.879798	 lr=0.061275
Epoch[103] Batch [0999]/[3760]	Speed: 116.591417 samples/sec	 accuracy=36.228125	 loss=2.879205	 lr=0.061263
Epoch[103] Batch [1049]/[3760]	Speed: 117.607411 samples/sec	 accuracy=36.211310	 loss=2.882331	 lr=0.061250
Epoch[103] Batch [1099]/[3760]	Speed: 117.872353 samples/sec	 accuracy=36.237216	 loss=2.882241	 lr=0.061238
Epoch[103] Batch [1149]/[3760]	Speed: 117.686046 samples/sec	 accuracy=36.229620	 loss=2.884302	 lr=0.061225
Epoch[103] Batch [1199]/[3760]	Speed: 117.493695 samples/sec	 accuracy=36.244792	 loss=2.882451	 lr=0.061212
Epoch[103] Batch [1249]/[3760]	Speed: 117.822035 samples/sec	 accuracy=36.245000	 loss=2.882678	 lr=0.061200
Epoch[103] Batch [1299]/[3760]	Speed: 117.741859 samples/sec	 accuracy=36.258413	 loss=2.882975	 lr=0.061187
Epoch[103] Batch [1349]/[3760]	Speed: 116.456859 samples/sec	 accuracy=36.254630	 loss=2.882010	 lr=0.061175
Epoch[103] Batch [1399]/[3760]	Speed: 116.910101 samples/sec	 accuracy=36.281250	 loss=2.880273	 lr=0.061162
Epoch[103] Batch [1449]/[3760]	Speed: 117.607595 samples/sec	 accuracy=36.240302	 loss=2.881159	 lr=0.061150
Epoch[103] Batch [1499]/[3760]	Speed: 117.126557 samples/sec	 accuracy=36.202083	 loss=2.881652	 lr=0.061137
Epoch[103] Batch [1549]/[3760]	Speed: 117.001500 samples/sec	 accuracy=36.185484	 loss=2.882947	 lr=0.061124
Epoch[103] Batch [1599]/[3760]	Speed: 117.841752 samples/sec	 accuracy=36.166992	 loss=2.882050	 lr=0.061112
Epoch[103] Batch [1649]/[3760]	Speed: 116.870008 samples/sec	 accuracy=36.189394	 loss=2.881849	 lr=0.061099
Epoch[103] Batch [1699]/[3760]	Speed: 117.316978 samples/sec	 accuracy=36.209559	 loss=2.881661	 lr=0.061087
Epoch[103] Batch [1749]/[3760]	Speed: 117.233767 samples/sec	 accuracy=36.216071	 loss=2.880514	 lr=0.061074
Epoch[103] Batch [1799]/[3760]	Speed: 117.105597 samples/sec	 accuracy=36.177951	 loss=2.882623	 lr=0.061062
Epoch[103] Batch [1849]/[3760]	Speed: 117.363111 samples/sec	 accuracy=36.192568	 loss=2.882979	 lr=0.061049
Epoch[103] Batch [1899]/[3760]	Speed: 117.816340 samples/sec	 accuracy=36.214638	 loss=2.882381	 lr=0.061036
Epoch[103] Batch [1949]/[3760]	Speed: 116.234297 samples/sec	 accuracy=36.236378	 loss=2.880819	 lr=0.061024
Epoch[103] Batch [1999]/[3760]	Speed: 117.781092 samples/sec	 accuracy=36.199219	 loss=2.882142	 lr=0.061011
Epoch[103] Batch [2049]/[3760]	Speed: 117.772118 samples/sec	 accuracy=36.163872	 loss=2.884113	 lr=0.060999
Epoch[103] Batch [2099]/[3760]	Speed: 117.741104 samples/sec	 accuracy=36.158482	 loss=2.884318	 lr=0.060986
Epoch[103] Batch [2149]/[3760]	Speed: 116.791711 samples/sec	 accuracy=36.195494	 loss=2.883711	 lr=0.060973
Epoch[103] Batch [2199]/[3760]	Speed: 118.139429 samples/sec	 accuracy=36.191051	 loss=2.883640	 lr=0.060961
Epoch[103] Batch [2249]/[3760]	Speed: 117.365006 samples/sec	 accuracy=36.161806	 loss=2.885277	 lr=0.060948
Epoch[103] Batch [2299]/[3760]	Speed: 117.618627 samples/sec	 accuracy=36.169158	 loss=2.885283	 lr=0.060936
Epoch[103] Batch [2349]/[3760]	Speed: 117.621870 samples/sec	 accuracy=36.174867	 loss=2.885866	 lr=0.060923
Epoch[103] Batch [2399]/[3760]	Speed: 117.252068 samples/sec	 accuracy=36.182943	 loss=2.886322	 lr=0.060911
Epoch[103] Batch [2449]/[3760]	Speed: 117.934260 samples/sec	 accuracy=36.199617	 loss=2.885088	 lr=0.060898
Epoch[103] Batch [2499]/[3760]	Speed: 116.893421 samples/sec	 accuracy=36.220625	 loss=2.884514	 lr=0.060885
Epoch[103] Batch [2549]/[3760]	Speed: 117.543019 samples/sec	 accuracy=36.240196	 loss=2.884105	 lr=0.060873
Epoch[103] Batch [2599]/[3760]	Speed: 117.671926 samples/sec	 accuracy=36.233774	 loss=2.884263	 lr=0.060860
Epoch[103] Batch [2649]/[3760]	Speed: 117.271905 samples/sec	 accuracy=36.232901	 loss=2.884548	 lr=0.060848
Epoch[103] Batch [2699]/[3760]	Speed: 118.404181 samples/sec	 accuracy=36.229745	 loss=2.883689	 lr=0.060835
Epoch[103] Batch [2749]/[3760]	Speed: 116.713922 samples/sec	 accuracy=36.236364	 loss=2.883117	 lr=0.060822
Epoch[103] Batch [2799]/[3760]	Speed: 117.781643 samples/sec	 accuracy=36.230469	 loss=2.883609	 lr=0.060810
Epoch[103] Batch [2849]/[3760]	Speed: 116.365823 samples/sec	 accuracy=36.239583	 loss=2.883943	 lr=0.060797
Epoch[103] Batch [2899]/[3760]	Speed: 117.495286 samples/sec	 accuracy=36.219828	 loss=2.885328	 lr=0.060785
Epoch[103] Batch [2949]/[3760]	Speed: 116.974259 samples/sec	 accuracy=36.235699	 loss=2.885082	 lr=0.060772
Epoch[103] Batch [2999]/[3760]	Speed: 117.509669 samples/sec	 accuracy=36.244271	 loss=2.885412	 lr=0.060759
Epoch[103] Batch [3049]/[3760]	Speed: 117.302135 samples/sec	 accuracy=36.248463	 loss=2.885312	 lr=0.060747
Epoch[103] Batch [3099]/[3760]	Speed: 117.901238 samples/sec	 accuracy=36.234375	 loss=2.885764	 lr=0.060734
Epoch[103] Batch [3149]/[3760]	Speed: 117.128479 samples/sec	 accuracy=36.249008	 loss=2.884815	 lr=0.060722
Epoch[103] Batch [3199]/[3760]	Speed: 117.548626 samples/sec	 accuracy=36.261230	 loss=2.884435	 lr=0.060709
Epoch[103] Batch [3249]/[3760]	Speed: 117.003694 samples/sec	 accuracy=36.262019	 loss=2.884221	 lr=0.060696
Epoch[103] Batch [3299]/[3760]	Speed: 116.943725 samples/sec	 accuracy=36.272254	 loss=2.884370	 lr=0.060684
Epoch[103] Batch [3349]/[3760]	Speed: 117.509673 samples/sec	 accuracy=36.285448	 loss=2.883591	 lr=0.060671
Epoch[103] Batch [3399]/[3760]	Speed: 118.116581 samples/sec	 accuracy=36.273897	 loss=2.884588	 lr=0.060659
Epoch[103] Batch [3449]/[3760]	Speed: 117.016334 samples/sec	 accuracy=36.251359	 loss=2.885861	 lr=0.060646
Epoch[103] Batch [3499]/[3760]	Speed: 117.747294 samples/sec	 accuracy=36.243750	 loss=2.886348	 lr=0.060633
Epoch[103] Batch [3549]/[3760]	Speed: 116.790016 samples/sec	 accuracy=36.261444	 loss=2.885623	 lr=0.060621
Epoch[103] Batch [3599]/[3760]	Speed: 117.526090 samples/sec	 accuracy=36.262153	 loss=2.885734	 lr=0.060608
Epoch[103] Batch [3649]/[3760]	Speed: 116.653902 samples/sec	 accuracy=36.260702	 loss=2.886078	 lr=0.060596
Epoch[103] Batch [3699]/[3760]	Speed: 117.473769 samples/sec	 accuracy=36.243666	 loss=2.886982	 lr=0.060583
Epoch[103] Batch [3749]/[3760]	Speed: 120.699547 samples/sec	 accuracy=36.266250	 loss=2.886304	 lr=0.060570
Batch [0049]/[0303]: acc-top1=37.468750 acc-top5=64.062500
Batch [0099]/[0303]: acc-top1=34.968750 acc-top5=63.078125
Batch [0149]/[0303]: acc-top1=36.510417 acc-top5=63.645833
Batch [0199]/[0303]: acc-top1=39.757812 acc-top5=65.554688
Batch [0249]/[0303]: acc-top1=39.087500 acc-top5=64.956250
Batch [0299]/[0303]: acc-top1=38.578125 acc-top5=64.927083
[Epoch 103] training: accuracy=36.267453	 loss=2.886065
[Epoch 103] speed: 116 samples/sec	time cost: 2177.815857
[Epoch 103] validation: acc-top1=38.567450 acc-top5=64.928837 loss=2.874636
Epoch[104] Batch [0049]/[3759]	Speed: 86.802254 samples/sec	 accuracy=37.312500	 loss=2.808625	 lr=0.060555
Epoch[104] Batch [0099]/[3759]	Speed: 117.864518 samples/sec	 accuracy=37.234375	 loss=2.822574	 lr=0.060543
Epoch[104] Batch [0149]/[3759]	Speed: 117.050267 samples/sec	 accuracy=37.208333	 loss=2.829536	 lr=0.060530
Epoch[104] Batch [0199]/[3759]	Speed: 117.996074 samples/sec	 accuracy=37.000000	 loss=2.837517	 lr=0.060518
Epoch[104] Batch [0249]/[3759]	Speed: 118.077740 samples/sec	 accuracy=36.943750	 loss=2.840010	 lr=0.060505
Epoch[104] Batch [0299]/[3759]	Speed: 116.742686 samples/sec	 accuracy=36.760417	 loss=2.848183	 lr=0.060492
Epoch[104] Batch [0349]/[3759]	Speed: 118.341598 samples/sec	 accuracy=36.714286	 loss=2.852241	 lr=0.060480
Epoch[104] Batch [0399]/[3759]	Speed: 117.331508 samples/sec	 accuracy=36.425781	 loss=2.858604	 lr=0.060467
Epoch[104] Batch [0449]/[3759]	Speed: 117.217086 samples/sec	 accuracy=36.451389	 loss=2.857166	 lr=0.060454
Epoch[104] Batch [0499]/[3759]	Speed: 117.281971 samples/sec	 accuracy=36.471875	 loss=2.861373	 lr=0.060442
Epoch[104] Batch [0549]/[3759]	Speed: 116.855308 samples/sec	 accuracy=36.471591	 loss=2.862309	 lr=0.060429
Epoch[104] Batch [0599]/[3759]	Speed: 117.263878 samples/sec	 accuracy=36.497396	 loss=2.862686	 lr=0.060417
Epoch[104] Batch [0649]/[3759]	Speed: 117.597184 samples/sec	 accuracy=36.475962	 loss=2.863992	 lr=0.060404
Epoch[104] Batch [0699]/[3759]	Speed: 117.999608 samples/sec	 accuracy=36.437500	 loss=2.867712	 lr=0.060391
Epoch[104] Batch [0749]/[3759]	Speed: 116.757190 samples/sec	 accuracy=36.402083	 loss=2.868496	 lr=0.060379
Epoch[104] Batch [0799]/[3759]	Speed: 117.608043 samples/sec	 accuracy=36.386719	 loss=2.871134	 lr=0.060366
Epoch[104] Batch [0849]/[3759]	Speed: 116.032186 samples/sec	 accuracy=36.373162	 loss=2.874342	 lr=0.060354
Epoch[104] Batch [0899]/[3759]	Speed: 118.157374 samples/sec	 accuracy=36.420139	 loss=2.871347	 lr=0.060341
Epoch[104] Batch [0949]/[3759]	Speed: 117.708465 samples/sec	 accuracy=36.445724	 loss=2.871992	 lr=0.060328
Epoch[104] Batch [0999]/[3759]	Speed: 117.064965 samples/sec	 accuracy=36.473438	 loss=2.870417	 lr=0.060316
Epoch[104] Batch [1049]/[3759]	Speed: 117.305771 samples/sec	 accuracy=36.498512	 loss=2.867636	 lr=0.060303
Epoch[104] Batch [1099]/[3759]	Speed: 117.450156 samples/sec	 accuracy=36.481534	 loss=2.868895	 lr=0.060290
Epoch[104] Batch [1149]/[3759]	Speed: 118.409058 samples/sec	 accuracy=36.459239	 loss=2.870089	 lr=0.060278
Epoch[104] Batch [1199]/[3759]	Speed: 117.736591 samples/sec	 accuracy=36.419271	 loss=2.871862	 lr=0.060265
Epoch[104] Batch [1249]/[3759]	Speed: 118.050746 samples/sec	 accuracy=36.411250	 loss=2.872876	 lr=0.060253
Epoch[104] Batch [1299]/[3759]	Speed: 116.850005 samples/sec	 accuracy=36.454327	 loss=2.872438	 lr=0.060240
Epoch[104] Batch [1349]/[3759]	Speed: 117.205088 samples/sec	 accuracy=36.459491	 loss=2.873770	 lr=0.060227
Epoch[104] Batch [1399]/[3759]	Speed: 117.341575 samples/sec	 accuracy=36.479911	 loss=2.875001	 lr=0.060215
Epoch[104] Batch [1449]/[3759]	Speed: 116.976933 samples/sec	 accuracy=36.500000	 loss=2.874406	 lr=0.060202
Epoch[104] Batch [1499]/[3759]	Speed: 118.065544 samples/sec	 accuracy=36.509375	 loss=2.875241	 lr=0.060189
Epoch[104] Batch [1549]/[3759]	Speed: 116.884946 samples/sec	 accuracy=36.518145	 loss=2.875308	 lr=0.060177
Epoch[104] Batch [1599]/[3759]	Speed: 117.368284 samples/sec	 accuracy=36.531250	 loss=2.874607	 lr=0.060164
Epoch[104] Batch [1649]/[3759]	Speed: 116.552964 samples/sec	 accuracy=36.544508	 loss=2.874483	 lr=0.060152
Epoch[104] Batch [1699]/[3759]	Speed: 117.269354 samples/sec	 accuracy=36.514706	 loss=2.875121	 lr=0.060139
Epoch[104] Batch [1749]/[3759]	Speed: 116.471321 samples/sec	 accuracy=36.519643	 loss=2.875383	 lr=0.060126
Epoch[104] Batch [1799]/[3759]	Speed: 117.623251 samples/sec	 accuracy=36.525174	 loss=2.875313	 lr=0.060114
Epoch[104] Batch [1849]/[3759]	Speed: 117.380063 samples/sec	 accuracy=36.515203	 loss=2.876265	 lr=0.060101
Epoch[104] Batch [1899]/[3759]	Speed: 118.169781 samples/sec	 accuracy=36.514803	 loss=2.876500	 lr=0.060088
Epoch[104] Batch [1949]/[3759]	Speed: 118.216614 samples/sec	 accuracy=36.514423	 loss=2.876528	 lr=0.060076
Epoch[104] Batch [1999]/[3759]	Speed: 116.865348 samples/sec	 accuracy=36.516406	 loss=2.877056	 lr=0.060063
Epoch[104] Batch [2049]/[3759]	Speed: 117.563391 samples/sec	 accuracy=36.499238	 loss=2.877644	 lr=0.060051
Epoch[104] Batch [2099]/[3759]	Speed: 117.363864 samples/sec	 accuracy=36.487351	 loss=2.877937	 lr=0.060038
Epoch[104] Batch [2149]/[3759]	Speed: 117.010165 samples/sec	 accuracy=36.497093	 loss=2.877722	 lr=0.060025
Epoch[104] Batch [2199]/[3759]	Speed: 117.087230 samples/sec	 accuracy=36.497159	 loss=2.878037	 lr=0.060013
Epoch[104] Batch [2249]/[3759]	Speed: 118.272335 samples/sec	 accuracy=36.507639	 loss=2.878104	 lr=0.060000
Epoch[104] Batch [2299]/[3759]	Speed: 117.116853 samples/sec	 accuracy=36.538043	 loss=2.876409	 lr=0.059987
Epoch[104] Batch [2349]/[3759]	Speed: 116.913220 samples/sec	 accuracy=36.518617	 loss=2.876536	 lr=0.059975
Epoch[104] Batch [2399]/[3759]	Speed: 117.519639 samples/sec	 accuracy=36.532552	 loss=2.876317	 lr=0.059962
Epoch[104] Batch [2449]/[3759]	Speed: 117.462058 samples/sec	 accuracy=36.536990	 loss=2.876209	 lr=0.059949
Epoch[104] Batch [2499]/[3759]	Speed: 117.656054 samples/sec	 accuracy=36.540000	 loss=2.877261	 lr=0.059937
Epoch[104] Batch [2549]/[3759]	Speed: 116.993695 samples/sec	 accuracy=36.540441	 loss=2.877372	 lr=0.059924
Epoch[104] Batch [2599]/[3759]	Speed: 117.663259 samples/sec	 accuracy=36.515024	 loss=2.878138	 lr=0.059911
Epoch[104] Batch [2649]/[3759]	Speed: 118.156971 samples/sec	 accuracy=36.528892	 loss=2.877246	 lr=0.059899
Epoch[104] Batch [2699]/[3759]	Speed: 116.951442 samples/sec	 accuracy=36.490741	 loss=2.878099	 lr=0.059886
Epoch[104] Batch [2749]/[3759]	Speed: 116.600583 samples/sec	 accuracy=36.461932	 loss=2.879609	 lr=0.059874
Epoch[104] Batch [2799]/[3759]	Speed: 117.203262 samples/sec	 accuracy=36.454241	 loss=2.880098	 lr=0.059861
Epoch[104] Batch [2849]/[3759]	Speed: 117.414956 samples/sec	 accuracy=36.512061	 loss=2.878349	 lr=0.059848
Epoch[104] Batch [2899]/[3759]	Speed: 117.311946 samples/sec	 accuracy=36.531250	 loss=2.877589	 lr=0.059836
Epoch[104] Batch [2949]/[3759]	Speed: 117.465004 samples/sec	 accuracy=36.539195	 loss=2.877872	 lr=0.059823
Epoch[104] Batch [2999]/[3759]	Speed: 117.458094 samples/sec	 accuracy=36.538542	 loss=2.877446	 lr=0.059810
Epoch[104] Batch [3049]/[3759]	Speed: 118.126536 samples/sec	 accuracy=36.516906	 loss=2.877522	 lr=0.059798
Epoch[104] Batch [3099]/[3759]	Speed: 117.051704 samples/sec	 accuracy=36.538810	 loss=2.876467	 lr=0.059785
Epoch[104] Batch [3149]/[3759]	Speed: 116.773872 samples/sec	 accuracy=36.551091	 loss=2.876035	 lr=0.059772
Epoch[104] Batch [3199]/[3759]	Speed: 117.013523 samples/sec	 accuracy=36.558105	 loss=2.875196	 lr=0.059760
Epoch[104] Batch [3249]/[3759]	Speed: 116.766336 samples/sec	 accuracy=36.563462	 loss=2.875314	 lr=0.059747
Epoch[104] Batch [3299]/[3759]	Speed: 118.184704 samples/sec	 accuracy=36.577178	 loss=2.874859	 lr=0.059734
Epoch[104] Batch [3349]/[3759]	Speed: 117.744977 samples/sec	 accuracy=36.573228	 loss=2.874455	 lr=0.059722
Epoch[104] Batch [3399]/[3759]	Speed: 117.397486 samples/sec	 accuracy=36.539982	 loss=2.875715	 lr=0.059709
Epoch[104] Batch [3449]/[3759]	Speed: 117.585506 samples/sec	 accuracy=36.556612	 loss=2.874508	 lr=0.059696
Epoch[104] Batch [3499]/[3759]	Speed: 117.668216 samples/sec	 accuracy=36.566518	 loss=2.874746	 lr=0.059684
Epoch[104] Batch [3549]/[3759]	Speed: 117.603679 samples/sec	 accuracy=36.580106	 loss=2.874412	 lr=0.059671
Epoch[104] Batch [3599]/[3759]	Speed: 116.014829 samples/sec	 accuracy=36.575521	 loss=2.875102	 lr=0.059659
Epoch[104] Batch [3649]/[3759]	Speed: 117.500585 samples/sec	 accuracy=36.590753	 loss=2.874862	 lr=0.059646
Epoch[104] Batch [3699]/[3759]	Speed: 117.974739 samples/sec	 accuracy=36.560389	 loss=2.875442	 lr=0.059633
Epoch[104] Batch [3749]/[3759]	Speed: 120.287302 samples/sec	 accuracy=36.557083	 loss=2.876515	 lr=0.059621
Batch [0049]/[0303]: acc-top1=33.531250 acc-top5=57.687500
Batch [0099]/[0303]: acc-top1=31.625000 acc-top5=56.718750
Batch [0149]/[0303]: acc-top1=32.885417 acc-top5=58.833333
Batch [0199]/[0303]: acc-top1=35.507812 acc-top5=60.882813
Batch [0249]/[0303]: acc-top1=34.450000 acc-top5=60.512500
Batch [0299]/[0303]: acc-top1=34.635417 acc-top5=61.062500
[Epoch 104] training: accuracy=36.547702	 loss=2.876948
[Epoch 104] speed: 116 samples/sec	time cost: 2176.441641
[Epoch 104] validation: acc-top1=34.637995 acc-top5=61.143771 loss=3.136478
Epoch[105] Batch [0049]/[3760]	Speed: 85.890300 samples/sec	 accuracy=38.468750	 loss=2.809970	 lr=0.059606
Epoch[105] Batch [0099]/[3760]	Speed: 118.411660 samples/sec	 accuracy=37.593750	 loss=2.829292	 lr=0.059593
Epoch[105] Batch [0149]/[3760]	Speed: 117.629264 samples/sec	 accuracy=37.572917	 loss=2.824719	 lr=0.059580
Epoch[105] Batch [0199]/[3760]	Speed: 117.532734 samples/sec	 accuracy=37.375000	 loss=2.827859	 lr=0.059568
Epoch[105] Batch [0249]/[3760]	Speed: 117.480332 samples/sec	 accuracy=37.237500	 loss=2.831506	 lr=0.059555
Epoch[105] Batch [0299]/[3760]	Speed: 117.398116 samples/sec	 accuracy=37.072917	 loss=2.840308	 lr=0.059542
Epoch[105] Batch [0349]/[3760]	Speed: 117.803753 samples/sec	 accuracy=37.080357	 loss=2.848628	 lr=0.059530
Epoch[105] Batch [0399]/[3760]	Speed: 116.684415 samples/sec	 accuracy=37.101562	 loss=2.847278	 lr=0.059517
Epoch[105] Batch [0449]/[3760]	Speed: 117.502466 samples/sec	 accuracy=37.131944	 loss=2.848926	 lr=0.059504
Epoch[105] Batch [0499]/[3760]	Speed: 117.662000 samples/sec	 accuracy=37.100000	 loss=2.850879	 lr=0.059492
Epoch[105] Batch [0549]/[3760]	Speed: 116.824137 samples/sec	 accuracy=37.088068	 loss=2.851128	 lr=0.059479
Epoch[105] Batch [0599]/[3760]	Speed: 117.964146 samples/sec	 accuracy=36.908854	 loss=2.854689	 lr=0.059466
Epoch[105] Batch [0649]/[3760]	Speed: 117.299952 samples/sec	 accuracy=36.947115	 loss=2.852191	 lr=0.059454
Epoch[105] Batch [0699]/[3760]	Speed: 117.102762 samples/sec	 accuracy=36.964286	 loss=2.854942	 lr=0.059441
Epoch[105] Batch [0749]/[3760]	Speed: 116.275015 samples/sec	 accuracy=36.954167	 loss=2.854353	 lr=0.059428
Epoch[105] Batch [0799]/[3760]	Speed: 117.449408 samples/sec	 accuracy=36.947266	 loss=2.854324	 lr=0.059416
Epoch[105] Batch [0849]/[3760]	Speed: 117.358375 samples/sec	 accuracy=36.786765	 loss=2.861359	 lr=0.059403
Epoch[105] Batch [0899]/[3760]	Speed: 118.184726 samples/sec	 accuracy=36.796875	 loss=2.858780	 lr=0.059390
Epoch[105] Batch [0949]/[3760]	Speed: 117.996939 samples/sec	 accuracy=36.886513	 loss=2.855966	 lr=0.059378
Epoch[105] Batch [0999]/[3760]	Speed: 117.344756 samples/sec	 accuracy=36.870313	 loss=2.856889	 lr=0.059365
Epoch[105] Batch [1049]/[3760]	Speed: 116.740865 samples/sec	 accuracy=36.867560	 loss=2.855244	 lr=0.059352
Epoch[105] Batch [1099]/[3760]	Speed: 117.165737 samples/sec	 accuracy=36.852273	 loss=2.857441	 lr=0.059340
Epoch[105] Batch [1149]/[3760]	Speed: 116.531865 samples/sec	 accuracy=36.849185	 loss=2.856604	 lr=0.059327
Epoch[105] Batch [1199]/[3760]	Speed: 116.628386 samples/sec	 accuracy=36.839844	 loss=2.858650	 lr=0.059314
Epoch[105] Batch [1249]/[3760]	Speed: 117.588752 samples/sec	 accuracy=36.821250	 loss=2.858758	 lr=0.059302
Epoch[105] Batch [1299]/[3760]	Speed: 116.942328 samples/sec	 accuracy=36.822115	 loss=2.859094	 lr=0.059289
Epoch[105] Batch [1349]/[3760]	Speed: 117.058203 samples/sec	 accuracy=36.799769	 loss=2.860171	 lr=0.059276
Epoch[105] Batch [1399]/[3760]	Speed: 117.043081 samples/sec	 accuracy=36.787946	 loss=2.860642	 lr=0.059264
Epoch[105] Batch [1449]/[3760]	Speed: 117.822530 samples/sec	 accuracy=36.811422	 loss=2.860504	 lr=0.059251
Epoch[105] Batch [1499]/[3760]	Speed: 117.745652 samples/sec	 accuracy=36.822917	 loss=2.859520	 lr=0.059238
Epoch[105] Batch [1549]/[3760]	Speed: 116.400493 samples/sec	 accuracy=36.811492	 loss=2.858977	 lr=0.059226
Epoch[105] Batch [1599]/[3760]	Speed: 117.755710 samples/sec	 accuracy=36.783203	 loss=2.859233	 lr=0.059213
Epoch[105] Batch [1649]/[3760]	Speed: 115.259400 samples/sec	 accuracy=36.782197	 loss=2.858525	 lr=0.059200
Epoch[105] Batch [1699]/[3760]	Speed: 117.542314 samples/sec	 accuracy=36.802390	 loss=2.859183	 lr=0.059188
Epoch[105] Batch [1749]/[3760]	Speed: 117.326167 samples/sec	 accuracy=36.825893	 loss=2.858791	 lr=0.059175
Epoch[105] Batch [1799]/[3760]	Speed: 117.002392 samples/sec	 accuracy=36.807292	 loss=2.859950	 lr=0.059162
Epoch[105] Batch [1849]/[3760]	Speed: 117.567238 samples/sec	 accuracy=36.807432	 loss=2.859715	 lr=0.059150
Epoch[105] Batch [1899]/[3760]	Speed: 117.152110 samples/sec	 accuracy=36.793586	 loss=2.859686	 lr=0.059137
Epoch[105] Batch [1949]/[3760]	Speed: 117.262799 samples/sec	 accuracy=36.783654	 loss=2.860423	 lr=0.059124
Epoch[105] Batch [1999]/[3760]	Speed: 117.086621 samples/sec	 accuracy=36.739844	 loss=2.861886	 lr=0.059112
Epoch[105] Batch [2049]/[3760]	Speed: 117.889469 samples/sec	 accuracy=36.734756	 loss=2.862010	 lr=0.059099
Epoch[105] Batch [2099]/[3760]	Speed: 117.475830 samples/sec	 accuracy=36.715774	 loss=2.862850	 lr=0.059086
Epoch[105] Batch [2149]/[3760]	Speed: 117.332392 samples/sec	 accuracy=36.714390	 loss=2.862031	 lr=0.059073
Epoch[105] Batch [2199]/[3760]	Speed: 117.404549 samples/sec	 accuracy=36.698864	 loss=2.861910	 lr=0.059061
Epoch[105] Batch [2249]/[3760]	Speed: 118.304409 samples/sec	 accuracy=36.681250	 loss=2.863844	 lr=0.059048
Epoch[105] Batch [2299]/[3760]	Speed: 116.861183 samples/sec	 accuracy=36.667120	 loss=2.864880	 lr=0.059035
Epoch[105] Batch [2349]/[3760]	Speed: 117.965118 samples/sec	 accuracy=36.650266	 loss=2.865877	 lr=0.059023
Epoch[105] Batch [2399]/[3760]	Speed: 117.085989 samples/sec	 accuracy=36.652344	 loss=2.866346	 lr=0.059010
Epoch[105] Batch [2449]/[3760]	Speed: 116.931431 samples/sec	 accuracy=36.666454	 loss=2.866646	 lr=0.058997
Epoch[105] Batch [2499]/[3760]	Speed: 117.218432 samples/sec	 accuracy=36.695625	 loss=2.865758	 lr=0.058985
Epoch[105] Batch [2549]/[3760]	Speed: 117.963840 samples/sec	 accuracy=36.694240	 loss=2.865383	 lr=0.058972
Epoch[105] Batch [2599]/[3760]	Speed: 116.364840 samples/sec	 accuracy=36.666466	 loss=2.866642	 lr=0.058959
Epoch[105] Batch [2649]/[3760]	Speed: 117.539180 samples/sec	 accuracy=36.676297	 loss=2.866717	 lr=0.058947
Epoch[105] Batch [2699]/[3760]	Speed: 117.392138 samples/sec	 accuracy=36.699074	 loss=2.866424	 lr=0.058934
Epoch[105] Batch [2749]/[3760]	Speed: 117.422659 samples/sec	 accuracy=36.719886	 loss=2.866548	 lr=0.058921
Epoch[105] Batch [2799]/[3760]	Speed: 117.749124 samples/sec	 accuracy=36.725446	 loss=2.867056	 lr=0.058909
Epoch[105] Batch [2849]/[3760]	Speed: 117.338160 samples/sec	 accuracy=36.726425	 loss=2.866666	 lr=0.058896
Epoch[105] Batch [2899]/[3760]	Speed: 117.210614 samples/sec	 accuracy=36.724138	 loss=2.866686	 lr=0.058883
Epoch[105] Batch [2949]/[3760]	Speed: 118.160652 samples/sec	 accuracy=36.736229	 loss=2.866715	 lr=0.058870
Epoch[105] Batch [2999]/[3760]	Speed: 118.025363 samples/sec	 accuracy=36.721875	 loss=2.866915	 lr=0.058858
Epoch[105] Batch [3049]/[3760]	Speed: 117.116578 samples/sec	 accuracy=36.718750	 loss=2.866888	 lr=0.058845
Epoch[105] Batch [3099]/[3760]	Speed: 116.733286 samples/sec	 accuracy=36.714718	 loss=2.867445	 lr=0.058832
Epoch[105] Batch [3149]/[3760]	Speed: 117.508230 samples/sec	 accuracy=36.716766	 loss=2.867069	 lr=0.058820
Epoch[105] Batch [3199]/[3760]	Speed: 117.727145 samples/sec	 accuracy=36.712891	 loss=2.867161	 lr=0.058807
Epoch[105] Batch [3249]/[3760]	Speed: 118.160098 samples/sec	 accuracy=36.701442	 loss=2.867650	 lr=0.058794
Epoch[105] Batch [3299]/[3760]	Speed: 117.409585 samples/sec	 accuracy=36.714489	 loss=2.866263	 lr=0.058782
Epoch[105] Batch [3349]/[3760]	Speed: 118.433946 samples/sec	 accuracy=36.701026	 loss=2.866448	 lr=0.058769
Epoch[105] Batch [3399]/[3760]	Speed: 117.539529 samples/sec	 accuracy=36.694393	 loss=2.866600	 lr=0.058756
Epoch[105] Batch [3449]/[3760]	Speed: 117.655479 samples/sec	 accuracy=36.705616	 loss=2.865955	 lr=0.058744
Epoch[105] Batch [3499]/[3760]	Speed: 118.433394 samples/sec	 accuracy=36.682589	 loss=2.866305	 lr=0.058731
Epoch[105] Batch [3549]/[3760]	Speed: 116.594477 samples/sec	 accuracy=36.679577	 loss=2.866366	 lr=0.058718
Epoch[105] Batch [3599]/[3760]	Speed: 118.390674 samples/sec	 accuracy=36.664497	 loss=2.866422	 lr=0.058705
Epoch[105] Batch [3649]/[3760]	Speed: 117.642548 samples/sec	 accuracy=36.685360	 loss=2.865357	 lr=0.058693
Epoch[105] Batch [3699]/[3760]	Speed: 117.421889 samples/sec	 accuracy=36.660473	 loss=2.866550	 lr=0.058680
Epoch[105] Batch [3749]/[3760]	Speed: 120.260318 samples/sec	 accuracy=36.670833	 loss=2.866407	 lr=0.058667
Batch [0049]/[0303]: acc-top1=35.656250 acc-top5=62.500000
Batch [0099]/[0303]: acc-top1=34.625000 acc-top5=62.343750
Batch [0149]/[0303]: acc-top1=35.739583 acc-top5=62.510417
Batch [0199]/[0303]: acc-top1=38.867188 acc-top5=65.171875
Batch [0249]/[0303]: acc-top1=38.106250 acc-top5=64.781250
Batch [0299]/[0303]: acc-top1=38.010417 acc-top5=65.031250
[Epoch 105] training: accuracy=36.682181	 loss=2.866072
[Epoch 105] speed: 116 samples/sec	time cost: 2176.966857
[Epoch 105] validation: acc-top1=37.953795 acc-top5=64.944307 loss=2.924250
Epoch[106] Batch [0049]/[3760]	Speed: 88.861545 samples/sec	 accuracy=38.031250	 loss=2.812805	 lr=0.058652
Epoch[106] Batch [0099]/[3760]	Speed: 117.436286 samples/sec	 accuracy=38.046875	 loss=2.789465	 lr=0.058639
Epoch[106] Batch [0149]/[3760]	Speed: 117.070357 samples/sec	 accuracy=37.906250	 loss=2.806560	 lr=0.058627
Epoch[106] Batch [0199]/[3760]	Speed: 116.711858 samples/sec	 accuracy=37.585938	 loss=2.822886	 lr=0.058614
Epoch[106] Batch [0249]/[3760]	Speed: 117.672387 samples/sec	 accuracy=37.643750	 loss=2.829785	 lr=0.058601
Epoch[106] Batch [0299]/[3760]	Speed: 116.238638 samples/sec	 accuracy=37.619792	 loss=2.824904	 lr=0.058589
Epoch[106] Batch [0349]/[3760]	Speed: 117.862281 samples/sec	 accuracy=37.526786	 loss=2.829265	 lr=0.058576
Epoch[106] Batch [0399]/[3760]	Speed: 117.072062 samples/sec	 accuracy=37.390625	 loss=2.831960	 lr=0.058563
Epoch[106] Batch [0449]/[3760]	Speed: 118.223472 samples/sec	 accuracy=37.305556	 loss=2.836323	 lr=0.058550
Epoch[106] Batch [0499]/[3760]	Speed: 116.891965 samples/sec	 accuracy=37.287500	 loss=2.833881	 lr=0.058538
Epoch[106] Batch [0549]/[3760]	Speed: 116.874344 samples/sec	 accuracy=37.315341	 loss=2.830352	 lr=0.058525
Epoch[106] Batch [0599]/[3760]	Speed: 117.725028 samples/sec	 accuracy=37.312500	 loss=2.830367	 lr=0.058512
Epoch[106] Batch [0649]/[3760]	Speed: 115.463806 samples/sec	 accuracy=37.223558	 loss=2.829664	 lr=0.058500
Epoch[106] Batch [0699]/[3760]	Speed: 117.501369 samples/sec	 accuracy=37.185268	 loss=2.834736	 lr=0.058487
Epoch[106] Batch [0749]/[3760]	Speed: 117.424115 samples/sec	 accuracy=37.137500	 loss=2.835786	 lr=0.058474
Epoch[106] Batch [0799]/[3760]	Speed: 116.976860 samples/sec	 accuracy=37.173828	 loss=2.831352	 lr=0.058461
Epoch[106] Batch [0849]/[3760]	Speed: 116.993623 samples/sec	 accuracy=37.097426	 loss=2.835740	 lr=0.058449
Epoch[106] Batch [0899]/[3760]	Speed: 117.047656 samples/sec	 accuracy=37.107639	 loss=2.832657	 lr=0.058436
Epoch[106] Batch [0949]/[3760]	Speed: 117.288510 samples/sec	 accuracy=37.059211	 loss=2.834742	 lr=0.058423
Epoch[106] Batch [0999]/[3760]	Speed: 117.399974 samples/sec	 accuracy=37.017187	 loss=2.834995	 lr=0.058411
Epoch[106] Batch [1049]/[3760]	Speed: 116.923460 samples/sec	 accuracy=36.989583	 loss=2.834435	 lr=0.058398
Epoch[106] Batch [1099]/[3760]	Speed: 117.470728 samples/sec	 accuracy=36.970170	 loss=2.834151	 lr=0.058385
Epoch[106] Batch [1149]/[3760]	Speed: 117.651413 samples/sec	 accuracy=36.953804	 loss=2.833258	 lr=0.058372
Epoch[106] Batch [1199]/[3760]	Speed: 116.931738 samples/sec	 accuracy=36.964844	 loss=2.833010	 lr=0.058360
Epoch[106] Batch [1249]/[3760]	Speed: 118.058223 samples/sec	 accuracy=36.978750	 loss=2.833314	 lr=0.058347
Epoch[106] Batch [1299]/[3760]	Speed: 117.033338 samples/sec	 accuracy=36.963942	 loss=2.834454	 lr=0.058334
Epoch[106] Batch [1349]/[3760]	Speed: 117.315390 samples/sec	 accuracy=36.983796	 loss=2.834530	 lr=0.058322
Epoch[106] Batch [1399]/[3760]	Speed: 117.688332 samples/sec	 accuracy=36.981027	 loss=2.836099	 lr=0.058309
Epoch[106] Batch [1449]/[3760]	Speed: 117.176059 samples/sec	 accuracy=36.955819	 loss=2.837358	 lr=0.058296
Epoch[106] Batch [1499]/[3760]	Speed: 116.982355 samples/sec	 accuracy=36.976042	 loss=2.837828	 lr=0.058283
Epoch[106] Batch [1549]/[3760]	Speed: 117.765358 samples/sec	 accuracy=36.957661	 loss=2.838973	 lr=0.058271
Epoch[106] Batch [1599]/[3760]	Speed: 116.137002 samples/sec	 accuracy=36.960938	 loss=2.840157	 lr=0.058258
Epoch[106] Batch [1649]/[3760]	Speed: 117.850455 samples/sec	 accuracy=36.925189	 loss=2.841675	 lr=0.058245
Epoch[106] Batch [1699]/[3760]	Speed: 118.022788 samples/sec	 accuracy=36.958640	 loss=2.840586	 lr=0.058233
Epoch[106] Batch [1749]/[3760]	Speed: 117.660305 samples/sec	 accuracy=36.939286	 loss=2.841113	 lr=0.058220
Epoch[106] Batch [1799]/[3760]	Speed: 117.635993 samples/sec	 accuracy=36.922743	 loss=2.842650	 lr=0.058207
Epoch[106] Batch [1849]/[3760]	Speed: 117.690705 samples/sec	 accuracy=36.895270	 loss=2.844904	 lr=0.058194
Epoch[106] Batch [1899]/[3760]	Speed: 118.521103 samples/sec	 accuracy=36.888980	 loss=2.844933	 lr=0.058182
Epoch[106] Batch [1949]/[3760]	Speed: 117.085983 samples/sec	 accuracy=36.886218	 loss=2.845964	 lr=0.058169
Epoch[106] Batch [1999]/[3760]	Speed: 117.252737 samples/sec	 accuracy=36.882812	 loss=2.846519	 lr=0.058156
Epoch[106] Batch [2049]/[3760]	Speed: 117.594206 samples/sec	 accuracy=36.881860	 loss=2.845973	 lr=0.058143
Epoch[106] Batch [2099]/[3760]	Speed: 117.801648 samples/sec	 accuracy=36.851935	 loss=2.847593	 lr=0.058131
Epoch[106] Batch [2149]/[3760]	Speed: 117.433720 samples/sec	 accuracy=36.868459	 loss=2.847744	 lr=0.058118
Epoch[106] Batch [2199]/[3760]	Speed: 117.880797 samples/sec	 accuracy=36.839489	 loss=2.850053	 lr=0.058105
Epoch[106] Batch [2249]/[3760]	Speed: 117.388051 samples/sec	 accuracy=36.829861	 loss=2.850831	 lr=0.058093
Epoch[106] Batch [2299]/[3760]	Speed: 117.205012 samples/sec	 accuracy=36.825408	 loss=2.852310	 lr=0.058080
Epoch[106] Batch [2349]/[3760]	Speed: 117.321026 samples/sec	 accuracy=36.805851	 loss=2.853552	 lr=0.058067
Epoch[106] Batch [2399]/[3760]	Speed: 117.188979 samples/sec	 accuracy=36.827474	 loss=2.853758	 lr=0.058054
Epoch[106] Batch [2449]/[3760]	Speed: 117.765556 samples/sec	 accuracy=36.832270	 loss=2.853999	 lr=0.058042
Epoch[106] Batch [2499]/[3760]	Speed: 118.020740 samples/sec	 accuracy=36.816875	 loss=2.854046	 lr=0.058029
Epoch[106] Batch [2549]/[3760]	Speed: 116.391055 samples/sec	 accuracy=36.808824	 loss=2.854877	 lr=0.058016
Epoch[106] Batch [2599]/[3760]	Speed: 117.136879 samples/sec	 accuracy=36.768630	 loss=2.855882	 lr=0.058003
Epoch[106] Batch [2649]/[3760]	Speed: 116.952966 samples/sec	 accuracy=36.760024	 loss=2.856227	 lr=0.057991
Epoch[106] Batch [2699]/[3760]	Speed: 117.707704 samples/sec	 accuracy=36.782407	 loss=2.855464	 lr=0.057978
Epoch[106] Batch [2749]/[3760]	Speed: 117.086926 samples/sec	 accuracy=36.772727	 loss=2.856497	 lr=0.057965
Epoch[106] Batch [2799]/[3760]	Speed: 117.632507 samples/sec	 accuracy=36.794643	 loss=2.856112	 lr=0.057953
Epoch[106] Batch [2849]/[3760]	Speed: 117.247629 samples/sec	 accuracy=36.805921	 loss=2.855875	 lr=0.057940
Epoch[106] Batch [2899]/[3760]	Speed: 117.361360 samples/sec	 accuracy=36.810345	 loss=2.855037	 lr=0.057927
Epoch[106] Batch [2949]/[3760]	Speed: 118.309114 samples/sec	 accuracy=36.828390	 loss=2.855306	 lr=0.057914
Epoch[106] Batch [2999]/[3760]	Speed: 117.335402 samples/sec	 accuracy=36.815625	 loss=2.856271	 lr=0.057902
Epoch[106] Batch [3049]/[3760]	Speed: 117.679616 samples/sec	 accuracy=36.788422	 loss=2.857371	 lr=0.057889
Epoch[106] Batch [3099]/[3760]	Speed: 117.542644 samples/sec	 accuracy=36.811996	 loss=2.856444	 lr=0.057876
Epoch[106] Batch [3149]/[3760]	Speed: 117.102313 samples/sec	 accuracy=36.786706	 loss=2.857286	 lr=0.057863
Epoch[106] Batch [3199]/[3760]	Speed: 116.765082 samples/sec	 accuracy=36.777832	 loss=2.857253	 lr=0.057851
Epoch[106] Batch [3249]/[3760]	Speed: 117.661101 samples/sec	 accuracy=36.782692	 loss=2.856879	 lr=0.057838
Epoch[106] Batch [3299]/[3760]	Speed: 117.675585 samples/sec	 accuracy=36.767992	 loss=2.857896	 lr=0.057825
Epoch[106] Batch [3349]/[3760]	Speed: 117.192069 samples/sec	 accuracy=36.780317	 loss=2.857660	 lr=0.057812
Epoch[106] Batch [3399]/[3760]	Speed: 117.230655 samples/sec	 accuracy=36.781710	 loss=2.857726	 lr=0.057800
Epoch[106] Batch [3449]/[3760]	Speed: 117.780787 samples/sec	 accuracy=36.755435	 loss=2.858798	 lr=0.057787
Epoch[106] Batch [3499]/[3760]	Speed: 115.887221 samples/sec	 accuracy=36.737500	 loss=2.859681	 lr=0.057774
Epoch[106] Batch [3549]/[3760]	Speed: 117.698722 samples/sec	 accuracy=36.741197	 loss=2.859746	 lr=0.057762
Epoch[106] Batch [3599]/[3760]	Speed: 117.489837 samples/sec	 accuracy=36.747396	 loss=2.859855	 lr=0.057749
Epoch[106] Batch [3649]/[3760]	Speed: 117.815020 samples/sec	 accuracy=36.735873	 loss=2.860468	 lr=0.057736
Epoch[106] Batch [3699]/[3760]	Speed: 117.816620 samples/sec	 accuracy=36.723395	 loss=2.861391	 lr=0.057723
Epoch[106] Batch [3749]/[3760]	Speed: 120.361770 samples/sec	 accuracy=36.736667	 loss=2.860929	 lr=0.057711
Batch [0049]/[0303]: acc-top1=37.437500 acc-top5=63.656250
Batch [0099]/[0303]: acc-top1=36.312500 acc-top5=62.500000
Batch [0149]/[0303]: acc-top1=37.697917 acc-top5=63.635417
Batch [0199]/[0303]: acc-top1=40.187500 acc-top5=65.898438
Batch [0249]/[0303]: acc-top1=39.206250 acc-top5=65.893750
Batch [0299]/[0303]: acc-top1=38.796875 acc-top5=65.848958
[Epoch 106] training: accuracy=36.743268	 loss=2.860648
[Epoch 106] speed: 116 samples/sec	time cost: 2176.102053
[Epoch 106] validation: acc-top1=38.784035 acc-top5=65.877682 loss=2.871571
Epoch[107] Batch [0049]/[3759]	Speed: 89.191900 samples/sec	 accuracy=37.062500	 loss=2.814864	 lr=0.057695
Epoch[107] Batch [0099]/[3759]	Speed: 117.686476 samples/sec	 accuracy=37.609375	 loss=2.815885	 lr=0.057682
Epoch[107] Batch [0149]/[3759]	Speed: 118.311084 samples/sec	 accuracy=37.656250	 loss=2.812339	 lr=0.057670
Epoch[107] Batch [0199]/[3759]	Speed: 116.907393 samples/sec	 accuracy=37.476562	 loss=2.818210	 lr=0.057657
Epoch[107] Batch [0249]/[3759]	Speed: 117.985901 samples/sec	 accuracy=37.437500	 loss=2.829609	 lr=0.057644
Epoch[107] Batch [0299]/[3759]	Speed: 117.959124 samples/sec	 accuracy=37.536458	 loss=2.816735	 lr=0.057632
Epoch[107] Batch [0349]/[3759]	Speed: 117.652617 samples/sec	 accuracy=37.508929	 loss=2.819816	 lr=0.057619
Epoch[107] Batch [0399]/[3759]	Speed: 117.408035 samples/sec	 accuracy=37.328125	 loss=2.822153	 lr=0.057606
Epoch[107] Batch [0449]/[3759]	Speed: 117.360953 samples/sec	 accuracy=37.305556	 loss=2.817747	 lr=0.057593
Epoch[107] Batch [0499]/[3759]	Speed: 116.821359 samples/sec	 accuracy=37.262500	 loss=2.821783	 lr=0.057581
Epoch[107] Batch [0549]/[3759]	Speed: 116.821127 samples/sec	 accuracy=37.150568	 loss=2.828035	 lr=0.057568
Epoch[107] Batch [0599]/[3759]	Speed: 117.076855 samples/sec	 accuracy=37.122396	 loss=2.827178	 lr=0.057555
Epoch[107] Batch [0649]/[3759]	Speed: 117.349068 samples/sec	 accuracy=37.038462	 loss=2.835256	 lr=0.057542
Epoch[107] Batch [0699]/[3759]	Speed: 116.756279 samples/sec	 accuracy=37.216518	 loss=2.827975	 lr=0.057530
Epoch[107] Batch [0749]/[3759]	Speed: 117.239945 samples/sec	 accuracy=37.175000	 loss=2.828746	 lr=0.057517
Epoch[107] Batch [0799]/[3759]	Speed: 118.208886 samples/sec	 accuracy=37.150391	 loss=2.830685	 lr=0.057504
Epoch[107] Batch [0849]/[3759]	Speed: 116.577312 samples/sec	 accuracy=37.240809	 loss=2.829819	 lr=0.057491
Epoch[107] Batch [0899]/[3759]	Speed: 117.204708 samples/sec	 accuracy=37.256944	 loss=2.832246	 lr=0.057479
Epoch[107] Batch [0949]/[3759]	Speed: 117.193473 samples/sec	 accuracy=37.296053	 loss=2.828419	 lr=0.057466
Epoch[107] Batch [0999]/[3759]	Speed: 117.556897 samples/sec	 accuracy=37.284375	 loss=2.828789	 lr=0.057453
Epoch[107] Batch [1049]/[3759]	Speed: 117.871186 samples/sec	 accuracy=37.296131	 loss=2.827483	 lr=0.057440
Epoch[107] Batch [1099]/[3759]	Speed: 116.966586 samples/sec	 accuracy=37.210227	 loss=2.830635	 lr=0.057428
Epoch[107] Batch [1149]/[3759]	Speed: 117.608931 samples/sec	 accuracy=37.203804	 loss=2.830444	 lr=0.057415
Epoch[107] Batch [1199]/[3759]	Speed: 118.193136 samples/sec	 accuracy=37.253906	 loss=2.829825	 lr=0.057402
Epoch[107] Batch [1249]/[3759]	Speed: 117.248725 samples/sec	 accuracy=37.225000	 loss=2.829685	 lr=0.057389
Epoch[107] Batch [1299]/[3759]	Speed: 117.594592 samples/sec	 accuracy=37.207933	 loss=2.830937	 lr=0.057376
Epoch[107] Batch [1349]/[3759]	Speed: 117.060926 samples/sec	 accuracy=37.228009	 loss=2.831307	 lr=0.057364
Epoch[107] Batch [1399]/[3759]	Speed: 116.889243 samples/sec	 accuracy=37.208705	 loss=2.832666	 lr=0.057351
Epoch[107] Batch [1449]/[3759]	Speed: 117.681319 samples/sec	 accuracy=37.201509	 loss=2.833799	 lr=0.057338
Epoch[107] Batch [1499]/[3759]	Speed: 117.929764 samples/sec	 accuracy=37.222917	 loss=2.834240	 lr=0.057325
Epoch[107] Batch [1549]/[3759]	Speed: 117.906486 samples/sec	 accuracy=37.190524	 loss=2.836528	 lr=0.057313
Epoch[107] Batch [1599]/[3759]	Speed: 118.035656 samples/sec	 accuracy=37.207031	 loss=2.836676	 lr=0.057300
Epoch[107] Batch [1649]/[3759]	Speed: 116.035812 samples/sec	 accuracy=37.196023	 loss=2.836824	 lr=0.057287
Epoch[107] Batch [1699]/[3759]	Speed: 117.501816 samples/sec	 accuracy=37.165441	 loss=2.837777	 lr=0.057274
Epoch[107] Batch [1749]/[3759]	Speed: 117.789021 samples/sec	 accuracy=37.178571	 loss=2.837026	 lr=0.057262
Epoch[107] Batch [1799]/[3759]	Speed: 117.704023 samples/sec	 accuracy=37.151042	 loss=2.837728	 lr=0.057249
Epoch[107] Batch [1849]/[3759]	Speed: 117.536280 samples/sec	 accuracy=37.168919	 loss=2.837979	 lr=0.057236
Epoch[107] Batch [1899]/[3759]	Speed: 116.576258 samples/sec	 accuracy=37.152138	 loss=2.839481	 lr=0.057223
Epoch[107] Batch [1949]/[3759]	Speed: 117.592971 samples/sec	 accuracy=37.175481	 loss=2.838557	 lr=0.057211
Epoch[107] Batch [1999]/[3759]	Speed: 116.929374 samples/sec	 accuracy=37.191406	 loss=2.838534	 lr=0.057198
Epoch[107] Batch [2049]/[3759]	Speed: 117.954534 samples/sec	 accuracy=37.217226	 loss=2.837665	 lr=0.057185
Epoch[107] Batch [2099]/[3759]	Speed: 117.409464 samples/sec	 accuracy=37.212798	 loss=2.837641	 lr=0.057172
Epoch[107] Batch [2149]/[3759]	Speed: 117.978387 samples/sec	 accuracy=37.207849	 loss=2.837339	 lr=0.057160
Epoch[107] Batch [2199]/[3759]	Speed: 117.578931 samples/sec	 accuracy=37.218040	 loss=2.838098	 lr=0.057147
Epoch[107] Batch [2249]/[3759]	Speed: 117.139847 samples/sec	 accuracy=37.203472	 loss=2.838114	 lr=0.057134
Epoch[107] Batch [2299]/[3759]	Speed: 117.369634 samples/sec	 accuracy=37.225543	 loss=2.837683	 lr=0.057121
Epoch[107] Batch [2349]/[3759]	Speed: 116.806466 samples/sec	 accuracy=37.210771	 loss=2.837820	 lr=0.057108
Epoch[107] Batch [2399]/[3759]	Speed: 117.990996 samples/sec	 accuracy=37.209635	 loss=2.837671	 lr=0.057096
Epoch[107] Batch [2449]/[3759]	Speed: 117.383264 samples/sec	 accuracy=37.203444	 loss=2.837370	 lr=0.057083
Epoch[107] Batch [2499]/[3759]	Speed: 117.879399 samples/sec	 accuracy=37.196250	 loss=2.837588	 lr=0.057070
Epoch[107] Batch [2549]/[3759]	Speed: 117.458614 samples/sec	 accuracy=37.189951	 loss=2.838739	 lr=0.057057
Epoch[107] Batch [2599]/[3759]	Speed: 117.945088 samples/sec	 accuracy=37.178486	 loss=2.839956	 lr=0.057045
Epoch[107] Batch [2649]/[3759]	Speed: 117.288941 samples/sec	 accuracy=37.142099	 loss=2.841201	 lr=0.057032
Epoch[107] Batch [2699]/[3759]	Speed: 117.428476 samples/sec	 accuracy=37.157986	 loss=2.841532	 lr=0.057019
Epoch[107] Batch [2749]/[3759]	Speed: 118.111770 samples/sec	 accuracy=37.154545	 loss=2.841527	 lr=0.057006
Epoch[107] Batch [2799]/[3759]	Speed: 117.556533 samples/sec	 accuracy=37.148438	 loss=2.841784	 lr=0.056994
Epoch[107] Batch [2849]/[3759]	Speed: 117.660260 samples/sec	 accuracy=37.142544	 loss=2.842615	 lr=0.056981
Epoch[107] Batch [2899]/[3759]	Speed: 117.504448 samples/sec	 accuracy=37.160560	 loss=2.841528	 lr=0.056968
Epoch[107] Batch [2949]/[3759]	Speed: 117.474254 samples/sec	 accuracy=37.146186	 loss=2.842411	 lr=0.056955
Epoch[107] Batch [2999]/[3759]	Speed: 117.935550 samples/sec	 accuracy=37.152604	 loss=2.841315	 lr=0.056942
Epoch[107] Batch [3049]/[3759]	Speed: 118.045073 samples/sec	 accuracy=37.164959	 loss=2.840372	 lr=0.056930
Epoch[107] Batch [3099]/[3759]	Speed: 117.536324 samples/sec	 accuracy=37.148185	 loss=2.841356	 lr=0.056917
Epoch[107] Batch [3149]/[3759]	Speed: 117.455929 samples/sec	 accuracy=37.136409	 loss=2.841895	 lr=0.056904
Epoch[107] Batch [3199]/[3759]	Speed: 117.706938 samples/sec	 accuracy=37.163574	 loss=2.841743	 lr=0.056891
Epoch[107] Batch [3249]/[3759]	Speed: 117.450894 samples/sec	 accuracy=37.167788	 loss=2.841780	 lr=0.056879
Epoch[107] Batch [3299]/[3759]	Speed: 117.160204 samples/sec	 accuracy=37.149148	 loss=2.842784	 lr=0.056866
Epoch[107] Batch [3349]/[3759]	Speed: 117.491489 samples/sec	 accuracy=37.159049	 loss=2.842044	 lr=0.056853
Epoch[107] Batch [3399]/[3759]	Speed: 117.813117 samples/sec	 accuracy=37.154412	 loss=2.842685	 lr=0.056840
Epoch[107] Batch [3449]/[3759]	Speed: 117.211065 samples/sec	 accuracy=37.151721	 loss=2.843343	 lr=0.056828
Epoch[107] Batch [3499]/[3759]	Speed: 117.061449 samples/sec	 accuracy=37.162500	 loss=2.843533	 lr=0.056815
Epoch[107] Batch [3549]/[3759]	Speed: 116.811959 samples/sec	 accuracy=37.144806	 loss=2.844336	 lr=0.056802
Epoch[107] Batch [3599]/[3759]	Speed: 116.466894 samples/sec	 accuracy=37.154948	 loss=2.843853	 lr=0.056789
Epoch[107] Batch [3649]/[3759]	Speed: 117.865262 samples/sec	 accuracy=37.160531	 loss=2.843963	 lr=0.056776
Epoch[107] Batch [3699]/[3759]	Speed: 116.736430 samples/sec	 accuracy=37.135980	 loss=2.845033	 lr=0.056764
Epoch[107] Batch [3749]/[3759]	Speed: 120.872926 samples/sec	 accuracy=37.123750	 loss=2.845376	 lr=0.056751
Batch [0049]/[0303]: acc-top1=37.500000 acc-top5=63.625000
Batch [0099]/[0303]: acc-top1=34.812500 acc-top5=62.125000
Batch [0149]/[0303]: acc-top1=36.447917 acc-top5=62.906250
Batch [0199]/[0303]: acc-top1=39.070312 acc-top5=64.757812
Batch [0249]/[0303]: acc-top1=38.325000 acc-top5=64.293750
Batch [0299]/[0303]: acc-top1=37.520833 acc-top5=63.984375
[Epoch 107] training: accuracy=37.123819	 loss=2.845295
[Epoch 107] speed: 116 samples/sec	time cost: 2173.935966
[Epoch 107] validation: acc-top1=37.474216 acc-top5=63.943894 loss=2.899394
Epoch[108] Batch [0049]/[3760]	Speed: 87.519994 samples/sec	 accuracy=37.625000	 loss=2.826199	 lr=0.056736
Epoch[108] Batch [0099]/[3760]	Speed: 117.294110 samples/sec	 accuracy=36.781250	 loss=2.852336	 lr=0.056723
Epoch[108] Batch [0149]/[3760]	Speed: 117.077430 samples/sec	 accuracy=37.156250	 loss=2.821664	 lr=0.056710
Epoch[108] Batch [0199]/[3760]	Speed: 117.176023 samples/sec	 accuracy=36.898438	 loss=2.835120	 lr=0.056697
Epoch[108] Batch [0249]/[3760]	Speed: 117.575175 samples/sec	 accuracy=36.881250	 loss=2.840004	 lr=0.056685
Epoch[108] Batch [0299]/[3760]	Speed: 117.366238 samples/sec	 accuracy=37.125000	 loss=2.831847	 lr=0.056672
Epoch[108] Batch [0349]/[3760]	Speed: 117.686969 samples/sec	 accuracy=37.236607	 loss=2.831635	 lr=0.056659
Epoch[108] Batch [0399]/[3760]	Speed: 117.836255 samples/sec	 accuracy=37.324219	 loss=2.829395	 lr=0.056646
Epoch[108] Batch [0449]/[3760]	Speed: 118.262413 samples/sec	 accuracy=37.166667	 loss=2.836810	 lr=0.056634
Epoch[108] Batch [0499]/[3760]	Speed: 117.112020 samples/sec	 accuracy=37.321875	 loss=2.828034	 lr=0.056621
Epoch[108] Batch [0549]/[3760]	Speed: 117.304591 samples/sec	 accuracy=37.468750	 loss=2.825603	 lr=0.056608
Epoch[108] Batch [0599]/[3760]	Speed: 117.402820 samples/sec	 accuracy=37.359375	 loss=2.826646	 lr=0.056595
Epoch[108] Batch [0649]/[3760]	Speed: 117.128899 samples/sec	 accuracy=37.358173	 loss=2.826837	 lr=0.056582
Epoch[108] Batch [0699]/[3760]	Speed: 117.223286 samples/sec	 accuracy=37.247768	 loss=2.828962	 lr=0.056570
Epoch[108] Batch [0749]/[3760]	Speed: 116.927926 samples/sec	 accuracy=37.214583	 loss=2.832387	 lr=0.056557
Epoch[108] Batch [0799]/[3760]	Speed: 116.023183 samples/sec	 accuracy=37.195312	 loss=2.832639	 lr=0.056544
Epoch[108] Batch [0849]/[3760]	Speed: 117.161907 samples/sec	 accuracy=37.174632	 loss=2.833357	 lr=0.056531
Epoch[108] Batch [0899]/[3760]	Speed: 117.417858 samples/sec	 accuracy=37.142361	 loss=2.836608	 lr=0.056518
Epoch[108] Batch [0949]/[3760]	Speed: 117.131894 samples/sec	 accuracy=37.126645	 loss=2.836606	 lr=0.056506
Epoch[108] Batch [0999]/[3760]	Speed: 117.355543 samples/sec	 accuracy=37.184375	 loss=2.835399	 lr=0.056493
Epoch[108] Batch [1049]/[3760]	Speed: 117.393051 samples/sec	 accuracy=37.162202	 loss=2.834860	 lr=0.056480
Epoch[108] Batch [1099]/[3760]	Speed: 117.446123 samples/sec	 accuracy=37.129261	 loss=2.835466	 lr=0.056467
Epoch[108] Batch [1149]/[3760]	Speed: 118.346749 samples/sec	 accuracy=37.105978	 loss=2.835016	 lr=0.056455
Epoch[108] Batch [1199]/[3760]	Speed: 117.356013 samples/sec	 accuracy=37.049479	 loss=2.838819	 lr=0.056442
Epoch[108] Batch [1249]/[3760]	Speed: 117.568898 samples/sec	 accuracy=37.047500	 loss=2.836977	 lr=0.056429
Epoch[108] Batch [1299]/[3760]	Speed: 117.877280 samples/sec	 accuracy=37.042067	 loss=2.837822	 lr=0.056416
Epoch[108] Batch [1349]/[3760]	Speed: 117.730200 samples/sec	 accuracy=37.032407	 loss=2.838030	 lr=0.056403
Epoch[108] Batch [1399]/[3760]	Speed: 116.946541 samples/sec	 accuracy=37.069196	 loss=2.838098	 lr=0.056391
Epoch[108] Batch [1449]/[3760]	Speed: 117.699929 samples/sec	 accuracy=37.100216	 loss=2.837482	 lr=0.056378
Epoch[108] Batch [1499]/[3760]	Speed: 117.168260 samples/sec	 accuracy=37.085417	 loss=2.838715	 lr=0.056365
Epoch[108] Batch [1549]/[3760]	Speed: 117.397264 samples/sec	 accuracy=37.120968	 loss=2.836833	 lr=0.056352
Epoch[108] Batch [1599]/[3760]	Speed: 117.237993 samples/sec	 accuracy=37.119141	 loss=2.836765	 lr=0.056339
Epoch[108] Batch [1649]/[3760]	Speed: 117.353852 samples/sec	 accuracy=37.098485	 loss=2.836936	 lr=0.056327
Epoch[108] Batch [1699]/[3760]	Speed: 117.405124 samples/sec	 accuracy=37.096507	 loss=2.837147	 lr=0.056314
Epoch[108] Batch [1749]/[3760]	Speed: 116.171577 samples/sec	 accuracy=37.087500	 loss=2.837580	 lr=0.056301
Epoch[108] Batch [1799]/[3760]	Speed: 117.324177 samples/sec	 accuracy=37.041667	 loss=2.838669	 lr=0.056288
Epoch[108] Batch [1849]/[3760]	Speed: 117.468263 samples/sec	 accuracy=37.051520	 loss=2.837764	 lr=0.056275
Epoch[108] Batch [1899]/[3760]	Speed: 117.295746 samples/sec	 accuracy=37.070724	 loss=2.837299	 lr=0.056263
Epoch[108] Batch [1949]/[3760]	Speed: 117.186113 samples/sec	 accuracy=37.092949	 loss=2.836964	 lr=0.056250
Epoch[108] Batch [1999]/[3760]	Speed: 117.602338 samples/sec	 accuracy=37.106250	 loss=2.836867	 lr=0.056237
Epoch[108] Batch [2049]/[3760]	Speed: 117.472688 samples/sec	 accuracy=37.152439	 loss=2.836022	 lr=0.056224
Epoch[108] Batch [2099]/[3760]	Speed: 117.453831 samples/sec	 accuracy=37.156994	 loss=2.835262	 lr=0.056211
Epoch[108] Batch [2149]/[3760]	Speed: 117.942180 samples/sec	 accuracy=37.195494	 loss=2.834050	 lr=0.056199
Epoch[108] Batch [2199]/[3760]	Speed: 118.317347 samples/sec	 accuracy=37.164062	 loss=2.835388	 lr=0.056186
Epoch[108] Batch [2249]/[3760]	Speed: 118.015272 samples/sec	 accuracy=37.153472	 loss=2.836166	 lr=0.056173
Epoch[108] Batch [2299]/[3760]	Speed: 117.246267 samples/sec	 accuracy=37.213995	 loss=2.834595	 lr=0.056160
Epoch[108] Batch [2349]/[3760]	Speed: 116.962975 samples/sec	 accuracy=37.226064	 loss=2.834302	 lr=0.056147
Epoch[108] Batch [2399]/[3760]	Speed: 118.059671 samples/sec	 accuracy=37.249349	 loss=2.833313	 lr=0.056135
Epoch[108] Batch [2449]/[3760]	Speed: 117.447690 samples/sec	 accuracy=37.245536	 loss=2.833571	 lr=0.056122
Epoch[108] Batch [2499]/[3760]	Speed: 117.826962 samples/sec	 accuracy=37.238125	 loss=2.833671	 lr=0.056109
Epoch[108] Batch [2549]/[3760]	Speed: 117.859667 samples/sec	 accuracy=37.227328	 loss=2.833897	 lr=0.056096
Epoch[108] Batch [2599]/[3760]	Speed: 117.311842 samples/sec	 accuracy=37.206731	 loss=2.834984	 lr=0.056083
Epoch[108] Batch [2649]/[3760]	Speed: 117.306798 samples/sec	 accuracy=37.213443	 loss=2.834232	 lr=0.056071
Epoch[108] Batch [2699]/[3760]	Speed: 117.263167 samples/sec	 accuracy=37.215856	 loss=2.834329	 lr=0.056058
Epoch[108] Batch [2749]/[3760]	Speed: 116.105566 samples/sec	 accuracy=37.218750	 loss=2.834274	 lr=0.056045
Epoch[108] Batch [2799]/[3760]	Speed: 118.941715 samples/sec	 accuracy=37.186384	 loss=2.836311	 lr=0.056032
Epoch[108] Batch [2849]/[3760]	Speed: 117.042011 samples/sec	 accuracy=37.191338	 loss=2.837343	 lr=0.056019
Epoch[108] Batch [2899]/[3760]	Speed: 117.734984 samples/sec	 accuracy=37.197198	 loss=2.837417	 lr=0.056007
Epoch[108] Batch [2949]/[3760]	Speed: 117.895273 samples/sec	 accuracy=37.217161	 loss=2.836791	 lr=0.055994
Epoch[108] Batch [2999]/[3760]	Speed: 116.937269 samples/sec	 accuracy=37.240625	 loss=2.836925	 lr=0.055981
Epoch[108] Batch [3049]/[3760]	Speed: 117.718538 samples/sec	 accuracy=37.215676	 loss=2.837335	 lr=0.055968
Epoch[108] Batch [3099]/[3760]	Speed: 117.231662 samples/sec	 accuracy=37.251008	 loss=2.837105	 lr=0.055955
Epoch[108] Batch [3149]/[3760]	Speed: 116.496208 samples/sec	 accuracy=37.232639	 loss=2.837473	 lr=0.055943
Epoch[108] Batch [3199]/[3760]	Speed: 117.235241 samples/sec	 accuracy=37.208496	 loss=2.837861	 lr=0.055930
Epoch[108] Batch [3249]/[3760]	Speed: 117.136647 samples/sec	 accuracy=37.201923	 loss=2.838002	 lr=0.055917
Epoch[108] Batch [3299]/[3760]	Speed: 117.622809 samples/sec	 accuracy=37.212595	 loss=2.837615	 lr=0.055904
Epoch[108] Batch [3349]/[3760]	Speed: 117.524313 samples/sec	 accuracy=37.213619	 loss=2.836943	 lr=0.055891
Epoch[108] Batch [3399]/[3760]	Speed: 117.859058 samples/sec	 accuracy=37.232537	 loss=2.836795	 lr=0.055879
Epoch[108] Batch [3449]/[3760]	Speed: 116.752144 samples/sec	 accuracy=37.246377	 loss=2.835843	 lr=0.055866
Epoch[108] Batch [3499]/[3760]	Speed: 117.635244 samples/sec	 accuracy=37.241964	 loss=2.835901	 lr=0.055853
Epoch[108] Batch [3549]/[3760]	Speed: 117.487292 samples/sec	 accuracy=37.237676	 loss=2.836216	 lr=0.055840
Epoch[108] Batch [3599]/[3760]	Speed: 117.420313 samples/sec	 accuracy=37.228299	 loss=2.836578	 lr=0.055827
Epoch[108] Batch [3649]/[3760]	Speed: 117.434999 samples/sec	 accuracy=37.203767	 loss=2.837384	 lr=0.055815
Epoch[108] Batch [3699]/[3760]	Speed: 115.290973 samples/sec	 accuracy=37.219595	 loss=2.837157	 lr=0.055802
Epoch[108] Batch [3749]/[3760]	Speed: 120.477390 samples/sec	 accuracy=37.233333	 loss=2.836998	 lr=0.055789
Batch [0049]/[0303]: acc-top1=36.968750 acc-top5=63.562500
Batch [0099]/[0303]: acc-top1=34.781250 acc-top5=60.765625
Batch [0149]/[0303]: acc-top1=37.177083 acc-top5=62.895833
Batch [0199]/[0303]: acc-top1=39.289062 acc-top5=64.796875
Batch [0249]/[0303]: acc-top1=38.837500 acc-top5=64.787500
Batch [0299]/[0303]: acc-top1=38.406250 acc-top5=65.088542
[Epoch 108] training: accuracy=37.226147	 loss=2.837284
[Epoch 108] speed: 116 samples/sec	time cost: 2176.864247
[Epoch 108] validation: acc-top1=38.330239 acc-top5=65.068069 loss=2.861498
Epoch[109] Batch [0049]/[3760]	Speed: 87.393201 samples/sec	 accuracy=37.406250	 loss=2.769663	 lr=0.055774
Epoch[109] Batch [0099]/[3760]	Speed: 117.735763 samples/sec	 accuracy=37.640625	 loss=2.796550	 lr=0.055761
Epoch[109] Batch [0149]/[3760]	Speed: 117.862899 samples/sec	 accuracy=37.802083	 loss=2.804738	 lr=0.055748
Epoch[109] Batch [0199]/[3760]	Speed: 117.408178 samples/sec	 accuracy=37.453125	 loss=2.819530	 lr=0.055735
Epoch[109] Batch [0249]/[3760]	Speed: 118.224132 samples/sec	 accuracy=37.437500	 loss=2.821316	 lr=0.055722
Epoch[109] Batch [0299]/[3760]	Speed: 118.016121 samples/sec	 accuracy=37.343750	 loss=2.824410	 lr=0.055709
Epoch[109] Batch [0349]/[3760]	Speed: 118.553159 samples/sec	 accuracy=37.281250	 loss=2.821625	 lr=0.055697
Epoch[109] Batch [0399]/[3760]	Speed: 118.602938 samples/sec	 accuracy=37.312500	 loss=2.816068	 lr=0.055684
Epoch[109] Batch [0449]/[3760]	Speed: 117.794077 samples/sec	 accuracy=37.368056	 loss=2.815884	 lr=0.055671
Epoch[109] Batch [0499]/[3760]	Speed: 118.497154 samples/sec	 accuracy=37.556250	 loss=2.812143	 lr=0.055658
Epoch[109] Batch [0549]/[3760]	Speed: 117.029684 samples/sec	 accuracy=37.531250	 loss=2.811957	 lr=0.055645
Epoch[109] Batch [0599]/[3760]	Speed: 117.938941 samples/sec	 accuracy=37.572917	 loss=2.812507	 lr=0.055633
Epoch[109] Batch [0649]/[3760]	Speed: 116.701846 samples/sec	 accuracy=37.579327	 loss=2.813931	 lr=0.055620
Epoch[109] Batch [0699]/[3760]	Speed: 118.101611 samples/sec	 accuracy=37.638393	 loss=2.809570	 lr=0.055607
Epoch[109] Batch [0749]/[3760]	Speed: 117.608151 samples/sec	 accuracy=37.693750	 loss=2.808175	 lr=0.055594
Epoch[109] Batch [0799]/[3760]	Speed: 117.812160 samples/sec	 accuracy=37.703125	 loss=2.808820	 lr=0.055581
Epoch[109] Batch [0849]/[3760]	Speed: 117.138777 samples/sec	 accuracy=37.757353	 loss=2.807808	 lr=0.055568
Epoch[109] Batch [0899]/[3760]	Speed: 117.077947 samples/sec	 accuracy=37.755208	 loss=2.805469	 lr=0.055556
Epoch[109] Batch [0949]/[3760]	Speed: 116.038959 samples/sec	 accuracy=37.722039	 loss=2.807646	 lr=0.055543
Epoch[109] Batch [0999]/[3760]	Speed: 118.087145 samples/sec	 accuracy=37.709375	 loss=2.809289	 lr=0.055530
Epoch[109] Batch [1049]/[3760]	Speed: 117.991668 samples/sec	 accuracy=37.697917	 loss=2.812324	 lr=0.055517
Epoch[109] Batch [1099]/[3760]	Speed: 117.125229 samples/sec	 accuracy=37.772727	 loss=2.810052	 lr=0.055504
Epoch[109] Batch [1149]/[3760]	Speed: 117.625230 samples/sec	 accuracy=37.793478	 loss=2.809821	 lr=0.055492
Epoch[109] Batch [1199]/[3760]	Speed: 117.193659 samples/sec	 accuracy=37.747396	 loss=2.811492	 lr=0.055479
Epoch[109] Batch [1249]/[3760]	Speed: 117.474263 samples/sec	 accuracy=37.777500	 loss=2.811303	 lr=0.055466
Epoch[109] Batch [1299]/[3760]	Speed: 116.924904 samples/sec	 accuracy=37.786058	 loss=2.811394	 lr=0.055453
Epoch[109] Batch [1349]/[3760]	Speed: 117.434751 samples/sec	 accuracy=37.844907	 loss=2.808701	 lr=0.055440
Epoch[109] Batch [1399]/[3760]	Speed: 117.713838 samples/sec	 accuracy=37.805804	 loss=2.810182	 lr=0.055427
Epoch[109] Batch [1449]/[3760]	Speed: 116.842740 samples/sec	 accuracy=37.814655	 loss=2.810523	 lr=0.055415
Epoch[109] Batch [1499]/[3760]	Speed: 116.974452 samples/sec	 accuracy=37.793750	 loss=2.812542	 lr=0.055402
Epoch[109] Batch [1549]/[3760]	Speed: 117.372306 samples/sec	 accuracy=37.811492	 loss=2.811093	 lr=0.055389
Epoch[109] Batch [1599]/[3760]	Speed: 118.199768 samples/sec	 accuracy=37.816406	 loss=2.812206	 lr=0.055376
Epoch[109] Batch [1649]/[3760]	Speed: 117.514862 samples/sec	 accuracy=37.790720	 loss=2.813270	 lr=0.055363
Epoch[109] Batch [1699]/[3760]	Speed: 117.124163 samples/sec	 accuracy=37.761949	 loss=2.814115	 lr=0.055351
Epoch[109] Batch [1749]/[3760]	Speed: 117.360766 samples/sec	 accuracy=37.745536	 loss=2.814443	 lr=0.055338
Epoch[109] Batch [1799]/[3760]	Speed: 117.073911 samples/sec	 accuracy=37.756076	 loss=2.813876	 lr=0.055325
Epoch[109] Batch [1849]/[3760]	Speed: 116.884696 samples/sec	 accuracy=37.734797	 loss=2.813823	 lr=0.055312
Epoch[109] Batch [1899]/[3760]	Speed: 115.961443 samples/sec	 accuracy=37.720395	 loss=2.815954	 lr=0.055299
Epoch[109] Batch [1949]/[3760]	Speed: 117.890863 samples/sec	 accuracy=37.703526	 loss=2.815896	 lr=0.055286
Epoch[109] Batch [1999]/[3760]	Speed: 116.838051 samples/sec	 accuracy=37.716406	 loss=2.815593	 lr=0.055274
Epoch[109] Batch [2049]/[3760]	Speed: 117.092358 samples/sec	 accuracy=37.734756	 loss=2.814153	 lr=0.055261
Epoch[109] Batch [2099]/[3760]	Speed: 117.059213 samples/sec	 accuracy=37.708333	 loss=2.815147	 lr=0.055248
Epoch[109] Batch [2149]/[3760]	Speed: 117.214212 samples/sec	 accuracy=37.681686	 loss=2.816806	 lr=0.055235
Epoch[109] Batch [2199]/[3760]	Speed: 117.042612 samples/sec	 accuracy=37.674716	 loss=2.817185	 lr=0.055222
Epoch[109] Batch [2249]/[3760]	Speed: 116.701784 samples/sec	 accuracy=37.659722	 loss=2.816619	 lr=0.055209
Epoch[109] Batch [2299]/[3760]	Speed: 116.877653 samples/sec	 accuracy=37.646060	 loss=2.816668	 lr=0.055197
Epoch[109] Batch [2349]/[3760]	Speed: 118.211343 samples/sec	 accuracy=37.660239	 loss=2.816400	 lr=0.055184
Epoch[109] Batch [2399]/[3760]	Speed: 117.051382 samples/sec	 accuracy=37.671224	 loss=2.816341	 lr=0.055171
Epoch[109] Batch [2449]/[3760]	Speed: 117.311770 samples/sec	 accuracy=37.674107	 loss=2.815719	 lr=0.055158
Epoch[109] Batch [2499]/[3760]	Speed: 117.969980 samples/sec	 accuracy=37.701250	 loss=2.814652	 lr=0.055145
Epoch[109] Batch [2549]/[3760]	Speed: 116.529343 samples/sec	 accuracy=37.667892	 loss=2.816057	 lr=0.055133
Epoch[109] Batch [2599]/[3760]	Speed: 117.101362 samples/sec	 accuracy=37.650240	 loss=2.816972	 lr=0.055120
Epoch[109] Batch [2649]/[3760]	Speed: 116.600708 samples/sec	 accuracy=37.649764	 loss=2.818021	 lr=0.055107
Epoch[109] Batch [2699]/[3760]	Speed: 117.349102 samples/sec	 accuracy=37.663773	 loss=2.817733	 lr=0.055094
Epoch[109] Batch [2749]/[3760]	Speed: 117.380677 samples/sec	 accuracy=37.630114	 loss=2.818910	 lr=0.055081
Epoch[109] Batch [2799]/[3760]	Speed: 117.619795 samples/sec	 accuracy=37.594308	 loss=2.820797	 lr=0.055068
Epoch[109] Batch [2849]/[3760]	Speed: 117.959374 samples/sec	 accuracy=37.604715	 loss=2.820551	 lr=0.055056
Epoch[109] Batch [2899]/[3760]	Speed: 117.202774 samples/sec	 accuracy=37.607220	 loss=2.821268	 lr=0.055043
Epoch[109] Batch [2949]/[3760]	Speed: 117.619213 samples/sec	 accuracy=37.590042	 loss=2.821360	 lr=0.055030
Epoch[109] Batch [2999]/[3760]	Speed: 117.435203 samples/sec	 accuracy=37.576563	 loss=2.821694	 lr=0.055017
Epoch[109] Batch [3049]/[3760]	Speed: 117.403891 samples/sec	 accuracy=37.565061	 loss=2.822419	 lr=0.055004
Epoch[109] Batch [3099]/[3760]	Speed: 116.837153 samples/sec	 accuracy=37.569556	 loss=2.822060	 lr=0.054991
Epoch[109] Batch [3149]/[3760]	Speed: 117.833984 samples/sec	 accuracy=37.564484	 loss=2.822408	 lr=0.054979
Epoch[109] Batch [3199]/[3760]	Speed: 116.967948 samples/sec	 accuracy=37.564453	 loss=2.822272	 lr=0.054966
Epoch[109] Batch [3249]/[3760]	Speed: 117.902169 samples/sec	 accuracy=37.552885	 loss=2.823013	 lr=0.054953
Epoch[109] Batch [3299]/[3760]	Speed: 117.874465 samples/sec	 accuracy=37.567235	 loss=2.822107	 lr=0.054940
Epoch[109] Batch [3349]/[3760]	Speed: 117.863619 samples/sec	 accuracy=37.549907	 loss=2.822759	 lr=0.054927
Epoch[109] Batch [3399]/[3760]	Speed: 116.803464 samples/sec	 accuracy=37.567096	 loss=2.821984	 lr=0.054914
Epoch[109] Batch [3449]/[3760]	Speed: 116.680691 samples/sec	 accuracy=37.588315	 loss=2.821012	 lr=0.054902
Epoch[109] Batch [3499]/[3760]	Speed: 118.273971 samples/sec	 accuracy=37.597321	 loss=2.820619	 lr=0.054889
Epoch[109] Batch [3549]/[3760]	Speed: 117.216049 samples/sec	 accuracy=37.597271	 loss=2.820854	 lr=0.054876
Epoch[109] Batch [3599]/[3760]	Speed: 117.617097 samples/sec	 accuracy=37.579861	 loss=2.821794	 lr=0.054863
Epoch[109] Batch [3649]/[3760]	Speed: 117.318368 samples/sec	 accuracy=37.577055	 loss=2.822064	 lr=0.054850
Epoch[109] Batch [3699]/[3760]	Speed: 117.734550 samples/sec	 accuracy=37.560389	 loss=2.823505	 lr=0.054837
Epoch[109] Batch [3749]/[3760]	Speed: 120.361567 samples/sec	 accuracy=37.547917	 loss=2.823556	 lr=0.054825
Batch [0049]/[0303]: acc-top1=38.750000 acc-top5=63.906250
Batch [0099]/[0303]: acc-top1=35.500000 acc-top5=62.500000
Batch [0149]/[0303]: acc-top1=37.916667 acc-top5=64.052083
Batch [0199]/[0303]: acc-top1=40.328125 acc-top5=65.750000
Batch [0249]/[0303]: acc-top1=39.575000 acc-top5=65.275000
Batch [0299]/[0303]: acc-top1=38.859375 acc-top5=65.119792
[Epoch 109] training: accuracy=37.545711	 loss=2.823786
[Epoch 109] speed: 116 samples/sec	time cost: 2176.731307
[Epoch 109] validation: acc-top1=38.825289 acc-top5=65.114480 loss=2.872728
Epoch[110] Batch [0049]/[3759]	Speed: 88.269063 samples/sec	 accuracy=38.562500	 loss=2.751133	 lr=0.054809
Epoch[110] Batch [0099]/[3759]	Speed: 116.706938 samples/sec	 accuracy=37.812500	 loss=2.784503	 lr=0.054796
Epoch[110] Batch [0149]/[3759]	Speed: 117.375662 samples/sec	 accuracy=37.989583	 loss=2.769841	 lr=0.054783
Epoch[110] Batch [0199]/[3759]	Speed: 117.589655 samples/sec	 accuracy=38.101562	 loss=2.763290	 lr=0.054771
Epoch[110] Batch [0249]/[3759]	Speed: 117.213717 samples/sec	 accuracy=38.075000	 loss=2.767199	 lr=0.054758
Epoch[110] Batch [0299]/[3759]	Speed: 118.128758 samples/sec	 accuracy=37.968750	 loss=2.775741	 lr=0.054745
Epoch[110] Batch [0349]/[3759]	Speed: 117.560303 samples/sec	 accuracy=38.174107	 loss=2.772960	 lr=0.054732
Epoch[110] Batch [0399]/[3759]	Speed: 118.494850 samples/sec	 accuracy=38.140625	 loss=2.776229	 lr=0.054719
Epoch[110] Batch [0449]/[3759]	Speed: 117.272118 samples/sec	 accuracy=38.059028	 loss=2.784365	 lr=0.054706
Epoch[110] Batch [0499]/[3759]	Speed: 117.484923 samples/sec	 accuracy=38.146875	 loss=2.784676	 lr=0.054694
Epoch[110] Batch [0549]/[3759]	Speed: 117.728266 samples/sec	 accuracy=38.161932	 loss=2.788678	 lr=0.054681
Epoch[110] Batch [0599]/[3759]	Speed: 118.072342 samples/sec	 accuracy=38.072917	 loss=2.792895	 lr=0.054668
Epoch[110] Batch [0649]/[3759]	Speed: 117.361407 samples/sec	 accuracy=38.057692	 loss=2.794749	 lr=0.054655
Epoch[110] Batch [0699]/[3759]	Speed: 117.092702 samples/sec	 accuracy=38.071429	 loss=2.795078	 lr=0.054642
Epoch[110] Batch [0749]/[3759]	Speed: 117.472909 samples/sec	 accuracy=38.116667	 loss=2.792408	 lr=0.054629
Epoch[110] Batch [0799]/[3759]	Speed: 118.226096 samples/sec	 accuracy=38.140625	 loss=2.788943	 lr=0.054617
Epoch[110] Batch [0849]/[3759]	Speed: 117.317029 samples/sec	 accuracy=38.071691	 loss=2.790760	 lr=0.054604
Epoch[110] Batch [0899]/[3759]	Speed: 117.650089 samples/sec	 accuracy=38.100694	 loss=2.792304	 lr=0.054591
Epoch[110] Batch [0949]/[3759]	Speed: 117.469288 samples/sec	 accuracy=38.074013	 loss=2.790816	 lr=0.054578
Epoch[110] Batch [0999]/[3759]	Speed: 116.592409 samples/sec	 accuracy=38.017187	 loss=2.794019	 lr=0.054565
Epoch[110] Batch [1049]/[3759]	Speed: 117.163199 samples/sec	 accuracy=38.001488	 loss=2.795052	 lr=0.054552
Epoch[110] Batch [1099]/[3759]	Speed: 116.925612 samples/sec	 accuracy=37.946023	 loss=2.797897	 lr=0.054539
Epoch[110] Batch [1149]/[3759]	Speed: 117.779829 samples/sec	 accuracy=37.927989	 loss=2.797391	 lr=0.054527
Epoch[110] Batch [1199]/[3759]	Speed: 117.418793 samples/sec	 accuracy=37.919271	 loss=2.798128	 lr=0.054514
Epoch[110] Batch [1249]/[3759]	Speed: 118.506931 samples/sec	 accuracy=37.907500	 loss=2.797977	 lr=0.054501
Epoch[110] Batch [1299]/[3759]	Speed: 116.599941 samples/sec	 accuracy=37.870192	 loss=2.799971	 lr=0.054488
Epoch[110] Batch [1349]/[3759]	Speed: 118.453596 samples/sec	 accuracy=37.896991	 loss=2.798915	 lr=0.054475
Epoch[110] Batch [1399]/[3759]	Speed: 117.965859 samples/sec	 accuracy=37.844866	 loss=2.800837	 lr=0.054462
Epoch[110] Batch [1449]/[3759]	Speed: 117.777670 samples/sec	 accuracy=37.877155	 loss=2.799783	 lr=0.054450
Epoch[110] Batch [1499]/[3759]	Speed: 117.855733 samples/sec	 accuracy=37.876042	 loss=2.800549	 lr=0.054437
Epoch[110] Batch [1549]/[3759]	Speed: 116.886715 samples/sec	 accuracy=37.870968	 loss=2.801335	 lr=0.054424
Epoch[110] Batch [1599]/[3759]	Speed: 118.137418 samples/sec	 accuracy=37.859375	 loss=2.803021	 lr=0.054411
Epoch[110] Batch [1649]/[3759]	Speed: 117.760954 samples/sec	 accuracy=37.839962	 loss=2.804546	 lr=0.054398
Epoch[110] Batch [1699]/[3759]	Speed: 117.658690 samples/sec	 accuracy=37.830882	 loss=2.805626	 lr=0.054385
Epoch[110] Batch [1749]/[3759]	Speed: 117.802739 samples/sec	 accuracy=37.853571	 loss=2.805097	 lr=0.054372
Epoch[110] Batch [1799]/[3759]	Speed: 118.090444 samples/sec	 accuracy=37.851562	 loss=2.804225	 lr=0.054360
Epoch[110] Batch [1849]/[3759]	Speed: 117.169364 samples/sec	 accuracy=37.822635	 loss=2.805349	 lr=0.054347
Epoch[110] Batch [1899]/[3759]	Speed: 117.241843 samples/sec	 accuracy=37.828947	 loss=2.805301	 lr=0.054334
Epoch[110] Batch [1949]/[3759]	Speed: 118.072294 samples/sec	 accuracy=37.823718	 loss=2.805673	 lr=0.054321
Epoch[110] Batch [1999]/[3759]	Speed: 116.755352 samples/sec	 accuracy=37.818750	 loss=2.805184	 lr=0.054308
Epoch[110] Batch [2049]/[3759]	Speed: 115.964660 samples/sec	 accuracy=37.827744	 loss=2.805346	 lr=0.054295
Epoch[110] Batch [2099]/[3759]	Speed: 117.322501 samples/sec	 accuracy=37.831101	 loss=2.805750	 lr=0.054282
Epoch[110] Batch [2149]/[3759]	Speed: 117.724500 samples/sec	 accuracy=37.793605	 loss=2.806782	 lr=0.054270
Epoch[110] Batch [2199]/[3759]	Speed: 117.138461 samples/sec	 accuracy=37.751420	 loss=2.808376	 lr=0.054257
Epoch[110] Batch [2249]/[3759]	Speed: 116.802153 samples/sec	 accuracy=37.713194	 loss=2.810161	 lr=0.054244
Epoch[110] Batch [2299]/[3759]	Speed: 117.156830 samples/sec	 accuracy=37.713315	 loss=2.810525	 lr=0.054231
Epoch[110] Batch [2349]/[3759]	Speed: 117.977792 samples/sec	 accuracy=37.724069	 loss=2.809564	 lr=0.054218
Epoch[110] Batch [2399]/[3759]	Speed: 116.795955 samples/sec	 accuracy=37.727865	 loss=2.809130	 lr=0.054205
Epoch[110] Batch [2449]/[3759]	Speed: 117.160608 samples/sec	 accuracy=37.746811	 loss=2.809266	 lr=0.054193
Epoch[110] Batch [2499]/[3759]	Speed: 117.652038 samples/sec	 accuracy=37.750000	 loss=2.808910	 lr=0.054180
Epoch[110] Batch [2549]/[3759]	Speed: 117.319246 samples/sec	 accuracy=37.735907	 loss=2.809546	 lr=0.054167
Epoch[110] Batch [2599]/[3759]	Speed: 116.964259 samples/sec	 accuracy=37.743389	 loss=2.809425	 lr=0.054154
Epoch[110] Batch [2649]/[3759]	Speed: 117.412742 samples/sec	 accuracy=37.765920	 loss=2.808964	 lr=0.054141
Epoch[110] Batch [2699]/[3759]	Speed: 117.413482 samples/sec	 accuracy=37.781250	 loss=2.808446	 lr=0.054128
Epoch[110] Batch [2749]/[3759]	Speed: 117.168502 samples/sec	 accuracy=37.759091	 loss=2.809496	 lr=0.054115
Epoch[110] Batch [2799]/[3759]	Speed: 116.909803 samples/sec	 accuracy=37.751674	 loss=2.810307	 lr=0.054103
Epoch[110] Batch [2849]/[3759]	Speed: 116.898994 samples/sec	 accuracy=37.756579	 loss=2.810068	 lr=0.054090
Epoch[110] Batch [2899]/[3759]	Speed: 117.296500 samples/sec	 accuracy=37.744612	 loss=2.810369	 lr=0.054077
Epoch[110] Batch [2949]/[3759]	Speed: 116.917151 samples/sec	 accuracy=37.753178	 loss=2.809296	 lr=0.054064
Epoch[110] Batch [2999]/[3759]	Speed: 117.830144 samples/sec	 accuracy=37.764063	 loss=2.809711	 lr=0.054051
Epoch[110] Batch [3049]/[3759]	Speed: 116.143620 samples/sec	 accuracy=37.758709	 loss=2.809602	 lr=0.054038
Epoch[110] Batch [3099]/[3759]	Speed: 117.413200 samples/sec	 accuracy=37.731351	 loss=2.809957	 lr=0.054025
Epoch[110] Batch [3149]/[3759]	Speed: 116.952977 samples/sec	 accuracy=37.717262	 loss=2.810833	 lr=0.054013
Epoch[110] Batch [3199]/[3759]	Speed: 117.753616 samples/sec	 accuracy=37.703125	 loss=2.811960	 lr=0.054000
Epoch[110] Batch [3249]/[3759]	Speed: 117.199922 samples/sec	 accuracy=37.696635	 loss=2.812052	 lr=0.053987
Epoch[110] Batch [3299]/[3759]	Speed: 117.268715 samples/sec	 accuracy=37.700758	 loss=2.812910	 lr=0.053974
Epoch[110] Batch [3349]/[3759]	Speed: 116.753091 samples/sec	 accuracy=37.681437	 loss=2.813383	 lr=0.053961
Epoch[110] Batch [3399]/[3759]	Speed: 117.838215 samples/sec	 accuracy=37.647059	 loss=2.814979	 lr=0.053948
Epoch[110] Batch [3449]/[3759]	Speed: 117.673592 samples/sec	 accuracy=37.660326	 loss=2.814795	 lr=0.053935
Epoch[110] Batch [3499]/[3759]	Speed: 118.292681 samples/sec	 accuracy=37.677232	 loss=2.813976	 lr=0.053923
Epoch[110] Batch [3549]/[3759]	Speed: 117.556090 samples/sec	 accuracy=37.665493	 loss=2.813819	 lr=0.053910
Epoch[110] Batch [3599]/[3759]	Speed: 117.824209 samples/sec	 accuracy=37.649306	 loss=2.814319	 lr=0.053897
Epoch[110] Batch [3649]/[3759]	Speed: 117.376713 samples/sec	 accuracy=37.654538	 loss=2.814091	 lr=0.053884
Epoch[110] Batch [3699]/[3759]	Speed: 117.312564 samples/sec	 accuracy=37.652872	 loss=2.813982	 lr=0.053871
Epoch[110] Batch [3749]/[3759]	Speed: 121.562226 samples/sec	 accuracy=37.657083	 loss=2.813497	 lr=0.053858
Batch [0049]/[0303]: acc-top1=38.812500 acc-top5=64.656250
Batch [0099]/[0303]: acc-top1=34.937500 acc-top5=60.546875
Batch [0149]/[0303]: acc-top1=36.156250 acc-top5=62.177083
Batch [0199]/[0303]: acc-top1=38.585938 acc-top5=64.289062
Batch [0249]/[0303]: acc-top1=37.856250 acc-top5=64.212500
Batch [0299]/[0303]: acc-top1=37.369792 acc-top5=64.130208
[Epoch 110] training: accuracy=37.655045	 loss=2.813830
[Epoch 110] speed: 116 samples/sec	time cost: 2175.239270
[Epoch 110] validation: acc-top1=37.422649 acc-top5=64.129538 loss=2.956629
Epoch[111] Batch [0049]/[3760]	Speed: 86.601266 samples/sec	 accuracy=37.156250	 loss=2.814685	 lr=0.053843
Epoch[111] Batch [0099]/[3760]	Speed: 117.317091 samples/sec	 accuracy=38.109375	 loss=2.781367	 lr=0.053830
Epoch[111] Batch [0149]/[3760]	Speed: 116.930237 samples/sec	 accuracy=37.895833	 loss=2.778390	 lr=0.053817
Epoch[111] Batch [0199]/[3760]	Speed: 116.705945 samples/sec	 accuracy=37.921875	 loss=2.773994	 lr=0.053805
Epoch[111] Batch [0249]/[3760]	Speed: 116.975800 samples/sec	 accuracy=37.950000	 loss=2.769859	 lr=0.053792
Epoch[111] Batch [0299]/[3760]	Speed: 117.114509 samples/sec	 accuracy=37.947917	 loss=2.771475	 lr=0.053779
Epoch[111] Batch [0349]/[3760]	Speed: 117.403399 samples/sec	 accuracy=38.040179	 loss=2.774190	 lr=0.053766
Epoch[111] Batch [0399]/[3760]	Speed: 117.384131 samples/sec	 accuracy=38.144531	 loss=2.769456	 lr=0.053753
Epoch[111] Batch [0449]/[3760]	Speed: 117.574797 samples/sec	 accuracy=38.128472	 loss=2.772766	 lr=0.053740
Epoch[111] Batch [0499]/[3760]	Speed: 116.995358 samples/sec	 accuracy=38.118750	 loss=2.771805	 lr=0.053727
Epoch[111] Batch [0549]/[3760]	Speed: 117.559007 samples/sec	 accuracy=38.167614	 loss=2.775737	 lr=0.053715
Epoch[111] Batch [0599]/[3760]	Speed: 116.627750 samples/sec	 accuracy=38.208333	 loss=2.777306	 lr=0.053702
Epoch[111] Batch [0649]/[3760]	Speed: 116.927502 samples/sec	 accuracy=38.105769	 loss=2.778976	 lr=0.053689
Epoch[111] Batch [0699]/[3760]	Speed: 117.847822 samples/sec	 accuracy=38.006696	 loss=2.782749	 lr=0.053676
Epoch[111] Batch [0749]/[3760]	Speed: 116.963406 samples/sec	 accuracy=37.885417	 loss=2.791382	 lr=0.053663
Epoch[111] Batch [0799]/[3760]	Speed: 116.741073 samples/sec	 accuracy=37.896484	 loss=2.792766	 lr=0.053650
Epoch[111] Batch [0849]/[3760]	Speed: 116.153476 samples/sec	 accuracy=37.908088	 loss=2.794868	 lr=0.053637
Epoch[111] Batch [0899]/[3760]	Speed: 117.541823 samples/sec	 accuracy=37.954861	 loss=2.794936	 lr=0.053624
Epoch[111] Batch [0949]/[3760]	Speed: 117.053824 samples/sec	 accuracy=38.023026	 loss=2.793573	 lr=0.053612
Epoch[111] Batch [0999]/[3760]	Speed: 116.920301 samples/sec	 accuracy=38.057812	 loss=2.793061	 lr=0.053599
Epoch[111] Batch [1049]/[3760]	Speed: 117.051629 samples/sec	 accuracy=38.047619	 loss=2.794445	 lr=0.053586
Epoch[111] Batch [1099]/[3760]	Speed: 117.761347 samples/sec	 accuracy=38.012784	 loss=2.797055	 lr=0.053573
Epoch[111] Batch [1149]/[3760]	Speed: 117.898546 samples/sec	 accuracy=38.036685	 loss=2.797603	 lr=0.053560
Epoch[111] Batch [1199]/[3760]	Speed: 115.117618 samples/sec	 accuracy=38.018229	 loss=2.797512	 lr=0.053547
Epoch[111] Batch [1249]/[3760]	Speed: 116.681914 samples/sec	 accuracy=38.001250	 loss=2.800876	 lr=0.053534
Epoch[111] Batch [1299]/[3760]	Speed: 116.502230 samples/sec	 accuracy=37.973558	 loss=2.800356	 lr=0.053522
Epoch[111] Batch [1349]/[3760]	Speed: 117.118844 samples/sec	 accuracy=37.958333	 loss=2.801127	 lr=0.053509
Epoch[111] Batch [1399]/[3760]	Speed: 116.993399 samples/sec	 accuracy=38.014509	 loss=2.799281	 lr=0.053496
Epoch[111] Batch [1449]/[3760]	Speed: 117.705273 samples/sec	 accuracy=37.985991	 loss=2.801235	 lr=0.053483
Epoch[111] Batch [1499]/[3760]	Speed: 117.501347 samples/sec	 accuracy=37.926042	 loss=2.802966	 lr=0.053470
Epoch[111] Batch [1549]/[3760]	Speed: 117.245981 samples/sec	 accuracy=37.904234	 loss=2.803052	 lr=0.053457
Epoch[111] Batch [1599]/[3760]	Speed: 117.332011 samples/sec	 accuracy=37.868164	 loss=2.802598	 lr=0.053444
Epoch[111] Batch [1649]/[3760]	Speed: 117.535472 samples/sec	 accuracy=37.851326	 loss=2.802553	 lr=0.053431
Epoch[111] Batch [1699]/[3760]	Speed: 117.443138 samples/sec	 accuracy=37.882353	 loss=2.800837	 lr=0.053419
Epoch[111] Batch [1749]/[3760]	Speed: 117.450919 samples/sec	 accuracy=37.876786	 loss=2.801854	 lr=0.053406
Epoch[111] Batch [1799]/[3760]	Speed: 117.261634 samples/sec	 accuracy=37.850694	 loss=2.802111	 lr=0.053393
Epoch[111] Batch [1849]/[3760]	Speed: 117.749192 samples/sec	 accuracy=37.831926	 loss=2.803286	 lr=0.053380
Epoch[111] Batch [1899]/[3760]	Speed: 117.752624 samples/sec	 accuracy=37.806743	 loss=2.804702	 lr=0.053367
Epoch[111] Batch [1949]/[3760]	Speed: 117.205719 samples/sec	 accuracy=37.765224	 loss=2.806296	 lr=0.053354
Epoch[111] Batch [1999]/[3760]	Speed: 117.507391 samples/sec	 accuracy=37.736719	 loss=2.806193	 lr=0.053341
Epoch[111] Batch [2049]/[3760]	Speed: 116.910090 samples/sec	 accuracy=37.740854	 loss=2.805423	 lr=0.053329
Epoch[111] Batch [2099]/[3760]	Speed: 116.736987 samples/sec	 accuracy=37.732143	 loss=2.805291	 lr=0.053316
Epoch[111] Batch [2149]/[3760]	Speed: 117.533109 samples/sec	 accuracy=37.710029	 loss=2.806371	 lr=0.053303
Epoch[111] Batch [2199]/[3760]	Speed: 115.657793 samples/sec	 accuracy=37.732955	 loss=2.806326	 lr=0.053290
Epoch[111] Batch [2249]/[3760]	Speed: 117.166413 samples/sec	 accuracy=37.742361	 loss=2.806027	 lr=0.053277
Epoch[111] Batch [2299]/[3760]	Speed: 116.397104 samples/sec	 accuracy=37.747962	 loss=2.805279	 lr=0.053264
Epoch[111] Batch [2349]/[3760]	Speed: 117.676009 samples/sec	 accuracy=37.739362	 loss=2.805106	 lr=0.053251
Epoch[111] Batch [2399]/[3760]	Speed: 117.164537 samples/sec	 accuracy=37.707031	 loss=2.806061	 lr=0.053238
Epoch[111] Batch [2449]/[3760]	Speed: 116.772223 samples/sec	 accuracy=37.745536	 loss=2.804960	 lr=0.053226
Epoch[111] Batch [2499]/[3760]	Speed: 116.935385 samples/sec	 accuracy=37.746250	 loss=2.805688	 lr=0.053213
Epoch[111] Batch [2549]/[3760]	Speed: 117.469021 samples/sec	 accuracy=37.738971	 loss=2.805988	 lr=0.053200
Epoch[111] Batch [2599]/[3760]	Speed: 117.349273 samples/sec	 accuracy=37.767428	 loss=2.805968	 lr=0.053187
Epoch[111] Batch [2649]/[3760]	Speed: 117.260279 samples/sec	 accuracy=37.764151	 loss=2.806065	 lr=0.053174
Epoch[111] Batch [2699]/[3760]	Speed: 117.612452 samples/sec	 accuracy=37.784722	 loss=2.805873	 lr=0.053161
Epoch[111] Batch [2749]/[3760]	Speed: 117.280891 samples/sec	 accuracy=37.762500	 loss=2.807340	 lr=0.053148
Epoch[111] Batch [2799]/[3760]	Speed: 116.814326 samples/sec	 accuracy=37.795201	 loss=2.806657	 lr=0.053135
Epoch[111] Batch [2849]/[3760]	Speed: 116.863324 samples/sec	 accuracy=37.768092	 loss=2.807148	 lr=0.053123
Epoch[111] Batch [2899]/[3760]	Speed: 116.904753 samples/sec	 accuracy=37.745690	 loss=2.807698	 lr=0.053110
Epoch[111] Batch [2949]/[3760]	Speed: 117.280945 samples/sec	 accuracy=37.728284	 loss=2.807886	 lr=0.053097
Epoch[111] Batch [2999]/[3760]	Speed: 116.679362 samples/sec	 accuracy=37.719792	 loss=2.809058	 lr=0.053084
Epoch[111] Batch [3049]/[3760]	Speed: 117.491558 samples/sec	 accuracy=37.728996	 loss=2.808468	 lr=0.053071
Epoch[111] Batch [3099]/[3760]	Speed: 117.343973 samples/sec	 accuracy=37.747984	 loss=2.807405	 lr=0.053058
Epoch[111] Batch [3149]/[3760]	Speed: 116.189067 samples/sec	 accuracy=37.735119	 loss=2.807557	 lr=0.053045
Epoch[111] Batch [3199]/[3760]	Speed: 118.002379 samples/sec	 accuracy=37.737793	 loss=2.807626	 lr=0.053033
Epoch[111] Batch [3249]/[3760]	Speed: 116.941222 samples/sec	 accuracy=37.737981	 loss=2.807694	 lr=0.053020
Epoch[111] Batch [3299]/[3760]	Speed: 117.080324 samples/sec	 accuracy=37.765152	 loss=2.806065	 lr=0.053007
Epoch[111] Batch [3349]/[3760]	Speed: 116.435320 samples/sec	 accuracy=37.770989	 loss=2.806034	 lr=0.052994
Epoch[111] Batch [3399]/[3760]	Speed: 117.889199 samples/sec	 accuracy=37.762868	 loss=2.806156	 lr=0.052981
Epoch[111] Batch [3449]/[3760]	Speed: 117.524031 samples/sec	 accuracy=37.752717	 loss=2.806951	 lr=0.052968
Epoch[111] Batch [3499]/[3760]	Speed: 116.949078 samples/sec	 accuracy=37.741964	 loss=2.807295	 lr=0.052955
Epoch[111] Batch [3549]/[3760]	Speed: 116.944379 samples/sec	 accuracy=37.742077	 loss=2.808090	 lr=0.052942
Epoch[111] Batch [3599]/[3760]	Speed: 117.694223 samples/sec	 accuracy=37.729601	 loss=2.808035	 lr=0.052930
Epoch[111] Batch [3649]/[3760]	Speed: 116.410396 samples/sec	 accuracy=37.733305	 loss=2.808132	 lr=0.052917
Epoch[111] Batch [3699]/[3760]	Speed: 116.473651 samples/sec	 accuracy=37.732264	 loss=2.807932	 lr=0.052904
Epoch[111] Batch [3749]/[3760]	Speed: 120.825954 samples/sec	 accuracy=37.750417	 loss=2.807449	 lr=0.052891
Batch [0049]/[0303]: acc-top1=39.031250 acc-top5=63.875000
Batch [0099]/[0303]: acc-top1=36.203125 acc-top5=62.765625
Batch [0149]/[0303]: acc-top1=37.906250 acc-top5=63.437500
Batch [0199]/[0303]: acc-top1=40.500000 acc-top5=65.679688
Batch [0249]/[0303]: acc-top1=39.918750 acc-top5=65.500000
Batch [0299]/[0303]: acc-top1=39.796875 acc-top5=65.838542
[Epoch 111] training: accuracy=37.754322	 loss=2.807133
[Epoch 111] speed: 116 samples/sec	time cost: 2180.796614
[Epoch 111] validation: acc-top1=39.887583 acc-top5=65.872525 loss=2.849196
Epoch[112] Batch [0049]/[3760]	Speed: 86.197457 samples/sec	 accuracy=38.156250	 loss=2.834509	 lr=0.052875
Epoch[112] Batch [0099]/[3760]	Speed: 117.721041 samples/sec	 accuracy=38.218750	 loss=2.817075	 lr=0.052863
Epoch[112] Batch [0149]/[3760]	Speed: 116.867909 samples/sec	 accuracy=38.104167	 loss=2.822328	 lr=0.052850
Epoch[112] Batch [0199]/[3760]	Speed: 117.454466 samples/sec	 accuracy=37.945312	 loss=2.811680	 lr=0.052837
Epoch[112] Batch [0249]/[3760]	Speed: 117.204126 samples/sec	 accuracy=37.825000	 loss=2.808287	 lr=0.052824
Epoch[112] Batch [0299]/[3760]	Speed: 117.638726 samples/sec	 accuracy=38.088542	 loss=2.798748	 lr=0.052811
Epoch[112] Batch [0349]/[3760]	Speed: 117.475834 samples/sec	 accuracy=38.071429	 loss=2.798867	 lr=0.052798
Epoch[112] Batch [0399]/[3760]	Speed: 115.939844 samples/sec	 accuracy=38.078125	 loss=2.804938	 lr=0.052785
Epoch[112] Batch [0449]/[3760]	Speed: 117.148346 samples/sec	 accuracy=37.930556	 loss=2.806191	 lr=0.052772
Epoch[112] Batch [0499]/[3760]	Speed: 117.349662 samples/sec	 accuracy=37.912500	 loss=2.808639	 lr=0.052760
Epoch[112] Batch [0549]/[3760]	Speed: 118.239237 samples/sec	 accuracy=37.980114	 loss=2.804064	 lr=0.052747
Epoch[112] Batch [0599]/[3760]	Speed: 117.010431 samples/sec	 accuracy=37.929688	 loss=2.799196	 lr=0.052734
Epoch[112] Batch [0649]/[3760]	Speed: 117.369293 samples/sec	 accuracy=37.824519	 loss=2.803594	 lr=0.052721
Epoch[112] Batch [0699]/[3760]	Speed: 117.800275 samples/sec	 accuracy=37.814732	 loss=2.803293	 lr=0.052708
Epoch[112] Batch [0749]/[3760]	Speed: 117.792155 samples/sec	 accuracy=37.797917	 loss=2.804599	 lr=0.052695
Epoch[112] Batch [0799]/[3760]	Speed: 117.395897 samples/sec	 accuracy=37.732422	 loss=2.806617	 lr=0.052682
Epoch[112] Batch [0849]/[3760]	Speed: 117.564288 samples/sec	 accuracy=37.834559	 loss=2.802198	 lr=0.052669
Epoch[112] Batch [0899]/[3760]	Speed: 117.956342 samples/sec	 accuracy=37.840278	 loss=2.801469	 lr=0.052657
Epoch[112] Batch [0949]/[3760]	Speed: 116.804679 samples/sec	 accuracy=37.860197	 loss=2.802794	 lr=0.052644
Epoch[112] Batch [0999]/[3760]	Speed: 117.268402 samples/sec	 accuracy=37.914062	 loss=2.801600	 lr=0.052631
Epoch[112] Batch [1049]/[3760]	Speed: 117.802059 samples/sec	 accuracy=37.931548	 loss=2.801430	 lr=0.052618
Epoch[112] Batch [1099]/[3760]	Speed: 117.588015 samples/sec	 accuracy=37.910511	 loss=2.804051	 lr=0.052605
Epoch[112] Batch [1149]/[3760]	Speed: 118.433533 samples/sec	 accuracy=37.923913	 loss=2.801368	 lr=0.052592
Epoch[112] Batch [1199]/[3760]	Speed: 117.228464 samples/sec	 accuracy=37.881510	 loss=2.801339	 lr=0.052579
Epoch[112] Batch [1249]/[3760]	Speed: 118.185406 samples/sec	 accuracy=37.921250	 loss=2.801299	 lr=0.052566
Epoch[112] Batch [1299]/[3760]	Speed: 118.327306 samples/sec	 accuracy=37.935096	 loss=2.798928	 lr=0.052553
Epoch[112] Batch [1349]/[3760]	Speed: 117.657199 samples/sec	 accuracy=37.943287	 loss=2.799321	 lr=0.052541
Epoch[112] Batch [1399]/[3760]	Speed: 116.705942 samples/sec	 accuracy=37.907366	 loss=2.799066	 lr=0.052528
Epoch[112] Batch [1449]/[3760]	Speed: 117.876817 samples/sec	 accuracy=37.898707	 loss=2.799655	 lr=0.052515
Epoch[112] Batch [1499]/[3760]	Speed: 117.259439 samples/sec	 accuracy=37.913542	 loss=2.798437	 lr=0.052502
Epoch[112] Batch [1549]/[3760]	Speed: 117.362296 samples/sec	 accuracy=37.881048	 loss=2.799890	 lr=0.052489
Epoch[112] Batch [1599]/[3760]	Speed: 117.527948 samples/sec	 accuracy=37.921875	 loss=2.798887	 lr=0.052476
Epoch[112] Batch [1649]/[3760]	Speed: 117.258683 samples/sec	 accuracy=37.979167	 loss=2.795804	 lr=0.052463
Epoch[112] Batch [1699]/[3760]	Speed: 117.932167 samples/sec	 accuracy=37.972426	 loss=2.796972	 lr=0.052450
Epoch[112] Batch [1749]/[3760]	Speed: 117.736328 samples/sec	 accuracy=37.969643	 loss=2.797203	 lr=0.052438
Epoch[112] Batch [1799]/[3760]	Speed: 117.555212 samples/sec	 accuracy=37.980903	 loss=2.797274	 lr=0.052425
Epoch[112] Batch [1849]/[3760]	Speed: 117.779307 samples/sec	 accuracy=37.974662	 loss=2.796354	 lr=0.052412
Epoch[112] Batch [1899]/[3760]	Speed: 117.194201 samples/sec	 accuracy=37.973684	 loss=2.796561	 lr=0.052399
Epoch[112] Batch [1949]/[3760]	Speed: 117.594047 samples/sec	 accuracy=37.971955	 loss=2.797555	 lr=0.052386
Epoch[112] Batch [1999]/[3760]	Speed: 117.778873 samples/sec	 accuracy=37.987500	 loss=2.796777	 lr=0.052373
Epoch[112] Batch [2049]/[3760]	Speed: 117.817467 samples/sec	 accuracy=38.006098	 loss=2.796568	 lr=0.052360
Epoch[112] Batch [2099]/[3760]	Speed: 117.416527 samples/sec	 accuracy=38.001488	 loss=2.796265	 lr=0.052347
Epoch[112] Batch [2149]/[3760]	Speed: 117.287422 samples/sec	 accuracy=38.009448	 loss=2.796761	 lr=0.052334
Epoch[112] Batch [2199]/[3760]	Speed: 118.195109 samples/sec	 accuracy=37.996449	 loss=2.796917	 lr=0.052322
Epoch[112] Batch [2249]/[3760]	Speed: 118.426961 samples/sec	 accuracy=37.993056	 loss=2.797378	 lr=0.052309
Epoch[112] Batch [2299]/[3760]	Speed: 116.599428 samples/sec	 accuracy=37.999321	 loss=2.796858	 lr=0.052296
Epoch[112] Batch [2349]/[3760]	Speed: 117.892377 samples/sec	 accuracy=37.993351	 loss=2.797358	 lr=0.052283
Epoch[112] Batch [2399]/[3760]	Speed: 115.907883 samples/sec	 accuracy=37.998698	 loss=2.796337	 lr=0.052270
Epoch[112] Batch [2449]/[3760]	Speed: 117.487555 samples/sec	 accuracy=37.965561	 loss=2.797345	 lr=0.052257
Epoch[112] Batch [2499]/[3760]	Speed: 117.260021 samples/sec	 accuracy=37.960000	 loss=2.797205	 lr=0.052244
Epoch[112] Batch [2549]/[3760]	Speed: 117.325631 samples/sec	 accuracy=37.943627	 loss=2.798680	 lr=0.052231
Epoch[112] Batch [2599]/[3760]	Speed: 116.707369 samples/sec	 accuracy=37.942308	 loss=2.798394	 lr=0.052219
Epoch[112] Batch [2649]/[3760]	Speed: 117.820908 samples/sec	 accuracy=37.963443	 loss=2.797801	 lr=0.052206
Epoch[112] Batch [2699]/[3760]	Speed: 117.237263 samples/sec	 accuracy=37.940394	 loss=2.799294	 lr=0.052193
Epoch[112] Batch [2749]/[3760]	Speed: 117.487372 samples/sec	 accuracy=37.963636	 loss=2.799421	 lr=0.052180
Epoch[112] Batch [2799]/[3760]	Speed: 117.717895 samples/sec	 accuracy=37.949219	 loss=2.800121	 lr=0.052167
Epoch[112] Batch [2849]/[3760]	Speed: 116.919275 samples/sec	 accuracy=37.925439	 loss=2.800557	 lr=0.052154
Epoch[112] Batch [2899]/[3760]	Speed: 117.547770 samples/sec	 accuracy=37.906789	 loss=2.801342	 lr=0.052141
Epoch[112] Batch [2949]/[3760]	Speed: 117.347259 samples/sec	 accuracy=37.882415	 loss=2.801821	 lr=0.052128
Epoch[112] Batch [2999]/[3760]	Speed: 117.614355 samples/sec	 accuracy=37.879167	 loss=2.801830	 lr=0.052115
Epoch[112] Batch [3049]/[3760]	Speed: 117.410873 samples/sec	 accuracy=37.894467	 loss=2.801353	 lr=0.052103
Epoch[112] Batch [3099]/[3760]	Speed: 117.500098 samples/sec	 accuracy=37.903226	 loss=2.800429	 lr=0.052090
Epoch[112] Batch [3149]/[3760]	Speed: 117.266238 samples/sec	 accuracy=37.904762	 loss=2.800113	 lr=0.052077
Epoch[112] Batch [3199]/[3760]	Speed: 117.830940 samples/sec	 accuracy=37.910645	 loss=2.799661	 lr=0.052064
Epoch[112] Batch [3249]/[3760]	Speed: 117.382608 samples/sec	 accuracy=37.896635	 loss=2.800395	 lr=0.052051
Epoch[112] Batch [3299]/[3760]	Speed: 117.119417 samples/sec	 accuracy=37.902462	 loss=2.800058	 lr=0.052038
Epoch[112] Batch [3349]/[3760]	Speed: 117.895853 samples/sec	 accuracy=37.894123	 loss=2.800005	 lr=0.052025
Epoch[112] Batch [3399]/[3760]	Speed: 115.600255 samples/sec	 accuracy=37.903952	 loss=2.799320	 lr=0.052012
Epoch[112] Batch [3449]/[3760]	Speed: 117.353913 samples/sec	 accuracy=37.908967	 loss=2.798584	 lr=0.051999
Epoch[112] Batch [3499]/[3760]	Speed: 117.999952 samples/sec	 accuracy=37.931696	 loss=2.798297	 lr=0.051987
Epoch[112] Batch [3549]/[3760]	Speed: 117.279740 samples/sec	 accuracy=37.939261	 loss=2.798332	 lr=0.051974
Epoch[112] Batch [3599]/[3760]	Speed: 116.846527 samples/sec	 accuracy=37.947049	 loss=2.797963	 lr=0.051961
Epoch[112] Batch [3649]/[3760]	Speed: 117.394667 samples/sec	 accuracy=37.952911	 loss=2.797343	 lr=0.051948
Epoch[112] Batch [3699]/[3760]	Speed: 117.147869 samples/sec	 accuracy=37.942145	 loss=2.797837	 lr=0.051935
Epoch[112] Batch [3749]/[3760]	Speed: 120.961697 samples/sec	 accuracy=37.950000	 loss=2.797639	 lr=0.051922
Batch [0049]/[0303]: acc-top1=39.750000 acc-top5=63.843750
Batch [0099]/[0303]: acc-top1=38.046875 acc-top5=63.218750
Batch [0149]/[0303]: acc-top1=39.583333 acc-top5=64.520833
Batch [0199]/[0303]: acc-top1=41.664062 acc-top5=66.132812
Batch [0249]/[0303]: acc-top1=40.500000 acc-top5=65.668750
Batch [0299]/[0303]: acc-top1=39.489583 acc-top5=65.625000
[Epoch 112] training: accuracy=37.952128	 loss=2.797612
[Epoch 112] speed: 116 samples/sec	time cost: 2176.256483
[Epoch 112] validation: acc-top1=39.428630 acc-top5=65.583746 loss=2.791038
Epoch[113] Batch [0049]/[3759]	Speed: 87.588333 samples/sec	 accuracy=38.750000	 loss=2.729380	 lr=0.051907
Epoch[113] Batch [0099]/[3759]	Speed: 116.325173 samples/sec	 accuracy=38.546875	 loss=2.750354	 lr=0.051894
Epoch[113] Batch [0149]/[3759]	Speed: 117.444930 samples/sec	 accuracy=38.770833	 loss=2.733512	 lr=0.051881
Epoch[113] Batch [0199]/[3759]	Speed: 116.399156 samples/sec	 accuracy=39.000000	 loss=2.726879	 lr=0.051868
Epoch[113] Batch [0249]/[3759]	Speed: 117.809442 samples/sec	 accuracy=38.800000	 loss=2.740854	 lr=0.051855
Epoch[113] Batch [0299]/[3759]	Speed: 117.220673 samples/sec	 accuracy=38.479167	 loss=2.754468	 lr=0.051842
Epoch[113] Batch [0349]/[3759]	Speed: 117.850911 samples/sec	 accuracy=38.406250	 loss=2.760423	 lr=0.051829
Epoch[113] Batch [0399]/[3759]	Speed: 117.581310 samples/sec	 accuracy=38.335938	 loss=2.761782	 lr=0.051816
Epoch[113] Batch [0449]/[3759]	Speed: 116.578089 samples/sec	 accuracy=38.340278	 loss=2.762972	 lr=0.051804
Epoch[113] Batch [0499]/[3759]	Speed: 116.436693 samples/sec	 accuracy=38.246875	 loss=2.768633	 lr=0.051791
Epoch[113] Batch [0549]/[3759]	Speed: 117.683996 samples/sec	 accuracy=38.198864	 loss=2.769817	 lr=0.051778
Epoch[113] Batch [0599]/[3759]	Speed: 118.139415 samples/sec	 accuracy=38.260417	 loss=2.767487	 lr=0.051765
Epoch[113] Batch [0649]/[3759]	Speed: 116.179133 samples/sec	 accuracy=38.242788	 loss=2.769491	 lr=0.051752
Epoch[113] Batch [0699]/[3759]	Speed: 117.309050 samples/sec	 accuracy=38.129464	 loss=2.772629	 lr=0.051739
Epoch[113] Batch [0749]/[3759]	Speed: 117.399911 samples/sec	 accuracy=38.110417	 loss=2.776212	 lr=0.051726
Epoch[113] Batch [0799]/[3759]	Speed: 117.002921 samples/sec	 accuracy=38.222656	 loss=2.770271	 lr=0.051713
Epoch[113] Batch [0849]/[3759]	Speed: 117.466366 samples/sec	 accuracy=38.244485	 loss=2.767855	 lr=0.051700
Epoch[113] Batch [0899]/[3759]	Speed: 117.798445 samples/sec	 accuracy=38.256944	 loss=2.766956	 lr=0.051688
Epoch[113] Batch [0949]/[3759]	Speed: 117.291438 samples/sec	 accuracy=38.322368	 loss=2.766819	 lr=0.051675
Epoch[113] Batch [0999]/[3759]	Speed: 117.647380 samples/sec	 accuracy=38.300000	 loss=2.770489	 lr=0.051662
Epoch[113] Batch [1049]/[3759]	Speed: 117.220335 samples/sec	 accuracy=38.285714	 loss=2.770500	 lr=0.051649
Epoch[113] Batch [1099]/[3759]	Speed: 117.686715 samples/sec	 accuracy=38.268466	 loss=2.771029	 lr=0.051636
Epoch[113] Batch [1149]/[3759]	Speed: 117.767373 samples/sec	 accuracy=38.248641	 loss=2.772689	 lr=0.051623
Epoch[113] Batch [1199]/[3759]	Speed: 117.814716 samples/sec	 accuracy=38.282552	 loss=2.772501	 lr=0.051610
Epoch[113] Batch [1249]/[3759]	Speed: 117.405299 samples/sec	 accuracy=38.295000	 loss=2.772527	 lr=0.051597
Epoch[113] Batch [1299]/[3759]	Speed: 117.972795 samples/sec	 accuracy=38.259615	 loss=2.775801	 lr=0.051584
Epoch[113] Batch [1349]/[3759]	Speed: 117.769796 samples/sec	 accuracy=38.208333	 loss=2.777240	 lr=0.051572
Epoch[113] Batch [1399]/[3759]	Speed: 117.991480 samples/sec	 accuracy=38.189732	 loss=2.777370	 lr=0.051559
Epoch[113] Batch [1449]/[3759]	Speed: 116.989195 samples/sec	 accuracy=38.185345	 loss=2.778419	 lr=0.051546
Epoch[113] Batch [1499]/[3759]	Speed: 117.558920 samples/sec	 accuracy=38.193750	 loss=2.779021	 lr=0.051533
Epoch[113] Batch [1549]/[3759]	Speed: 118.072203 samples/sec	 accuracy=38.196573	 loss=2.780056	 lr=0.051520
Epoch[113] Batch [1599]/[3759]	Speed: 117.035502 samples/sec	 accuracy=38.198242	 loss=2.780201	 lr=0.051507
Epoch[113] Batch [1649]/[3759]	Speed: 115.972274 samples/sec	 accuracy=38.207386	 loss=2.780927	 lr=0.051494
Epoch[113] Batch [1699]/[3759]	Speed: 117.904490 samples/sec	 accuracy=38.230699	 loss=2.780279	 lr=0.051481
Epoch[113] Batch [1749]/[3759]	Speed: 117.584395 samples/sec	 accuracy=38.235714	 loss=2.780170	 lr=0.051468
Epoch[113] Batch [1799]/[3759]	Speed: 117.968102 samples/sec	 accuracy=38.240451	 loss=2.780533	 lr=0.051456
Epoch[113] Batch [1849]/[3759]	Speed: 117.258406 samples/sec	 accuracy=38.228885	 loss=2.781061	 lr=0.051443
Epoch[113] Batch [1899]/[3759]	Speed: 117.652091 samples/sec	 accuracy=38.178454	 loss=2.782363	 lr=0.051430
Epoch[113] Batch [1949]/[3759]	Speed: 117.305842 samples/sec	 accuracy=38.129808	 loss=2.784673	 lr=0.051417
Epoch[113] Batch [1999]/[3759]	Speed: 117.716058 samples/sec	 accuracy=38.132812	 loss=2.784833	 lr=0.051404
Epoch[113] Batch [2049]/[3759]	Speed: 118.047289 samples/sec	 accuracy=38.145579	 loss=2.783581	 lr=0.051391
Epoch[113] Batch [2099]/[3759]	Speed: 116.908377 samples/sec	 accuracy=38.134673	 loss=2.784504	 lr=0.051378
Epoch[113] Batch [2149]/[3759]	Speed: 118.002201 samples/sec	 accuracy=38.119186	 loss=2.786025	 lr=0.051365
Epoch[113] Batch [2199]/[3759]	Speed: 117.109456 samples/sec	 accuracy=38.091619	 loss=2.787379	 lr=0.051352
Epoch[113] Batch [2249]/[3759]	Speed: 118.058909 samples/sec	 accuracy=38.085417	 loss=2.786954	 lr=0.051339
Epoch[113] Batch [2299]/[3759]	Speed: 117.664426 samples/sec	 accuracy=38.067935	 loss=2.787223	 lr=0.051327
Epoch[113] Batch [2349]/[3759]	Speed: 117.785549 samples/sec	 accuracy=38.063165	 loss=2.787014	 lr=0.051314
Epoch[113] Batch [2399]/[3759]	Speed: 117.785172 samples/sec	 accuracy=38.087891	 loss=2.786399	 lr=0.051301
Epoch[113] Batch [2449]/[3759]	Speed: 117.815014 samples/sec	 accuracy=38.078444	 loss=2.785990	 lr=0.051288
Epoch[113] Batch [2499]/[3759]	Speed: 117.220022 samples/sec	 accuracy=38.093125	 loss=2.784911	 lr=0.051275
Epoch[113] Batch [2549]/[3759]	Speed: 117.155979 samples/sec	 accuracy=38.107843	 loss=2.784260	 lr=0.051262
Epoch[113] Batch [2599]/[3759]	Speed: 117.558315 samples/sec	 accuracy=38.118990	 loss=2.783863	 lr=0.051249
Epoch[113] Batch [2649]/[3759]	Speed: 116.346263 samples/sec	 accuracy=38.122052	 loss=2.783875	 lr=0.051236
Epoch[113] Batch [2699]/[3759]	Speed: 116.918261 samples/sec	 accuracy=38.122685	 loss=2.784000	 lr=0.051223
Epoch[113] Batch [2749]/[3759]	Speed: 117.521721 samples/sec	 accuracy=38.114205	 loss=2.784594	 lr=0.051211
Epoch[113] Batch [2799]/[3759]	Speed: 117.188154 samples/sec	 accuracy=38.125000	 loss=2.784583	 lr=0.051198
Epoch[113] Batch [2849]/[3759]	Speed: 118.082397 samples/sec	 accuracy=38.172697	 loss=2.783198	 lr=0.051185
Epoch[113] Batch [2899]/[3759]	Speed: 116.691510 samples/sec	 accuracy=38.183190	 loss=2.783922	 lr=0.051172
Epoch[113] Batch [2949]/[3759]	Speed: 116.732296 samples/sec	 accuracy=38.169492	 loss=2.784086	 lr=0.051159
Epoch[113] Batch [2999]/[3759]	Speed: 117.105805 samples/sec	 accuracy=38.175521	 loss=2.783422	 lr=0.051146
Epoch[113] Batch [3049]/[3759]	Speed: 118.148269 samples/sec	 accuracy=38.170594	 loss=2.783833	 lr=0.051133
Epoch[113] Batch [3099]/[3759]	Speed: 118.145492 samples/sec	 accuracy=38.153226	 loss=2.785020	 lr=0.051120
Epoch[113] Batch [3149]/[3759]	Speed: 117.391042 samples/sec	 accuracy=38.168155	 loss=2.784222	 lr=0.051107
Epoch[113] Batch [3199]/[3759]	Speed: 117.035294 samples/sec	 accuracy=38.155273	 loss=2.784589	 lr=0.051095
Epoch[113] Batch [3249]/[3759]	Speed: 117.175951 samples/sec	 accuracy=38.165865	 loss=2.784376	 lr=0.051082
Epoch[113] Batch [3299]/[3759]	Speed: 117.875853 samples/sec	 accuracy=38.164299	 loss=2.784567	 lr=0.051069
Epoch[113] Batch [3349]/[3759]	Speed: 117.193217 samples/sec	 accuracy=38.161381	 loss=2.785026	 lr=0.051056
Epoch[113] Batch [3399]/[3759]	Speed: 118.292208 samples/sec	 accuracy=38.170496	 loss=2.784692	 lr=0.051043
Epoch[113] Batch [3449]/[3759]	Speed: 116.847186 samples/sec	 accuracy=38.185688	 loss=2.783694	 lr=0.051030
Epoch[113] Batch [3499]/[3759]	Speed: 117.317755 samples/sec	 accuracy=38.168304	 loss=2.784354	 lr=0.051017
Epoch[113] Batch [3549]/[3759]	Speed: 117.264058 samples/sec	 accuracy=38.147887	 loss=2.785294	 lr=0.051004
Epoch[113] Batch [3599]/[3759]	Speed: 117.912802 samples/sec	 accuracy=38.142361	 loss=2.785383	 lr=0.050991
Epoch[113] Batch [3649]/[3759]	Speed: 115.824476 samples/sec	 accuracy=38.150685	 loss=2.784819	 lr=0.050978
Epoch[113] Batch [3699]/[3759]	Speed: 116.934702 samples/sec	 accuracy=38.134291	 loss=2.785662	 lr=0.050966
Epoch[113] Batch [3749]/[3759]	Speed: 120.647368 samples/sec	 accuracy=38.128750	 loss=2.785727	 lr=0.050953
Batch [0049]/[0303]: acc-top1=38.937500 acc-top5=63.968750
Batch [0099]/[0303]: acc-top1=36.281250 acc-top5=61.640625
Batch [0149]/[0303]: acc-top1=36.906250 acc-top5=62.218750
Batch [0199]/[0303]: acc-top1=39.546875 acc-top5=64.617188
Batch [0249]/[0303]: acc-top1=38.925000 acc-top5=64.625000
Batch [0299]/[0303]: acc-top1=38.197917 acc-top5=64.473958
[Epoch 113] training: accuracy=38.130154	 loss=2.785433
[Epoch 113] speed: 116 samples/sec	time cost: 2175.522451
[Epoch 113] validation: acc-top1=38.165223 acc-top5=64.485355 loss=2.938168
Epoch[114] Batch [0049]/[3760]	Speed: 88.201165 samples/sec	 accuracy=38.875000	 loss=2.744959	 lr=0.050937
Epoch[114] Batch [0099]/[3760]	Speed: 117.038290 samples/sec	 accuracy=38.546875	 loss=2.769045	 lr=0.050925
Epoch[114] Batch [0149]/[3760]	Speed: 116.969206 samples/sec	 accuracy=39.083333	 loss=2.745250	 lr=0.050912
Epoch[114] Batch [0199]/[3760]	Speed: 117.419504 samples/sec	 accuracy=39.031250	 loss=2.736668	 lr=0.050899
Epoch[114] Batch [0249]/[3760]	Speed: 117.414435 samples/sec	 accuracy=39.175000	 loss=2.739189	 lr=0.050886
Epoch[114] Batch [0299]/[3760]	Speed: 117.447902 samples/sec	 accuracy=39.270833	 loss=2.734600	 lr=0.050873
Epoch[114] Batch [0349]/[3760]	Speed: 117.594707 samples/sec	 accuracy=39.200893	 loss=2.741344	 lr=0.050860
Epoch[114] Batch [0399]/[3760]	Speed: 117.598811 samples/sec	 accuracy=39.117188	 loss=2.741566	 lr=0.050847
Epoch[114] Batch [0449]/[3760]	Speed: 117.893016 samples/sec	 accuracy=38.899306	 loss=2.752685	 lr=0.050834
Epoch[114] Batch [0499]/[3760]	Speed: 116.846871 samples/sec	 accuracy=38.921875	 loss=2.752926	 lr=0.050821
Epoch[114] Batch [0549]/[3760]	Speed: 117.117462 samples/sec	 accuracy=38.852273	 loss=2.757113	 lr=0.050809
Epoch[114] Batch [0599]/[3760]	Speed: 117.620096 samples/sec	 accuracy=38.742188	 loss=2.761504	 lr=0.050796
Epoch[114] Batch [0649]/[3760]	Speed: 117.039775 samples/sec	 accuracy=38.822115	 loss=2.754986	 lr=0.050783
Epoch[114] Batch [0699]/[3760]	Speed: 117.708173 samples/sec	 accuracy=38.750000	 loss=2.756554	 lr=0.050770
Epoch[114] Batch [0749]/[3760]	Speed: 117.758859 samples/sec	 accuracy=38.797917	 loss=2.754520	 lr=0.050757
Epoch[114] Batch [0799]/[3760]	Speed: 117.416377 samples/sec	 accuracy=38.714844	 loss=2.759850	 lr=0.050744
Epoch[114] Batch [0849]/[3760]	Speed: 117.570129 samples/sec	 accuracy=38.812500	 loss=2.757174	 lr=0.050731
Epoch[114] Batch [0899]/[3760]	Speed: 115.698881 samples/sec	 accuracy=38.869792	 loss=2.756293	 lr=0.050718
Epoch[114] Batch [0949]/[3760]	Speed: 116.706079 samples/sec	 accuracy=38.858553	 loss=2.759376	 lr=0.050705
Epoch[114] Batch [0999]/[3760]	Speed: 117.632995 samples/sec	 accuracy=38.834375	 loss=2.761695	 lr=0.050692
Epoch[114] Batch [1049]/[3760]	Speed: 117.683006 samples/sec	 accuracy=38.812500	 loss=2.762326	 lr=0.050680
Epoch[114] Batch [1099]/[3760]	Speed: 117.949655 samples/sec	 accuracy=38.765625	 loss=2.763453	 lr=0.050667
Epoch[114] Batch [1149]/[3760]	Speed: 117.841230 samples/sec	 accuracy=38.750000	 loss=2.763965	 lr=0.050654
Epoch[114] Batch [1199]/[3760]	Speed: 117.589815 samples/sec	 accuracy=38.700521	 loss=2.766068	 lr=0.050641
Epoch[114] Batch [1249]/[3760]	Speed: 117.133045 samples/sec	 accuracy=38.702500	 loss=2.766690	 lr=0.050628
Epoch[114] Batch [1299]/[3760]	Speed: 118.059653 samples/sec	 accuracy=38.725962	 loss=2.765415	 lr=0.050615
Epoch[114] Batch [1349]/[3760]	Speed: 117.400726 samples/sec	 accuracy=38.738426	 loss=2.761607	 lr=0.050602
Epoch[114] Batch [1399]/[3760]	Speed: 117.987772 samples/sec	 accuracy=38.739955	 loss=2.761848	 lr=0.050589
Epoch[114] Batch [1449]/[3760]	Speed: 117.024421 samples/sec	 accuracy=38.741379	 loss=2.760608	 lr=0.050576
Epoch[114] Batch [1499]/[3760]	Speed: 117.426460 samples/sec	 accuracy=38.711458	 loss=2.761759	 lr=0.050563
Epoch[114] Batch [1549]/[3760]	Speed: 116.997308 samples/sec	 accuracy=38.720766	 loss=2.760777	 lr=0.050551
Epoch[114] Batch [1599]/[3760]	Speed: 117.237311 samples/sec	 accuracy=38.729492	 loss=2.759739	 lr=0.050538
Epoch[114] Batch [1649]/[3760]	Speed: 117.628513 samples/sec	 accuracy=38.740530	 loss=2.759283	 lr=0.050525
Epoch[114] Batch [1699]/[3760]	Speed: 117.542999 samples/sec	 accuracy=38.693934	 loss=2.759231	 lr=0.050512
Epoch[114] Batch [1749]/[3760]	Speed: 117.348902 samples/sec	 accuracy=38.664286	 loss=2.760835	 lr=0.050499
Epoch[114] Batch [1799]/[3760]	Speed: 117.978748 samples/sec	 accuracy=38.611111	 loss=2.762561	 lr=0.050486
Epoch[114] Batch [1849]/[3760]	Speed: 117.579579 samples/sec	 accuracy=38.607264	 loss=2.762799	 lr=0.050473
Epoch[114] Batch [1899]/[3760]	Speed: 117.243031 samples/sec	 accuracy=38.604441	 loss=2.763071	 lr=0.050460
Epoch[114] Batch [1949]/[3760]	Speed: 116.076031 samples/sec	 accuracy=38.665064	 loss=2.762482	 lr=0.050447
Epoch[114] Batch [1999]/[3760]	Speed: 117.688455 samples/sec	 accuracy=38.648438	 loss=2.762445	 lr=0.050435
Epoch[114] Batch [2049]/[3760]	Speed: 117.216241 samples/sec	 accuracy=38.646341	 loss=2.762295	 lr=0.050422
Epoch[114] Batch [2099]/[3760]	Speed: 117.323599 samples/sec	 accuracy=38.651786	 loss=2.761694	 lr=0.050409
Epoch[114] Batch [2149]/[3760]	Speed: 117.564040 samples/sec	 accuracy=38.683140	 loss=2.760312	 lr=0.050396
Epoch[114] Batch [2199]/[3760]	Speed: 117.800672 samples/sec	 accuracy=38.683949	 loss=2.760507	 lr=0.050383
Epoch[114] Batch [2249]/[3760]	Speed: 117.331988 samples/sec	 accuracy=38.700694	 loss=2.760439	 lr=0.050370
Epoch[114] Batch [2299]/[3760]	Speed: 117.569320 samples/sec	 accuracy=38.707880	 loss=2.760431	 lr=0.050357
Epoch[114] Batch [2349]/[3760]	Speed: 116.847495 samples/sec	 accuracy=38.695479	 loss=2.762284	 lr=0.050344
Epoch[114] Batch [2399]/[3760]	Speed: 118.084478 samples/sec	 accuracy=38.647135	 loss=2.764265	 lr=0.050331
Epoch[114] Batch [2449]/[3760]	Speed: 117.344899 samples/sec	 accuracy=38.604592	 loss=2.766597	 lr=0.050318
Epoch[114] Batch [2499]/[3760]	Speed: 117.102388 samples/sec	 accuracy=38.618750	 loss=2.767241	 lr=0.050306
Epoch[114] Batch [2549]/[3760]	Speed: 117.285980 samples/sec	 accuracy=38.606618	 loss=2.767205	 lr=0.050293
Epoch[114] Batch [2599]/[3760]	Speed: 117.500795 samples/sec	 accuracy=38.595553	 loss=2.767072	 lr=0.050280
Epoch[114] Batch [2649]/[3760]	Speed: 117.546835 samples/sec	 accuracy=38.603184	 loss=2.765712	 lr=0.050267
Epoch[114] Batch [2699]/[3760]	Speed: 117.413605 samples/sec	 accuracy=38.580440	 loss=2.767809	 lr=0.050254
Epoch[114] Batch [2749]/[3760]	Speed: 116.935793 samples/sec	 accuracy=38.575000	 loss=2.767852	 lr=0.050241
Epoch[114] Batch [2799]/[3760]	Speed: 117.249192 samples/sec	 accuracy=38.562500	 loss=2.767966	 lr=0.050228
Epoch[114] Batch [2849]/[3760]	Speed: 117.438366 samples/sec	 accuracy=38.560855	 loss=2.767789	 lr=0.050215
Epoch[114] Batch [2899]/[3760]	Speed: 117.940378 samples/sec	 accuracy=38.556034	 loss=2.768375	 lr=0.050202
Epoch[114] Batch [2949]/[3760]	Speed: 114.339797 samples/sec	 accuracy=38.544492	 loss=2.768695	 lr=0.050189
Epoch[114] Batch [2999]/[3760]	Speed: 118.133956 samples/sec	 accuracy=38.538021	 loss=2.768893	 lr=0.050177
Epoch[114] Batch [3049]/[3760]	Speed: 117.472416 samples/sec	 accuracy=38.505123	 loss=2.769607	 lr=0.050164
Epoch[114] Batch [3099]/[3760]	Speed: 117.780994 samples/sec	 accuracy=38.489919	 loss=2.770186	 lr=0.050151
Epoch[114] Batch [3149]/[3760]	Speed: 116.964984 samples/sec	 accuracy=38.489087	 loss=2.770803	 lr=0.050138
Epoch[114] Batch [3199]/[3760]	Speed: 117.689667 samples/sec	 accuracy=38.465332	 loss=2.771957	 lr=0.050125
Epoch[114] Batch [3249]/[3760]	Speed: 117.993710 samples/sec	 accuracy=38.471635	 loss=2.771639	 lr=0.050112
Epoch[114] Batch [3299]/[3760]	Speed: 118.622011 samples/sec	 accuracy=38.451705	 loss=2.772712	 lr=0.050099
Epoch[114] Batch [3349]/[3760]	Speed: 117.048803 samples/sec	 accuracy=38.465019	 loss=2.771791	 lr=0.050086
Epoch[114] Batch [3399]/[3760]	Speed: 118.141167 samples/sec	 accuracy=38.451287	 loss=2.772307	 lr=0.050073
Epoch[114] Batch [3449]/[3760]	Speed: 117.174244 samples/sec	 accuracy=38.436141	 loss=2.773294	 lr=0.050060
Epoch[114] Batch [3499]/[3760]	Speed: 117.998341 samples/sec	 accuracy=38.451786	 loss=2.772368	 lr=0.050048
Epoch[114] Batch [3549]/[3760]	Speed: 118.005749 samples/sec	 accuracy=38.463468	 loss=2.771490	 lr=0.050035
Epoch[114] Batch [3599]/[3760]	Speed: 116.507386 samples/sec	 accuracy=38.455295	 loss=2.771810	 lr=0.050022
Epoch[114] Batch [3649]/[3760]	Speed: 118.120665 samples/sec	 accuracy=38.464469	 loss=2.771778	 lr=0.050009
Epoch[114] Batch [3699]/[3760]	Speed: 117.840515 samples/sec	 accuracy=38.445101	 loss=2.772360	 lr=0.049996
Epoch[114] Batch [3749]/[3760]	Speed: 120.873743 samples/sec	 accuracy=38.450417	 loss=2.772351	 lr=0.049983
Batch [0049]/[0303]: acc-top1=38.687500 acc-top5=64.937500
Batch [0099]/[0303]: acc-top1=36.437500 acc-top5=63.078125
Batch [0149]/[0303]: acc-top1=38.229167 acc-top5=63.979167
Batch [0199]/[0303]: acc-top1=40.726562 acc-top5=66.093750
Batch [0249]/[0303]: acc-top1=40.231250 acc-top5=66.362500
Batch [0299]/[0303]: acc-top1=39.432292 acc-top5=66.442708
[Epoch 114] training: accuracy=38.456200	 loss=2.772358
[Epoch 114] speed: 116 samples/sec	time cost: 2176.904324
[Epoch 114] validation: acc-top1=39.449257 acc-top5=66.460396 loss=2.780111
Epoch[115] Batch [0049]/[3760]	Speed: 88.671200 samples/sec	 accuracy=39.875000	 loss=2.679437	 lr=0.049968
Epoch[115] Batch [0099]/[3760]	Speed: 118.031655 samples/sec	 accuracy=39.312500	 loss=2.706524	 lr=0.049955
Epoch[115] Batch [0149]/[3760]	Speed: 116.968468 samples/sec	 accuracy=39.531250	 loss=2.698164	 lr=0.049942
Epoch[115] Batch [0199]/[3760]	Speed: 115.237107 samples/sec	 accuracy=39.414062	 loss=2.703381	 lr=0.049929
Epoch[115] Batch [0249]/[3760]	Speed: 116.694000 samples/sec	 accuracy=39.475000	 loss=2.705370	 lr=0.049916
Epoch[115] Batch [0299]/[3760]	Speed: 117.874089 samples/sec	 accuracy=39.281250	 loss=2.709789	 lr=0.049903
Epoch[115] Batch [0349]/[3760]	Speed: 117.013665 samples/sec	 accuracy=39.196429	 loss=2.710842	 lr=0.049890
Epoch[115] Batch [0399]/[3760]	Speed: 117.385921 samples/sec	 accuracy=39.179688	 loss=2.719247	 lr=0.049877
Epoch[115] Batch [0449]/[3760]	Speed: 117.671530 samples/sec	 accuracy=39.006944	 loss=2.726382	 lr=0.049864
Epoch[115] Batch [0499]/[3760]	Speed: 117.239136 samples/sec	 accuracy=38.906250	 loss=2.729013	 lr=0.049852
Epoch[115] Batch [0549]/[3760]	Speed: 117.659132 samples/sec	 accuracy=38.923295	 loss=2.732745	 lr=0.049839
Epoch[115] Batch [0599]/[3760]	Speed: 117.906269 samples/sec	 accuracy=38.934896	 loss=2.733188	 lr=0.049826
Epoch[115] Batch [0649]/[3760]	Speed: 117.426146 samples/sec	 accuracy=38.786058	 loss=2.736750	 lr=0.049813
Epoch[115] Batch [0699]/[3760]	Speed: 117.619850 samples/sec	 accuracy=38.816964	 loss=2.735320	 lr=0.049800
Epoch[115] Batch [0749]/[3760]	Speed: 116.896767 samples/sec	 accuracy=38.787500	 loss=2.738818	 lr=0.049787
Epoch[115] Batch [0799]/[3760]	Speed: 117.170923 samples/sec	 accuracy=38.816406	 loss=2.737770	 lr=0.049774
Epoch[115] Batch [0849]/[3760]	Speed: 117.877751 samples/sec	 accuracy=38.818015	 loss=2.738754	 lr=0.049761
Epoch[115] Batch [0899]/[3760]	Speed: 117.797052 samples/sec	 accuracy=38.803819	 loss=2.740850	 lr=0.049748
Epoch[115] Batch [0949]/[3760]	Speed: 117.770558 samples/sec	 accuracy=38.728618	 loss=2.743832	 lr=0.049735
Epoch[115] Batch [0999]/[3760]	Speed: 118.096978 samples/sec	 accuracy=38.709375	 loss=2.745785	 lr=0.049723
Epoch[115] Batch [1049]/[3760]	Speed: 117.050644 samples/sec	 accuracy=38.666667	 loss=2.750546	 lr=0.049710
Epoch[115] Batch [1099]/[3760]	Speed: 117.631118 samples/sec	 accuracy=38.696023	 loss=2.748799	 lr=0.049697
Epoch[115] Batch [1149]/[3760]	Speed: 117.378175 samples/sec	 accuracy=38.724185	 loss=2.750212	 lr=0.049684
Epoch[115] Batch [1199]/[3760]	Speed: 115.293717 samples/sec	 accuracy=38.721354	 loss=2.752536	 lr=0.049671
Epoch[115] Batch [1249]/[3760]	Speed: 118.144699 samples/sec	 accuracy=38.711250	 loss=2.753455	 lr=0.049658
Epoch[115] Batch [1299]/[3760]	Speed: 117.414526 samples/sec	 accuracy=38.688702	 loss=2.755382	 lr=0.049645
Epoch[115] Batch [1349]/[3760]	Speed: 117.722164 samples/sec	 accuracy=38.652778	 loss=2.755789	 lr=0.049632
Epoch[115] Batch [1399]/[3760]	Speed: 116.951009 samples/sec	 accuracy=38.622768	 loss=2.756318	 lr=0.049619
Epoch[115] Batch [1449]/[3760]	Speed: 116.982133 samples/sec	 accuracy=38.649784	 loss=2.756514	 lr=0.049607
Epoch[115] Batch [1499]/[3760]	Speed: 117.355827 samples/sec	 accuracy=38.679167	 loss=2.756341	 lr=0.049594
Epoch[115] Batch [1549]/[3760]	Speed: 117.713447 samples/sec	 accuracy=38.646169	 loss=2.756878	 lr=0.049581
Epoch[115] Batch [1599]/[3760]	Speed: 117.632079 samples/sec	 accuracy=38.649414	 loss=2.756791	 lr=0.049568
Epoch[115] Batch [1649]/[3760]	Speed: 117.910898 samples/sec	 accuracy=38.657197	 loss=2.755875	 lr=0.049555
Epoch[115] Batch [1699]/[3760]	Speed: 116.995479 samples/sec	 accuracy=38.694853	 loss=2.754194	 lr=0.049542
Epoch[115] Batch [1749]/[3760]	Speed: 116.911133 samples/sec	 accuracy=38.699107	 loss=2.754908	 lr=0.049529
Epoch[115] Batch [1799]/[3760]	Speed: 116.973388 samples/sec	 accuracy=38.736111	 loss=2.754375	 lr=0.049516
Epoch[115] Batch [1849]/[3760]	Speed: 117.289729 samples/sec	 accuracy=38.718750	 loss=2.754951	 lr=0.049503
Epoch[115] Batch [1899]/[3760]	Speed: 117.644923 samples/sec	 accuracy=38.694079	 loss=2.756028	 lr=0.049490
Epoch[115] Batch [1949]/[3760]	Speed: 118.456355 samples/sec	 accuracy=38.686699	 loss=2.756328	 lr=0.049478
Epoch[115] Batch [1999]/[3760]	Speed: 117.614684 samples/sec	 accuracy=38.739844	 loss=2.754946	 lr=0.049465
Epoch[115] Batch [2049]/[3760]	Speed: 117.665938 samples/sec	 accuracy=38.750762	 loss=2.754702	 lr=0.049452
Epoch[115] Batch [2099]/[3760]	Speed: 118.277475 samples/sec	 accuracy=38.735119	 loss=2.754197	 lr=0.049439
Epoch[115] Batch [2149]/[3760]	Speed: 117.497511 samples/sec	 accuracy=38.690407	 loss=2.755746	 lr=0.049426
Epoch[115] Batch [2199]/[3760]	Speed: 117.641385 samples/sec	 accuracy=38.719460	 loss=2.756164	 lr=0.049413
Epoch[115] Batch [2249]/[3760]	Speed: 116.511892 samples/sec	 accuracy=38.727083	 loss=2.755537	 lr=0.049400
Epoch[115] Batch [2299]/[3760]	Speed: 117.656588 samples/sec	 accuracy=38.719429	 loss=2.756899	 lr=0.049387
Epoch[115] Batch [2349]/[3760]	Speed: 117.381200 samples/sec	 accuracy=38.715426	 loss=2.756972	 lr=0.049374
Epoch[115] Batch [2399]/[3760]	Speed: 117.415536 samples/sec	 accuracy=38.711589	 loss=2.756989	 lr=0.049361
Epoch[115] Batch [2449]/[3760]	Speed: 117.548444 samples/sec	 accuracy=38.686862	 loss=2.757803	 lr=0.049349
Epoch[115] Batch [2499]/[3760]	Speed: 117.483677 samples/sec	 accuracy=38.661875	 loss=2.758039	 lr=0.049336
Epoch[115] Batch [2549]/[3760]	Speed: 117.247120 samples/sec	 accuracy=38.689338	 loss=2.757328	 lr=0.049323
Epoch[115] Batch [2599]/[3760]	Speed: 117.419259 samples/sec	 accuracy=38.676683	 loss=2.758923	 lr=0.049310
Epoch[115] Batch [2649]/[3760]	Speed: 116.931941 samples/sec	 accuracy=38.679245	 loss=2.759239	 lr=0.049297
Epoch[115] Batch [2699]/[3760]	Speed: 117.163572 samples/sec	 accuracy=38.672454	 loss=2.759754	 lr=0.049284
Epoch[115] Batch [2749]/[3760]	Speed: 117.152964 samples/sec	 accuracy=38.687500	 loss=2.759860	 lr=0.049271
Epoch[115] Batch [2799]/[3760]	Speed: 117.097879 samples/sec	 accuracy=38.686942	 loss=2.760309	 lr=0.049258
Epoch[115] Batch [2849]/[3760]	Speed: 117.811180 samples/sec	 accuracy=38.722039	 loss=2.759274	 lr=0.049245
Epoch[115] Batch [2899]/[3760]	Speed: 117.377128 samples/sec	 accuracy=38.708513	 loss=2.759707	 lr=0.049233
Epoch[115] Batch [2949]/[3760]	Speed: 117.653025 samples/sec	 accuracy=38.716102	 loss=2.759397	 lr=0.049220
Epoch[115] Batch [2999]/[3760]	Speed: 117.106300 samples/sec	 accuracy=38.729687	 loss=2.759135	 lr=0.049207
Epoch[115] Batch [3049]/[3760]	Speed: 117.897785 samples/sec	 accuracy=38.704406	 loss=2.760066	 lr=0.049194
Epoch[115] Batch [3099]/[3760]	Speed: 118.349988 samples/sec	 accuracy=38.693548	 loss=2.759899	 lr=0.049181
Epoch[115] Batch [3149]/[3760]	Speed: 117.175009 samples/sec	 accuracy=38.698413	 loss=2.759476	 lr=0.049168
Epoch[115] Batch [3199]/[3760]	Speed: 117.296697 samples/sec	 accuracy=38.700195	 loss=2.760206	 lr=0.049155
Epoch[115] Batch [3249]/[3760]	Speed: 115.917604 samples/sec	 accuracy=38.686538	 loss=2.760330	 lr=0.049142
Epoch[115] Batch [3299]/[3760]	Speed: 116.974462 samples/sec	 accuracy=38.686553	 loss=2.760074	 lr=0.049129
Epoch[115] Batch [3349]/[3760]	Speed: 116.632892 samples/sec	 accuracy=38.681437	 loss=2.759819	 lr=0.049116
Epoch[115] Batch [3399]/[3760]	Speed: 117.266636 samples/sec	 accuracy=38.649357	 loss=2.760801	 lr=0.049104
Epoch[115] Batch [3449]/[3760]	Speed: 117.864886 samples/sec	 accuracy=38.655797	 loss=2.760953	 lr=0.049091
Epoch[115] Batch [3499]/[3760]	Speed: 118.118771 samples/sec	 accuracy=38.647768	 loss=2.760890	 lr=0.049078
Epoch[115] Batch [3549]/[3760]	Speed: 117.546745 samples/sec	 accuracy=38.630722	 loss=2.762101	 lr=0.049065
Epoch[115] Batch [3599]/[3760]	Speed: 117.080386 samples/sec	 accuracy=38.615451	 loss=2.762756	 lr=0.049052
Epoch[115] Batch [3649]/[3760]	Speed: 117.138707 samples/sec	 accuracy=38.625000	 loss=2.762326	 lr=0.049039
Epoch[115] Batch [3699]/[3760]	Speed: 117.440241 samples/sec	 accuracy=38.627534	 loss=2.762453	 lr=0.049026
Epoch[115] Batch [3749]/[3760]	Speed: 120.382644 samples/sec	 accuracy=38.619583	 loss=2.763213	 lr=0.049013
Batch [0049]/[0303]: acc-top1=40.250000 acc-top5=64.562500
Batch [0099]/[0303]: acc-top1=37.625000 acc-top5=64.140625
Batch [0149]/[0303]: acc-top1=39.072917 acc-top5=64.927083
Batch [0199]/[0303]: acc-top1=42.148438 acc-top5=67.187500
Batch [0249]/[0303]: acc-top1=41.681250 acc-top5=67.131250
Batch [0299]/[0303]: acc-top1=41.187500 acc-top5=67.270833
[Epoch 115] training: accuracy=38.617437	 loss=2.763462
[Epoch 115] speed: 116 samples/sec	time cost: 2176.572106
[Epoch 115] validation: acc-top1=41.032384 acc-top5=67.192657 loss=2.732594
Epoch[116] Batch [0049]/[3759]	Speed: 88.632308 samples/sec	 accuracy=38.875000	 loss=2.710248	 lr=0.048998
Epoch[116] Batch [0099]/[3759]	Speed: 116.932218 samples/sec	 accuracy=37.984375	 loss=2.754975	 lr=0.048985
Epoch[116] Batch [0149]/[3759]	Speed: 117.503858 samples/sec	 accuracy=38.291667	 loss=2.756592	 lr=0.048972
Epoch[116] Batch [0199]/[3759]	Speed: 116.993966 samples/sec	 accuracy=38.492188	 loss=2.747186	 lr=0.048959
Epoch[116] Batch [0249]/[3759]	Speed: 117.597962 samples/sec	 accuracy=38.437500	 loss=2.759679	 lr=0.048946
Epoch[116] Batch [0299]/[3759]	Speed: 116.911229 samples/sec	 accuracy=38.432292	 loss=2.762050	 lr=0.048933
Epoch[116] Batch [0349]/[3759]	Speed: 117.188340 samples/sec	 accuracy=38.285714	 loss=2.763445	 lr=0.048920
Epoch[116] Batch [0399]/[3759]	Speed: 117.687465 samples/sec	 accuracy=38.566406	 loss=2.756414	 lr=0.048908
Epoch[116] Batch [0449]/[3759]	Speed: 116.998235 samples/sec	 accuracy=38.687500	 loss=2.749863	 lr=0.048895
Epoch[116] Batch [0499]/[3759]	Speed: 116.301648 samples/sec	 accuracy=38.671875	 loss=2.751510	 lr=0.048882
Epoch[116] Batch [0549]/[3759]	Speed: 117.389098 samples/sec	 accuracy=38.565341	 loss=2.750799	 lr=0.048869
Epoch[116] Batch [0599]/[3759]	Speed: 117.169926 samples/sec	 accuracy=38.622396	 loss=2.750968	 lr=0.048856
Epoch[116] Batch [0649]/[3759]	Speed: 117.262946 samples/sec	 accuracy=38.615385	 loss=2.754940	 lr=0.048843
Epoch[116] Batch [0699]/[3759]	Speed: 117.593981 samples/sec	 accuracy=38.562500	 loss=2.754497	 lr=0.048830
Epoch[116] Batch [0749]/[3759]	Speed: 117.743239 samples/sec	 accuracy=38.572917	 loss=2.753693	 lr=0.048817
Epoch[116] Batch [0799]/[3759]	Speed: 117.084066 samples/sec	 accuracy=38.630859	 loss=2.748568	 lr=0.048804
Epoch[116] Batch [0849]/[3759]	Speed: 117.260892 samples/sec	 accuracy=38.593750	 loss=2.752360	 lr=0.048791
Epoch[116] Batch [0899]/[3759]	Speed: 117.880995 samples/sec	 accuracy=38.604167	 loss=2.753331	 lr=0.048779
Epoch[116] Batch [0949]/[3759]	Speed: 117.287449 samples/sec	 accuracy=38.671053	 loss=2.752031	 lr=0.048766
Epoch[116] Batch [0999]/[3759]	Speed: 117.517983 samples/sec	 accuracy=38.673437	 loss=2.752640	 lr=0.048753
Epoch[116] Batch [1049]/[3759]	Speed: 118.287102 samples/sec	 accuracy=38.709821	 loss=2.748531	 lr=0.048740
Epoch[116] Batch [1099]/[3759]	Speed: 116.919606 samples/sec	 accuracy=38.700284	 loss=2.748971	 lr=0.048727
Epoch[116] Batch [1149]/[3759]	Speed: 117.021284 samples/sec	 accuracy=38.669837	 loss=2.749666	 lr=0.048714
Epoch[116] Batch [1199]/[3759]	Speed: 116.874166 samples/sec	 accuracy=38.631510	 loss=2.753073	 lr=0.048701
Epoch[116] Batch [1249]/[3759]	Speed: 117.363109 samples/sec	 accuracy=38.627500	 loss=2.754130	 lr=0.048688
Epoch[116] Batch [1299]/[3759]	Speed: 116.989063 samples/sec	 accuracy=38.628606	 loss=2.753965	 lr=0.048675
Epoch[116] Batch [1349]/[3759]	Speed: 116.863531 samples/sec	 accuracy=38.625000	 loss=2.753808	 lr=0.048663
Epoch[116] Batch [1399]/[3759]	Speed: 117.823797 samples/sec	 accuracy=38.657366	 loss=2.753498	 lr=0.048650
Epoch[116] Batch [1449]/[3759]	Speed: 116.687356 samples/sec	 accuracy=38.689655	 loss=2.752091	 lr=0.048637
Epoch[116] Batch [1499]/[3759]	Speed: 117.344296 samples/sec	 accuracy=38.676042	 loss=2.751612	 lr=0.048624
Epoch[116] Batch [1549]/[3759]	Speed: 115.564621 samples/sec	 accuracy=38.677419	 loss=2.750369	 lr=0.048611
Epoch[116] Batch [1599]/[3759]	Speed: 117.842128 samples/sec	 accuracy=38.708008	 loss=2.748462	 lr=0.048598
Epoch[116] Batch [1649]/[3759]	Speed: 117.488470 samples/sec	 accuracy=38.705492	 loss=2.748656	 lr=0.048585
Epoch[116] Batch [1699]/[3759]	Speed: 118.034673 samples/sec	 accuracy=38.747243	 loss=2.747148	 lr=0.048572
Epoch[116] Batch [1749]/[3759]	Speed: 117.674465 samples/sec	 accuracy=38.785714	 loss=2.746943	 lr=0.048559
Epoch[116] Batch [1799]/[3759]	Speed: 117.673020 samples/sec	 accuracy=38.788194	 loss=2.748416	 lr=0.048547
Epoch[116] Batch [1849]/[3759]	Speed: 117.894722 samples/sec	 accuracy=38.769426	 loss=2.749884	 lr=0.048534
Epoch[116] Batch [1899]/[3759]	Speed: 117.456920 samples/sec	 accuracy=38.731086	 loss=2.751033	 lr=0.048521
Epoch[116] Batch [1949]/[3759]	Speed: 117.965428 samples/sec	 accuracy=38.766827	 loss=2.749167	 lr=0.048508
Epoch[116] Batch [1999]/[3759]	Speed: 116.779984 samples/sec	 accuracy=38.789844	 loss=2.747672	 lr=0.048495
Epoch[116] Batch [2049]/[3759]	Speed: 116.622902 samples/sec	 accuracy=38.803354	 loss=2.746887	 lr=0.048482
Epoch[116] Batch [2099]/[3759]	Speed: 117.377265 samples/sec	 accuracy=38.775298	 loss=2.748106	 lr=0.048469
Epoch[116] Batch [2149]/[3759]	Speed: 117.124100 samples/sec	 accuracy=38.779797	 loss=2.747490	 lr=0.048456
Epoch[116] Batch [2199]/[3759]	Speed: 117.193176 samples/sec	 accuracy=38.737216	 loss=2.748467	 lr=0.048443
Epoch[116] Batch [2249]/[3759]	Speed: 117.025895 samples/sec	 accuracy=38.720139	 loss=2.749599	 lr=0.048431
Epoch[116] Batch [2299]/[3759]	Speed: 117.651389 samples/sec	 accuracy=38.719429	 loss=2.749943	 lr=0.048418
Epoch[116] Batch [2349]/[3759]	Speed: 117.828542 samples/sec	 accuracy=38.680186	 loss=2.752801	 lr=0.048405
Epoch[116] Batch [2399]/[3759]	Speed: 117.718355 samples/sec	 accuracy=38.695312	 loss=2.751522	 lr=0.048392
Epoch[116] Batch [2449]/[3759]	Speed: 117.634836 samples/sec	 accuracy=38.700893	 loss=2.750876	 lr=0.048379
Epoch[116] Batch [2499]/[3759]	Speed: 117.907096 samples/sec	 accuracy=38.673750	 loss=2.751717	 lr=0.048366
Epoch[116] Batch [2549]/[3759]	Speed: 116.786001 samples/sec	 accuracy=38.661765	 loss=2.752824	 lr=0.048353
Epoch[116] Batch [2599]/[3759]	Speed: 116.005066 samples/sec	 accuracy=38.639423	 loss=2.752773	 lr=0.048340
Epoch[116] Batch [2649]/[3759]	Speed: 117.372579 samples/sec	 accuracy=38.639151	 loss=2.753689	 lr=0.048327
Epoch[116] Batch [2699]/[3759]	Speed: 117.534232 samples/sec	 accuracy=38.639468	 loss=2.754061	 lr=0.048314
Epoch[116] Batch [2749]/[3759]	Speed: 117.762801 samples/sec	 accuracy=38.621591	 loss=2.754710	 lr=0.048302
Epoch[116] Batch [2799]/[3759]	Speed: 117.407808 samples/sec	 accuracy=38.635603	 loss=2.753646	 lr=0.048289
Epoch[116] Batch [2849]/[3759]	Speed: 118.590280 samples/sec	 accuracy=38.630482	 loss=2.754686	 lr=0.048276
Epoch[116] Batch [2899]/[3759]	Speed: 117.637636 samples/sec	 accuracy=38.617457	 loss=2.755019	 lr=0.048263
Epoch[116] Batch [2949]/[3759]	Speed: 117.422961 samples/sec	 accuracy=38.607521	 loss=2.755065	 lr=0.048250
Epoch[116] Batch [2999]/[3759]	Speed: 117.054619 samples/sec	 accuracy=38.614063	 loss=2.755562	 lr=0.048237
Epoch[116] Batch [3049]/[3759]	Speed: 117.700977 samples/sec	 accuracy=38.605533	 loss=2.755620	 lr=0.048224
Epoch[116] Batch [3099]/[3759]	Speed: 117.317945 samples/sec	 accuracy=38.591230	 loss=2.755021	 lr=0.048211
Epoch[116] Batch [3149]/[3759]	Speed: 117.768185 samples/sec	 accuracy=38.581349	 loss=2.755045	 lr=0.048198
Epoch[116] Batch [3199]/[3759]	Speed: 116.875426 samples/sec	 accuracy=38.580078	 loss=2.755198	 lr=0.048186
Epoch[116] Batch [3249]/[3759]	Speed: 117.148660 samples/sec	 accuracy=38.598077	 loss=2.754866	 lr=0.048173
Epoch[116] Batch [3299]/[3759]	Speed: 117.056186 samples/sec	 accuracy=38.618845	 loss=2.754377	 lr=0.048160
Epoch[116] Batch [3349]/[3759]	Speed: 117.938697 samples/sec	 accuracy=38.629198	 loss=2.753446	 lr=0.048147
Epoch[116] Batch [3399]/[3759]	Speed: 118.018337 samples/sec	 accuracy=38.640625	 loss=2.752979	 lr=0.048134
Epoch[116] Batch [3449]/[3759]	Speed: 117.786926 samples/sec	 accuracy=38.635870	 loss=2.753605	 lr=0.048121
Epoch[116] Batch [3499]/[3759]	Speed: 117.383209 samples/sec	 accuracy=38.644643	 loss=2.753087	 lr=0.048108
Epoch[116] Batch [3549]/[3759]	Speed: 117.262537 samples/sec	 accuracy=38.651849	 loss=2.752660	 lr=0.048095
Epoch[116] Batch [3599]/[3759]	Speed: 117.552843 samples/sec	 accuracy=38.648438	 loss=2.753700	 lr=0.048082
Epoch[116] Batch [3649]/[3759]	Speed: 115.328644 samples/sec	 accuracy=38.640411	 loss=2.753957	 lr=0.048070
Epoch[116] Batch [3699]/[3759]	Speed: 117.224996 samples/sec	 accuracy=38.658361	 loss=2.753258	 lr=0.048057
Epoch[116] Batch [3749]/[3759]	Speed: 121.006324 samples/sec	 accuracy=38.648750	 loss=2.753853	 lr=0.048044
Batch [0049]/[0303]: acc-top1=39.218750 acc-top5=65.281250
Batch [0099]/[0303]: acc-top1=37.125000 acc-top5=63.921875
Batch [0149]/[0303]: acc-top1=39.197917 acc-top5=65.270833
Batch [0199]/[0303]: acc-top1=41.507812 acc-top5=67.179688
Batch [0249]/[0303]: acc-top1=40.781250 acc-top5=66.981250
Batch [0299]/[0303]: acc-top1=40.385417 acc-top5=66.822917
[Epoch 116] training: accuracy=38.658054	 loss=2.753563
[Epoch 116] speed: 116 samples/sec	time cost: 2176.490556
[Epoch 116] validation: acc-top1=40.367162 acc-top5=66.867781 loss=2.770941
Epoch[117] Batch [0049]/[3760]	Speed: 87.271584 samples/sec	 accuracy=38.656250	 loss=2.719391	 lr=0.048029
Epoch[117] Batch [0099]/[3760]	Speed: 117.480616 samples/sec	 accuracy=39.218750	 loss=2.710884	 lr=0.048016
Epoch[117] Batch [0149]/[3760]	Speed: 117.668765 samples/sec	 accuracy=39.322917	 loss=2.717725	 lr=0.048003
Epoch[117] Batch [0199]/[3760]	Speed: 116.537559 samples/sec	 accuracy=39.093750	 loss=2.729758	 lr=0.047990
Epoch[117] Batch [0249]/[3760]	Speed: 117.128141 samples/sec	 accuracy=39.568750	 loss=2.712242	 lr=0.047977
Epoch[117] Batch [0299]/[3760]	Speed: 117.362823 samples/sec	 accuracy=39.744792	 loss=2.706474	 lr=0.047964
Epoch[117] Batch [0349]/[3760]	Speed: 117.332793 samples/sec	 accuracy=39.714286	 loss=2.709382	 lr=0.047951
Epoch[117] Batch [0399]/[3760]	Speed: 117.720128 samples/sec	 accuracy=39.695312	 loss=2.712378	 lr=0.047938
Epoch[117] Batch [0449]/[3760]	Speed: 118.181706 samples/sec	 accuracy=39.607639	 loss=2.713127	 lr=0.047926
Epoch[117] Batch [0499]/[3760]	Speed: 116.335887 samples/sec	 accuracy=39.609375	 loss=2.712771	 lr=0.047913
Epoch[117] Batch [0549]/[3760]	Speed: 117.707678 samples/sec	 accuracy=39.588068	 loss=2.713462	 lr=0.047900
Epoch[117] Batch [0599]/[3760]	Speed: 118.000296 samples/sec	 accuracy=39.570312	 loss=2.712794	 lr=0.047887
Epoch[117] Batch [0649]/[3760]	Speed: 116.703982 samples/sec	 accuracy=39.545673	 loss=2.715248	 lr=0.047874
Epoch[117] Batch [0699]/[3760]	Speed: 117.245766 samples/sec	 accuracy=39.328125	 loss=2.722253	 lr=0.047861
Epoch[117] Batch [0749]/[3760]	Speed: 116.929982 samples/sec	 accuracy=39.327083	 loss=2.721753	 lr=0.047848
Epoch[117] Batch [0799]/[3760]	Speed: 118.390381 samples/sec	 accuracy=39.291016	 loss=2.723196	 lr=0.047835
Epoch[117] Batch [0849]/[3760]	Speed: 116.746057 samples/sec	 accuracy=39.259191	 loss=2.724674	 lr=0.047822
Epoch[117] Batch [0899]/[3760]	Speed: 114.748703 samples/sec	 accuracy=39.246528	 loss=2.723443	 lr=0.047810
Epoch[117] Batch [0949]/[3760]	Speed: 117.658455 samples/sec	 accuracy=39.231908	 loss=2.724323	 lr=0.047797
Epoch[117] Batch [0999]/[3760]	Speed: 117.277370 samples/sec	 accuracy=39.185938	 loss=2.726916	 lr=0.047784
Epoch[117] Batch [1049]/[3760]	Speed: 117.341762 samples/sec	 accuracy=39.193452	 loss=2.727107	 lr=0.047771
Epoch[117] Batch [1099]/[3760]	Speed: 117.414900 samples/sec	 accuracy=39.174716	 loss=2.726868	 lr=0.047758
Epoch[117] Batch [1149]/[3760]	Speed: 118.030360 samples/sec	 accuracy=39.161685	 loss=2.729393	 lr=0.047745
Epoch[117] Batch [1199]/[3760]	Speed: 117.575188 samples/sec	 accuracy=39.256510	 loss=2.723737	 lr=0.047732
Epoch[117] Batch [1249]/[3760]	Speed: 116.305040 samples/sec	 accuracy=39.328750	 loss=2.720771	 lr=0.047719
Epoch[117] Batch [1299]/[3760]	Speed: 117.471308 samples/sec	 accuracy=39.302885	 loss=2.722040	 lr=0.047706
Epoch[117] Batch [1349]/[3760]	Speed: 116.563830 samples/sec	 accuracy=39.314815	 loss=2.722390	 lr=0.047694
Epoch[117] Batch [1399]/[3760]	Speed: 116.904535 samples/sec	 accuracy=39.293527	 loss=2.723441	 lr=0.047681
Epoch[117] Batch [1449]/[3760]	Speed: 117.066491 samples/sec	 accuracy=39.279095	 loss=2.725352	 lr=0.047668
Epoch[117] Batch [1499]/[3760]	Speed: 118.007548 samples/sec	 accuracy=39.272917	 loss=2.725464	 lr=0.047655
Epoch[117] Batch [1549]/[3760]	Speed: 117.533191 samples/sec	 accuracy=39.230847	 loss=2.727999	 lr=0.047642
Epoch[117] Batch [1599]/[3760]	Speed: 117.690327 samples/sec	 accuracy=39.201172	 loss=2.731000	 lr=0.047629
Epoch[117] Batch [1649]/[3760]	Speed: 117.200379 samples/sec	 accuracy=39.141098	 loss=2.733369	 lr=0.047616
Epoch[117] Batch [1699]/[3760]	Speed: 117.255814 samples/sec	 accuracy=39.169118	 loss=2.733730	 lr=0.047603
Epoch[117] Batch [1749]/[3760]	Speed: 117.712186 samples/sec	 accuracy=39.150000	 loss=2.734075	 lr=0.047591
Epoch[117] Batch [1799]/[3760]	Speed: 117.471937 samples/sec	 accuracy=39.138021	 loss=2.734209	 lr=0.047578
Epoch[117] Batch [1849]/[3760]	Speed: 116.679255 samples/sec	 accuracy=39.173142	 loss=2.732976	 lr=0.047565
Epoch[117] Batch [1899]/[3760]	Speed: 116.705573 samples/sec	 accuracy=39.181743	 loss=2.733536	 lr=0.047552
Epoch[117] Batch [1949]/[3760]	Speed: 117.597176 samples/sec	 accuracy=39.162660	 loss=2.733964	 lr=0.047539
Epoch[117] Batch [1999]/[3760]	Speed: 117.133002 samples/sec	 accuracy=39.197656	 loss=2.732680	 lr=0.047526
Epoch[117] Batch [2049]/[3760]	Speed: 117.098985 samples/sec	 accuracy=39.169207	 loss=2.732409	 lr=0.047513
Epoch[117] Batch [2099]/[3760]	Speed: 118.087298 samples/sec	 accuracy=39.173363	 loss=2.731897	 lr=0.047500
Epoch[117] Batch [2149]/[3760]	Speed: 117.976217 samples/sec	 accuracy=39.146076	 loss=2.733480	 lr=0.047487
Epoch[117] Batch [2199]/[3760]	Speed: 117.792271 samples/sec	 accuracy=39.151989	 loss=2.733260	 lr=0.047475
Epoch[117] Batch [2249]/[3760]	Speed: 117.441740 samples/sec	 accuracy=39.164583	 loss=2.734269	 lr=0.047462
Epoch[117] Batch [2299]/[3760]	Speed: 117.414467 samples/sec	 accuracy=39.160326	 loss=2.734622	 lr=0.047449
Epoch[117] Batch [2349]/[3760]	Speed: 117.292648 samples/sec	 accuracy=39.134309	 loss=2.735971	 lr=0.047436
Epoch[117] Batch [2399]/[3760]	Speed: 117.893061 samples/sec	 accuracy=39.151042	 loss=2.736528	 lr=0.047423
Epoch[117] Batch [2449]/[3760]	Speed: 117.946443 samples/sec	 accuracy=39.135204	 loss=2.736179	 lr=0.047410
Epoch[117] Batch [2499]/[3760]	Speed: 117.473670 samples/sec	 accuracy=39.113125	 loss=2.737358	 lr=0.047397
Epoch[117] Batch [2549]/[3760]	Speed: 117.264092 samples/sec	 accuracy=39.122549	 loss=2.737531	 lr=0.047384
Epoch[117] Batch [2599]/[3760]	Speed: 117.495826 samples/sec	 accuracy=39.106971	 loss=2.737517	 lr=0.047372
Epoch[117] Batch [2649]/[3760]	Speed: 117.822515 samples/sec	 accuracy=39.107901	 loss=2.737272	 lr=0.047359
Epoch[117] Batch [2699]/[3760]	Speed: 118.093849 samples/sec	 accuracy=39.101852	 loss=2.738283	 lr=0.047346
Epoch[117] Batch [2749]/[3760]	Speed: 117.674450 samples/sec	 accuracy=39.111932	 loss=2.737718	 lr=0.047333
Epoch[117] Batch [2799]/[3760]	Speed: 118.004549 samples/sec	 accuracy=39.092634	 loss=2.738884	 lr=0.047320
Epoch[117] Batch [2849]/[3760]	Speed: 117.420763 samples/sec	 accuracy=39.082237	 loss=2.739025	 lr=0.047307
Epoch[117] Batch [2899]/[3760]	Speed: 118.098410 samples/sec	 accuracy=39.065194	 loss=2.739908	 lr=0.047294
Epoch[117] Batch [2949]/[3760]	Speed: 115.301842 samples/sec	 accuracy=39.094280	 loss=2.739492	 lr=0.047281
Epoch[117] Batch [2999]/[3760]	Speed: 117.905607 samples/sec	 accuracy=39.094271	 loss=2.739378	 lr=0.047269
Epoch[117] Batch [3049]/[3760]	Speed: 117.618401 samples/sec	 accuracy=39.096311	 loss=2.739274	 lr=0.047256
Epoch[117] Batch [3099]/[3760]	Speed: 117.254638 samples/sec	 accuracy=39.120968	 loss=2.738906	 lr=0.047243
Epoch[117] Batch [3149]/[3760]	Speed: 117.646210 samples/sec	 accuracy=39.131944	 loss=2.739259	 lr=0.047230
Epoch[117] Batch [3199]/[3760]	Speed: 118.491902 samples/sec	 accuracy=39.136230	 loss=2.739068	 lr=0.047217
Epoch[117] Batch [3249]/[3760]	Speed: 117.425872 samples/sec	 accuracy=39.137019	 loss=2.738354	 lr=0.047204
Epoch[117] Batch [3299]/[3760]	Speed: 117.142815 samples/sec	 accuracy=39.152936	 loss=2.737531	 lr=0.047191
Epoch[117] Batch [3349]/[3760]	Speed: 117.309090 samples/sec	 accuracy=39.155784	 loss=2.737434	 lr=0.047178
Epoch[117] Batch [3399]/[3760]	Speed: 117.588562 samples/sec	 accuracy=39.155331	 loss=2.737899	 lr=0.047166
Epoch[117] Batch [3449]/[3760]	Speed: 116.847536 samples/sec	 accuracy=39.182065	 loss=2.737145	 lr=0.047153
Epoch[117] Batch [3499]/[3760]	Speed: 118.150256 samples/sec	 accuracy=39.173661	 loss=2.737278	 lr=0.047140
Epoch[117] Batch [3549]/[3760]	Speed: 118.364566 samples/sec	 accuracy=39.175176	 loss=2.737276	 lr=0.047127
Epoch[117] Batch [3599]/[3760]	Speed: 116.671773 samples/sec	 accuracy=39.180122	 loss=2.737211	 lr=0.047114
Epoch[117] Batch [3649]/[3760]	Speed: 117.146167 samples/sec	 accuracy=39.208904	 loss=2.735950	 lr=0.047101
Epoch[117] Batch [3699]/[3760]	Speed: 117.121596 samples/sec	 accuracy=39.206926	 loss=2.736726	 lr=0.047088
Epoch[117] Batch [3749]/[3760]	Speed: 120.298372 samples/sec	 accuracy=39.181250	 loss=2.738042	 lr=0.047075
Batch [0049]/[0303]: acc-top1=39.687500 acc-top5=65.031250
Batch [0099]/[0303]: acc-top1=36.187500 acc-top5=62.531250
Batch [0149]/[0303]: acc-top1=38.270833 acc-top5=64.062500
Batch [0199]/[0303]: acc-top1=40.742188 acc-top5=65.781250
Batch [0249]/[0303]: acc-top1=40.512500 acc-top5=65.825000
Batch [0299]/[0303]: acc-top1=40.015625 acc-top5=66.020833
[Epoch 117] training: accuracy=39.183428	 loss=2.737890
[Epoch 117] speed: 116 samples/sec	time cost: 2176.641356
[Epoch 117] validation: acc-top1=40.057756 acc-top5=66.083952 loss=2.808520
Epoch[118] Batch [0049]/[3760]	Speed: 87.626868 samples/sec	 accuracy=39.218750	 loss=2.747363	 lr=0.047060
Epoch[118] Batch [0099]/[3760]	Speed: 117.465882 samples/sec	 accuracy=39.484375	 loss=2.733495	 lr=0.047047
Epoch[118] Batch [0149]/[3760]	Speed: 117.045191 samples/sec	 accuracy=39.885417	 loss=2.701875	 lr=0.047034
Epoch[118] Batch [0199]/[3760]	Speed: 114.866772 samples/sec	 accuracy=39.531250	 loss=2.711619	 lr=0.047021
Epoch[118] Batch [0249]/[3760]	Speed: 117.702029 samples/sec	 accuracy=39.356250	 loss=2.707798	 lr=0.047008
Epoch[118] Batch [0299]/[3760]	Speed: 117.052404 samples/sec	 accuracy=39.453125	 loss=2.700879	 lr=0.046996
Epoch[118] Batch [0349]/[3760]	Speed: 117.286631 samples/sec	 accuracy=39.526786	 loss=2.696785	 lr=0.046983
Epoch[118] Batch [0399]/[3760]	Speed: 116.367832 samples/sec	 accuracy=39.488281	 loss=2.701148	 lr=0.046970
Epoch[118] Batch [0449]/[3760]	Speed: 116.646186 samples/sec	 accuracy=39.472222	 loss=2.702345	 lr=0.046957
Epoch[118] Batch [0499]/[3760]	Speed: 117.353238 samples/sec	 accuracy=39.531250	 loss=2.702510	 lr=0.046944
Epoch[118] Batch [0549]/[3760]	Speed: 117.374306 samples/sec	 accuracy=39.448864	 loss=2.705638	 lr=0.046931
Epoch[118] Batch [0599]/[3760]	Speed: 117.133386 samples/sec	 accuracy=39.289062	 loss=2.712569	 lr=0.046918
Epoch[118] Batch [0649]/[3760]	Speed: 117.158241 samples/sec	 accuracy=39.125000	 loss=2.719235	 lr=0.046905
Epoch[118] Batch [0699]/[3760]	Speed: 117.258994 samples/sec	 accuracy=39.145089	 loss=2.721488	 lr=0.046893
Epoch[118] Batch [0749]/[3760]	Speed: 116.998273 samples/sec	 accuracy=39.152083	 loss=2.719294	 lr=0.046880
Epoch[118] Batch [0799]/[3760]	Speed: 117.952142 samples/sec	 accuracy=39.175781	 loss=2.719129	 lr=0.046867
Epoch[118] Batch [0849]/[3760]	Speed: 117.052637 samples/sec	 accuracy=39.238971	 loss=2.718345	 lr=0.046854
Epoch[118] Batch [0899]/[3760]	Speed: 117.296004 samples/sec	 accuracy=39.248264	 loss=2.718775	 lr=0.046841
Epoch[118] Batch [0949]/[3760]	Speed: 117.473255 samples/sec	 accuracy=39.340461	 loss=2.718487	 lr=0.046828
Epoch[118] Batch [0999]/[3760]	Speed: 117.754806 samples/sec	 accuracy=39.400000	 loss=2.717259	 lr=0.046815
Epoch[118] Batch [1049]/[3760]	Speed: 116.639898 samples/sec	 accuracy=39.398810	 loss=2.718665	 lr=0.046802
Epoch[118] Batch [1099]/[3760]	Speed: 116.896465 samples/sec	 accuracy=39.394886	 loss=2.718264	 lr=0.046790
Epoch[118] Batch [1149]/[3760]	Speed: 118.163750 samples/sec	 accuracy=39.313859	 loss=2.719937	 lr=0.046777
Epoch[118] Batch [1199]/[3760]	Speed: 115.863190 samples/sec	 accuracy=39.313802	 loss=2.720286	 lr=0.046764
Epoch[118] Batch [1249]/[3760]	Speed: 118.177132 samples/sec	 accuracy=39.362500	 loss=2.719382	 lr=0.046751
Epoch[118] Batch [1299]/[3760]	Speed: 117.148027 samples/sec	 accuracy=39.307692	 loss=2.721931	 lr=0.046738
Epoch[118] Batch [1349]/[3760]	Speed: 117.445699 samples/sec	 accuracy=39.273148	 loss=2.723150	 lr=0.046725
Epoch[118] Batch [1399]/[3760]	Speed: 117.420433 samples/sec	 accuracy=39.256696	 loss=2.722600	 lr=0.046712
Epoch[118] Batch [1449]/[3760]	Speed: 117.310182 samples/sec	 accuracy=39.245690	 loss=2.722837	 lr=0.046699
Epoch[118] Batch [1499]/[3760]	Speed: 117.980361 samples/sec	 accuracy=39.256250	 loss=2.722411	 lr=0.046687
Epoch[118] Batch [1549]/[3760]	Speed: 117.789304 samples/sec	 accuracy=39.228831	 loss=2.722975	 lr=0.046674
Epoch[118] Batch [1599]/[3760]	Speed: 116.375733 samples/sec	 accuracy=39.198242	 loss=2.725677	 lr=0.046661
Epoch[118] Batch [1649]/[3760]	Speed: 116.822778 samples/sec	 accuracy=39.156250	 loss=2.726154	 lr=0.046648
Epoch[118] Batch [1699]/[3760]	Speed: 117.818948 samples/sec	 accuracy=39.111213	 loss=2.726695	 lr=0.046635
Epoch[118] Batch [1749]/[3760]	Speed: 117.043058 samples/sec	 accuracy=39.112500	 loss=2.725815	 lr=0.046622
Epoch[118] Batch [1799]/[3760]	Speed: 118.218954 samples/sec	 accuracy=39.137153	 loss=2.726401	 lr=0.046609
Epoch[118] Batch [1849]/[3760]	Speed: 117.465257 samples/sec	 accuracy=39.179899	 loss=2.726242	 lr=0.046597
Epoch[118] Batch [1899]/[3760]	Speed: 117.101524 samples/sec	 accuracy=39.210526	 loss=2.724253	 lr=0.046584
Epoch[118] Batch [1949]/[3760]	Speed: 117.945746 samples/sec	 accuracy=39.192308	 loss=2.725217	 lr=0.046571
Epoch[118] Batch [1999]/[3760]	Speed: 117.927203 samples/sec	 accuracy=39.213281	 loss=2.724318	 lr=0.046558
Epoch[118] Batch [2049]/[3760]	Speed: 117.672674 samples/sec	 accuracy=39.224085	 loss=2.724410	 lr=0.046545
Epoch[118] Batch [2099]/[3760]	Speed: 117.289593 samples/sec	 accuracy=39.241071	 loss=2.722998	 lr=0.046532
Epoch[118] Batch [2149]/[3760]	Speed: 116.405987 samples/sec	 accuracy=39.262355	 loss=2.722395	 lr=0.046519
Epoch[118] Batch [2199]/[3760]	Speed: 117.316231 samples/sec	 accuracy=39.272727	 loss=2.721537	 lr=0.046506
Epoch[118] Batch [2249]/[3760]	Speed: 116.069623 samples/sec	 accuracy=39.277778	 loss=2.722891	 lr=0.046494
Epoch[118] Batch [2299]/[3760]	Speed: 118.016694 samples/sec	 accuracy=39.281929	 loss=2.722341	 lr=0.046481
Epoch[118] Batch [2349]/[3760]	Speed: 117.807244 samples/sec	 accuracy=39.287234	 loss=2.722013	 lr=0.046468
Epoch[118] Batch [2399]/[3760]	Speed: 117.862991 samples/sec	 accuracy=39.264974	 loss=2.723517	 lr=0.046455
Epoch[118] Batch [2449]/[3760]	Speed: 117.311844 samples/sec	 accuracy=39.223852	 loss=2.725644	 lr=0.046442
Epoch[118] Batch [2499]/[3760]	Speed: 117.715969 samples/sec	 accuracy=39.226250	 loss=2.725561	 lr=0.046429
Epoch[118] Batch [2549]/[3760]	Speed: 116.652087 samples/sec	 accuracy=39.227328	 loss=2.725573	 lr=0.046416
Epoch[118] Batch [2599]/[3760]	Speed: 117.643487 samples/sec	 accuracy=39.239784	 loss=2.725653	 lr=0.046404
Epoch[118] Batch [2649]/[3760]	Speed: 117.412067 samples/sec	 accuracy=39.254127	 loss=2.724590	 lr=0.046391
Epoch[118] Batch [2699]/[3760]	Speed: 116.898794 samples/sec	 accuracy=39.266782	 loss=2.725571	 lr=0.046378
Epoch[118] Batch [2749]/[3760]	Speed: 117.509690 samples/sec	 accuracy=39.257386	 loss=2.725173	 lr=0.046365
Epoch[118] Batch [2799]/[3760]	Speed: 116.821165 samples/sec	 accuracy=39.252790	 loss=2.724728	 lr=0.046352
Epoch[118] Batch [2849]/[3760]	Speed: 117.246171 samples/sec	 accuracy=39.234101	 loss=2.724852	 lr=0.046339
Epoch[118] Batch [2899]/[3760]	Speed: 117.784969 samples/sec	 accuracy=39.221983	 loss=2.725993	 lr=0.046326
Epoch[118] Batch [2949]/[3760]	Speed: 118.200402 samples/sec	 accuracy=39.203919	 loss=2.726696	 lr=0.046314
Epoch[118] Batch [2999]/[3760]	Speed: 117.915383 samples/sec	 accuracy=39.187500	 loss=2.727960	 lr=0.046301
Epoch[118] Batch [3049]/[3760]	Speed: 117.668896 samples/sec	 accuracy=39.188525	 loss=2.727834	 lr=0.046288
Epoch[118] Batch [3099]/[3760]	Speed: 117.663724 samples/sec	 accuracy=39.197077	 loss=2.727919	 lr=0.046275
Epoch[118] Batch [3149]/[3760]	Speed: 117.556225 samples/sec	 accuracy=39.178075	 loss=2.729325	 lr=0.046262
Epoch[118] Batch [3199]/[3760]	Speed: 117.937062 samples/sec	 accuracy=39.165527	 loss=2.730206	 lr=0.046249
Epoch[118] Batch [3249]/[3760]	Speed: 116.643694 samples/sec	 accuracy=39.169712	 loss=2.729890	 lr=0.046236
Epoch[118] Batch [3299]/[3760]	Speed: 116.689111 samples/sec	 accuracy=39.178977	 loss=2.729555	 lr=0.046223
Epoch[118] Batch [3349]/[3760]	Speed: 117.692605 samples/sec	 accuracy=39.153451	 loss=2.731120	 lr=0.046211
Epoch[118] Batch [3399]/[3760]	Speed: 117.304219 samples/sec	 accuracy=39.163603	 loss=2.729888	 lr=0.046198
Epoch[118] Batch [3449]/[3760]	Speed: 117.278726 samples/sec	 accuracy=39.173913	 loss=2.729095	 lr=0.046185
Epoch[118] Batch [3499]/[3760]	Speed: 117.909436 samples/sec	 accuracy=39.206696	 loss=2.727739	 lr=0.046172
Epoch[118] Batch [3549]/[3760]	Speed: 117.360345 samples/sec	 accuracy=39.196743	 loss=2.727692	 lr=0.046159
Epoch[118] Batch [3599]/[3760]	Speed: 117.416391 samples/sec	 accuracy=39.195312	 loss=2.726967	 lr=0.046146
Epoch[118] Batch [3649]/[3760]	Speed: 117.997438 samples/sec	 accuracy=39.200342	 loss=2.727594	 lr=0.046133
Epoch[118] Batch [3699]/[3760]	Speed: 116.578073 samples/sec	 accuracy=39.220861	 loss=2.726837	 lr=0.046121
Epoch[118] Batch [3749]/[3760]	Speed: 120.369486 samples/sec	 accuracy=39.197083	 loss=2.728117	 lr=0.046108
Batch [0049]/[0303]: acc-top1=41.187500 acc-top5=66.062500
Batch [0099]/[0303]: acc-top1=38.875000 acc-top5=65.187500
Batch [0149]/[0303]: acc-top1=39.791667 acc-top5=65.625000
Batch [0199]/[0303]: acc-top1=42.203125 acc-top5=67.695312
Batch [0249]/[0303]: acc-top1=41.056250 acc-top5=67.231250
Batch [0299]/[0303]: acc-top1=40.692708 acc-top5=67.255208
[Epoch 118] training: accuracy=39.197557	 loss=2.728036
[Epoch 118] speed: 116 samples/sec	time cost: 2177.455907
[Epoch 118] validation: acc-top1=40.650784 acc-top5=67.290635 loss=2.757282
Epoch[119] Batch [0049]/[3759]	Speed: 89.622572 samples/sec	 accuracy=38.937500	 loss=2.710467	 lr=0.046092
Epoch[119] Batch [0099]/[3759]	Speed: 117.246669 samples/sec	 accuracy=38.453125	 loss=2.773968	 lr=0.046079
Epoch[119] Batch [0149]/[3759]	Speed: 117.513223 samples/sec	 accuracy=39.187500	 loss=2.735164	 lr=0.046067
Epoch[119] Batch [0199]/[3759]	Speed: 117.072791 samples/sec	 accuracy=39.070312	 loss=2.738568	 lr=0.046054
Epoch[119] Batch [0249]/[3759]	Speed: 117.444500 samples/sec	 accuracy=39.112500	 loss=2.734404	 lr=0.046041
Epoch[119] Batch [0299]/[3759]	Speed: 117.905208 samples/sec	 accuracy=39.213542	 loss=2.718905	 lr=0.046028
Epoch[119] Batch [0349]/[3759]	Speed: 117.154937 samples/sec	 accuracy=39.151786	 loss=2.728139	 lr=0.046015
Epoch[119] Batch [0399]/[3759]	Speed: 116.904379 samples/sec	 accuracy=39.097656	 loss=2.732532	 lr=0.046002
Epoch[119] Batch [0449]/[3759]	Speed: 117.135517 samples/sec	 accuracy=39.381944	 loss=2.719633	 lr=0.045989
Epoch[119] Batch [0499]/[3759]	Speed: 115.874799 samples/sec	 accuracy=39.343750	 loss=2.718615	 lr=0.045977
Epoch[119] Batch [0549]/[3759]	Speed: 117.075147 samples/sec	 accuracy=39.411932	 loss=2.718012	 lr=0.045964
Epoch[119] Batch [0599]/[3759]	Speed: 117.891753 samples/sec	 accuracy=39.455729	 loss=2.719325	 lr=0.045951
Epoch[119] Batch [0649]/[3759]	Speed: 116.972370 samples/sec	 accuracy=39.526442	 loss=2.714070	 lr=0.045938
Epoch[119] Batch [0699]/[3759]	Speed: 117.775694 samples/sec	 accuracy=39.504464	 loss=2.714211	 lr=0.045925
Epoch[119] Batch [0749]/[3759]	Speed: 116.780437 samples/sec	 accuracy=39.614583	 loss=2.708198	 lr=0.045912
Epoch[119] Batch [0799]/[3759]	Speed: 117.221599 samples/sec	 accuracy=39.527344	 loss=2.712036	 lr=0.045899
Epoch[119] Batch [0849]/[3759]	Speed: 117.117910 samples/sec	 accuracy=39.568015	 loss=2.710082	 lr=0.045887
Epoch[119] Batch [0899]/[3759]	Speed: 117.232637 samples/sec	 accuracy=39.592014	 loss=2.709369	 lr=0.045874
Epoch[119] Batch [0949]/[3759]	Speed: 117.388211 samples/sec	 accuracy=39.570724	 loss=2.709177	 lr=0.045861
Epoch[119] Batch [0999]/[3759]	Speed: 117.651765 samples/sec	 accuracy=39.562500	 loss=2.710406	 lr=0.045848
Epoch[119] Batch [1049]/[3759]	Speed: 117.283744 samples/sec	 accuracy=39.529762	 loss=2.710890	 lr=0.045835
Epoch[119] Batch [1099]/[3759]	Speed: 117.071128 samples/sec	 accuracy=39.549716	 loss=2.710445	 lr=0.045822
Epoch[119] Batch [1149]/[3759]	Speed: 116.970485 samples/sec	 accuracy=39.437500	 loss=2.715003	 lr=0.045810
Epoch[119] Batch [1199]/[3759]	Speed: 117.461188 samples/sec	 accuracy=39.378906	 loss=2.716259	 lr=0.045797
Epoch[119] Batch [1249]/[3759]	Speed: 117.864108 samples/sec	 accuracy=39.387500	 loss=2.715742	 lr=0.045784
Epoch[119] Batch [1299]/[3759]	Speed: 116.992706 samples/sec	 accuracy=39.377404	 loss=2.715049	 lr=0.045771
Epoch[119] Batch [1349]/[3759]	Speed: 117.344156 samples/sec	 accuracy=39.310185	 loss=2.717142	 lr=0.045758
Epoch[119] Batch [1399]/[3759]	Speed: 117.658948 samples/sec	 accuracy=39.380580	 loss=2.716165	 lr=0.045745
Epoch[119] Batch [1449]/[3759]	Speed: 117.474835 samples/sec	 accuracy=39.363147	 loss=2.715836	 lr=0.045732
Epoch[119] Batch [1499]/[3759]	Speed: 117.546480 samples/sec	 accuracy=39.370833	 loss=2.714642	 lr=0.045720
Epoch[119] Batch [1549]/[3759]	Speed: 116.580911 samples/sec	 accuracy=39.419355	 loss=2.713887	 lr=0.045707
Epoch[119] Batch [1599]/[3759]	Speed: 117.486711 samples/sec	 accuracy=39.407227	 loss=2.714540	 lr=0.045694
Epoch[119] Batch [1649]/[3759]	Speed: 115.604936 samples/sec	 accuracy=39.382576	 loss=2.715440	 lr=0.045681
Epoch[119] Batch [1699]/[3759]	Speed: 117.110238 samples/sec	 accuracy=39.379596	 loss=2.715347	 lr=0.045668
Epoch[119] Batch [1749]/[3759]	Speed: 116.774606 samples/sec	 accuracy=39.367857	 loss=2.716134	 lr=0.045655
Epoch[119] Batch [1799]/[3759]	Speed: 117.408873 samples/sec	 accuracy=39.381944	 loss=2.716408	 lr=0.045642
Epoch[119] Batch [1849]/[3759]	Speed: 117.205396 samples/sec	 accuracy=39.423142	 loss=2.715787	 lr=0.045630
Epoch[119] Batch [1899]/[3759]	Speed: 116.772774 samples/sec	 accuracy=39.432566	 loss=2.715418	 lr=0.045617
Epoch[119] Batch [1949]/[3759]	Speed: 117.247305 samples/sec	 accuracy=39.455128	 loss=2.714011	 lr=0.045604
Epoch[119] Batch [1999]/[3759]	Speed: 117.000338 samples/sec	 accuracy=39.432812	 loss=2.715279	 lr=0.045591
Epoch[119] Batch [2049]/[3759]	Speed: 117.156372 samples/sec	 accuracy=39.445884	 loss=2.715194	 lr=0.045578
Epoch[119] Batch [2099]/[3759]	Speed: 117.727843 samples/sec	 accuracy=39.438988	 loss=2.715017	 lr=0.045565
Epoch[119] Batch [2149]/[3759]	Speed: 118.366924 samples/sec	 accuracy=39.472384	 loss=2.714020	 lr=0.045553
Epoch[119] Batch [2199]/[3759]	Speed: 116.623198 samples/sec	 accuracy=39.482955	 loss=2.713954	 lr=0.045540
Epoch[119] Batch [2249]/[3759]	Speed: 118.254691 samples/sec	 accuracy=39.485417	 loss=2.714266	 lr=0.045527
Epoch[119] Batch [2299]/[3759]	Speed: 117.669525 samples/sec	 accuracy=39.474185	 loss=2.714142	 lr=0.045514
Epoch[119] Batch [2349]/[3759]	Speed: 116.956041 samples/sec	 accuracy=39.473404	 loss=2.713747	 lr=0.045501
Epoch[119] Batch [2399]/[3759]	Speed: 117.480046 samples/sec	 accuracy=39.447917	 loss=2.714306	 lr=0.045488
Epoch[119] Batch [2449]/[3759]	Speed: 117.339093 samples/sec	 accuracy=39.449617	 loss=2.713185	 lr=0.045475
Epoch[119] Batch [2499]/[3759]	Speed: 118.250594 samples/sec	 accuracy=39.412500	 loss=2.713967	 lr=0.045463
Epoch[119] Batch [2549]/[3759]	Speed: 117.092489 samples/sec	 accuracy=39.413603	 loss=2.714857	 lr=0.045450
Epoch[119] Batch [2599]/[3759]	Speed: 116.076306 samples/sec	 accuracy=39.414062	 loss=2.715018	 lr=0.045437
Epoch[119] Batch [2649]/[3759]	Speed: 117.456926 samples/sec	 accuracy=39.410377	 loss=2.714915	 lr=0.045424
Epoch[119] Batch [2699]/[3759]	Speed: 116.991686 samples/sec	 accuracy=39.423611	 loss=2.715229	 lr=0.045411
Epoch[119] Batch [2749]/[3759]	Speed: 117.079407 samples/sec	 accuracy=39.408523	 loss=2.715568	 lr=0.045398
Epoch[119] Batch [2799]/[3759]	Speed: 117.949782 samples/sec	 accuracy=39.407924	 loss=2.715751	 lr=0.045386
Epoch[119] Batch [2849]/[3759]	Speed: 117.552696 samples/sec	 accuracy=39.401316	 loss=2.715877	 lr=0.045373
Epoch[119] Batch [2899]/[3759]	Speed: 117.437480 samples/sec	 accuracy=39.403017	 loss=2.715717	 lr=0.045360
Epoch[119] Batch [2949]/[3759]	Speed: 117.422820 samples/sec	 accuracy=39.397246	 loss=2.715412	 lr=0.045347
Epoch[119] Batch [2999]/[3759]	Speed: 118.020557 samples/sec	 accuracy=39.396354	 loss=2.716098	 lr=0.045334
Epoch[119] Batch [3049]/[3759]	Speed: 117.127753 samples/sec	 accuracy=39.368852	 loss=2.717209	 lr=0.045321
Epoch[119] Batch [3099]/[3759]	Speed: 116.761467 samples/sec	 accuracy=39.376008	 loss=2.717178	 lr=0.045309
Epoch[119] Batch [3149]/[3759]	Speed: 117.778542 samples/sec	 accuracy=39.386905	 loss=2.717133	 lr=0.045296
Epoch[119] Batch [3199]/[3759]	Speed: 117.151354 samples/sec	 accuracy=39.385254	 loss=2.717388	 lr=0.045283
Epoch[119] Batch [3249]/[3759]	Speed: 117.317890 samples/sec	 accuracy=39.381250	 loss=2.717802	 lr=0.045270
Epoch[119] Batch [3299]/[3759]	Speed: 118.164750 samples/sec	 accuracy=39.390152	 loss=2.717875	 lr=0.045257
Epoch[119] Batch [3349]/[3759]	Speed: 116.989374 samples/sec	 accuracy=39.377799	 loss=2.718290	 lr=0.045244
Epoch[119] Batch [3399]/[3759]	Speed: 114.425129 samples/sec	 accuracy=39.375460	 loss=2.718931	 lr=0.045231
Epoch[119] Batch [3449]/[3759]	Speed: 118.081461 samples/sec	 accuracy=39.383152	 loss=2.718683	 lr=0.045219
Epoch[119] Batch [3499]/[3759]	Speed: 117.760768 samples/sec	 accuracy=39.380804	 loss=2.719347	 lr=0.045206
Epoch[119] Batch [3549]/[3759]	Speed: 117.537341 samples/sec	 accuracy=39.367958	 loss=2.719246	 lr=0.045193
Epoch[119] Batch [3599]/[3759]	Speed: 117.332377 samples/sec	 accuracy=39.399306	 loss=2.718254	 lr=0.045180
Epoch[119] Batch [3649]/[3759]	Speed: 116.131960 samples/sec	 accuracy=39.414384	 loss=2.717560	 lr=0.045167
Epoch[119] Batch [3699]/[3759]	Speed: 117.374816 samples/sec	 accuracy=39.422297	 loss=2.717178	 lr=0.045154
Epoch[119] Batch [3749]/[3759]	Speed: 120.756498 samples/sec	 accuracy=39.436250	 loss=2.717037	 lr=0.045142
Batch [0049]/[0303]: acc-top1=41.843750 acc-top5=66.812500
Batch [0099]/[0303]: acc-top1=38.500000 acc-top5=64.328125
Batch [0149]/[0303]: acc-top1=39.510417 acc-top5=65.260417
Batch [0199]/[0303]: acc-top1=41.984375 acc-top5=67.234375
Batch [0249]/[0303]: acc-top1=41.318750 acc-top5=66.787500
Batch [0299]/[0303]: acc-top1=40.916667 acc-top5=66.734375
[Epoch 119] training: accuracy=39.436602	 loss=2.716913
[Epoch 119] speed: 116 samples/sec	time cost: 2177.041969
[Epoch 119] validation: acc-top1=40.846741 acc-top5=66.713078 loss=2.735907
Epoch[120] Batch [0049]/[3760]	Speed: 88.283085 samples/sec	 accuracy=40.937500	 loss=2.663900	 lr=0.045126
Epoch[120] Batch [0099]/[3760]	Speed: 117.800062 samples/sec	 accuracy=40.375000	 loss=2.694778	 lr=0.045114
Epoch[120] Batch [0149]/[3760]	Speed: 117.239269 samples/sec	 accuracy=40.250000	 loss=2.679679	 lr=0.045101
Epoch[120] Batch [0199]/[3760]	Speed: 117.487317 samples/sec	 accuracy=40.429688	 loss=2.669760	 lr=0.045088
Epoch[120] Batch [0249]/[3760]	Speed: 118.103629 samples/sec	 accuracy=40.387500	 loss=2.671690	 lr=0.045075
Epoch[120] Batch [0299]/[3760]	Speed: 116.568799 samples/sec	 accuracy=40.208333	 loss=2.685100	 lr=0.045062
Epoch[120] Batch [0349]/[3760]	Speed: 117.658449 samples/sec	 accuracy=40.397321	 loss=2.670831	 lr=0.045049
Epoch[120] Batch [0399]/[3760]	Speed: 116.893906 samples/sec	 accuracy=40.367188	 loss=2.669533	 lr=0.045037
Epoch[120] Batch [0449]/[3760]	Speed: 117.456517 samples/sec	 accuracy=40.229167	 loss=2.674667	 lr=0.045024
Epoch[120] Batch [0499]/[3760]	Speed: 117.179702 samples/sec	 accuracy=40.221875	 loss=2.674308	 lr=0.045011
Epoch[120] Batch [0549]/[3760]	Speed: 117.396273 samples/sec	 accuracy=40.346591	 loss=2.670558	 lr=0.044998
Epoch[120] Batch [0599]/[3760]	Speed: 116.702244 samples/sec	 accuracy=40.505208	 loss=2.663039	 lr=0.044985
Epoch[120] Batch [0649]/[3760]	Speed: 117.873180 samples/sec	 accuracy=40.459135	 loss=2.664097	 lr=0.044972
Epoch[120] Batch [0699]/[3760]	Speed: 118.692791 samples/sec	 accuracy=40.515625	 loss=2.663255	 lr=0.044960
Epoch[120] Batch [0749]/[3760]	Speed: 117.348657 samples/sec	 accuracy=40.472917	 loss=2.663966	 lr=0.044947
Epoch[120] Batch [0799]/[3760]	Speed: 116.959176 samples/sec	 accuracy=40.396484	 loss=2.668501	 lr=0.044934
Epoch[120] Batch [0849]/[3760]	Speed: 117.085477 samples/sec	 accuracy=40.325368	 loss=2.671732	 lr=0.044921
Epoch[120] Batch [0899]/[3760]	Speed: 116.890308 samples/sec	 accuracy=40.251736	 loss=2.674401	 lr=0.044908
Epoch[120] Batch [0949]/[3760]	Speed: 117.520925 samples/sec	 accuracy=40.241776	 loss=2.673580	 lr=0.044895
Epoch[120] Batch [0999]/[3760]	Speed: 117.998424 samples/sec	 accuracy=40.190625	 loss=2.674712	 lr=0.044883
Epoch[120] Batch [1049]/[3760]	Speed: 117.755391 samples/sec	 accuracy=40.144345	 loss=2.674432	 lr=0.044870
Epoch[120] Batch [1099]/[3760]	Speed: 118.084380 samples/sec	 accuracy=40.129261	 loss=2.676431	 lr=0.044857
Epoch[120] Batch [1149]/[3760]	Speed: 117.492866 samples/sec	 accuracy=40.073370	 loss=2.678986	 lr=0.044844
Epoch[120] Batch [1199]/[3760]	Speed: 116.263125 samples/sec	 accuracy=40.003906	 loss=2.682842	 lr=0.044831
Epoch[120] Batch [1249]/[3760]	Speed: 116.878602 samples/sec	 accuracy=39.968750	 loss=2.684062	 lr=0.044818
Epoch[120] Batch [1299]/[3760]	Speed: 117.680427 samples/sec	 accuracy=39.950721	 loss=2.685670	 lr=0.044806
Epoch[120] Batch [1349]/[3760]	Speed: 117.393062 samples/sec	 accuracy=39.924769	 loss=2.686786	 lr=0.044793
Epoch[120] Batch [1399]/[3760]	Speed: 117.658350 samples/sec	 accuracy=39.918527	 loss=2.687632	 lr=0.044780
Epoch[120] Batch [1449]/[3760]	Speed: 115.867746 samples/sec	 accuracy=39.935345	 loss=2.687992	 lr=0.044767
Epoch[120] Batch [1499]/[3760]	Speed: 117.967553 samples/sec	 accuracy=39.945833	 loss=2.687902	 lr=0.044754
Epoch[120] Batch [1549]/[3760]	Speed: 117.513574 samples/sec	 accuracy=39.916331	 loss=2.689362	 lr=0.044742
Epoch[120] Batch [1599]/[3760]	Speed: 116.716682 samples/sec	 accuracy=39.904297	 loss=2.691396	 lr=0.044729
Epoch[120] Batch [1649]/[3760]	Speed: 117.312445 samples/sec	 accuracy=39.870265	 loss=2.693498	 lr=0.044716
Epoch[120] Batch [1699]/[3760]	Speed: 117.872065 samples/sec	 accuracy=39.901654	 loss=2.691656	 lr=0.044703
Epoch[120] Batch [1749]/[3760]	Speed: 117.823354 samples/sec	 accuracy=39.886607	 loss=2.692216	 lr=0.044690
Epoch[120] Batch [1799]/[3760]	Speed: 116.581999 samples/sec	 accuracy=39.857639	 loss=2.693383	 lr=0.044677
Epoch[120] Batch [1849]/[3760]	Speed: 118.068944 samples/sec	 accuracy=39.848818	 loss=2.693295	 lr=0.044665
Epoch[120] Batch [1899]/[3760]	Speed: 117.383766 samples/sec	 accuracy=39.844572	 loss=2.694190	 lr=0.044652
Epoch[120] Batch [1949]/[3760]	Speed: 117.362338 samples/sec	 accuracy=39.822115	 loss=2.694431	 lr=0.044639
Epoch[120] Batch [1999]/[3760]	Speed: 117.437563 samples/sec	 accuracy=39.821875	 loss=2.695267	 lr=0.044626
Epoch[120] Batch [2049]/[3760]	Speed: 117.453543 samples/sec	 accuracy=39.791159	 loss=2.697453	 lr=0.044613
Epoch[120] Batch [2099]/[3760]	Speed: 116.845339 samples/sec	 accuracy=39.786458	 loss=2.697393	 lr=0.044600
Epoch[120] Batch [2149]/[3760]	Speed: 117.398551 samples/sec	 accuracy=39.805233	 loss=2.696491	 lr=0.044588
Epoch[120] Batch [2199]/[3760]	Speed: 117.741719 samples/sec	 accuracy=39.808239	 loss=2.695905	 lr=0.044575
Epoch[120] Batch [2249]/[3760]	Speed: 116.998466 samples/sec	 accuracy=39.778472	 loss=2.697288	 lr=0.044562
Epoch[120] Batch [2299]/[3760]	Speed: 116.683992 samples/sec	 accuracy=39.797554	 loss=2.696268	 lr=0.044549
Epoch[120] Batch [2349]/[3760]	Speed: 117.825393 samples/sec	 accuracy=39.825133	 loss=2.695569	 lr=0.044536
Epoch[120] Batch [2399]/[3760]	Speed: 117.873010 samples/sec	 accuracy=39.808594	 loss=2.696832	 lr=0.044524
Epoch[120] Batch [2449]/[3760]	Speed: 117.759281 samples/sec	 accuracy=39.809311	 loss=2.696638	 lr=0.044511
Epoch[120] Batch [2499]/[3760]	Speed: 117.394277 samples/sec	 accuracy=39.840625	 loss=2.696071	 lr=0.044498
Epoch[120] Batch [2549]/[3760]	Speed: 115.978945 samples/sec	 accuracy=39.840074	 loss=2.695652	 lr=0.044485
Epoch[120] Batch [2599]/[3760]	Speed: 117.748046 samples/sec	 accuracy=39.825120	 loss=2.696460	 lr=0.044472
Epoch[120] Batch [2649]/[3760]	Speed: 117.117561 samples/sec	 accuracy=39.814858	 loss=2.696812	 lr=0.044459
Epoch[120] Batch [2699]/[3760]	Speed: 117.748814 samples/sec	 accuracy=39.811343	 loss=2.696879	 lr=0.044447
Epoch[120] Batch [2749]/[3760]	Speed: 117.182710 samples/sec	 accuracy=39.790909	 loss=2.698150	 lr=0.044434
Epoch[120] Batch [2799]/[3760]	Speed: 117.037253 samples/sec	 accuracy=39.791295	 loss=2.698090	 lr=0.044421
Epoch[120] Batch [2849]/[3760]	Speed: 117.448119 samples/sec	 accuracy=39.783443	 loss=2.697773	 lr=0.044408
Epoch[120] Batch [2899]/[3760]	Speed: 117.763538 samples/sec	 accuracy=39.796875	 loss=2.697904	 lr=0.044395
Epoch[120] Batch [2949]/[3760]	Speed: 117.225585 samples/sec	 accuracy=39.821504	 loss=2.697334	 lr=0.044383
Epoch[120] Batch [2999]/[3760]	Speed: 117.285885 samples/sec	 accuracy=39.797917	 loss=2.698611	 lr=0.044370
Epoch[120] Batch [3049]/[3760]	Speed: 117.459590 samples/sec	 accuracy=39.787398	 loss=2.698542	 lr=0.044357
Epoch[120] Batch [3099]/[3760]	Speed: 117.630839 samples/sec	 accuracy=39.779234	 loss=2.699345	 lr=0.044344
Epoch[120] Batch [3149]/[3760]	Speed: 117.735713 samples/sec	 accuracy=39.781250	 loss=2.698779	 lr=0.044331
Epoch[120] Batch [3199]/[3760]	Speed: 117.089184 samples/sec	 accuracy=39.771973	 loss=2.699241	 lr=0.044318
Epoch[120] Batch [3249]/[3760]	Speed: 117.532175 samples/sec	 accuracy=39.772115	 loss=2.699089	 lr=0.044306
Epoch[120] Batch [3299]/[3760]	Speed: 116.857586 samples/sec	 accuracy=39.752367	 loss=2.700500	 lr=0.044293
Epoch[120] Batch [3349]/[3760]	Speed: 117.395564 samples/sec	 accuracy=39.770989	 loss=2.699436	 lr=0.044280
Epoch[120] Batch [3399]/[3760]	Speed: 117.649359 samples/sec	 accuracy=39.772978	 loss=2.699442	 lr=0.044267
Epoch[120] Batch [3449]/[3760]	Speed: 117.402263 samples/sec	 accuracy=39.780344	 loss=2.699188	 lr=0.044254
Epoch[120] Batch [3499]/[3760]	Speed: 116.770146 samples/sec	 accuracy=39.797321	 loss=2.698857	 lr=0.044242
Epoch[120] Batch [3549]/[3760]	Speed: 117.298046 samples/sec	 accuracy=39.791813	 loss=2.699737	 lr=0.044229
Epoch[120] Batch [3599]/[3760]	Speed: 115.962695 samples/sec	 accuracy=39.798611	 loss=2.699406	 lr=0.044216
Epoch[120] Batch [3649]/[3760]	Speed: 117.632055 samples/sec	 accuracy=39.784675	 loss=2.700150	 lr=0.044203
Epoch[120] Batch [3699]/[3760]	Speed: 116.859201 samples/sec	 accuracy=39.807855	 loss=2.699281	 lr=0.044190
Epoch[120] Batch [3749]/[3760]	Speed: 120.426184 samples/sec	 accuracy=39.788750	 loss=2.700724	 lr=0.044178
Batch [0049]/[0303]: acc-top1=41.750000 acc-top5=67.000000
Batch [0099]/[0303]: acc-top1=38.375000 acc-top5=64.515625
Batch [0149]/[0303]: acc-top1=39.354167 acc-top5=65.354167
Batch [0199]/[0303]: acc-top1=41.718750 acc-top5=67.195312
Batch [0249]/[0303]: acc-top1=41.187500 acc-top5=67.187500
Batch [0299]/[0303]: acc-top1=40.526042 acc-top5=67.072917
[Epoch 120] training: accuracy=39.786818	 loss=2.700747
[Epoch 120] speed: 116 samples/sec	time cost: 2176.472208
[Epoch 120] validation: acc-top1=40.527021 acc-top5=67.094678 loss=2.762391
Epoch[121] Batch [0049]/[3760]	Speed: 87.537152 samples/sec	 accuracy=40.875000	 loss=2.643638	 lr=0.044162
Epoch[121] Batch [0099]/[3760]	Speed: 117.127167 samples/sec	 accuracy=42.062500	 loss=2.622045	 lr=0.044149
Epoch[121] Batch [0149]/[3760]	Speed: 118.009365 samples/sec	 accuracy=41.979167	 loss=2.625068	 lr=0.044137
Epoch[121] Batch [0199]/[3760]	Speed: 117.388106 samples/sec	 accuracy=41.898438	 loss=2.623831	 lr=0.044124
Epoch[121] Batch [0249]/[3760]	Speed: 118.100475 samples/sec	 accuracy=41.512500	 loss=2.635458	 lr=0.044111
Epoch[121] Batch [0299]/[3760]	Speed: 117.189861 samples/sec	 accuracy=41.447917	 loss=2.641060	 lr=0.044098
Epoch[121] Batch [0349]/[3760]	Speed: 117.946808 samples/sec	 accuracy=41.352679	 loss=2.642763	 lr=0.044085
Epoch[121] Batch [0399]/[3760]	Speed: 117.200835 samples/sec	 accuracy=41.328125	 loss=2.643941	 lr=0.044073
Epoch[121] Batch [0449]/[3760]	Speed: 117.709811 samples/sec	 accuracy=41.187500	 loss=2.648234	 lr=0.044060
Epoch[121] Batch [0499]/[3760]	Speed: 116.665886 samples/sec	 accuracy=41.178125	 loss=2.648380	 lr=0.044047
Epoch[121] Batch [0549]/[3760]	Speed: 118.224471 samples/sec	 accuracy=41.107955	 loss=2.651420	 lr=0.044034
Epoch[121] Batch [0599]/[3760]	Speed: 117.565444 samples/sec	 accuracy=41.059896	 loss=2.650526	 lr=0.044021
Epoch[121] Batch [0649]/[3760]	Speed: 117.089708 samples/sec	 accuracy=40.990385	 loss=2.652157	 lr=0.044008
Epoch[121] Batch [0699]/[3760]	Speed: 116.965161 samples/sec	 accuracy=40.908482	 loss=2.656331	 lr=0.043996
Epoch[121] Batch [0749]/[3760]	Speed: 117.203808 samples/sec	 accuracy=40.881250	 loss=2.655931	 lr=0.043983
Epoch[121] Batch [0799]/[3760]	Speed: 117.927509 samples/sec	 accuracy=40.865234	 loss=2.656812	 lr=0.043970
Epoch[121] Batch [0849]/[3760]	Speed: 117.976736 samples/sec	 accuracy=40.737132	 loss=2.659734	 lr=0.043957
Epoch[121] Batch [0899]/[3760]	Speed: 115.573541 samples/sec	 accuracy=40.689236	 loss=2.659220	 lr=0.043944
Epoch[121] Batch [0949]/[3760]	Speed: 116.969901 samples/sec	 accuracy=40.590461	 loss=2.661025	 lr=0.043932
Epoch[121] Batch [0999]/[3760]	Speed: 117.351430 samples/sec	 accuracy=40.587500	 loss=2.666148	 lr=0.043919
Epoch[121] Batch [1049]/[3760]	Speed: 117.894032 samples/sec	 accuracy=40.535714	 loss=2.668589	 lr=0.043906
Epoch[121] Batch [1099]/[3760]	Speed: 117.385760 samples/sec	 accuracy=40.419034	 loss=2.674203	 lr=0.043893
Epoch[121] Batch [1149]/[3760]	Speed: 116.990265 samples/sec	 accuracy=40.418478	 loss=2.675527	 lr=0.043880
Epoch[121] Batch [1199]/[3760]	Speed: 117.538839 samples/sec	 accuracy=40.441406	 loss=2.673259	 lr=0.043868
Epoch[121] Batch [1249]/[3760]	Speed: 117.706139 samples/sec	 accuracy=40.405000	 loss=2.675317	 lr=0.043855
Epoch[121] Batch [1299]/[3760]	Speed: 117.380753 samples/sec	 accuracy=40.406250	 loss=2.675281	 lr=0.043842
Epoch[121] Batch [1349]/[3760]	Speed: 117.321757 samples/sec	 accuracy=40.372685	 loss=2.676230	 lr=0.043829
Epoch[121] Batch [1399]/[3760]	Speed: 117.766610 samples/sec	 accuracy=40.337054	 loss=2.677193	 lr=0.043816
Epoch[121] Batch [1449]/[3760]	Speed: 116.945389 samples/sec	 accuracy=40.330819	 loss=2.677944	 lr=0.043804
Epoch[121] Batch [1499]/[3760]	Speed: 117.607940 samples/sec	 accuracy=40.297917	 loss=2.678185	 lr=0.043791
Epoch[121] Batch [1549]/[3760]	Speed: 117.261815 samples/sec	 accuracy=40.270161	 loss=2.679663	 lr=0.043778
Epoch[121] Batch [1599]/[3760]	Speed: 117.444429 samples/sec	 accuracy=40.333008	 loss=2.676672	 lr=0.043765
Epoch[121] Batch [1649]/[3760]	Speed: 117.637065 samples/sec	 accuracy=40.276515	 loss=2.678673	 lr=0.043752
Epoch[121] Batch [1699]/[3760]	Speed: 117.286616 samples/sec	 accuracy=40.286765	 loss=2.678131	 lr=0.043740
Epoch[121] Batch [1749]/[3760]	Speed: 117.071500 samples/sec	 accuracy=40.266964	 loss=2.678680	 lr=0.043727
Epoch[121] Batch [1799]/[3760]	Speed: 118.310009 samples/sec	 accuracy=40.269097	 loss=2.677709	 lr=0.043714
Epoch[121] Batch [1849]/[3760]	Speed: 116.745194 samples/sec	 accuracy=40.298142	 loss=2.678014	 lr=0.043701
Epoch[121] Batch [1899]/[3760]	Speed: 117.245435 samples/sec	 accuracy=40.285362	 loss=2.679526	 lr=0.043688
Epoch[121] Batch [1949]/[3760]	Speed: 114.462420 samples/sec	 accuracy=40.313301	 loss=2.677688	 lr=0.043676
Epoch[121] Batch [1999]/[3760]	Speed: 116.806240 samples/sec	 accuracy=40.322656	 loss=2.676897	 lr=0.043663
Epoch[121] Batch [2049]/[3760]	Speed: 117.436411 samples/sec	 accuracy=40.309451	 loss=2.677488	 lr=0.043650
Epoch[121] Batch [2099]/[3760]	Speed: 117.334274 samples/sec	 accuracy=40.297619	 loss=2.678107	 lr=0.043637
Epoch[121] Batch [2149]/[3760]	Speed: 117.533048 samples/sec	 accuracy=40.303779	 loss=2.678448	 lr=0.043625
Epoch[121] Batch [2199]/[3760]	Speed: 117.286612 samples/sec	 accuracy=40.303977	 loss=2.678102	 lr=0.043612
Epoch[121] Batch [2249]/[3760]	Speed: 117.498808 samples/sec	 accuracy=40.306944	 loss=2.678357	 lr=0.043599
Epoch[121] Batch [2299]/[3760]	Speed: 117.232606 samples/sec	 accuracy=40.311141	 loss=2.678605	 lr=0.043586
Epoch[121] Batch [2349]/[3760]	Speed: 117.637907 samples/sec	 accuracy=40.311835	 loss=2.678814	 lr=0.043573
Epoch[121] Batch [2399]/[3760]	Speed: 117.474260 samples/sec	 accuracy=40.303385	 loss=2.678707	 lr=0.043561
Epoch[121] Batch [2449]/[3760]	Speed: 116.646962 samples/sec	 accuracy=40.304847	 loss=2.677802	 lr=0.043548
Epoch[121] Batch [2499]/[3760]	Speed: 117.319825 samples/sec	 accuracy=40.302500	 loss=2.677349	 lr=0.043535
Epoch[121] Batch [2549]/[3760]	Speed: 117.865944 samples/sec	 accuracy=40.281863	 loss=2.677358	 lr=0.043522
Epoch[121] Batch [2599]/[3760]	Speed: 117.540674 samples/sec	 accuracy=40.287861	 loss=2.676903	 lr=0.043509
Epoch[121] Batch [2649]/[3760]	Speed: 117.475521 samples/sec	 accuracy=40.251769	 loss=2.678185	 lr=0.043497
Epoch[121] Batch [2699]/[3760]	Speed: 117.612612 samples/sec	 accuracy=40.251157	 loss=2.678678	 lr=0.043484
Epoch[121] Batch [2749]/[3760]	Speed: 118.040272 samples/sec	 accuracy=40.239205	 loss=2.679096	 lr=0.043471
Epoch[121] Batch [2799]/[3760]	Speed: 117.777619 samples/sec	 accuracy=40.208147	 loss=2.679305	 lr=0.043458
Epoch[121] Batch [2849]/[3760]	Speed: 117.303914 samples/sec	 accuracy=40.167763	 loss=2.680961	 lr=0.043445
Epoch[121] Batch [2899]/[3760]	Speed: 117.257725 samples/sec	 accuracy=40.144397	 loss=2.681856	 lr=0.043433
Epoch[121] Batch [2949]/[3760]	Speed: 117.291798 samples/sec	 accuracy=40.150424	 loss=2.681416	 lr=0.043420
Epoch[121] Batch [2999]/[3760]	Speed: 116.017579 samples/sec	 accuracy=40.161458	 loss=2.681123	 lr=0.043407
Epoch[121] Batch [3049]/[3760]	Speed: 117.969753 samples/sec	 accuracy=40.149078	 loss=2.682226	 lr=0.043394
Epoch[121] Batch [3099]/[3760]	Speed: 116.959023 samples/sec	 accuracy=40.147681	 loss=2.682132	 lr=0.043382
Epoch[121] Batch [3149]/[3760]	Speed: 116.935802 samples/sec	 accuracy=40.126984	 loss=2.683069	 lr=0.043369
Epoch[121] Batch [3199]/[3760]	Speed: 117.412650 samples/sec	 accuracy=40.128418	 loss=2.683217	 lr=0.043356
Epoch[121] Batch [3249]/[3760]	Speed: 117.645696 samples/sec	 accuracy=40.121154	 loss=2.683893	 lr=0.043343
Epoch[121] Batch [3299]/[3760]	Speed: 117.686398 samples/sec	 accuracy=40.114583	 loss=2.684329	 lr=0.043330
Epoch[121] Batch [3349]/[3760]	Speed: 117.310591 samples/sec	 accuracy=40.111474	 loss=2.684730	 lr=0.043318
Epoch[121] Batch [3399]/[3760]	Speed: 117.291311 samples/sec	 accuracy=40.100184	 loss=2.685376	 lr=0.043305
Epoch[121] Batch [3449]/[3760]	Speed: 116.830646 samples/sec	 accuracy=40.082428	 loss=2.686018	 lr=0.043292
Epoch[121] Batch [3499]/[3760]	Speed: 117.264279 samples/sec	 accuracy=40.080804	 loss=2.686105	 lr=0.043279
Epoch[121] Batch [3549]/[3760]	Speed: 117.456619 samples/sec	 accuracy=40.077025	 loss=2.686183	 lr=0.043267
Epoch[121] Batch [3599]/[3760]	Speed: 117.286360 samples/sec	 accuracy=40.090278	 loss=2.685692	 lr=0.043254
Epoch[121] Batch [3649]/[3760]	Speed: 118.203944 samples/sec	 accuracy=40.079623	 loss=2.686120	 lr=0.043241
Epoch[121] Batch [3699]/[3760]	Speed: 116.647034 samples/sec	 accuracy=40.080659	 loss=2.686527	 lr=0.043228
Epoch[121] Batch [3749]/[3760]	Speed: 120.989045 samples/sec	 accuracy=40.079167	 loss=2.687269	 lr=0.043215
Batch [0049]/[0303]: acc-top1=40.906250 acc-top5=67.250000
Batch [0099]/[0303]: acc-top1=38.406250 acc-top5=65.156250
Batch [0149]/[0303]: acc-top1=40.041667 acc-top5=65.947917
Batch [0199]/[0303]: acc-top1=42.656250 acc-top5=68.039062
Batch [0249]/[0303]: acc-top1=42.231250 acc-top5=68.218750
Batch [0299]/[0303]: acc-top1=41.713542 acc-top5=68.093750
[Epoch 121] training: accuracy=40.077709	 loss=2.687499
[Epoch 121] speed: 116 samples/sec	time cost: 2177.344638
[Epoch 121] validation: acc-top1=41.671823 acc-top5=68.162129 loss=2.649938
Epoch[122] Batch [0049]/[3759]	Speed: 86.693501 samples/sec	 accuracy=40.625000	 loss=2.661261	 lr=0.043200
Epoch[122] Batch [0099]/[3759]	Speed: 117.445664 samples/sec	 accuracy=41.625000	 loss=2.628220	 lr=0.043187
Epoch[122] Batch [0149]/[3759]	Speed: 117.803528 samples/sec	 accuracy=41.333333	 loss=2.630686	 lr=0.043175
Epoch[122] Batch [0199]/[3759]	Speed: 117.106010 samples/sec	 accuracy=41.218750	 loss=2.640253	 lr=0.043162
Epoch[122] Batch [0249]/[3759]	Speed: 117.121097 samples/sec	 accuracy=41.187500	 loss=2.642224	 lr=0.043149
Epoch[122] Batch [0299]/[3759]	Speed: 115.435891 samples/sec	 accuracy=41.046875	 loss=2.643151	 lr=0.043136
Epoch[122] Batch [0349]/[3759]	Speed: 117.472415 samples/sec	 accuracy=40.825893	 loss=2.644668	 lr=0.043123
Epoch[122] Batch [0399]/[3759]	Speed: 117.660478 samples/sec	 accuracy=40.574219	 loss=2.651762	 lr=0.043111
Epoch[122] Batch [0449]/[3759]	Speed: 117.418075 samples/sec	 accuracy=40.510417	 loss=2.653105	 lr=0.043098
Epoch[122] Batch [0499]/[3759]	Speed: 116.879005 samples/sec	 accuracy=40.506250	 loss=2.654336	 lr=0.043085
Epoch[122] Batch [0549]/[3759]	Speed: 116.910256 samples/sec	 accuracy=40.465909	 loss=2.662310	 lr=0.043072
Epoch[122] Batch [0599]/[3759]	Speed: 117.659154 samples/sec	 accuracy=40.434896	 loss=2.664544	 lr=0.043060
Epoch[122] Batch [0649]/[3759]	Speed: 116.659755 samples/sec	 accuracy=40.399038	 loss=2.667582	 lr=0.043047
Epoch[122] Batch [0699]/[3759]	Speed: 118.519912 samples/sec	 accuracy=40.504464	 loss=2.668048	 lr=0.043034
Epoch[122] Batch [0749]/[3759]	Speed: 116.935773 samples/sec	 accuracy=40.489583	 loss=2.667516	 lr=0.043021
Epoch[122] Batch [0799]/[3759]	Speed: 117.576771 samples/sec	 accuracy=40.402344	 loss=2.668869	 lr=0.043008
Epoch[122] Batch [0849]/[3759]	Speed: 117.469444 samples/sec	 accuracy=40.443015	 loss=2.667728	 lr=0.042996
Epoch[122] Batch [0899]/[3759]	Speed: 117.065575 samples/sec	 accuracy=40.411458	 loss=2.670908	 lr=0.042983
Epoch[122] Batch [0949]/[3759]	Speed: 117.934286 samples/sec	 accuracy=40.476974	 loss=2.668268	 lr=0.042970
Epoch[122] Batch [0999]/[3759]	Speed: 117.976718 samples/sec	 accuracy=40.431250	 loss=2.669908	 lr=0.042957
Epoch[122] Batch [1049]/[3759]	Speed: 116.762505 samples/sec	 accuracy=40.436012	 loss=2.669667	 lr=0.042945
Epoch[122] Batch [1099]/[3759]	Speed: 117.409880 samples/sec	 accuracy=40.515625	 loss=2.666373	 lr=0.042932
Epoch[122] Batch [1149]/[3759]	Speed: 117.501188 samples/sec	 accuracy=40.494565	 loss=2.667354	 lr=0.042919
Epoch[122] Batch [1199]/[3759]	Speed: 116.130910 samples/sec	 accuracy=40.554688	 loss=2.663614	 lr=0.042906
Epoch[122] Batch [1249]/[3759]	Speed: 118.079580 samples/sec	 accuracy=40.488750	 loss=2.666114	 lr=0.042894
Epoch[122] Batch [1299]/[3759]	Speed: 118.606040 samples/sec	 accuracy=40.495192	 loss=2.669286	 lr=0.042881
Epoch[122] Batch [1349]/[3759]	Speed: 117.391933 samples/sec	 accuracy=40.460648	 loss=2.670495	 lr=0.042868
Epoch[122] Batch [1399]/[3759]	Speed: 116.659248 samples/sec	 accuracy=40.481027	 loss=2.670499	 lr=0.042855
Epoch[122] Batch [1449]/[3759]	Speed: 117.611384 samples/sec	 accuracy=40.477371	 loss=2.671530	 lr=0.042842
Epoch[122] Batch [1499]/[3759]	Speed: 117.171796 samples/sec	 accuracy=40.484375	 loss=2.671272	 lr=0.042830
Epoch[122] Batch [1549]/[3759]	Speed: 117.641303 samples/sec	 accuracy=40.513105	 loss=2.670608	 lr=0.042817
Epoch[122] Batch [1599]/[3759]	Speed: 117.902655 samples/sec	 accuracy=40.493164	 loss=2.672801	 lr=0.042804
Epoch[122] Batch [1649]/[3759]	Speed: 117.227412 samples/sec	 accuracy=40.512311	 loss=2.672720	 lr=0.042791
Epoch[122] Batch [1699]/[3759]	Speed: 117.927365 samples/sec	 accuracy=40.508272	 loss=2.673372	 lr=0.042779
Epoch[122] Batch [1749]/[3759]	Speed: 116.761439 samples/sec	 accuracy=40.486607	 loss=2.673395	 lr=0.042766
Epoch[122] Batch [1799]/[3759]	Speed: 117.513265 samples/sec	 accuracy=40.489583	 loss=2.673937	 lr=0.042753
Epoch[122] Batch [1849]/[3759]	Speed: 117.088171 samples/sec	 accuracy=40.476351	 loss=2.675194	 lr=0.042740
Epoch[122] Batch [1899]/[3759]	Speed: 117.413182 samples/sec	 accuracy=40.481908	 loss=2.674457	 lr=0.042728
Epoch[122] Batch [1949]/[3759]	Speed: 118.774031 samples/sec	 accuracy=40.467147	 loss=2.675073	 lr=0.042715
Epoch[122] Batch [1999]/[3759]	Speed: 117.606046 samples/sec	 accuracy=40.422656	 loss=2.676354	 lr=0.042702
Epoch[122] Batch [2049]/[3759]	Speed: 117.304682 samples/sec	 accuracy=40.391768	 loss=2.677384	 lr=0.042689
Epoch[122] Batch [2099]/[3759]	Speed: 117.001119 samples/sec	 accuracy=40.378720	 loss=2.678364	 lr=0.042677
Epoch[122] Batch [2149]/[3759]	Speed: 116.285993 samples/sec	 accuracy=40.365552	 loss=2.679797	 lr=0.042664
Epoch[122] Batch [2199]/[3759]	Speed: 117.324602 samples/sec	 accuracy=40.382812	 loss=2.679422	 lr=0.042651
Epoch[122] Batch [2249]/[3759]	Speed: 117.218319 samples/sec	 accuracy=40.381250	 loss=2.678488	 lr=0.042638
Epoch[122] Batch [2299]/[3759]	Speed: 117.244699 samples/sec	 accuracy=40.378397	 loss=2.678337	 lr=0.042626
Epoch[122] Batch [2349]/[3759]	Speed: 116.824792 samples/sec	 accuracy=40.351729	 loss=2.678905	 lr=0.042613
Epoch[122] Batch [2399]/[3759]	Speed: 118.045605 samples/sec	 accuracy=40.349609	 loss=2.678587	 lr=0.042600
Epoch[122] Batch [2449]/[3759]	Speed: 117.899746 samples/sec	 accuracy=40.336097	 loss=2.679273	 lr=0.042587
Epoch[122] Batch [2499]/[3759]	Speed: 115.879178 samples/sec	 accuracy=40.336875	 loss=2.678591	 lr=0.042575
Epoch[122] Batch [2549]/[3759]	Speed: 116.630823 samples/sec	 accuracy=40.329657	 loss=2.678467	 lr=0.042562
Epoch[122] Batch [2599]/[3759]	Speed: 117.394613 samples/sec	 accuracy=40.304087	 loss=2.679485	 lr=0.042549
Epoch[122] Batch [2649]/[3759]	Speed: 117.615953 samples/sec	 accuracy=40.304835	 loss=2.679469	 lr=0.042536
Epoch[122] Batch [2699]/[3759]	Speed: 117.701538 samples/sec	 accuracy=40.266204	 loss=2.681244	 lr=0.042524
Epoch[122] Batch [2749]/[3759]	Speed: 116.469112 samples/sec	 accuracy=40.278409	 loss=2.681349	 lr=0.042511
Epoch[122] Batch [2799]/[3759]	Speed: 117.270860 samples/sec	 accuracy=40.265067	 loss=2.682505	 lr=0.042498
Epoch[122] Batch [2849]/[3759]	Speed: 116.393483 samples/sec	 accuracy=40.277412	 loss=2.682169	 lr=0.042485
Epoch[122] Batch [2899]/[3759]	Speed: 117.339679 samples/sec	 accuracy=40.269397	 loss=2.682722	 lr=0.042473
Epoch[122] Batch [2949]/[3759]	Speed: 117.255066 samples/sec	 accuracy=40.291843	 loss=2.681239	 lr=0.042460
Epoch[122] Batch [2999]/[3759]	Speed: 116.599939 samples/sec	 accuracy=40.294792	 loss=2.681414	 lr=0.042447
Epoch[122] Batch [3049]/[3759]	Speed: 117.438347 samples/sec	 accuracy=40.279713	 loss=2.681542	 lr=0.042434
Epoch[122] Batch [3099]/[3759]	Speed: 117.553201 samples/sec	 accuracy=40.271169	 loss=2.682021	 lr=0.042422
Epoch[122] Batch [3149]/[3759]	Speed: 117.316949 samples/sec	 accuracy=40.289683	 loss=2.681624	 lr=0.042409
Epoch[122] Batch [3199]/[3759]	Speed: 116.981124 samples/sec	 accuracy=40.311523	 loss=2.681085	 lr=0.042396
Epoch[122] Batch [3249]/[3759]	Speed: 117.609587 samples/sec	 accuracy=40.324038	 loss=2.681093	 lr=0.042383
Epoch[122] Batch [3299]/[3759]	Speed: 117.285370 samples/sec	 accuracy=40.343277	 loss=2.680358	 lr=0.042371
Epoch[122] Batch [3349]/[3759]	Speed: 117.660859 samples/sec	 accuracy=40.334422	 loss=2.681126	 lr=0.042358
Epoch[122] Batch [3399]/[3759]	Speed: 117.308085 samples/sec	 accuracy=40.315717	 loss=2.681971	 lr=0.042345
Epoch[122] Batch [3449]/[3759]	Speed: 117.557497 samples/sec	 accuracy=40.314764	 loss=2.682588	 lr=0.042332
Epoch[122] Batch [3499]/[3759]	Speed: 117.341593 samples/sec	 accuracy=40.315179	 loss=2.681990	 lr=0.042320
Epoch[122] Batch [3549]/[3759]	Speed: 117.633274 samples/sec	 accuracy=40.312500	 loss=2.682205	 lr=0.042307
Epoch[122] Batch [3599]/[3759]	Speed: 115.718446 samples/sec	 accuracy=40.305556	 loss=2.682502	 lr=0.042294
Epoch[122] Batch [3649]/[3759]	Speed: 118.000803 samples/sec	 accuracy=40.316353	 loss=2.682276	 lr=0.042281
Epoch[122] Batch [3699]/[3759]	Speed: 116.954880 samples/sec	 accuracy=40.315878	 loss=2.682319	 lr=0.042269
Epoch[122] Batch [3749]/[3759]	Speed: 120.431183 samples/sec	 accuracy=40.333333	 loss=2.681746	 lr=0.042256
Batch [0049]/[0303]: acc-top1=42.437500 acc-top5=67.062500
Batch [0099]/[0303]: acc-top1=39.515625 acc-top5=65.203125
Batch [0149]/[0303]: acc-top1=41.447917 acc-top5=66.781250
Batch [0199]/[0303]: acc-top1=43.648438 acc-top5=68.210938
Batch [0249]/[0303]: acc-top1=42.912500 acc-top5=68.300000
Batch [0299]/[0303]: acc-top1=42.229167 acc-top5=68.156250
[Epoch 122] training: accuracy=40.323640	 loss=2.682148
[Epoch 122] speed: 116 samples/sec	time cost: 2176.917795
[Epoch 122] validation: acc-top1=42.264851 acc-top5=68.187913 loss=2.643585
Epoch[123] Batch [0049]/[3760]	Speed: 86.814424 samples/sec	 accuracy=39.812500	 loss=2.629746	 lr=0.042241
Epoch[123] Batch [0099]/[3760]	Speed: 117.463277 samples/sec	 accuracy=40.437500	 loss=2.634502	 lr=0.042228
Epoch[123] Batch [0149]/[3760]	Speed: 117.623132 samples/sec	 accuracy=40.718750	 loss=2.637835	 lr=0.042215
Epoch[123] Batch [0199]/[3760]	Speed: 117.753047 samples/sec	 accuracy=40.578125	 loss=2.649602	 lr=0.042203
Epoch[123] Batch [0249]/[3760]	Speed: 117.350431 samples/sec	 accuracy=40.625000	 loss=2.652215	 lr=0.042190
Epoch[123] Batch [0299]/[3760]	Speed: 117.134512 samples/sec	 accuracy=40.755208	 loss=2.654445	 lr=0.042177
Epoch[123] Batch [0349]/[3760]	Speed: 117.287403 samples/sec	 accuracy=40.571429	 loss=2.654366	 lr=0.042164
Epoch[123] Batch [0399]/[3760]	Speed: 117.471215 samples/sec	 accuracy=40.535156	 loss=2.651521	 lr=0.042152
Epoch[123] Batch [0449]/[3760]	Speed: 117.438815 samples/sec	 accuracy=40.565972	 loss=2.648162	 lr=0.042139
Epoch[123] Batch [0499]/[3760]	Speed: 118.028564 samples/sec	 accuracy=40.440625	 loss=2.651895	 lr=0.042126
Epoch[123] Batch [0549]/[3760]	Speed: 117.755808 samples/sec	 accuracy=40.491477	 loss=2.648688	 lr=0.042113
Epoch[123] Batch [0599]/[3760]	Speed: 116.800357 samples/sec	 accuracy=40.453125	 loss=2.651580	 lr=0.042101
Epoch[123] Batch [0649]/[3760]	Speed: 117.682124 samples/sec	 accuracy=40.500000	 loss=2.647973	 lr=0.042088
Epoch[123] Batch [0699]/[3760]	Speed: 117.601361 samples/sec	 accuracy=40.558036	 loss=2.648008	 lr=0.042075
Epoch[123] Batch [0749]/[3760]	Speed: 117.220393 samples/sec	 accuracy=40.568750	 loss=2.647023	 lr=0.042062
Epoch[123] Batch [0799]/[3760]	Speed: 117.514326 samples/sec	 accuracy=40.708984	 loss=2.643689	 lr=0.042050
Epoch[123] Batch [0849]/[3760]	Speed: 115.913546 samples/sec	 accuracy=40.735294	 loss=2.644700	 lr=0.042037
Epoch[123] Batch [0899]/[3760]	Speed: 117.314568 samples/sec	 accuracy=40.685764	 loss=2.646825	 lr=0.042024
Epoch[123] Batch [0949]/[3760]	Speed: 117.418837 samples/sec	 accuracy=40.666118	 loss=2.647809	 lr=0.042012
Epoch[123] Batch [0999]/[3760]	Speed: 117.182683 samples/sec	 accuracy=40.707813	 loss=2.646610	 lr=0.041999
Epoch[123] Batch [1049]/[3760]	Speed: 117.463472 samples/sec	 accuracy=40.696429	 loss=2.647287	 lr=0.041986
Epoch[123] Batch [1099]/[3760]	Speed: 117.573075 samples/sec	 accuracy=40.704545	 loss=2.648884	 lr=0.041973
Epoch[123] Batch [1149]/[3760]	Speed: 117.946376 samples/sec	 accuracy=40.779891	 loss=2.647724	 lr=0.041961
Epoch[123] Batch [1199]/[3760]	Speed: 117.547040 samples/sec	 accuracy=40.739583	 loss=2.649737	 lr=0.041948
Epoch[123] Batch [1249]/[3760]	Speed: 117.804793 samples/sec	 accuracy=40.740000	 loss=2.650861	 lr=0.041935
Epoch[123] Batch [1299]/[3760]	Speed: 117.823068 samples/sec	 accuracy=40.727163	 loss=2.651381	 lr=0.041922
Epoch[123] Batch [1349]/[3760]	Speed: 116.782240 samples/sec	 accuracy=40.707176	 loss=2.652069	 lr=0.041910
Epoch[123] Batch [1399]/[3760]	Speed: 117.617352 samples/sec	 accuracy=40.704241	 loss=2.651819	 lr=0.041897
Epoch[123] Batch [1449]/[3760]	Speed: 117.336293 samples/sec	 accuracy=40.713362	 loss=2.652764	 lr=0.041884
Epoch[123] Batch [1499]/[3760]	Speed: 118.275607 samples/sec	 accuracy=40.667708	 loss=2.654693	 lr=0.041872
Epoch[123] Batch [1549]/[3760]	Speed: 117.971448 samples/sec	 accuracy=40.692540	 loss=2.653292	 lr=0.041859
Epoch[123] Batch [1599]/[3760]	Speed: 117.783923 samples/sec	 accuracy=40.728516	 loss=2.652896	 lr=0.041846
Epoch[123] Batch [1649]/[3760]	Speed: 117.724825 samples/sec	 accuracy=40.737689	 loss=2.652749	 lr=0.041833
Epoch[123] Batch [1699]/[3760]	Speed: 117.923863 samples/sec	 accuracy=40.715993	 loss=2.652636	 lr=0.041821
Epoch[123] Batch [1749]/[3760]	Speed: 117.626566 samples/sec	 accuracy=40.708036	 loss=2.651891	 lr=0.041808
Epoch[123] Batch [1799]/[3760]	Speed: 117.571068 samples/sec	 accuracy=40.701389	 loss=2.653271	 lr=0.041795
Epoch[123] Batch [1849]/[3760]	Speed: 118.237534 samples/sec	 accuracy=40.700169	 loss=2.653628	 lr=0.041782
Epoch[123] Batch [1899]/[3760]	Speed: 117.418505 samples/sec	 accuracy=40.731086	 loss=2.653973	 lr=0.041770
Epoch[123] Batch [1949]/[3760]	Speed: 116.514486 samples/sec	 accuracy=40.745994	 loss=2.654124	 lr=0.041757
Epoch[123] Batch [1999]/[3760]	Speed: 117.084423 samples/sec	 accuracy=40.735156	 loss=2.653025	 lr=0.041744
Epoch[123] Batch [2049]/[3760]	Speed: 117.451038 samples/sec	 accuracy=40.693598	 loss=2.654350	 lr=0.041732
Epoch[123] Batch [2099]/[3760]	Speed: 117.623049 samples/sec	 accuracy=40.657738	 loss=2.656102	 lr=0.041719
Epoch[123] Batch [2149]/[3760]	Speed: 117.934256 samples/sec	 accuracy=40.660610	 loss=2.656649	 lr=0.041706
Epoch[123] Batch [2199]/[3760]	Speed: 117.747327 samples/sec	 accuracy=40.650568	 loss=2.656686	 lr=0.041693
Epoch[123] Batch [2249]/[3760]	Speed: 117.885023 samples/sec	 accuracy=40.667361	 loss=2.655697	 lr=0.041681
Epoch[123] Batch [2299]/[3760]	Speed: 118.274459 samples/sec	 accuracy=40.639946	 loss=2.657695	 lr=0.041668
Epoch[123] Batch [2349]/[3760]	Speed: 117.537453 samples/sec	 accuracy=40.656915	 loss=2.657166	 lr=0.041655
Epoch[123] Batch [2399]/[3760]	Speed: 116.671130 samples/sec	 accuracy=40.635417	 loss=2.658685	 lr=0.041643
Epoch[123] Batch [2449]/[3760]	Speed: 117.701573 samples/sec	 accuracy=40.600765	 loss=2.659948	 lr=0.041630
Epoch[123] Batch [2499]/[3760]	Speed: 117.104083 samples/sec	 accuracy=40.598125	 loss=2.659702	 lr=0.041617
Epoch[123] Batch [2549]/[3760]	Speed: 117.096511 samples/sec	 accuracy=40.615809	 loss=2.659148	 lr=0.041604
Epoch[123] Batch [2599]/[3760]	Speed: 117.024106 samples/sec	 accuracy=40.604567	 loss=2.659950	 lr=0.041592
Epoch[123] Batch [2649]/[3760]	Speed: 116.629738 samples/sec	 accuracy=40.596698	 loss=2.660983	 lr=0.041579
Epoch[123] Batch [2699]/[3760]	Speed: 117.400946 samples/sec	 accuracy=40.590278	 loss=2.661434	 lr=0.041566
Epoch[123] Batch [2749]/[3760]	Speed: 117.169392 samples/sec	 accuracy=40.584091	 loss=2.660905	 lr=0.041554
Epoch[123] Batch [2799]/[3760]	Speed: 117.716535 samples/sec	 accuracy=40.590960	 loss=2.661412	 lr=0.041541
Epoch[123] Batch [2849]/[3760]	Speed: 117.337967 samples/sec	 accuracy=40.608553	 loss=2.661392	 lr=0.041528
Epoch[123] Batch [2899]/[3760]	Speed: 117.767383 samples/sec	 accuracy=40.596444	 loss=2.661591	 lr=0.041515
Epoch[123] Batch [2949]/[3760]	Speed: 118.233947 samples/sec	 accuracy=40.620763	 loss=2.660469	 lr=0.041503
Epoch[123] Batch [2999]/[3760]	Speed: 116.373910 samples/sec	 accuracy=40.615104	 loss=2.661022	 lr=0.041490
Epoch[123] Batch [3049]/[3760]	Speed: 117.178544 samples/sec	 accuracy=40.578381	 loss=2.662307	 lr=0.041477
Epoch[123] Batch [3099]/[3760]	Speed: 117.172910 samples/sec	 accuracy=40.559476	 loss=2.663337	 lr=0.041465
Epoch[123] Batch [3149]/[3760]	Speed: 118.010232 samples/sec	 accuracy=40.571429	 loss=2.663286	 lr=0.041452
Epoch[123] Batch [3199]/[3760]	Speed: 117.589083 samples/sec	 accuracy=40.568848	 loss=2.662792	 lr=0.041439
Epoch[123] Batch [3249]/[3760]	Speed: 117.727686 samples/sec	 accuracy=40.548558	 loss=2.663372	 lr=0.041426
Epoch[123] Batch [3299]/[3760]	Speed: 117.579973 samples/sec	 accuracy=40.517992	 loss=2.664961	 lr=0.041414
Epoch[123] Batch [3349]/[3760]	Speed: 117.353422 samples/sec	 accuracy=40.522388	 loss=2.664763	 lr=0.041401
Epoch[123] Batch [3399]/[3760]	Speed: 117.258406 samples/sec	 accuracy=40.519301	 loss=2.665134	 lr=0.041388
Epoch[123] Batch [3449]/[3760]	Speed: 117.589165 samples/sec	 accuracy=40.515851	 loss=2.665099	 lr=0.041376
Epoch[123] Batch [3499]/[3760]	Speed: 116.563935 samples/sec	 accuracy=40.508482	 loss=2.665443	 lr=0.041363
Epoch[123] Batch [3549]/[3760]	Speed: 117.206749 samples/sec	 accuracy=40.497359	 loss=2.665890	 lr=0.041350
Epoch[123] Batch [3599]/[3760]	Speed: 116.867487 samples/sec	 accuracy=40.501736	 loss=2.665303	 lr=0.041338
Epoch[123] Batch [3649]/[3760]	Speed: 117.424969 samples/sec	 accuracy=40.488870	 loss=2.665620	 lr=0.041325
Epoch[123] Batch [3699]/[3760]	Speed: 117.805901 samples/sec	 accuracy=40.495355	 loss=2.665178	 lr=0.041312
Epoch[123] Batch [3749]/[3760]	Speed: 120.745730 samples/sec	 accuracy=40.500417	 loss=2.665432	 lr=0.041299
Batch [0049]/[0303]: acc-top1=40.312500 acc-top5=66.093750
Batch [0099]/[0303]: acc-top1=37.750000 acc-top5=64.156250
Batch [0149]/[0303]: acc-top1=39.916667 acc-top5=65.791667
Batch [0199]/[0303]: acc-top1=42.718750 acc-top5=67.960938
Batch [0249]/[0303]: acc-top1=42.187500 acc-top5=67.818750
Batch [0299]/[0303]: acc-top1=41.484375 acc-top5=67.776042
[Epoch 123] training: accuracy=40.497839	 loss=2.665745
[Epoch 123] speed: 116 samples/sec	time cost: 2174.429915
[Epoch 123] validation: acc-top1=41.403672 acc-top5=67.749587 loss=2.689133
Epoch[124] Batch [0049]/[3760]	Speed: 86.831120 samples/sec	 accuracy=41.531250	 loss=2.621807	 lr=0.041284
Epoch[124] Batch [0099]/[3760]	Speed: 118.254134 samples/sec	 accuracy=41.406250	 loss=2.627155	 lr=0.041271
Epoch[124] Batch [0149]/[3760]	Speed: 117.256791 samples/sec	 accuracy=41.270833	 loss=2.619169	 lr=0.041259
Epoch[124] Batch [0199]/[3760]	Speed: 117.100379 samples/sec	 accuracy=41.351562	 loss=2.613937	 lr=0.041246
Epoch[124] Batch [0249]/[3760]	Speed: 117.066617 samples/sec	 accuracy=41.306250	 loss=2.620464	 lr=0.041233
Epoch[124] Batch [0299]/[3760]	Speed: 116.127075 samples/sec	 accuracy=41.453125	 loss=2.614114	 lr=0.041221
Epoch[124] Batch [0349]/[3760]	Speed: 117.084105 samples/sec	 accuracy=41.482143	 loss=2.610469	 lr=0.041208
Epoch[124] Batch [0399]/[3760]	Speed: 117.262779 samples/sec	 accuracy=41.316406	 loss=2.618656	 lr=0.041195
Epoch[124] Batch [0449]/[3760]	Speed: 117.391190 samples/sec	 accuracy=41.218750	 loss=2.619348	 lr=0.041183
Epoch[124] Batch [0499]/[3760]	Speed: 117.348169 samples/sec	 accuracy=41.287500	 loss=2.619264	 lr=0.041170
Epoch[124] Batch [0549]/[3760]	Speed: 117.615999 samples/sec	 accuracy=41.355114	 loss=2.618986	 lr=0.041157
Epoch[124] Batch [0599]/[3760]	Speed: 116.750273 samples/sec	 accuracy=41.270833	 loss=2.623148	 lr=0.041145
Epoch[124] Batch [0649]/[3760]	Speed: 117.102854 samples/sec	 accuracy=41.322115	 loss=2.621674	 lr=0.041132
Epoch[124] Batch [0699]/[3760]	Speed: 117.183010 samples/sec	 accuracy=41.238839	 loss=2.625453	 lr=0.041119
Epoch[124] Batch [0749]/[3760]	Speed: 117.493257 samples/sec	 accuracy=41.181250	 loss=2.628791	 lr=0.041106
Epoch[124] Batch [0799]/[3760]	Speed: 117.349781 samples/sec	 accuracy=41.113281	 loss=2.631139	 lr=0.041094
Epoch[124] Batch [0849]/[3760]	Speed: 117.281592 samples/sec	 accuracy=41.112132	 loss=2.633003	 lr=0.041081
Epoch[124] Batch [0899]/[3760]	Speed: 117.396414 samples/sec	 accuracy=41.020833	 loss=2.637281	 lr=0.041068
Epoch[124] Batch [0949]/[3760]	Speed: 117.316339 samples/sec	 accuracy=40.986842	 loss=2.636438	 lr=0.041056
Epoch[124] Batch [0999]/[3760]	Speed: 117.242004 samples/sec	 accuracy=40.990625	 loss=2.635771	 lr=0.041043
Epoch[124] Batch [1049]/[3760]	Speed: 117.704783 samples/sec	 accuracy=40.947917	 loss=2.636451	 lr=0.041030
Epoch[124] Batch [1099]/[3760]	Speed: 116.965446 samples/sec	 accuracy=40.913352	 loss=2.637826	 lr=0.041018
Epoch[124] Batch [1149]/[3760]	Speed: 117.124864 samples/sec	 accuracy=40.861413	 loss=2.638657	 lr=0.041005
Epoch[124] Batch [1199]/[3760]	Speed: 117.554749 samples/sec	 accuracy=40.816406	 loss=2.640264	 lr=0.040992
Epoch[124] Batch [1249]/[3760]	Speed: 117.351179 samples/sec	 accuracy=40.896250	 loss=2.637583	 lr=0.040980
Epoch[124] Batch [1299]/[3760]	Speed: 117.743351 samples/sec	 accuracy=40.854567	 loss=2.639397	 lr=0.040967
Epoch[124] Batch [1349]/[3760]	Speed: 115.084673 samples/sec	 accuracy=40.859954	 loss=2.638132	 lr=0.040954
Epoch[124] Batch [1399]/[3760]	Speed: 117.502324 samples/sec	 accuracy=40.805804	 loss=2.640234	 lr=0.040941
Epoch[124] Batch [1449]/[3760]	Speed: 117.308186 samples/sec	 accuracy=40.770474	 loss=2.643623	 lr=0.040929
Epoch[124] Batch [1499]/[3760]	Speed: 117.315379 samples/sec	 accuracy=40.746875	 loss=2.644705	 lr=0.040916
Epoch[124] Batch [1549]/[3760]	Speed: 117.391754 samples/sec	 accuracy=40.706653	 loss=2.647496	 lr=0.040903
Epoch[124] Batch [1599]/[3760]	Speed: 116.967080 samples/sec	 accuracy=40.726562	 loss=2.646845	 lr=0.040891
Epoch[124] Batch [1649]/[3760]	Speed: 116.711362 samples/sec	 accuracy=40.742424	 loss=2.647003	 lr=0.040878
Epoch[124] Batch [1699]/[3760]	Speed: 117.185487 samples/sec	 accuracy=40.738051	 loss=2.647655	 lr=0.040865
Epoch[124] Batch [1749]/[3760]	Speed: 116.632261 samples/sec	 accuracy=40.730357	 loss=2.650093	 lr=0.040853
Epoch[124] Batch [1799]/[3760]	Speed: 117.109991 samples/sec	 accuracy=40.731771	 loss=2.649735	 lr=0.040840
Epoch[124] Batch [1849]/[3760]	Speed: 117.561103 samples/sec	 accuracy=40.729730	 loss=2.648791	 lr=0.040827
Epoch[124] Batch [1899]/[3760]	Speed: 116.906756 samples/sec	 accuracy=40.744243	 loss=2.648640	 lr=0.040815
Epoch[124] Batch [1949]/[3760]	Speed: 117.566299 samples/sec	 accuracy=40.721955	 loss=2.649350	 lr=0.040802
Epoch[124] Batch [1999]/[3760]	Speed: 117.210661 samples/sec	 accuracy=40.735938	 loss=2.648069	 lr=0.040789
Epoch[124] Batch [2049]/[3760]	Speed: 117.285381 samples/sec	 accuracy=40.736280	 loss=2.647701	 lr=0.040777
Epoch[124] Batch [2099]/[3760]	Speed: 117.298896 samples/sec	 accuracy=40.732143	 loss=2.648182	 lr=0.040764
Epoch[124] Batch [2149]/[3760]	Speed: 117.742549 samples/sec	 accuracy=40.740552	 loss=2.648489	 lr=0.040751
Epoch[124] Batch [2199]/[3760]	Speed: 117.796641 samples/sec	 accuracy=40.735795	 loss=2.649075	 lr=0.040739
Epoch[124] Batch [2249]/[3760]	Speed: 117.665716 samples/sec	 accuracy=40.712500	 loss=2.649249	 lr=0.040726
Epoch[124] Batch [2299]/[3760]	Speed: 117.874048 samples/sec	 accuracy=40.702446	 loss=2.650267	 lr=0.040713
Epoch[124] Batch [2349]/[3760]	Speed: 117.279302 samples/sec	 accuracy=40.708112	 loss=2.651171	 lr=0.040701
Epoch[124] Batch [2399]/[3760]	Speed: 117.965943 samples/sec	 accuracy=40.692708	 loss=2.653046	 lr=0.040688
Epoch[124] Batch [2449]/[3760]	Speed: 115.683848 samples/sec	 accuracy=40.714286	 loss=2.653078	 lr=0.040675
Epoch[124] Batch [2499]/[3760]	Speed: 117.352265 samples/sec	 accuracy=40.740625	 loss=2.652533	 lr=0.040663
Epoch[124] Batch [2549]/[3760]	Speed: 117.128247 samples/sec	 accuracy=40.752451	 loss=2.652295	 lr=0.040650
Epoch[124] Batch [2599]/[3760]	Speed: 117.990394 samples/sec	 accuracy=40.764423	 loss=2.652123	 lr=0.040637
Epoch[124] Batch [2649]/[3760]	Speed: 116.996248 samples/sec	 accuracy=40.765330	 loss=2.651202	 lr=0.040625
Epoch[124] Batch [2699]/[3760]	Speed: 118.035765 samples/sec	 accuracy=40.802662	 loss=2.650352	 lr=0.040612
Epoch[124] Batch [2749]/[3760]	Speed: 117.694304 samples/sec	 accuracy=40.806818	 loss=2.649664	 lr=0.040599
Epoch[124] Batch [2799]/[3760]	Speed: 116.941184 samples/sec	 accuracy=40.806362	 loss=2.650101	 lr=0.040587
Epoch[124] Batch [2849]/[3760]	Speed: 117.431713 samples/sec	 accuracy=40.797149	 loss=2.649707	 lr=0.040574
Epoch[124] Batch [2899]/[3760]	Speed: 117.485767 samples/sec	 accuracy=40.793642	 loss=2.650041	 lr=0.040561
Epoch[124] Batch [2949]/[3760]	Speed: 116.783257 samples/sec	 accuracy=40.786547	 loss=2.650703	 lr=0.040549
Epoch[124] Batch [2999]/[3760]	Speed: 116.918189 samples/sec	 accuracy=40.776563	 loss=2.650918	 lr=0.040536
Epoch[124] Batch [3049]/[3760]	Speed: 117.040032 samples/sec	 accuracy=40.772029	 loss=2.651224	 lr=0.040523
Epoch[124] Batch [3099]/[3760]	Speed: 117.850527 samples/sec	 accuracy=40.763105	 loss=2.652112	 lr=0.040511
Epoch[124] Batch [3149]/[3760]	Speed: 117.001158 samples/sec	 accuracy=40.751488	 loss=2.653006	 lr=0.040498
Epoch[124] Batch [3199]/[3760]	Speed: 117.038616 samples/sec	 accuracy=40.764160	 loss=2.652930	 lr=0.040485
Epoch[124] Batch [3249]/[3760]	Speed: 117.282325 samples/sec	 accuracy=40.782212	 loss=2.651781	 lr=0.040473
Epoch[124] Batch [3299]/[3760]	Speed: 117.513612 samples/sec	 accuracy=40.793087	 loss=2.651932	 lr=0.040460
Epoch[124] Batch [3349]/[3760]	Speed: 117.050600 samples/sec	 accuracy=40.817164	 loss=2.650921	 lr=0.040447
Epoch[124] Batch [3399]/[3760]	Speed: 117.356257 samples/sec	 accuracy=40.824449	 loss=2.650738	 lr=0.040435
Epoch[124] Batch [3449]/[3760]	Speed: 117.675606 samples/sec	 accuracy=40.829710	 loss=2.650548	 lr=0.040422
Epoch[124] Batch [3499]/[3760]	Speed: 117.401637 samples/sec	 accuracy=40.833929	 loss=2.650181	 lr=0.040409
Epoch[124] Batch [3549]/[3760]	Speed: 115.714622 samples/sec	 accuracy=40.827905	 loss=2.650812	 lr=0.040397
Epoch[124] Batch [3599]/[3760]	Speed: 118.026502 samples/sec	 accuracy=40.813802	 loss=2.651328	 lr=0.040384
Epoch[124] Batch [3649]/[3760]	Speed: 116.901738 samples/sec	 accuracy=40.825771	 loss=2.650937	 lr=0.040371
Epoch[124] Batch [3699]/[3760]	Speed: 117.696319 samples/sec	 accuracy=40.846284	 loss=2.650361	 lr=0.040359
Epoch[124] Batch [3749]/[3760]	Speed: 120.445811 samples/sec	 accuracy=40.854167	 loss=2.650247	 lr=0.040346
Batch [0049]/[0303]: acc-top1=40.093750 acc-top5=64.281250
Batch [0099]/[0303]: acc-top1=36.843750 acc-top5=62.671875
Batch [0149]/[0303]: acc-top1=39.906250 acc-top5=64.802083
Batch [0199]/[0303]: acc-top1=41.867188 acc-top5=66.695312
Batch [0249]/[0303]: acc-top1=41.168750 acc-top5=66.800000
Batch [0299]/[0303]: acc-top1=40.656250 acc-top5=67.078125
[Epoch 124] training: accuracy=40.848155	 loss=2.650472
[Epoch 124] speed: 116 samples/sec	time cost: 2179.382046
[Epoch 124] validation: acc-top1=40.645627 acc-top5=67.156559 loss=2.739740
Epoch[125] Batch [0049]/[3759]	Speed: 87.727677 samples/sec	 accuracy=41.281250	 loss=2.632003	 lr=0.040331
Epoch[125] Batch [0099]/[3759]	Speed: 116.753509 samples/sec	 accuracy=41.109375	 loss=2.621466	 lr=0.040318
Epoch[125] Batch [0149]/[3759]	Speed: 116.781160 samples/sec	 accuracy=41.177083	 loss=2.615713	 lr=0.040306
Epoch[125] Batch [0199]/[3759]	Speed: 117.418124 samples/sec	 accuracy=41.125000	 loss=2.621105	 lr=0.040293
Epoch[125] Batch [0249]/[3759]	Speed: 116.812097 samples/sec	 accuracy=40.906250	 loss=2.629476	 lr=0.040280
Epoch[125] Batch [0299]/[3759]	Speed: 117.489591 samples/sec	 accuracy=40.656250	 loss=2.640938	 lr=0.040268
Epoch[125] Batch [0349]/[3759]	Speed: 116.692298 samples/sec	 accuracy=40.955357	 loss=2.628010	 lr=0.040255
Epoch[125] Batch [0399]/[3759]	Speed: 117.623595 samples/sec	 accuracy=41.121094	 loss=2.624672	 lr=0.040242
Epoch[125] Batch [0449]/[3759]	Speed: 117.132673 samples/sec	 accuracy=41.104167	 loss=2.628516	 lr=0.040230
Epoch[125] Batch [0499]/[3759]	Speed: 116.556874 samples/sec	 accuracy=41.143750	 loss=2.626641	 lr=0.040217
Epoch[125] Batch [0549]/[3759]	Speed: 118.132985 samples/sec	 accuracy=41.247159	 loss=2.624196	 lr=0.040204
Epoch[125] Batch [0599]/[3759]	Speed: 116.901628 samples/sec	 accuracy=41.257812	 loss=2.625094	 lr=0.040192
Epoch[125] Batch [0649]/[3759]	Speed: 117.569255 samples/sec	 accuracy=41.286058	 loss=2.622699	 lr=0.040179
Epoch[125] Batch [0699]/[3759]	Speed: 117.836769 samples/sec	 accuracy=41.305804	 loss=2.623273	 lr=0.040166
Epoch[125] Batch [0749]/[3759]	Speed: 115.457664 samples/sec	 accuracy=41.314583	 loss=2.624853	 lr=0.040154
Epoch[125] Batch [0799]/[3759]	Speed: 117.924557 samples/sec	 accuracy=41.255859	 loss=2.627890	 lr=0.040141
Epoch[125] Batch [0849]/[3759]	Speed: 117.011821 samples/sec	 accuracy=41.165441	 loss=2.631781	 lr=0.040128
Epoch[125] Batch [0899]/[3759]	Speed: 117.129813 samples/sec	 accuracy=41.215278	 loss=2.633201	 lr=0.040116
Epoch[125] Batch [0949]/[3759]	Speed: 117.042275 samples/sec	 accuracy=41.175987	 loss=2.634762	 lr=0.040103
Epoch[125] Batch [0999]/[3759]	Speed: 117.160298 samples/sec	 accuracy=41.154687	 loss=2.635328	 lr=0.040091
Epoch[125] Batch [1049]/[3759]	Speed: 117.118345 samples/sec	 accuracy=41.212798	 loss=2.632639	 lr=0.040078
Epoch[125] Batch [1099]/[3759]	Speed: 117.160983 samples/sec	 accuracy=41.198864	 loss=2.633575	 lr=0.040065
Epoch[125] Batch [1149]/[3759]	Speed: 117.923166 samples/sec	 accuracy=41.144022	 loss=2.636139	 lr=0.040053
Epoch[125] Batch [1199]/[3759]	Speed: 117.500794 samples/sec	 accuracy=41.104167	 loss=2.636492	 lr=0.040040
Epoch[125] Batch [1249]/[3759]	Speed: 117.530675 samples/sec	 accuracy=41.121250	 loss=2.636682	 lr=0.040027
Epoch[125] Batch [1299]/[3759]	Speed: 117.354350 samples/sec	 accuracy=41.067308	 loss=2.638436	 lr=0.040015
Epoch[125] Batch [1349]/[3759]	Speed: 118.488547 samples/sec	 accuracy=41.097222	 loss=2.636655	 lr=0.040002
Epoch[125] Batch [1399]/[3759]	Speed: 116.901663 samples/sec	 accuracy=41.059152	 loss=2.638117	 lr=0.039989
Epoch[125] Batch [1449]/[3759]	Speed: 117.563822 samples/sec	 accuracy=41.056034	 loss=2.638002	 lr=0.039977
Epoch[125] Batch [1499]/[3759]	Speed: 116.921720 samples/sec	 accuracy=41.023958	 loss=2.639794	 lr=0.039964
Epoch[125] Batch [1549]/[3759]	Speed: 117.238557 samples/sec	 accuracy=41.029234	 loss=2.639783	 lr=0.039952
Epoch[125] Batch [1599]/[3759]	Speed: 117.641557 samples/sec	 accuracy=41.047852	 loss=2.639578	 lr=0.039939
Epoch[125] Batch [1649]/[3759]	Speed: 118.034553 samples/sec	 accuracy=41.000000	 loss=2.642023	 lr=0.039926
Epoch[125] Batch [1699]/[3759]	Speed: 117.308127 samples/sec	 accuracy=40.989890	 loss=2.642304	 lr=0.039914
Epoch[125] Batch [1749]/[3759]	Speed: 117.422093 samples/sec	 accuracy=41.000000	 loss=2.641386	 lr=0.039901
Epoch[125] Batch [1799]/[3759]	Speed: 116.751543 samples/sec	 accuracy=41.017361	 loss=2.642380	 lr=0.039888
Epoch[125] Batch [1849]/[3759]	Speed: 115.569638 samples/sec	 accuracy=41.039696	 loss=2.641786	 lr=0.039876
Epoch[125] Batch [1899]/[3759]	Speed: 117.691348 samples/sec	 accuracy=41.040296	 loss=2.641972	 lr=0.039863
Epoch[125] Batch [1949]/[3759]	Speed: 117.620656 samples/sec	 accuracy=41.021635	 loss=2.641204	 lr=0.039850
Epoch[125] Batch [1999]/[3759]	Speed: 117.539006 samples/sec	 accuracy=41.005469	 loss=2.642253	 lr=0.039838
Epoch[125] Batch [2049]/[3759]	Speed: 117.337337 samples/sec	 accuracy=41.020579	 loss=2.642725	 lr=0.039825
Epoch[125] Batch [2099]/[3759]	Speed: 117.234865 samples/sec	 accuracy=41.010417	 loss=2.642851	 lr=0.039813
Epoch[125] Batch [2149]/[3759]	Speed: 117.554372 samples/sec	 accuracy=41.055233	 loss=2.641358	 lr=0.039800
Epoch[125] Batch [2199]/[3759]	Speed: 117.930665 samples/sec	 accuracy=41.029830	 loss=2.642684	 lr=0.039787
Epoch[125] Batch [2249]/[3759]	Speed: 117.329149 samples/sec	 accuracy=41.034028	 loss=2.642330	 lr=0.039775
Epoch[125] Batch [2299]/[3759]	Speed: 117.509373 samples/sec	 accuracy=41.043478	 loss=2.643275	 lr=0.039762
Epoch[125] Batch [2349]/[3759]	Speed: 117.495057 samples/sec	 accuracy=41.055851	 loss=2.643344	 lr=0.039749
Epoch[125] Batch [2399]/[3759]	Speed: 117.619338 samples/sec	 accuracy=41.046875	 loss=2.643046	 lr=0.039737
Epoch[125] Batch [2449]/[3759]	Speed: 117.385436 samples/sec	 accuracy=41.058036	 loss=2.642535	 lr=0.039724
Epoch[125] Batch [2499]/[3759]	Speed: 116.947328 samples/sec	 accuracy=41.061250	 loss=2.642004	 lr=0.039712
Epoch[125] Batch [2549]/[3759]	Speed: 117.421547 samples/sec	 accuracy=41.062500	 loss=2.642475	 lr=0.039699
Epoch[125] Batch [2599]/[3759]	Speed: 118.302337 samples/sec	 accuracy=41.062500	 loss=2.642516	 lr=0.039686
Epoch[125] Batch [2649]/[3759]	Speed: 117.301517 samples/sec	 accuracy=41.057193	 loss=2.643054	 lr=0.039674
Epoch[125] Batch [2699]/[3759]	Speed: 117.324973 samples/sec	 accuracy=41.050926	 loss=2.644547	 lr=0.039661
Epoch[125] Batch [2749]/[3759]	Speed: 117.121122 samples/sec	 accuracy=41.102841	 loss=2.642663	 lr=0.039648
Epoch[125] Batch [2799]/[3759]	Speed: 117.071331 samples/sec	 accuracy=41.082589	 loss=2.642916	 lr=0.039636
Epoch[125] Batch [2849]/[3759]	Speed: 117.580935 samples/sec	 accuracy=41.089364	 loss=2.642974	 lr=0.039623
Epoch[125] Batch [2899]/[3759]	Speed: 117.367771 samples/sec	 accuracy=41.088901	 loss=2.642667	 lr=0.039611
Epoch[125] Batch [2949]/[3759]	Speed: 115.811996 samples/sec	 accuracy=41.074682	 loss=2.643272	 lr=0.039598
Epoch[125] Batch [2999]/[3759]	Speed: 117.354506 samples/sec	 accuracy=41.056771	 loss=2.644240	 lr=0.039585
Epoch[125] Batch [3049]/[3759]	Speed: 116.967945 samples/sec	 accuracy=41.050717	 loss=2.644217	 lr=0.039573
Epoch[125] Batch [3099]/[3759]	Speed: 117.122016 samples/sec	 accuracy=41.071069	 loss=2.644086	 lr=0.039560
Epoch[125] Batch [3149]/[3759]	Speed: 117.158924 samples/sec	 accuracy=41.090278	 loss=2.642851	 lr=0.039548
Epoch[125] Batch [3199]/[3759]	Speed: 117.649084 samples/sec	 accuracy=41.077148	 loss=2.643878	 lr=0.039535
Epoch[125] Batch [3249]/[3759]	Speed: 117.361128 samples/sec	 accuracy=41.050962	 loss=2.645755	 lr=0.039522
Epoch[125] Batch [3299]/[3759]	Speed: 116.957596 samples/sec	 accuracy=41.046402	 loss=2.645828	 lr=0.039510
Epoch[125] Batch [3349]/[3759]	Speed: 117.011809 samples/sec	 accuracy=41.047108	 loss=2.645944	 lr=0.039497
Epoch[125] Batch [3399]/[3759]	Speed: 117.475899 samples/sec	 accuracy=41.021599	 loss=2.647335	 lr=0.039485
Epoch[125] Batch [3449]/[3759]	Speed: 117.322551 samples/sec	 accuracy=41.023098	 loss=2.646970	 lr=0.039472
Epoch[125] Batch [3499]/[3759]	Speed: 117.407880 samples/sec	 accuracy=41.031250	 loss=2.646348	 lr=0.039459
Epoch[125] Batch [3549]/[3759]	Speed: 117.306440 samples/sec	 accuracy=41.022887	 loss=2.646321	 lr=0.039447
Epoch[125] Batch [3599]/[3759]	Speed: 117.544817 samples/sec	 accuracy=41.015625	 loss=2.646413	 lr=0.039434
Epoch[125] Batch [3649]/[3759]	Speed: 118.154880 samples/sec	 accuracy=41.046233	 loss=2.644648	 lr=0.039421
Epoch[125] Batch [3699]/[3759]	Speed: 117.477789 samples/sec	 accuracy=41.059966	 loss=2.644390	 lr=0.039409
Epoch[125] Batch [3749]/[3759]	Speed: 120.535351 samples/sec	 accuracy=41.057917	 loss=2.644480	 lr=0.039396
Batch [0049]/[0303]: acc-top1=43.281250 acc-top5=66.843750
Batch [0099]/[0303]: acc-top1=39.312500 acc-top5=65.406250
Batch [0149]/[0303]: acc-top1=41.697917 acc-top5=67.041667
Batch [0199]/[0303]: acc-top1=44.351562 acc-top5=69.109375
Batch [0249]/[0303]: acc-top1=43.431250 acc-top5=68.662500
Batch [0299]/[0303]: acc-top1=42.697917 acc-top5=68.671875
[Epoch 125] training: accuracy=41.050229	 loss=2.644645
[Epoch 125] speed: 116 samples/sec	time cost: 2177.703343
[Epoch 125] validation: acc-top1=42.682550 acc-top5=68.693276 loss=2.670130
Epoch[126] Batch [0049]/[3760]	Speed: 88.307944 samples/sec	 accuracy=41.968750	 loss=2.632854	 lr=0.039381
Epoch[126] Batch [0099]/[3760]	Speed: 117.377914 samples/sec	 accuracy=41.921875	 loss=2.630087	 lr=0.039369
Epoch[126] Batch [0149]/[3760]	Speed: 116.913604 samples/sec	 accuracy=41.927083	 loss=2.613030	 lr=0.039356
Epoch[126] Batch [0199]/[3760]	Speed: 117.722177 samples/sec	 accuracy=41.765625	 loss=2.612909	 lr=0.039344
Epoch[126] Batch [0249]/[3760]	Speed: 117.880088 samples/sec	 accuracy=41.850000	 loss=2.602931	 lr=0.039331
Epoch[126] Batch [0299]/[3760]	Speed: 115.975249 samples/sec	 accuracy=41.947917	 loss=2.596993	 lr=0.039318
Epoch[126] Batch [0349]/[3760]	Speed: 117.842330 samples/sec	 accuracy=42.022321	 loss=2.590633	 lr=0.039306
Epoch[126] Batch [0399]/[3760]	Speed: 116.782129 samples/sec	 accuracy=42.058594	 loss=2.592558	 lr=0.039293
Epoch[126] Batch [0449]/[3760]	Speed: 116.879149 samples/sec	 accuracy=42.041667	 loss=2.593356	 lr=0.039281
Epoch[126] Batch [0499]/[3760]	Speed: 116.865505 samples/sec	 accuracy=41.846875	 loss=2.603503	 lr=0.039268
Epoch[126] Batch [0549]/[3760]	Speed: 117.029540 samples/sec	 accuracy=41.735795	 loss=2.609366	 lr=0.039255
Epoch[126] Batch [0599]/[3760]	Speed: 117.270994 samples/sec	 accuracy=41.736979	 loss=2.611358	 lr=0.039243
Epoch[126] Batch [0649]/[3760]	Speed: 117.219377 samples/sec	 accuracy=41.723558	 loss=2.609227	 lr=0.039230
Epoch[126] Batch [0699]/[3760]	Speed: 117.037658 samples/sec	 accuracy=41.741071	 loss=2.608677	 lr=0.039218
Epoch[126] Batch [0749]/[3760]	Speed: 116.850488 samples/sec	 accuracy=41.677083	 loss=2.612831	 lr=0.039205
Epoch[126] Batch [0799]/[3760]	Speed: 117.342939 samples/sec	 accuracy=41.714844	 loss=2.612101	 lr=0.039192
Epoch[126] Batch [0849]/[3760]	Speed: 117.420734 samples/sec	 accuracy=41.727941	 loss=2.609380	 lr=0.039180
Epoch[126] Batch [0899]/[3760]	Speed: 117.811972 samples/sec	 accuracy=41.585069	 loss=2.616270	 lr=0.039167
Epoch[126] Batch [0949]/[3760]	Speed: 117.324248 samples/sec	 accuracy=41.600329	 loss=2.614792	 lr=0.039155
Epoch[126] Batch [0999]/[3760]	Speed: 116.995634 samples/sec	 accuracy=41.623438	 loss=2.612088	 lr=0.039142
Epoch[126] Batch [1049]/[3760]	Speed: 117.706324 samples/sec	 accuracy=41.647321	 loss=2.613685	 lr=0.039129
Epoch[126] Batch [1099]/[3760]	Speed: 117.361290 samples/sec	 accuracy=41.613636	 loss=2.614814	 lr=0.039117
Epoch[126] Batch [1149]/[3760]	Speed: 117.643547 samples/sec	 accuracy=41.548913	 loss=2.617604	 lr=0.039104
Epoch[126] Batch [1199]/[3760]	Speed: 117.606390 samples/sec	 accuracy=41.492188	 loss=2.620542	 lr=0.039092
Epoch[126] Batch [1249]/[3760]	Speed: 117.337673 samples/sec	 accuracy=41.507500	 loss=2.618672	 lr=0.039079
Epoch[126] Batch [1299]/[3760]	Speed: 117.528672 samples/sec	 accuracy=41.537260	 loss=2.617921	 lr=0.039067
Epoch[126] Batch [1349]/[3760]	Speed: 117.610290 samples/sec	 accuracy=41.539352	 loss=2.616634	 lr=0.039054
Epoch[126] Batch [1399]/[3760]	Speed: 116.049823 samples/sec	 accuracy=41.505580	 loss=2.617661	 lr=0.039041
Epoch[126] Batch [1449]/[3760]	Speed: 116.837097 samples/sec	 accuracy=41.478448	 loss=2.616364	 lr=0.039029
Epoch[126] Batch [1499]/[3760]	Speed: 117.685821 samples/sec	 accuracy=41.446875	 loss=2.615651	 lr=0.039016
Epoch[126] Batch [1549]/[3760]	Speed: 116.936443 samples/sec	 accuracy=41.460685	 loss=2.614468	 lr=0.039004
Epoch[126] Batch [1599]/[3760]	Speed: 117.307430 samples/sec	 accuracy=41.488281	 loss=2.613977	 lr=0.038991
Epoch[126] Batch [1649]/[3760]	Speed: 116.373659 samples/sec	 accuracy=41.503788	 loss=2.614059	 lr=0.038978
Epoch[126] Batch [1699]/[3760]	Speed: 118.049307 samples/sec	 accuracy=41.500919	 loss=2.614709	 lr=0.038966
Epoch[126] Batch [1749]/[3760]	Speed: 117.503383 samples/sec	 accuracy=41.479464	 loss=2.616223	 lr=0.038953
Epoch[126] Batch [1799]/[3760]	Speed: 117.250677 samples/sec	 accuracy=41.453993	 loss=2.616834	 lr=0.038941
Epoch[126] Batch [1849]/[3760]	Speed: 117.684578 samples/sec	 accuracy=41.431588	 loss=2.618743	 lr=0.038928
Epoch[126] Batch [1899]/[3760]	Speed: 117.587535 samples/sec	 accuracy=41.443257	 loss=2.619309	 lr=0.038916
Epoch[126] Batch [1949]/[3760]	Speed: 117.236659 samples/sec	 accuracy=41.431090	 loss=2.620367	 lr=0.038903
Epoch[126] Batch [1999]/[3760]	Speed: 117.342078 samples/sec	 accuracy=41.388281	 loss=2.621409	 lr=0.038890
Epoch[126] Batch [2049]/[3760]	Speed: 117.334913 samples/sec	 accuracy=41.385671	 loss=2.620711	 lr=0.038878
Epoch[126] Batch [2099]/[3760]	Speed: 117.715356 samples/sec	 accuracy=41.343750	 loss=2.622451	 lr=0.038865
Epoch[126] Batch [2149]/[3760]	Speed: 118.028901 samples/sec	 accuracy=41.340116	 loss=2.622162	 lr=0.038853
Epoch[126] Batch [2199]/[3760]	Speed: 116.462395 samples/sec	 accuracy=41.335938	 loss=2.622166	 lr=0.038840
Epoch[126] Batch [2249]/[3760]	Speed: 117.281627 samples/sec	 accuracy=41.337500	 loss=2.622569	 lr=0.038828
Epoch[126] Batch [2299]/[3760]	Speed: 117.325279 samples/sec	 accuracy=41.328804	 loss=2.622630	 lr=0.038815
Epoch[126] Batch [2349]/[3760]	Speed: 117.196485 samples/sec	 accuracy=41.336436	 loss=2.621645	 lr=0.038802
Epoch[126] Batch [2399]/[3760]	Speed: 117.241016 samples/sec	 accuracy=41.351562	 loss=2.622023	 lr=0.038790
Epoch[126] Batch [2449]/[3760]	Speed: 116.850416 samples/sec	 accuracy=41.343112	 loss=2.622099	 lr=0.038777
Epoch[126] Batch [2499]/[3760]	Speed: 117.689790 samples/sec	 accuracy=41.366250	 loss=2.620992	 lr=0.038765
Epoch[126] Batch [2549]/[3760]	Speed: 116.255907 samples/sec	 accuracy=41.375000	 loss=2.621426	 lr=0.038752
Epoch[126] Batch [2599]/[3760]	Speed: 117.849654 samples/sec	 accuracy=41.362380	 loss=2.621387	 lr=0.038740
Epoch[126] Batch [2649]/[3760]	Speed: 117.204421 samples/sec	 accuracy=41.370873	 loss=2.620905	 lr=0.038727
Epoch[126] Batch [2699]/[3760]	Speed: 116.695444 samples/sec	 accuracy=41.369792	 loss=2.620771	 lr=0.038714
Epoch[126] Batch [2749]/[3760]	Speed: 117.232840 samples/sec	 accuracy=41.388636	 loss=2.620398	 lr=0.038702
Epoch[126] Batch [2799]/[3760]	Speed: 117.789017 samples/sec	 accuracy=41.390067	 loss=2.619898	 lr=0.038689
Epoch[126] Batch [2849]/[3760]	Speed: 117.268553 samples/sec	 accuracy=41.385417	 loss=2.619937	 lr=0.038677
Epoch[126] Batch [2899]/[3760]	Speed: 117.600803 samples/sec	 accuracy=41.380927	 loss=2.620701	 lr=0.038664
Epoch[126] Batch [2949]/[3760]	Speed: 117.419172 samples/sec	 accuracy=41.382415	 loss=2.621420	 lr=0.038652
Epoch[126] Batch [2999]/[3760]	Speed: 117.700127 samples/sec	 accuracy=41.358333	 loss=2.622311	 lr=0.038639
Epoch[126] Batch [3049]/[3760]	Speed: 117.462628 samples/sec	 accuracy=41.344775	 loss=2.622660	 lr=0.038626
Epoch[126] Batch [3099]/[3760]	Speed: 117.411135 samples/sec	 accuracy=41.317036	 loss=2.623386	 lr=0.038614
Epoch[126] Batch [3149]/[3760]	Speed: 117.455624 samples/sec	 accuracy=41.291667	 loss=2.624690	 lr=0.038601
Epoch[126] Batch [3199]/[3760]	Speed: 117.612383 samples/sec	 accuracy=41.296387	 loss=2.624781	 lr=0.038589
Epoch[126] Batch [3249]/[3760]	Speed: 117.460928 samples/sec	 accuracy=41.264904	 loss=2.625680	 lr=0.038576
Epoch[126] Batch [3299]/[3760]	Speed: 117.341583 samples/sec	 accuracy=41.283144	 loss=2.625308	 lr=0.038564
Epoch[126] Batch [3349]/[3760]	Speed: 117.179424 samples/sec	 accuracy=41.256063	 loss=2.625990	 lr=0.038551
Epoch[126] Batch [3399]/[3760]	Speed: 117.219522 samples/sec	 accuracy=41.270680	 loss=2.625223	 lr=0.038539
Epoch[126] Batch [3449]/[3760]	Speed: 117.558058 samples/sec	 accuracy=41.274004	 loss=2.625134	 lr=0.038526
Epoch[126] Batch [3499]/[3760]	Speed: 118.146176 samples/sec	 accuracy=41.283036	 loss=2.625025	 lr=0.038513
Epoch[126] Batch [3549]/[3760]	Speed: 118.316444 samples/sec	 accuracy=41.286972	 loss=2.624963	 lr=0.038501
Epoch[126] Batch [3599]/[3760]	Speed: 117.843196 samples/sec	 accuracy=41.284288	 loss=2.624928	 lr=0.038488
Epoch[126] Batch [3649]/[3760]	Speed: 116.346873 samples/sec	 accuracy=41.275685	 loss=2.625946	 lr=0.038476
Epoch[126] Batch [3699]/[3760]	Speed: 117.324949 samples/sec	 accuracy=41.279139	 loss=2.626040	 lr=0.038463
Epoch[126] Batch [3749]/[3760]	Speed: 119.859160 samples/sec	 accuracy=41.248750	 loss=2.627243	 lr=0.038451
Batch [0049]/[0303]: acc-top1=41.687500 acc-top5=66.250000
Batch [0099]/[0303]: acc-top1=38.859375 acc-top5=64.875000
Batch [0149]/[0303]: acc-top1=41.020833 acc-top5=66.687500
Batch [0199]/[0303]: acc-top1=43.156250 acc-top5=68.140625
Batch [0249]/[0303]: acc-top1=42.237500 acc-top5=67.662500
Batch [0299]/[0303]: acc-top1=41.625000 acc-top5=68.031250
[Epoch 126] training: accuracy=41.247091	 loss=2.627302
[Epoch 126] speed: 116 samples/sec	time cost: 2176.736992
[Epoch 126] validation: acc-top1=41.640883 acc-top5=68.043523 loss=2.632012
Epoch[127] Batch [0049]/[3760]	Speed: 86.809531 samples/sec	 accuracy=41.750000	 loss=2.594165	 lr=0.038436
Epoch[127] Batch [0099]/[3760]	Speed: 117.616078 samples/sec	 accuracy=41.703125	 loss=2.613103	 lr=0.038423
Epoch[127] Batch [0149]/[3760]	Speed: 117.741487 samples/sec	 accuracy=41.635417	 loss=2.601956	 lr=0.038411
Epoch[127] Batch [0199]/[3760]	Speed: 117.684441 samples/sec	 accuracy=41.632812	 loss=2.593392	 lr=0.038398
Epoch[127] Batch [0249]/[3760]	Speed: 116.787675 samples/sec	 accuracy=41.600000	 loss=2.595514	 lr=0.038385
Epoch[127] Batch [0299]/[3760]	Speed: 117.216416 samples/sec	 accuracy=41.739583	 loss=2.594377	 lr=0.038373
Epoch[127] Batch [0349]/[3760]	Speed: 117.101929 samples/sec	 accuracy=41.723214	 loss=2.591125	 lr=0.038360
Epoch[127] Batch [0399]/[3760]	Speed: 116.916780 samples/sec	 accuracy=41.628906	 loss=2.597169	 lr=0.038348
Epoch[127] Batch [0449]/[3760]	Speed: 117.933696 samples/sec	 accuracy=41.531250	 loss=2.599840	 lr=0.038335
Epoch[127] Batch [0499]/[3760]	Speed: 116.912631 samples/sec	 accuracy=41.615625	 loss=2.600276	 lr=0.038323
Epoch[127] Batch [0549]/[3760]	Speed: 116.931351 samples/sec	 accuracy=41.582386	 loss=2.600337	 lr=0.038310
Epoch[127] Batch [0599]/[3760]	Speed: 117.811934 samples/sec	 accuracy=41.567708	 loss=2.601265	 lr=0.038298
Epoch[127] Batch [0649]/[3760]	Speed: 117.727791 samples/sec	 accuracy=41.497596	 loss=2.605480	 lr=0.038285
Epoch[127] Batch [0699]/[3760]	Speed: 117.726888 samples/sec	 accuracy=41.441964	 loss=2.607422	 lr=0.038273
Epoch[127] Batch [0749]/[3760]	Speed: 116.702932 samples/sec	 accuracy=41.518750	 loss=2.602464	 lr=0.038260
Epoch[127] Batch [0799]/[3760]	Speed: 117.556933 samples/sec	 accuracy=41.572266	 loss=2.599678	 lr=0.038248
Epoch[127] Batch [0849]/[3760]	Speed: 117.714369 samples/sec	 accuracy=41.560662	 loss=2.600735	 lr=0.038235
Epoch[127] Batch [0899]/[3760]	Speed: 117.300402 samples/sec	 accuracy=41.513889	 loss=2.602686	 lr=0.038222
Epoch[127] Batch [0949]/[3760]	Speed: 117.012953 samples/sec	 accuracy=41.509868	 loss=2.606217	 lr=0.038210
Epoch[127] Batch [0999]/[3760]	Speed: 116.075628 samples/sec	 accuracy=41.579687	 loss=2.606306	 lr=0.038197
Epoch[127] Batch [1049]/[3760]	Speed: 117.808690 samples/sec	 accuracy=41.552083	 loss=2.610058	 lr=0.038185
Epoch[127] Batch [1099]/[3760]	Speed: 117.135850 samples/sec	 accuracy=41.607955	 loss=2.607458	 lr=0.038172
Epoch[127] Batch [1149]/[3760]	Speed: 118.039080 samples/sec	 accuracy=41.581522	 loss=2.609029	 lr=0.038160
Epoch[127] Batch [1199]/[3760]	Speed: 117.141166 samples/sec	 accuracy=41.578125	 loss=2.608686	 lr=0.038147
Epoch[127] Batch [1249]/[3760]	Speed: 118.045814 samples/sec	 accuracy=41.566250	 loss=2.610620	 lr=0.038135
Epoch[127] Batch [1299]/[3760]	Speed: 117.951889 samples/sec	 accuracy=41.585337	 loss=2.610972	 lr=0.038122
Epoch[127] Batch [1349]/[3760]	Speed: 118.142315 samples/sec	 accuracy=41.570602	 loss=2.612513	 lr=0.038110
Epoch[127] Batch [1399]/[3760]	Speed: 117.550923 samples/sec	 accuracy=41.623884	 loss=2.609993	 lr=0.038097
Epoch[127] Batch [1449]/[3760]	Speed: 118.189674 samples/sec	 accuracy=41.589440	 loss=2.610527	 lr=0.038085
Epoch[127] Batch [1499]/[3760]	Speed: 117.313733 samples/sec	 accuracy=41.561458	 loss=2.611348	 lr=0.038072
Epoch[127] Batch [1549]/[3760]	Speed: 117.411777 samples/sec	 accuracy=41.562500	 loss=2.611943	 lr=0.038060
Epoch[127] Batch [1599]/[3760]	Speed: 117.691560 samples/sec	 accuracy=41.599609	 loss=2.611573	 lr=0.038047
Epoch[127] Batch [1649]/[3760]	Speed: 117.197985 samples/sec	 accuracy=41.578598	 loss=2.612538	 lr=0.038035
Epoch[127] Batch [1699]/[3760]	Speed: 117.595236 samples/sec	 accuracy=41.573529	 loss=2.611951	 lr=0.038022
Epoch[127] Batch [1749]/[3760]	Speed: 118.219573 samples/sec	 accuracy=41.590179	 loss=2.612348	 lr=0.038010
Epoch[127] Batch [1799]/[3760]	Speed: 117.209072 samples/sec	 accuracy=41.584201	 loss=2.612996	 lr=0.037997
Epoch[127] Batch [1849]/[3760]	Speed: 118.278140 samples/sec	 accuracy=41.580236	 loss=2.612722	 lr=0.037984
Epoch[127] Batch [1899]/[3760]	Speed: 117.162719 samples/sec	 accuracy=41.594572	 loss=2.611540	 lr=0.037972
Epoch[127] Batch [1949]/[3760]	Speed: 117.959154 samples/sec	 accuracy=41.584936	 loss=2.611717	 lr=0.037959
Epoch[127] Batch [1999]/[3760]	Speed: 117.600151 samples/sec	 accuracy=41.558594	 loss=2.612094	 lr=0.037947
Epoch[127] Batch [2049]/[3760]	Speed: 117.352677 samples/sec	 accuracy=41.502287	 loss=2.614332	 lr=0.037934
Epoch[127] Batch [2099]/[3760]	Speed: 116.474268 samples/sec	 accuracy=41.488839	 loss=2.615615	 lr=0.037922
Epoch[127] Batch [2149]/[3760]	Speed: 117.440717 samples/sec	 accuracy=41.479651	 loss=2.616101	 lr=0.037909
Epoch[127] Batch [2199]/[3760]	Speed: 116.930770 samples/sec	 accuracy=41.456676	 loss=2.617680	 lr=0.037897
Epoch[127] Batch [2249]/[3760]	Speed: 117.782507 samples/sec	 accuracy=41.409028	 loss=2.618676	 lr=0.037884
Epoch[127] Batch [2299]/[3760]	Speed: 118.347636 samples/sec	 accuracy=41.439538	 loss=2.617521	 lr=0.037872
Epoch[127] Batch [2349]/[3760]	Speed: 117.948002 samples/sec	 accuracy=41.435505	 loss=2.617982	 lr=0.037859
Epoch[127] Batch [2399]/[3760]	Speed: 117.821210 samples/sec	 accuracy=41.473307	 loss=2.617520	 lr=0.037847
Epoch[127] Batch [2449]/[3760]	Speed: 118.069300 samples/sec	 accuracy=41.464286	 loss=2.616876	 lr=0.037834
Epoch[127] Batch [2499]/[3760]	Speed: 117.895130 samples/sec	 accuracy=41.428125	 loss=2.618545	 lr=0.037822
Epoch[127] Batch [2549]/[3760]	Speed: 117.116785 samples/sec	 accuracy=41.449755	 loss=2.617717	 lr=0.037809
Epoch[127] Batch [2599]/[3760]	Speed: 118.251481 samples/sec	 accuracy=41.441707	 loss=2.618097	 lr=0.037797
Epoch[127] Batch [2649]/[3760]	Speed: 118.210031 samples/sec	 accuracy=41.416863	 loss=2.619720	 lr=0.037784
Epoch[127] Batch [2699]/[3760]	Speed: 117.341027 samples/sec	 accuracy=41.440972	 loss=2.618464	 lr=0.037772
Epoch[127] Batch [2749]/[3760]	Speed: 117.826309 samples/sec	 accuracy=41.430114	 loss=2.619855	 lr=0.037759
Epoch[127] Batch [2799]/[3760]	Speed: 117.870104 samples/sec	 accuracy=41.423549	 loss=2.620502	 lr=0.037747
Epoch[127] Batch [2849]/[3760]	Speed: 117.337799 samples/sec	 accuracy=41.413377	 loss=2.620758	 lr=0.037734
Epoch[127] Batch [2899]/[3760]	Speed: 117.774873 samples/sec	 accuracy=41.452047	 loss=2.619545	 lr=0.037722
Epoch[127] Batch [2949]/[3760]	Speed: 117.126734 samples/sec	 accuracy=41.451271	 loss=2.619188	 lr=0.037709
Epoch[127] Batch [2999]/[3760]	Speed: 117.218258 samples/sec	 accuracy=41.441146	 loss=2.619151	 lr=0.037697
Epoch[127] Batch [3049]/[3760]	Speed: 117.596086 samples/sec	 accuracy=41.449795	 loss=2.618900	 lr=0.037684
Epoch[127] Batch [3099]/[3760]	Speed: 117.274078 samples/sec	 accuracy=41.409274	 loss=2.620494	 lr=0.037672
Epoch[127] Batch [3149]/[3760]	Speed: 117.347165 samples/sec	 accuracy=41.419147	 loss=2.620411	 lr=0.037659
Epoch[127] Batch [3199]/[3760]	Speed: 115.945678 samples/sec	 accuracy=41.418457	 loss=2.619829	 lr=0.037647
Epoch[127] Batch [3249]/[3760]	Speed: 117.882462 samples/sec	 accuracy=41.420192	 loss=2.619754	 lr=0.037634
Epoch[127] Batch [3299]/[3760]	Speed: 118.232502 samples/sec	 accuracy=41.430871	 loss=2.618688	 lr=0.037622
Epoch[127] Batch [3349]/[3760]	Speed: 116.667147 samples/sec	 accuracy=41.429571	 loss=2.617955	 lr=0.037609
Epoch[127] Batch [3399]/[3760]	Speed: 117.179941 samples/sec	 accuracy=41.454504	 loss=2.616976	 lr=0.037597
Epoch[127] Batch [3449]/[3760]	Speed: 117.833743 samples/sec	 accuracy=41.472826	 loss=2.615687	 lr=0.037584
Epoch[127] Batch [3499]/[3760]	Speed: 118.085882 samples/sec	 accuracy=41.484821	 loss=2.615604	 lr=0.037572
Epoch[127] Batch [3549]/[3760]	Speed: 117.487588 samples/sec	 accuracy=41.495599	 loss=2.614288	 lr=0.037559
Epoch[127] Batch [3599]/[3760]	Speed: 116.957318 samples/sec	 accuracy=41.522569	 loss=2.613188	 lr=0.037547
Epoch[127] Batch [3649]/[3760]	Speed: 117.606783 samples/sec	 accuracy=41.511986	 loss=2.613513	 lr=0.037534
Epoch[127] Batch [3699]/[3760]	Speed: 117.342592 samples/sec	 accuracy=41.523226	 loss=2.613369	 lr=0.037522
Epoch[127] Batch [3749]/[3760]	Speed: 120.274259 samples/sec	 accuracy=41.513750	 loss=2.613949	 lr=0.037509
Batch [0049]/[0303]: acc-top1=43.593750 acc-top5=68.437500
Batch [0099]/[0303]: acc-top1=40.187500 acc-top5=66.171875
Batch [0149]/[0303]: acc-top1=41.125000 acc-top5=67.187500
Batch [0199]/[0303]: acc-top1=43.687500 acc-top5=68.890625
Batch [0249]/[0303]: acc-top1=43.093750 acc-top5=68.612500
Batch [0299]/[0303]: acc-top1=43.062500 acc-top5=68.953125
[Epoch 127] training: accuracy=41.512217	 loss=2.613756
[Epoch 127] speed: 116 samples/sec	time cost: 2175.046158
[Epoch 127] validation: acc-top1=43.012583 acc-top5=68.951114 loss=2.605731
Epoch[128] Batch [0049]/[3759]	Speed: 86.167256 samples/sec	 accuracy=42.531250	 loss=2.590353	 lr=0.037494
Epoch[128] Batch [0099]/[3759]	Speed: 117.486615 samples/sec	 accuracy=41.968750	 loss=2.615700	 lr=0.037482
Epoch[128] Batch [0149]/[3759]	Speed: 117.809872 samples/sec	 accuracy=42.072917	 loss=2.594118	 lr=0.037469
Epoch[128] Batch [0199]/[3759]	Speed: 117.306365 samples/sec	 accuracy=41.882812	 loss=2.602995	 lr=0.037457
Epoch[128] Batch [0249]/[3759]	Speed: 118.026679 samples/sec	 accuracy=41.962500	 loss=2.598662	 lr=0.037444
Epoch[128] Batch [0299]/[3759]	Speed: 116.995976 samples/sec	 accuracy=42.010417	 loss=2.596298	 lr=0.037432
Epoch[128] Batch [0349]/[3759]	Speed: 117.573337 samples/sec	 accuracy=41.973214	 loss=2.598673	 lr=0.037419
Epoch[128] Batch [0399]/[3759]	Speed: 118.162073 samples/sec	 accuracy=42.097656	 loss=2.594856	 lr=0.037407
Epoch[128] Batch [0449]/[3759]	Speed: 117.731192 samples/sec	 accuracy=42.052083	 loss=2.604481	 lr=0.037394
Epoch[128] Batch [0499]/[3759]	Speed: 116.103423 samples/sec	 accuracy=42.134375	 loss=2.598478	 lr=0.037382
Epoch[128] Batch [0549]/[3759]	Speed: 117.732882 samples/sec	 accuracy=42.122159	 loss=2.599005	 lr=0.037369
Epoch[128] Batch [0599]/[3759]	Speed: 116.955872 samples/sec	 accuracy=42.166667	 loss=2.594329	 lr=0.037357
Epoch[128] Batch [0649]/[3759]	Speed: 117.205917 samples/sec	 accuracy=42.194712	 loss=2.590091	 lr=0.037345
Epoch[128] Batch [0699]/[3759]	Speed: 117.355144 samples/sec	 accuracy=42.238839	 loss=2.588805	 lr=0.037332
Epoch[128] Batch [0749]/[3759]	Speed: 117.224669 samples/sec	 accuracy=42.283333	 loss=2.586063	 lr=0.037320
Epoch[128] Batch [0799]/[3759]	Speed: 116.842623 samples/sec	 accuracy=42.242188	 loss=2.586988	 lr=0.037307
Epoch[128] Batch [0849]/[3759]	Speed: 116.607634 samples/sec	 accuracy=42.121324	 loss=2.589753	 lr=0.037295
Epoch[128] Batch [0899]/[3759]	Speed: 118.181128 samples/sec	 accuracy=42.071181	 loss=2.587291	 lr=0.037282
Epoch[128] Batch [0949]/[3759]	Speed: 117.268794 samples/sec	 accuracy=42.115132	 loss=2.584596	 lr=0.037270
Epoch[128] Batch [0999]/[3759]	Speed: 117.544165 samples/sec	 accuracy=42.046875	 loss=2.586962	 lr=0.037257
Epoch[128] Batch [1049]/[3759]	Speed: 116.943075 samples/sec	 accuracy=42.049107	 loss=2.587565	 lr=0.037245
Epoch[128] Batch [1099]/[3759]	Speed: 117.447470 samples/sec	 accuracy=42.068182	 loss=2.586072	 lr=0.037232
Epoch[128] Batch [1149]/[3759]	Speed: 117.619895 samples/sec	 accuracy=42.001359	 loss=2.588462	 lr=0.037220
Epoch[128] Batch [1199]/[3759]	Speed: 116.825628 samples/sec	 accuracy=42.035156	 loss=2.588601	 lr=0.037207
Epoch[128] Batch [1249]/[3759]	Speed: 118.192385 samples/sec	 accuracy=42.020000	 loss=2.588396	 lr=0.037195
Epoch[128] Batch [1299]/[3759]	Speed: 117.435057 samples/sec	 accuracy=41.956731	 loss=2.590516	 lr=0.037182
Epoch[128] Batch [1349]/[3759]	Speed: 118.171737 samples/sec	 accuracy=41.929398	 loss=2.593519	 lr=0.037170
Epoch[128] Batch [1399]/[3759]	Speed: 116.912965 samples/sec	 accuracy=41.963170	 loss=2.593124	 lr=0.037157
Epoch[128] Batch [1449]/[3759]	Speed: 117.609066 samples/sec	 accuracy=41.954741	 loss=2.594143	 lr=0.037145
Epoch[128] Batch [1499]/[3759]	Speed: 117.663555 samples/sec	 accuracy=41.958333	 loss=2.594316	 lr=0.037133
Epoch[128] Batch [1549]/[3759]	Speed: 116.928230 samples/sec	 accuracy=41.965726	 loss=2.594889	 lr=0.037120
Epoch[128] Batch [1599]/[3759]	Speed: 116.059389 samples/sec	 accuracy=41.989258	 loss=2.593619	 lr=0.037108
Epoch[128] Batch [1649]/[3759]	Speed: 117.536896 samples/sec	 accuracy=41.992424	 loss=2.594306	 lr=0.037095
Epoch[128] Batch [1699]/[3759]	Speed: 117.951956 samples/sec	 accuracy=41.946691	 loss=2.596574	 lr=0.037083
Epoch[128] Batch [1749]/[3759]	Speed: 117.253984 samples/sec	 accuracy=41.937500	 loss=2.595306	 lr=0.037070
Epoch[128] Batch [1799]/[3759]	Speed: 117.377312 samples/sec	 accuracy=41.940972	 loss=2.595197	 lr=0.037058
Epoch[128] Batch [1849]/[3759]	Speed: 117.542596 samples/sec	 accuracy=41.941723	 loss=2.595229	 lr=0.037045
Epoch[128] Batch [1899]/[3759]	Speed: 118.080240 samples/sec	 accuracy=41.948191	 loss=2.594948	 lr=0.037033
Epoch[128] Batch [1949]/[3759]	Speed: 117.743096 samples/sec	 accuracy=41.954327	 loss=2.594657	 lr=0.037020
Epoch[128] Batch [1999]/[3759]	Speed: 117.678686 samples/sec	 accuracy=41.960156	 loss=2.593394	 lr=0.037008
Epoch[128] Batch [2049]/[3759]	Speed: 116.865875 samples/sec	 accuracy=41.991616	 loss=2.592225	 lr=0.036996
Epoch[128] Batch [2099]/[3759]	Speed: 117.611913 samples/sec	 accuracy=42.002976	 loss=2.592027	 lr=0.036983
Epoch[128] Batch [2149]/[3759]	Speed: 118.219137 samples/sec	 accuracy=41.952035	 loss=2.594018	 lr=0.036971
Epoch[128] Batch [2199]/[3759]	Speed: 117.834549 samples/sec	 accuracy=41.987216	 loss=2.592119	 lr=0.036958
Epoch[128] Batch [2249]/[3759]	Speed: 117.591627 samples/sec	 accuracy=42.019444	 loss=2.592265	 lr=0.036946
Epoch[128] Batch [2299]/[3759]	Speed: 117.793732 samples/sec	 accuracy=42.015625	 loss=2.593673	 lr=0.036933
Epoch[128] Batch [2349]/[3759]	Speed: 117.515149 samples/sec	 accuracy=42.019282	 loss=2.593319	 lr=0.036921
Epoch[128] Batch [2399]/[3759]	Speed: 118.033963 samples/sec	 accuracy=41.975260	 loss=2.593787	 lr=0.036908
Epoch[128] Batch [2449]/[3759]	Speed: 116.946973 samples/sec	 accuracy=41.987245	 loss=2.593431	 lr=0.036896
Epoch[128] Batch [2499]/[3759]	Speed: 117.935712 samples/sec	 accuracy=42.008750	 loss=2.593035	 lr=0.036883
Epoch[128] Batch [2549]/[3759]	Speed: 117.450997 samples/sec	 accuracy=42.025735	 loss=2.592944	 lr=0.036871
Epoch[128] Batch [2599]/[3759]	Speed: 116.540890 samples/sec	 accuracy=42.002404	 loss=2.592579	 lr=0.036859
Epoch[128] Batch [2649]/[3759]	Speed: 117.902553 samples/sec	 accuracy=42.009434	 loss=2.591506	 lr=0.036846
Epoch[128] Batch [2699]/[3759]	Speed: 117.498717 samples/sec	 accuracy=41.997685	 loss=2.592843	 lr=0.036834
Epoch[128] Batch [2749]/[3759]	Speed: 116.658410 samples/sec	 accuracy=41.993750	 loss=2.592815	 lr=0.036821
Epoch[128] Batch [2799]/[3759]	Speed: 117.588102 samples/sec	 accuracy=41.969866	 loss=2.594193	 lr=0.036809
Epoch[128] Batch [2849]/[3759]	Speed: 117.704771 samples/sec	 accuracy=41.953399	 loss=2.595118	 lr=0.036796
Epoch[128] Batch [2899]/[3759]	Speed: 117.469576 samples/sec	 accuracy=41.961746	 loss=2.595533	 lr=0.036784
Epoch[128] Batch [2949]/[3759]	Speed: 118.176980 samples/sec	 accuracy=41.956038	 loss=2.595600	 lr=0.036771
Epoch[128] Batch [2999]/[3759]	Speed: 117.569202 samples/sec	 accuracy=41.976042	 loss=2.594607	 lr=0.036759
Epoch[128] Batch [3049]/[3759]	Speed: 118.051540 samples/sec	 accuracy=41.975410	 loss=2.594770	 lr=0.036747
Epoch[128] Batch [3099]/[3759]	Speed: 117.528122 samples/sec	 accuracy=41.941028	 loss=2.596196	 lr=0.036734
Epoch[128] Batch [3149]/[3759]	Speed: 117.761485 samples/sec	 accuracy=41.936012	 loss=2.596506	 lr=0.036722
Epoch[128] Batch [3199]/[3759]	Speed: 117.633129 samples/sec	 accuracy=41.928223	 loss=2.596610	 lr=0.036709
Epoch[128] Batch [3249]/[3759]	Speed: 117.906295 samples/sec	 accuracy=41.922596	 loss=2.596824	 lr=0.036697
Epoch[128] Batch [3299]/[3759]	Speed: 117.411367 samples/sec	 accuracy=41.928977	 loss=2.596300	 lr=0.036684
Epoch[128] Batch [3349]/[3759]	Speed: 117.993465 samples/sec	 accuracy=41.934235	 loss=2.596742	 lr=0.036672
Epoch[128] Batch [3399]/[3759]	Speed: 117.694619 samples/sec	 accuracy=41.938419	 loss=2.597122	 lr=0.036660
Epoch[128] Batch [3449]/[3759]	Speed: 117.942683 samples/sec	 accuracy=41.937953	 loss=2.597099	 lr=0.036647
Epoch[128] Batch [3499]/[3759]	Speed: 117.641761 samples/sec	 accuracy=41.948661	 loss=2.597582	 lr=0.036635
Epoch[128] Batch [3549]/[3759]	Speed: 118.008745 samples/sec	 accuracy=41.962148	 loss=2.597197	 lr=0.036622
Epoch[128] Batch [3599]/[3759]	Speed: 117.688465 samples/sec	 accuracy=41.976128	 loss=2.596469	 lr=0.036610
Epoch[128] Batch [3649]/[3759]	Speed: 117.447822 samples/sec	 accuracy=41.963613	 loss=2.596349	 lr=0.036597
Epoch[128] Batch [3699]/[3759]	Speed: 117.718042 samples/sec	 accuracy=41.972128	 loss=2.595632	 lr=0.036585
Epoch[128] Batch [3749]/[3759]	Speed: 120.440999 samples/sec	 accuracy=41.962917	 loss=2.596235	 lr=0.036573
Batch [0049]/[0303]: acc-top1=41.250000 acc-top5=67.000000
Batch [0099]/[0303]: acc-top1=39.765625 acc-top5=66.234375
Batch [0149]/[0303]: acc-top1=41.541667 acc-top5=67.197917
Batch [0199]/[0303]: acc-top1=44.453125 acc-top5=69.460938
Batch [0249]/[0303]: acc-top1=43.587500 acc-top5=69.093750
Batch [0299]/[0303]: acc-top1=42.817708 acc-top5=68.979167
[Epoch 128] training: accuracy=41.969689	 loss=2.596054
[Epoch 128] speed: 116 samples/sec	time cost: 2174.067894
[Epoch 128] validation: acc-top1=42.677393 acc-top5=68.951114 loss=2.654515
Epoch[129] Batch [0049]/[3760]	Speed: 87.626008 samples/sec	 accuracy=44.218750	 loss=2.498345	 lr=0.036558
Epoch[129] Batch [0099]/[3760]	Speed: 116.038154 samples/sec	 accuracy=43.734375	 loss=2.522326	 lr=0.036546
Epoch[129] Batch [0149]/[3760]	Speed: 118.100795 samples/sec	 accuracy=42.864583	 loss=2.571569	 lr=0.036533
Epoch[129] Batch [0199]/[3760]	Speed: 117.526323 samples/sec	 accuracy=42.867188	 loss=2.572160	 lr=0.036521
Epoch[129] Batch [0249]/[3760]	Speed: 117.678346 samples/sec	 accuracy=42.812500	 loss=2.570125	 lr=0.036508
Epoch[129] Batch [0299]/[3760]	Speed: 117.709265 samples/sec	 accuracy=42.802083	 loss=2.569637	 lr=0.036496
Epoch[129] Batch [0349]/[3760]	Speed: 118.325704 samples/sec	 accuracy=42.437500	 loss=2.583567	 lr=0.036483
Epoch[129] Batch [0399]/[3760]	Speed: 117.084848 samples/sec	 accuracy=42.414062	 loss=2.578148	 lr=0.036471
Epoch[129] Batch [0449]/[3760]	Speed: 117.303311 samples/sec	 accuracy=42.364583	 loss=2.579903	 lr=0.036459
Epoch[129] Batch [0499]/[3760]	Speed: 116.481302 samples/sec	 accuracy=42.381250	 loss=2.575086	 lr=0.036446
Epoch[129] Batch [0549]/[3760]	Speed: 117.666980 samples/sec	 accuracy=42.414773	 loss=2.573775	 lr=0.036434
Epoch[129] Batch [0599]/[3760]	Speed: 116.721891 samples/sec	 accuracy=42.322917	 loss=2.578103	 lr=0.036421
Epoch[129] Batch [0649]/[3760]	Speed: 117.192329 samples/sec	 accuracy=42.365385	 loss=2.574996	 lr=0.036409
Epoch[129] Batch [0699]/[3760]	Speed: 117.150686 samples/sec	 accuracy=42.337054	 loss=2.576005	 lr=0.036397
Epoch[129] Batch [0749]/[3760]	Speed: 117.629801 samples/sec	 accuracy=42.231250	 loss=2.578395	 lr=0.036384
Epoch[129] Batch [0799]/[3760]	Speed: 116.990695 samples/sec	 accuracy=42.220703	 loss=2.579231	 lr=0.036372
Epoch[129] Batch [0849]/[3760]	Speed: 118.005467 samples/sec	 accuracy=42.266544	 loss=2.579813	 lr=0.036359
Epoch[129] Batch [0899]/[3760]	Speed: 118.022687 samples/sec	 accuracy=42.234375	 loss=2.578324	 lr=0.036347
Epoch[129] Batch [0949]/[3760]	Speed: 117.957365 samples/sec	 accuracy=42.261513	 loss=2.578969	 lr=0.036334
Epoch[129] Batch [0999]/[3760]	Speed: 117.562600 samples/sec	 accuracy=42.304688	 loss=2.577337	 lr=0.036322
Epoch[129] Batch [1049]/[3760]	Speed: 117.628759 samples/sec	 accuracy=42.230655	 loss=2.581272	 lr=0.036310
Epoch[129] Batch [1099]/[3760]	Speed: 117.770979 samples/sec	 accuracy=42.268466	 loss=2.580664	 lr=0.036297
Epoch[129] Batch [1149]/[3760]	Speed: 117.012721 samples/sec	 accuracy=42.224185	 loss=2.581439	 lr=0.036285
Epoch[129] Batch [1199]/[3760]	Speed: 115.390640 samples/sec	 accuracy=42.208333	 loss=2.582961	 lr=0.036272
Epoch[129] Batch [1249]/[3760]	Speed: 117.724796 samples/sec	 accuracy=42.251250	 loss=2.580387	 lr=0.036260
Epoch[129] Batch [1299]/[3760]	Speed: 117.383850 samples/sec	 accuracy=42.252404	 loss=2.579573	 lr=0.036248
Epoch[129] Batch [1349]/[3760]	Speed: 117.778190 samples/sec	 accuracy=42.300926	 loss=2.578357	 lr=0.036235
Epoch[129] Batch [1399]/[3760]	Speed: 117.659124 samples/sec	 accuracy=42.313616	 loss=2.577185	 lr=0.036223
Epoch[129] Batch [1449]/[3760]	Speed: 117.389595 samples/sec	 accuracy=42.280172	 loss=2.577994	 lr=0.036210
Epoch[129] Batch [1499]/[3760]	Speed: 117.384165 samples/sec	 accuracy=42.289583	 loss=2.577542	 lr=0.036198
Epoch[129] Batch [1549]/[3760]	Speed: 117.154632 samples/sec	 accuracy=42.304435	 loss=2.577798	 lr=0.036186
Epoch[129] Batch [1599]/[3760]	Speed: 117.578608 samples/sec	 accuracy=42.349609	 loss=2.577054	 lr=0.036173
Epoch[129] Batch [1649]/[3760]	Speed: 117.740824 samples/sec	 accuracy=42.306818	 loss=2.579355	 lr=0.036161
Epoch[129] Batch [1699]/[3760]	Speed: 117.966208 samples/sec	 accuracy=42.352022	 loss=2.578343	 lr=0.036148
Epoch[129] Batch [1749]/[3760]	Speed: 117.537309 samples/sec	 accuracy=42.359821	 loss=2.577661	 lr=0.036136
Epoch[129] Batch [1799]/[3760]	Speed: 118.078551 samples/sec	 accuracy=42.361111	 loss=2.577063	 lr=0.036124
Epoch[129] Batch [1849]/[3760]	Speed: 117.428746 samples/sec	 accuracy=42.353041	 loss=2.577148	 lr=0.036111
Epoch[129] Batch [1899]/[3760]	Speed: 117.700340 samples/sec	 accuracy=42.348684	 loss=2.577246	 lr=0.036099
Epoch[129] Batch [1949]/[3760]	Speed: 117.106770 samples/sec	 accuracy=42.352564	 loss=2.575923	 lr=0.036087
Epoch[129] Batch [1999]/[3760]	Speed: 117.072380 samples/sec	 accuracy=42.293750	 loss=2.578830	 lr=0.036074
Epoch[129] Batch [2049]/[3760]	Speed: 117.977159 samples/sec	 accuracy=42.341463	 loss=2.577515	 lr=0.036062
Epoch[129] Batch [2099]/[3760]	Speed: 117.257486 samples/sec	 accuracy=42.343750	 loss=2.578166	 lr=0.036049
Epoch[129] Batch [2149]/[3760]	Speed: 118.224663 samples/sec	 accuracy=42.358285	 loss=2.577284	 lr=0.036037
Epoch[129] Batch [2199]/[3760]	Speed: 117.270787 samples/sec	 accuracy=42.306818	 loss=2.579704	 lr=0.036025
Epoch[129] Batch [2249]/[3760]	Speed: 118.047513 samples/sec	 accuracy=42.277083	 loss=2.582120	 lr=0.036012
Epoch[129] Batch [2299]/[3760]	Speed: 117.302005 samples/sec	 accuracy=42.271739	 loss=2.581688	 lr=0.036000
Epoch[129] Batch [2349]/[3760]	Speed: 116.719201 samples/sec	 accuracy=42.255984	 loss=2.582945	 lr=0.035987
Epoch[129] Batch [2399]/[3760]	Speed: 117.436483 samples/sec	 accuracy=42.242839	 loss=2.584108	 lr=0.035975
Epoch[129] Batch [2449]/[3760]	Speed: 117.100597 samples/sec	 accuracy=42.259566	 loss=2.582241	 lr=0.035963
Epoch[129] Batch [2499]/[3760]	Speed: 117.036859 samples/sec	 accuracy=42.231250	 loss=2.582939	 lr=0.035950
Epoch[129] Batch [2549]/[3760]	Speed: 117.913332 samples/sec	 accuracy=42.208333	 loss=2.583565	 lr=0.035938
Epoch[129] Batch [2599]/[3760]	Speed: 118.006538 samples/sec	 accuracy=42.174880	 loss=2.583871	 lr=0.035926
Epoch[129] Batch [2649]/[3760]	Speed: 117.808019 samples/sec	 accuracy=42.167453	 loss=2.583838	 lr=0.035913
Epoch[129] Batch [2699]/[3760]	Speed: 118.091782 samples/sec	 accuracy=42.160301	 loss=2.584001	 lr=0.035901
Epoch[129] Batch [2749]/[3760]	Speed: 117.576183 samples/sec	 accuracy=42.157386	 loss=2.583226	 lr=0.035888
Epoch[129] Batch [2799]/[3760]	Speed: 117.374689 samples/sec	 accuracy=42.137835	 loss=2.583348	 lr=0.035876
Epoch[129] Batch [2849]/[3760]	Speed: 117.885439 samples/sec	 accuracy=42.116776	 loss=2.584269	 lr=0.035864
Epoch[129] Batch [2899]/[3760]	Speed: 117.962776 samples/sec	 accuracy=42.146552	 loss=2.582676	 lr=0.035851
Epoch[129] Batch [2949]/[3760]	Speed: 117.284295 samples/sec	 accuracy=42.141419	 loss=2.583152	 lr=0.035839
Epoch[129] Batch [2999]/[3760]	Speed: 117.825272 samples/sec	 accuracy=42.147396	 loss=2.582240	 lr=0.035827
Epoch[129] Batch [3049]/[3760]	Speed: 118.036575 samples/sec	 accuracy=42.150615	 loss=2.581993	 lr=0.035814
Epoch[129] Batch [3099]/[3760]	Speed: 117.100291 samples/sec	 accuracy=42.163306	 loss=2.581746	 lr=0.035802
Epoch[129] Batch [3149]/[3760]	Speed: 117.935921 samples/sec	 accuracy=42.162698	 loss=2.582534	 lr=0.035789
Epoch[129] Batch [3199]/[3760]	Speed: 117.762206 samples/sec	 accuracy=42.171875	 loss=2.582513	 lr=0.035777
Epoch[129] Batch [3249]/[3760]	Speed: 117.224028 samples/sec	 accuracy=42.166346	 loss=2.582897	 lr=0.035765
Epoch[129] Batch [3299]/[3760]	Speed: 117.252208 samples/sec	 accuracy=42.181345	 loss=2.582363	 lr=0.035752
Epoch[129] Batch [3349]/[3760]	Speed: 117.866299 samples/sec	 accuracy=42.161381	 loss=2.583359	 lr=0.035740
Epoch[129] Batch [3399]/[3760]	Speed: 117.484719 samples/sec	 accuracy=42.174632	 loss=2.583460	 lr=0.035728
Epoch[129] Batch [3449]/[3760]	Speed: 117.502005 samples/sec	 accuracy=42.156250	 loss=2.584128	 lr=0.035715
Epoch[129] Batch [3499]/[3760]	Speed: 115.983142 samples/sec	 accuracy=42.145982	 loss=2.584581	 lr=0.035703
Epoch[129] Batch [3549]/[3760]	Speed: 117.781140 samples/sec	 accuracy=42.132042	 loss=2.584936	 lr=0.035691
Epoch[129] Batch [3599]/[3760]	Speed: 117.533548 samples/sec	 accuracy=42.134983	 loss=2.585275	 lr=0.035678
Epoch[129] Batch [3649]/[3760]	Speed: 117.811921 samples/sec	 accuracy=42.121575	 loss=2.585836	 lr=0.035666
Epoch[129] Batch [3699]/[3760]	Speed: 117.576179 samples/sec	 accuracy=42.119088	 loss=2.586228	 lr=0.035654
Epoch[129] Batch [3749]/[3760]	Speed: 120.209216 samples/sec	 accuracy=42.107083	 loss=2.585816	 lr=0.035641
Batch [0049]/[0303]: acc-top1=43.093750 acc-top5=68.781250
Batch [0099]/[0303]: acc-top1=41.046875 acc-top5=67.609375
Batch [0149]/[0303]: acc-top1=42.572917 acc-top5=68.541667
Batch [0199]/[0303]: acc-top1=45.015625 acc-top5=69.968750
Batch [0249]/[0303]: acc-top1=44.106250 acc-top5=69.243750
Batch [0299]/[0303]: acc-top1=43.458333 acc-top5=69.354167
[Epoch 129] training: accuracy=42.104388	 loss=2.585913
[Epoch 129] speed: 116 samples/sec	time cost: 2174.804461
[Epoch 129] validation: acc-top1=43.419967 acc-top5=69.379125 loss=2.605481
Epoch[130] Batch [0049]/[3759]	Speed: 88.055329 samples/sec	 accuracy=43.562500	 loss=2.530385	 lr=0.035626
Epoch[130] Batch [0099]/[3759]	Speed: 117.641182 samples/sec	 accuracy=42.703125	 loss=2.537593	 lr=0.035614
Epoch[130] Batch [0149]/[3759]	Speed: 117.537310 samples/sec	 accuracy=43.125000	 loss=2.547633	 lr=0.035602
Epoch[130] Batch [0199]/[3759]	Speed: 118.244895 samples/sec	 accuracy=43.007812	 loss=2.550381	 lr=0.035589
Epoch[130] Batch [0249]/[3759]	Speed: 117.658375 samples/sec	 accuracy=42.737500	 loss=2.547417	 lr=0.035577
Epoch[130] Batch [0299]/[3759]	Speed: 118.278083 samples/sec	 accuracy=42.848958	 loss=2.540384	 lr=0.035565
Epoch[130] Batch [0349]/[3759]	Speed: 117.018224 samples/sec	 accuracy=42.866071	 loss=2.537976	 lr=0.035552
Epoch[130] Batch [0399]/[3759]	Speed: 116.914906 samples/sec	 accuracy=42.906250	 loss=2.537339	 lr=0.035540
Epoch[130] Batch [0449]/[3759]	Speed: 117.610552 samples/sec	 accuracy=42.819444	 loss=2.542289	 lr=0.035528
Epoch[130] Batch [0499]/[3759]	Speed: 117.271229 samples/sec	 accuracy=42.771875	 loss=2.543916	 lr=0.035515
Epoch[130] Batch [0549]/[3759]	Speed: 118.501031 samples/sec	 accuracy=42.650568	 loss=2.548813	 lr=0.035503
Epoch[130] Batch [0599]/[3759]	Speed: 116.916984 samples/sec	 accuracy=42.601562	 loss=2.550941	 lr=0.035491
Epoch[130] Batch [0649]/[3759]	Speed: 118.182076 samples/sec	 accuracy=42.639423	 loss=2.546919	 lr=0.035478
Epoch[130] Batch [0699]/[3759]	Speed: 117.030310 samples/sec	 accuracy=42.694196	 loss=2.549606	 lr=0.035466
Epoch[130] Batch [0749]/[3759]	Speed: 118.387703 samples/sec	 accuracy=42.583333	 loss=2.551269	 lr=0.035453
Epoch[130] Batch [0799]/[3759]	Speed: 117.242140 samples/sec	 accuracy=42.558594	 loss=2.555428	 lr=0.035441
Epoch[130] Batch [0849]/[3759]	Speed: 115.715814 samples/sec	 accuracy=42.658088	 loss=2.553023	 lr=0.035429
Epoch[130] Batch [0899]/[3759]	Speed: 117.760733 samples/sec	 accuracy=42.666667	 loss=2.553710	 lr=0.035416
Epoch[130] Batch [0949]/[3759]	Speed: 117.133858 samples/sec	 accuracy=42.682566	 loss=2.552998	 lr=0.035404
Epoch[130] Batch [0999]/[3759]	Speed: 116.779151 samples/sec	 accuracy=42.739063	 loss=2.552442	 lr=0.035392
Epoch[130] Batch [1049]/[3759]	Speed: 117.772257 samples/sec	 accuracy=42.697917	 loss=2.553562	 lr=0.035379
Epoch[130] Batch [1099]/[3759]	Speed: 118.192439 samples/sec	 accuracy=42.744318	 loss=2.551404	 lr=0.035367
Epoch[130] Batch [1149]/[3759]	Speed: 117.542535 samples/sec	 accuracy=42.724185	 loss=2.550915	 lr=0.035355
Epoch[130] Batch [1199]/[3759]	Speed: 117.339607 samples/sec	 accuracy=42.638021	 loss=2.553563	 lr=0.035342
Epoch[130] Batch [1249]/[3759]	Speed: 117.308807 samples/sec	 accuracy=42.636250	 loss=2.555705	 lr=0.035330
Epoch[130] Batch [1299]/[3759]	Speed: 117.433477 samples/sec	 accuracy=42.633413	 loss=2.556296	 lr=0.035318
Epoch[130] Batch [1349]/[3759]	Speed: 117.577507 samples/sec	 accuracy=42.635417	 loss=2.556079	 lr=0.035305
Epoch[130] Batch [1399]/[3759]	Speed: 117.269145 samples/sec	 accuracy=42.602679	 loss=2.558035	 lr=0.035293
Epoch[130] Batch [1449]/[3759]	Speed: 117.547339 samples/sec	 accuracy=42.607759	 loss=2.558458	 lr=0.035281
Epoch[130] Batch [1499]/[3759]	Speed: 116.888287 samples/sec	 accuracy=42.576042	 loss=2.559080	 lr=0.035269
Epoch[130] Batch [1549]/[3759]	Speed: 116.753342 samples/sec	 accuracy=42.565524	 loss=2.557974	 lr=0.035256
Epoch[130] Batch [1599]/[3759]	Speed: 117.317034 samples/sec	 accuracy=42.560547	 loss=2.559091	 lr=0.035244
Epoch[130] Batch [1649]/[3759]	Speed: 117.756109 samples/sec	 accuracy=42.572917	 loss=2.558125	 lr=0.035232
Epoch[130] Batch [1699]/[3759]	Speed: 117.932810 samples/sec	 accuracy=42.568934	 loss=2.558608	 lr=0.035219
Epoch[130] Batch [1749]/[3759]	Speed: 117.809359 samples/sec	 accuracy=42.541964	 loss=2.560707	 lr=0.035207
Epoch[130] Batch [1799]/[3759]	Speed: 117.985845 samples/sec	 accuracy=42.527778	 loss=2.560844	 lr=0.035195
Epoch[130] Batch [1849]/[3759]	Speed: 117.364142 samples/sec	 accuracy=42.485642	 loss=2.562066	 lr=0.035182
Epoch[130] Batch [1899]/[3759]	Speed: 117.473334 samples/sec	 accuracy=42.489309	 loss=2.561942	 lr=0.035170
Epoch[130] Batch [1949]/[3759]	Speed: 117.109352 samples/sec	 accuracy=42.480769	 loss=2.561959	 lr=0.035158
Epoch[130] Batch [1999]/[3759]	Speed: 115.912408 samples/sec	 accuracy=42.498438	 loss=2.560892	 lr=0.035145
Epoch[130] Batch [2049]/[3759]	Speed: 118.094227 samples/sec	 accuracy=42.519055	 loss=2.561034	 lr=0.035133
Epoch[130] Batch [2099]/[3759]	Speed: 117.248725 samples/sec	 accuracy=42.548363	 loss=2.558785	 lr=0.035121
Epoch[130] Batch [2149]/[3759]	Speed: 116.849344 samples/sec	 accuracy=42.547965	 loss=2.559970	 lr=0.035108
Epoch[130] Batch [2199]/[3759]	Speed: 117.950747 samples/sec	 accuracy=42.564631	 loss=2.559323	 lr=0.035096
Epoch[130] Batch [2249]/[3759]	Speed: 117.185795 samples/sec	 accuracy=42.568056	 loss=2.558872	 lr=0.035084
Epoch[130] Batch [2299]/[3759]	Speed: 117.488268 samples/sec	 accuracy=42.552989	 loss=2.560041	 lr=0.035071
Epoch[130] Batch [2349]/[3759]	Speed: 117.900131 samples/sec	 accuracy=42.557846	 loss=2.560245	 lr=0.035059
Epoch[130] Batch [2399]/[3759]	Speed: 117.659595 samples/sec	 accuracy=42.544922	 loss=2.560870	 lr=0.035047
Epoch[130] Batch [2449]/[3759]	Speed: 117.346425 samples/sec	 accuracy=42.529974	 loss=2.561016	 lr=0.035035
Epoch[130] Batch [2499]/[3759]	Speed: 116.977391 samples/sec	 accuracy=42.508125	 loss=2.562295	 lr=0.035022
Epoch[130] Batch [2549]/[3759]	Speed: 117.428837 samples/sec	 accuracy=42.491422	 loss=2.562681	 lr=0.035010
Epoch[130] Batch [2599]/[3759]	Speed: 116.879221 samples/sec	 accuracy=42.492188	 loss=2.563259	 lr=0.034998
Epoch[130] Batch [2649]/[3759]	Speed: 117.228911 samples/sec	 accuracy=42.503538	 loss=2.562571	 lr=0.034985
Epoch[130] Batch [2699]/[3759]	Speed: 116.747358 samples/sec	 accuracy=42.491319	 loss=2.562756	 lr=0.034973
Epoch[130] Batch [2749]/[3759]	Speed: 117.719597 samples/sec	 accuracy=42.473864	 loss=2.563602	 lr=0.034961
Epoch[130] Batch [2799]/[3759]	Speed: 116.990582 samples/sec	 accuracy=42.454799	 loss=2.564038	 lr=0.034948
Epoch[130] Batch [2849]/[3759]	Speed: 117.838786 samples/sec	 accuracy=42.445175	 loss=2.564878	 lr=0.034936
Epoch[130] Batch [2899]/[3759]	Speed: 117.631687 samples/sec	 accuracy=42.427263	 loss=2.565291	 lr=0.034924
Epoch[130] Batch [2949]/[3759]	Speed: 117.109160 samples/sec	 accuracy=42.405720	 loss=2.567069	 lr=0.034911
Epoch[130] Batch [2999]/[3759]	Speed: 117.402074 samples/sec	 accuracy=42.395833	 loss=2.567639	 lr=0.034899
Epoch[130] Batch [3049]/[3759]	Speed: 117.139439 samples/sec	 accuracy=42.403689	 loss=2.567615	 lr=0.034887
Epoch[130] Batch [3099]/[3759]	Speed: 117.216321 samples/sec	 accuracy=42.405746	 loss=2.567261	 lr=0.034875
Epoch[130] Batch [3149]/[3759]	Speed: 117.181792 samples/sec	 accuracy=42.399802	 loss=2.567898	 lr=0.034862
Epoch[130] Batch [3199]/[3759]	Speed: 115.572518 samples/sec	 accuracy=42.407715	 loss=2.567998	 lr=0.034850
Epoch[130] Batch [3249]/[3759]	Speed: 117.331622 samples/sec	 accuracy=42.398077	 loss=2.568425	 lr=0.034838
Epoch[130] Batch [3299]/[3759]	Speed: 117.633934 samples/sec	 accuracy=42.391098	 loss=2.568222	 lr=0.034825
Epoch[130] Batch [3349]/[3759]	Speed: 117.260318 samples/sec	 accuracy=42.386660	 loss=2.568085	 lr=0.034813
Epoch[130] Batch [3399]/[3759]	Speed: 116.820501 samples/sec	 accuracy=42.365349	 loss=2.570005	 lr=0.034801
Epoch[130] Batch [3449]/[3759]	Speed: 117.301896 samples/sec	 accuracy=42.345109	 loss=2.570904	 lr=0.034789
Epoch[130] Batch [3499]/[3759]	Speed: 117.564529 samples/sec	 accuracy=42.362054	 loss=2.571037	 lr=0.034776
Epoch[130] Batch [3549]/[3759]	Speed: 117.773786 samples/sec	 accuracy=42.367077	 loss=2.570830	 lr=0.034764
Epoch[130] Batch [3599]/[3759]	Speed: 116.766366 samples/sec	 accuracy=42.353299	 loss=2.571164	 lr=0.034752
Epoch[130] Batch [3649]/[3759]	Speed: 118.343808 samples/sec	 accuracy=42.357021	 loss=2.571238	 lr=0.034739
Epoch[130] Batch [3699]/[3759]	Speed: 117.586368 samples/sec	 accuracy=42.341216	 loss=2.571906	 lr=0.034727
Epoch[130] Batch [3749]/[3759]	Speed: 120.379051 samples/sec	 accuracy=42.350833	 loss=2.571610	 lr=0.034715
Batch [0049]/[0303]: acc-top1=43.437500 acc-top5=68.250000
Batch [0099]/[0303]: acc-top1=40.796875 acc-top5=66.843750
Batch [0149]/[0303]: acc-top1=42.343750 acc-top5=68.031250
Batch [0199]/[0303]: acc-top1=44.382812 acc-top5=69.562500
Batch [0249]/[0303]: acc-top1=43.662500 acc-top5=69.512500
Batch [0299]/[0303]: acc-top1=43.286458 acc-top5=69.447917
[Epoch 130] training: accuracy=42.362081	 loss=2.571201
[Epoch 130] speed: 116 samples/sec	time cost: 2175.312745
[Epoch 130] validation: acc-top1=43.327145 acc-top5=69.446163 loss=2.642174
Epoch[131] Batch [0049]/[3760]	Speed: 89.120803 samples/sec	 accuracy=42.281250	 loss=2.513938	 lr=0.034700
Epoch[131] Batch [0099]/[3760]	Speed: 117.742526 samples/sec	 accuracy=42.437500	 loss=2.543289	 lr=0.034688
Epoch[131] Batch [0149]/[3760]	Speed: 117.455965 samples/sec	 accuracy=42.281250	 loss=2.538945	 lr=0.034676
Epoch[131] Batch [0199]/[3760]	Speed: 117.594581 samples/sec	 accuracy=42.429688	 loss=2.529445	 lr=0.034664
Epoch[131] Batch [0249]/[3760]	Speed: 117.348112 samples/sec	 accuracy=42.368750	 loss=2.533412	 lr=0.034651
Epoch[131] Batch [0299]/[3760]	Speed: 117.110354 samples/sec	 accuracy=42.395833	 loss=2.542247	 lr=0.034639
Epoch[131] Batch [0349]/[3760]	Speed: 117.502390 samples/sec	 accuracy=42.593750	 loss=2.546455	 lr=0.034627
Epoch[131] Batch [0399]/[3760]	Speed: 117.274922 samples/sec	 accuracy=42.628906	 loss=2.548980	 lr=0.034614
Epoch[131] Batch [0449]/[3760]	Speed: 117.655644 samples/sec	 accuracy=42.677083	 loss=2.548519	 lr=0.034602
Epoch[131] Batch [0499]/[3760]	Speed: 117.747407 samples/sec	 accuracy=42.881250	 loss=2.542371	 lr=0.034590
Epoch[131] Batch [0549]/[3760]	Speed: 116.918916 samples/sec	 accuracy=42.883523	 loss=2.540611	 lr=0.034578
Epoch[131] Batch [0599]/[3760]	Speed: 116.484343 samples/sec	 accuracy=42.861979	 loss=2.539745	 lr=0.034565
Epoch[131] Batch [0649]/[3760]	Speed: 117.256036 samples/sec	 accuracy=42.879808	 loss=2.539472	 lr=0.034553
Epoch[131] Batch [0699]/[3760]	Speed: 117.891838 samples/sec	 accuracy=42.834821	 loss=2.539539	 lr=0.034541
Epoch[131] Batch [0749]/[3760]	Speed: 117.228623 samples/sec	 accuracy=42.785417	 loss=2.542177	 lr=0.034529
Epoch[131] Batch [0799]/[3760]	Speed: 117.204509 samples/sec	 accuracy=42.742188	 loss=2.546486	 lr=0.034516
Epoch[131] Batch [0849]/[3760]	Speed: 117.138429 samples/sec	 accuracy=42.794118	 loss=2.544221	 lr=0.034504
Epoch[131] Batch [0899]/[3760]	Speed: 118.233064 samples/sec	 accuracy=42.817708	 loss=2.543677	 lr=0.034492
Epoch[131] Batch [0949]/[3760]	Speed: 117.395048 samples/sec	 accuracy=42.786184	 loss=2.544055	 lr=0.034480
Epoch[131] Batch [0999]/[3760]	Speed: 116.845130 samples/sec	 accuracy=42.804688	 loss=2.543570	 lr=0.034467
Epoch[131] Batch [1049]/[3760]	Speed: 118.146036 samples/sec	 accuracy=42.809524	 loss=2.542833	 lr=0.034455
Epoch[131] Batch [1099]/[3760]	Speed: 117.940513 samples/sec	 accuracy=42.813920	 loss=2.541843	 lr=0.034443
Epoch[131] Batch [1149]/[3760]	Speed: 117.223009 samples/sec	 accuracy=42.841033	 loss=2.539634	 lr=0.034431
Epoch[131] Batch [1199]/[3760]	Speed: 117.444286 samples/sec	 accuracy=42.790365	 loss=2.542421	 lr=0.034418
Epoch[131] Batch [1249]/[3760]	Speed: 117.956139 samples/sec	 accuracy=42.795000	 loss=2.542860	 lr=0.034406
Epoch[131] Batch [1299]/[3760]	Speed: 117.641524 samples/sec	 accuracy=42.759615	 loss=2.545555	 lr=0.034394
Epoch[131] Batch [1349]/[3760]	Speed: 117.953707 samples/sec	 accuracy=42.739583	 loss=2.544421	 lr=0.034381
Epoch[131] Batch [1399]/[3760]	Speed: 117.547273 samples/sec	 accuracy=42.691964	 loss=2.545112	 lr=0.034369
Epoch[131] Batch [1449]/[3760]	Speed: 117.606775 samples/sec	 accuracy=42.713362	 loss=2.544874	 lr=0.034357
Epoch[131] Batch [1499]/[3760]	Speed: 117.385778 samples/sec	 accuracy=42.659375	 loss=2.546777	 lr=0.034345
Epoch[131] Batch [1549]/[3760]	Speed: 117.190623 samples/sec	 accuracy=42.698589	 loss=2.547776	 lr=0.034332
Epoch[131] Batch [1599]/[3760]	Speed: 117.850290 samples/sec	 accuracy=42.710938	 loss=2.546793	 lr=0.034320
Epoch[131] Batch [1649]/[3760]	Speed: 117.377134 samples/sec	 accuracy=42.705492	 loss=2.546537	 lr=0.034308
Epoch[131] Batch [1699]/[3760]	Speed: 117.173148 samples/sec	 accuracy=42.751838	 loss=2.547266	 lr=0.034296
Epoch[131] Batch [1749]/[3760]	Speed: 115.961635 samples/sec	 accuracy=42.755357	 loss=2.547314	 lr=0.034284
Epoch[131] Batch [1799]/[3760]	Speed: 117.377686 samples/sec	 accuracy=42.730035	 loss=2.548896	 lr=0.034271
Epoch[131] Batch [1849]/[3760]	Speed: 117.232521 samples/sec	 accuracy=42.749155	 loss=2.548860	 lr=0.034259
Epoch[131] Batch [1899]/[3760]	Speed: 117.414220 samples/sec	 accuracy=42.701480	 loss=2.551230	 lr=0.034247
Epoch[131] Batch [1949]/[3760]	Speed: 117.808288 samples/sec	 accuracy=42.754808	 loss=2.549474	 lr=0.034235
Epoch[131] Batch [1999]/[3760]	Speed: 117.527664 samples/sec	 accuracy=42.750781	 loss=2.549965	 lr=0.034222
Epoch[131] Batch [2049]/[3760]	Speed: 117.482727 samples/sec	 accuracy=42.727134	 loss=2.552060	 lr=0.034210
Epoch[131] Batch [2099]/[3760]	Speed: 117.401241 samples/sec	 accuracy=42.748512	 loss=2.550987	 lr=0.034198
Epoch[131] Batch [2149]/[3760]	Speed: 117.343249 samples/sec	 accuracy=42.760901	 loss=2.550844	 lr=0.034186
Epoch[131] Batch [2199]/[3760]	Speed: 117.605118 samples/sec	 accuracy=42.770597	 loss=2.550409	 lr=0.034173
Epoch[131] Batch [2249]/[3760]	Speed: 117.030548 samples/sec	 accuracy=42.736111	 loss=2.551569	 lr=0.034161
Epoch[131] Batch [2299]/[3760]	Speed: 116.967647 samples/sec	 accuracy=42.750000	 loss=2.551428	 lr=0.034149
Epoch[131] Batch [2349]/[3760]	Speed: 117.176930 samples/sec	 accuracy=42.712766	 loss=2.552140	 lr=0.034137
Epoch[131] Batch [2399]/[3760]	Speed: 117.535112 samples/sec	 accuracy=42.738281	 loss=2.551803	 lr=0.034124
Epoch[131] Batch [2449]/[3760]	Speed: 116.660641 samples/sec	 accuracy=42.697066	 loss=2.552681	 lr=0.034112
Epoch[131] Batch [2499]/[3760]	Speed: 117.085458 samples/sec	 accuracy=42.709375	 loss=2.552270	 lr=0.034100
Epoch[131] Batch [2549]/[3760]	Speed: 116.866309 samples/sec	 accuracy=42.689338	 loss=2.552411	 lr=0.034088
Epoch[131] Batch [2599]/[3760]	Speed: 117.930402 samples/sec	 accuracy=42.668870	 loss=2.553458	 lr=0.034076
Epoch[131] Batch [2649]/[3760]	Speed: 116.844488 samples/sec	 accuracy=42.668042	 loss=2.553542	 lr=0.034063
Epoch[131] Batch [2699]/[3760]	Speed: 117.750282 samples/sec	 accuracy=42.657986	 loss=2.553500	 lr=0.034051
Epoch[131] Batch [2749]/[3760]	Speed: 117.617102 samples/sec	 accuracy=42.667614	 loss=2.553286	 lr=0.034039
Epoch[131] Batch [2799]/[3760]	Speed: 116.973070 samples/sec	 accuracy=42.652902	 loss=2.554189	 lr=0.034027
Epoch[131] Batch [2849]/[3760]	Speed: 117.668304 samples/sec	 accuracy=42.617873	 loss=2.555280	 lr=0.034014
Epoch[131] Batch [2899]/[3760]	Speed: 115.739841 samples/sec	 accuracy=42.643858	 loss=2.554665	 lr=0.034002
Epoch[131] Batch [2949]/[3760]	Speed: 118.045057 samples/sec	 accuracy=42.657839	 loss=2.554026	 lr=0.033990
Epoch[131] Batch [2999]/[3760]	Speed: 117.357208 samples/sec	 accuracy=42.655729	 loss=2.553836	 lr=0.033978
Epoch[131] Batch [3049]/[3760]	Speed: 117.073424 samples/sec	 accuracy=42.650102	 loss=2.554379	 lr=0.033966
Epoch[131] Batch [3099]/[3760]	Speed: 117.408364 samples/sec	 accuracy=42.654738	 loss=2.554901	 lr=0.033953
Epoch[131] Batch [3149]/[3760]	Speed: 117.403357 samples/sec	 accuracy=42.639881	 loss=2.555351	 lr=0.033941
Epoch[131] Batch [3199]/[3760]	Speed: 117.432540 samples/sec	 accuracy=42.618652	 loss=2.555935	 lr=0.033929
Epoch[131] Batch [3249]/[3760]	Speed: 117.650442 samples/sec	 accuracy=42.614423	 loss=2.556337	 lr=0.033917
Epoch[131] Batch [3299]/[3760]	Speed: 117.850331 samples/sec	 accuracy=42.634470	 loss=2.555505	 lr=0.033904
Epoch[131] Batch [3349]/[3760]	Speed: 117.062020 samples/sec	 accuracy=42.621269	 loss=2.555762	 lr=0.033892
Epoch[131] Batch [3399]/[3760]	Speed: 117.754976 samples/sec	 accuracy=42.633272	 loss=2.555478	 lr=0.033880
Epoch[131] Batch [3449]/[3760]	Speed: 117.455455 samples/sec	 accuracy=42.605978	 loss=2.556446	 lr=0.033868
Epoch[131] Batch [3499]/[3760]	Speed: 117.730758 samples/sec	 accuracy=42.597321	 loss=2.556598	 lr=0.033856
Epoch[131] Batch [3549]/[3760]	Speed: 117.748811 samples/sec	 accuracy=42.604754	 loss=2.556349	 lr=0.033843
Epoch[131] Batch [3599]/[3760]	Speed: 117.603946 samples/sec	 accuracy=42.602431	 loss=2.556770	 lr=0.033831
Epoch[131] Batch [3649]/[3760]	Speed: 116.987259 samples/sec	 accuracy=42.609589	 loss=2.556698	 lr=0.033819
Epoch[131] Batch [3699]/[3760]	Speed: 117.374911 samples/sec	 accuracy=42.615709	 loss=2.556689	 lr=0.033807
Epoch[131] Batch [3749]/[3760]	Speed: 120.580791 samples/sec	 accuracy=42.588333	 loss=2.557478	 lr=0.033795
Batch [0049]/[0303]: acc-top1=41.750000 acc-top5=65.343750
Batch [0099]/[0303]: acc-top1=40.078125 acc-top5=65.156250
Batch [0149]/[0303]: acc-top1=41.968750 acc-top5=66.739583
Batch [0199]/[0303]: acc-top1=44.406250 acc-top5=68.718750
Batch [0249]/[0303]: acc-top1=43.856250 acc-top5=68.743750
Batch [0299]/[0303]: acc-top1=43.229167 acc-top5=68.640625
[Epoch 131] training: accuracy=42.593085	 loss=2.557461
[Epoch 131] speed: 116 samples/sec	time cost: 2175.441133
[Epoch 131] validation: acc-top1=43.224010 acc-top5=68.646865 loss=2.615808
Epoch[132] Batch [0049]/[3760]	Speed: 86.914645 samples/sec	 accuracy=43.062500	 loss=2.523072	 lr=0.033780
Epoch[132] Batch [0099]/[3760]	Speed: 117.705069 samples/sec	 accuracy=42.484375	 loss=2.517225	 lr=0.033768
Epoch[132] Batch [0149]/[3760]	Speed: 117.880051 samples/sec	 accuracy=42.291667	 loss=2.531677	 lr=0.033756
Epoch[132] Batch [0199]/[3760]	Speed: 117.462767 samples/sec	 accuracy=42.664062	 loss=2.519156	 lr=0.033743
Epoch[132] Batch [0249]/[3760]	Speed: 117.595644 samples/sec	 accuracy=42.856250	 loss=2.514193	 lr=0.033731
Epoch[132] Batch [0299]/[3760]	Speed: 115.212299 samples/sec	 accuracy=42.927083	 loss=2.518189	 lr=0.033719
Epoch[132] Batch [0349]/[3760]	Speed: 117.423457 samples/sec	 accuracy=43.035714	 loss=2.521388	 lr=0.033707
Epoch[132] Batch [0399]/[3760]	Speed: 117.120407 samples/sec	 accuracy=43.019531	 loss=2.520857	 lr=0.033695
Epoch[132] Batch [0449]/[3760]	Speed: 117.995930 samples/sec	 accuracy=43.059028	 loss=2.520097	 lr=0.033682
Epoch[132] Batch [0499]/[3760]	Speed: 117.035703 samples/sec	 accuracy=43.168750	 loss=2.519818	 lr=0.033670
Epoch[132] Batch [0549]/[3760]	Speed: 117.530080 samples/sec	 accuracy=43.159091	 loss=2.522752	 lr=0.033658
Epoch[132] Batch [0599]/[3760]	Speed: 117.630800 samples/sec	 accuracy=43.192708	 loss=2.519857	 lr=0.033646
Epoch[132] Batch [0649]/[3760]	Speed: 116.734521 samples/sec	 accuracy=43.019231	 loss=2.526327	 lr=0.033634
Epoch[132] Batch [0699]/[3760]	Speed: 118.125464 samples/sec	 accuracy=42.968750	 loss=2.530700	 lr=0.033621
Epoch[132] Batch [0749]/[3760]	Speed: 117.828375 samples/sec	 accuracy=42.993750	 loss=2.530354	 lr=0.033609
Epoch[132] Batch [0799]/[3760]	Speed: 118.278411 samples/sec	 accuracy=42.943359	 loss=2.532949	 lr=0.033597
Epoch[132] Batch [0849]/[3760]	Speed: 116.436172 samples/sec	 accuracy=42.966912	 loss=2.531876	 lr=0.033585
Epoch[132] Batch [0899]/[3760]	Speed: 117.670348 samples/sec	 accuracy=42.935764	 loss=2.532017	 lr=0.033573
Epoch[132] Batch [0949]/[3760]	Speed: 117.148629 samples/sec	 accuracy=42.847039	 loss=2.535746	 lr=0.033561
Epoch[132] Batch [0999]/[3760]	Speed: 117.428042 samples/sec	 accuracy=42.868750	 loss=2.536633	 lr=0.033548
Epoch[132] Batch [1049]/[3760]	Speed: 117.071641 samples/sec	 accuracy=42.857143	 loss=2.533398	 lr=0.033536
Epoch[132] Batch [1099]/[3760]	Speed: 117.481047 samples/sec	 accuracy=42.884943	 loss=2.534654	 lr=0.033524
Epoch[132] Batch [1149]/[3760]	Speed: 118.177682 samples/sec	 accuracy=42.927989	 loss=2.533384	 lr=0.033512
Epoch[132] Batch [1199]/[3760]	Speed: 117.340200 samples/sec	 accuracy=42.920573	 loss=2.534308	 lr=0.033500
Epoch[132] Batch [1249]/[3760]	Speed: 117.435927 samples/sec	 accuracy=42.938750	 loss=2.535373	 lr=0.033487
Epoch[132] Batch [1299]/[3760]	Speed: 118.041498 samples/sec	 accuracy=42.885817	 loss=2.537310	 lr=0.033475
Epoch[132] Batch [1349]/[3760]	Speed: 117.347114 samples/sec	 accuracy=42.875000	 loss=2.537351	 lr=0.033463
Epoch[132] Batch [1399]/[3760]	Speed: 117.655618 samples/sec	 accuracy=42.892857	 loss=2.534914	 lr=0.033451
Epoch[132] Batch [1449]/[3760]	Speed: 117.688348 samples/sec	 accuracy=42.891164	 loss=2.535698	 lr=0.033439
Epoch[132] Batch [1499]/[3760]	Speed: 115.294139 samples/sec	 accuracy=42.881250	 loss=2.537192	 lr=0.033427
Epoch[132] Batch [1549]/[3760]	Speed: 117.165436 samples/sec	 accuracy=42.869960	 loss=2.537329	 lr=0.033414
Epoch[132] Batch [1599]/[3760]	Speed: 117.399811 samples/sec	 accuracy=42.895508	 loss=2.536445	 lr=0.033402
Epoch[132] Batch [1649]/[3760]	Speed: 118.397067 samples/sec	 accuracy=42.924242	 loss=2.535550	 lr=0.033390
Epoch[132] Batch [1699]/[3760]	Speed: 117.350502 samples/sec	 accuracy=42.909007	 loss=2.535245	 lr=0.033378
Epoch[132] Batch [1749]/[3760]	Speed: 118.143058 samples/sec	 accuracy=42.917857	 loss=2.535074	 lr=0.033366
Epoch[132] Batch [1799]/[3760]	Speed: 117.609363 samples/sec	 accuracy=42.928819	 loss=2.534745	 lr=0.033354
Epoch[132] Batch [1849]/[3760]	Speed: 117.332621 samples/sec	 accuracy=42.928209	 loss=2.534707	 lr=0.033341
Epoch[132] Batch [1899]/[3760]	Speed: 117.256378 samples/sec	 accuracy=42.899671	 loss=2.536102	 lr=0.033329
Epoch[132] Batch [1949]/[3760]	Speed: 118.070557 samples/sec	 accuracy=42.864583	 loss=2.537340	 lr=0.033317
Epoch[132] Batch [1999]/[3760]	Speed: 116.981953 samples/sec	 accuracy=42.835938	 loss=2.538312	 lr=0.033305
Epoch[132] Batch [2049]/[3760]	Speed: 117.930523 samples/sec	 accuracy=42.809451	 loss=2.539119	 lr=0.033293
Epoch[132] Batch [2099]/[3760]	Speed: 116.914455 samples/sec	 accuracy=42.831845	 loss=2.539276	 lr=0.033281
Epoch[132] Batch [2149]/[3760]	Speed: 118.112625 samples/sec	 accuracy=42.856831	 loss=2.537798	 lr=0.033269
Epoch[132] Batch [2199]/[3760]	Speed: 116.894637 samples/sec	 accuracy=42.877841	 loss=2.538096	 lr=0.033256
Epoch[132] Batch [2249]/[3760]	Speed: 117.444716 samples/sec	 accuracy=42.863194	 loss=2.538649	 lr=0.033244
Epoch[132] Batch [2299]/[3760]	Speed: 117.129479 samples/sec	 accuracy=42.862772	 loss=2.540273	 lr=0.033232
Epoch[132] Batch [2349]/[3760]	Speed: 117.545623 samples/sec	 accuracy=42.870346	 loss=2.539458	 lr=0.033220
Epoch[132] Batch [2399]/[3760]	Speed: 117.446914 samples/sec	 accuracy=42.887370	 loss=2.537677	 lr=0.033208
Epoch[132] Batch [2449]/[3760]	Speed: 117.158761 samples/sec	 accuracy=42.878827	 loss=2.538599	 lr=0.033196
Epoch[132] Batch [2499]/[3760]	Speed: 117.723075 samples/sec	 accuracy=42.870000	 loss=2.538922	 lr=0.033183
Epoch[132] Batch [2549]/[3760]	Speed: 117.280022 samples/sec	 accuracy=42.868873	 loss=2.539103	 lr=0.033171
Epoch[132] Batch [2599]/[3760]	Speed: 118.039075 samples/sec	 accuracy=42.861779	 loss=2.538934	 lr=0.033159
Epoch[132] Batch [2649]/[3760]	Speed: 116.438174 samples/sec	 accuracy=42.839033	 loss=2.539953	 lr=0.033147
Epoch[132] Batch [2699]/[3760]	Speed: 116.554172 samples/sec	 accuracy=42.824074	 loss=2.540657	 lr=0.033135
Epoch[132] Batch [2749]/[3760]	Speed: 117.595261 samples/sec	 accuracy=42.832955	 loss=2.540219	 lr=0.033123
Epoch[132] Batch [2799]/[3760]	Speed: 117.752584 samples/sec	 accuracy=42.821429	 loss=2.541303	 lr=0.033111
Epoch[132] Batch [2849]/[3760]	Speed: 117.564021 samples/sec	 accuracy=42.810855	 loss=2.542102	 lr=0.033098
Epoch[132] Batch [2899]/[3760]	Speed: 117.279752 samples/sec	 accuracy=42.779095	 loss=2.543835	 lr=0.033086
Epoch[132] Batch [2949]/[3760]	Speed: 117.191177 samples/sec	 accuracy=42.796610	 loss=2.544037	 lr=0.033074
Epoch[132] Batch [2999]/[3760]	Speed: 117.561764 samples/sec	 accuracy=42.800000	 loss=2.544170	 lr=0.033062
Epoch[132] Batch [3049]/[3760]	Speed: 117.825725 samples/sec	 accuracy=42.816086	 loss=2.543952	 lr=0.033050
Epoch[132] Batch [3099]/[3760]	Speed: 117.687823 samples/sec	 accuracy=42.800907	 loss=2.544294	 lr=0.033038
Epoch[132] Batch [3149]/[3760]	Speed: 117.942350 samples/sec	 accuracy=42.809028	 loss=2.544249	 lr=0.033026
Epoch[132] Batch [3199]/[3760]	Speed: 117.313586 samples/sec	 accuracy=42.814453	 loss=2.544151	 lr=0.033014
Epoch[132] Batch [3249]/[3760]	Speed: 117.151307 samples/sec	 accuracy=42.832212	 loss=2.543315	 lr=0.033001
Epoch[132] Batch [3299]/[3760]	Speed: 117.415549 samples/sec	 accuracy=42.842330	 loss=2.543452	 lr=0.032989
Epoch[132] Batch [3349]/[3760]	Speed: 118.245061 samples/sec	 accuracy=42.843284	 loss=2.543644	 lr=0.032977
Epoch[132] Batch [3399]/[3760]	Speed: 117.604703 samples/sec	 accuracy=42.865349	 loss=2.543311	 lr=0.032965
Epoch[132] Batch [3449]/[3760]	Speed: 116.964549 samples/sec	 accuracy=42.880435	 loss=2.542893	 lr=0.032953
Epoch[132] Batch [3499]/[3760]	Speed: 118.242659 samples/sec	 accuracy=42.864732	 loss=2.543290	 lr=0.032941
Epoch[132] Batch [3549]/[3760]	Speed: 117.534572 samples/sec	 accuracy=42.856514	 loss=2.543448	 lr=0.032929
Epoch[132] Batch [3599]/[3760]	Speed: 118.417687 samples/sec	 accuracy=42.861111	 loss=2.543876	 lr=0.032917
Epoch[132] Batch [3649]/[3760]	Speed: 117.225436 samples/sec	 accuracy=42.840325	 loss=2.544244	 lr=0.032904
Epoch[132] Batch [3699]/[3760]	Speed: 118.253229 samples/sec	 accuracy=42.829814	 loss=2.545148	 lr=0.032892
Epoch[132] Batch [3749]/[3760]	Speed: 120.511580 samples/sec	 accuracy=42.820417	 loss=2.545238	 lr=0.032880
Batch [0049]/[0303]: acc-top1=44.812500 acc-top5=70.906250
Batch [0099]/[0303]: acc-top1=42.546875 acc-top5=68.796875
Batch [0149]/[0303]: acc-top1=44.114583 acc-top5=69.947917
Batch [0199]/[0303]: acc-top1=45.937500 acc-top5=71.195312
Batch [0249]/[0303]: acc-top1=45.050000 acc-top5=70.768750
Batch [0299]/[0303]: acc-top1=44.750000 acc-top5=70.651042
[Epoch 132] training: accuracy=42.824551	 loss=2.545128
[Epoch 132] speed: 116 samples/sec	time cost: 2174.101442
[Epoch 132] validation: acc-top1=44.698845 acc-top5=70.621906 loss=2.528124
Epoch[133] Batch [0049]/[3759]	Speed: 88.111542 samples/sec	 accuracy=44.750000	 loss=2.476629	 lr=0.032866
Epoch[133] Batch [0099]/[3759]	Speed: 117.589572 samples/sec	 accuracy=43.859375	 loss=2.486362	 lr=0.032854
Epoch[133] Batch [0149]/[3759]	Speed: 117.426060 samples/sec	 accuracy=44.104167	 loss=2.488271	 lr=0.032841
Epoch[133] Batch [0199]/[3759]	Speed: 117.453393 samples/sec	 accuracy=44.257812	 loss=2.478672	 lr=0.032829
Epoch[133] Batch [0249]/[3759]	Speed: 117.115736 samples/sec	 accuracy=44.081250	 loss=2.475355	 lr=0.032817
Epoch[133] Batch [0299]/[3759]	Speed: 117.181205 samples/sec	 accuracy=43.786458	 loss=2.480591	 lr=0.032805
Epoch[133] Batch [0349]/[3759]	Speed: 117.647689 samples/sec	 accuracy=43.767857	 loss=2.482905	 lr=0.032793
Epoch[133] Batch [0399]/[3759]	Speed: 116.914323 samples/sec	 accuracy=43.781250	 loss=2.485508	 lr=0.032781
Epoch[133] Batch [0449]/[3759]	Speed: 117.373512 samples/sec	 accuracy=43.600694	 loss=2.491496	 lr=0.032769
Epoch[133] Batch [0499]/[3759]	Speed: 117.318618 samples/sec	 accuracy=43.600000	 loss=2.492934	 lr=0.032757
Epoch[133] Batch [0549]/[3759]	Speed: 117.508439 samples/sec	 accuracy=43.633523	 loss=2.494154	 lr=0.032745
Epoch[133] Batch [0599]/[3759]	Speed: 117.704564 samples/sec	 accuracy=43.643229	 loss=2.495397	 lr=0.032732
Epoch[133] Batch [0649]/[3759]	Speed: 117.032380 samples/sec	 accuracy=43.637019	 loss=2.494330	 lr=0.032720
Epoch[133] Batch [0699]/[3759]	Speed: 117.962617 samples/sec	 accuracy=43.571429	 loss=2.496530	 lr=0.032708
Epoch[133] Batch [0749]/[3759]	Speed: 117.343472 samples/sec	 accuracy=43.585417	 loss=2.498749	 lr=0.032696
Epoch[133] Batch [0799]/[3759]	Speed: 117.526244 samples/sec	 accuracy=43.492188	 loss=2.500310	 lr=0.032684
Epoch[133] Batch [0849]/[3759]	Speed: 117.268068 samples/sec	 accuracy=43.474265	 loss=2.503778	 lr=0.032672
Epoch[133] Batch [0899]/[3759]	Speed: 117.683283 samples/sec	 accuracy=43.493056	 loss=2.501501	 lr=0.032660
Epoch[133] Batch [0949]/[3759]	Speed: 117.584004 samples/sec	 accuracy=43.442434	 loss=2.501549	 lr=0.032648
Epoch[133] Batch [0999]/[3759]	Speed: 117.060216 samples/sec	 accuracy=43.432813	 loss=2.502634	 lr=0.032636
Epoch[133] Batch [1049]/[3759]	Speed: 117.701779 samples/sec	 accuracy=43.489583	 loss=2.502849	 lr=0.032624
Epoch[133] Batch [1099]/[3759]	Speed: 117.046859 samples/sec	 accuracy=43.494318	 loss=2.503529	 lr=0.032611
Epoch[133] Batch [1149]/[3759]	Speed: 117.256114 samples/sec	 accuracy=43.461957	 loss=2.506133	 lr=0.032599
Epoch[133] Batch [1199]/[3759]	Speed: 116.205005 samples/sec	 accuracy=43.375000	 loss=2.509923	 lr=0.032587
Epoch[133] Batch [1249]/[3759]	Speed: 117.345490 samples/sec	 accuracy=43.410000	 loss=2.508291	 lr=0.032575
Epoch[133] Batch [1299]/[3759]	Speed: 117.101548 samples/sec	 accuracy=43.442308	 loss=2.509452	 lr=0.032563
Epoch[133] Batch [1349]/[3759]	Speed: 117.323067 samples/sec	 accuracy=43.462963	 loss=2.509666	 lr=0.032551
Epoch[133] Batch [1399]/[3759]	Speed: 117.844734 samples/sec	 accuracy=43.417411	 loss=2.513017	 lr=0.032539
Epoch[133] Batch [1449]/[3759]	Speed: 117.063295 samples/sec	 accuracy=43.359914	 loss=2.514287	 lr=0.032527
Epoch[133] Batch [1499]/[3759]	Speed: 117.662076 samples/sec	 accuracy=43.373958	 loss=2.513302	 lr=0.032515
Epoch[133] Batch [1549]/[3759]	Speed: 117.561086 samples/sec	 accuracy=43.341734	 loss=2.514115	 lr=0.032503
Epoch[133] Batch [1599]/[3759]	Speed: 118.233366 samples/sec	 accuracy=43.365234	 loss=2.512994	 lr=0.032491
Epoch[133] Batch [1649]/[3759]	Speed: 118.081699 samples/sec	 accuracy=43.378788	 loss=2.513178	 lr=0.032479
Epoch[133] Batch [1699]/[3759]	Speed: 117.960320 samples/sec	 accuracy=43.383272	 loss=2.513281	 lr=0.032466
Epoch[133] Batch [1749]/[3759]	Speed: 117.386024 samples/sec	 accuracy=43.390179	 loss=2.513132	 lr=0.032454
Epoch[133] Batch [1799]/[3759]	Speed: 116.932315 samples/sec	 accuracy=43.358507	 loss=2.514932	 lr=0.032442
Epoch[133] Batch [1849]/[3759]	Speed: 117.519510 samples/sec	 accuracy=43.312500	 loss=2.516655	 lr=0.032430
Epoch[133] Batch [1899]/[3759]	Speed: 117.577309 samples/sec	 accuracy=43.302632	 loss=2.517684	 lr=0.032418
Epoch[133] Batch [1949]/[3759]	Speed: 117.858667 samples/sec	 accuracy=43.301282	 loss=2.518455	 lr=0.032406
Epoch[133] Batch [1999]/[3759]	Speed: 117.681888 samples/sec	 accuracy=43.311719	 loss=2.518834	 lr=0.032394
Epoch[133] Batch [2049]/[3759]	Speed: 118.414391 samples/sec	 accuracy=43.314024	 loss=2.518538	 lr=0.032382
Epoch[133] Batch [2099]/[3759]	Speed: 117.354304 samples/sec	 accuracy=43.330357	 loss=2.518018	 lr=0.032370
Epoch[133] Batch [2149]/[3759]	Speed: 116.389734 samples/sec	 accuracy=43.320494	 loss=2.517946	 lr=0.032358
Epoch[133] Batch [2199]/[3759]	Speed: 117.654522 samples/sec	 accuracy=43.306818	 loss=2.518643	 lr=0.032346
Epoch[133] Batch [2249]/[3759]	Speed: 117.240209 samples/sec	 accuracy=43.317361	 loss=2.518693	 lr=0.032334
Epoch[133] Batch [2299]/[3759]	Speed: 116.418595 samples/sec	 accuracy=43.321332	 loss=2.518947	 lr=0.032322
Epoch[133] Batch [2349]/[3759]	Speed: 115.931867 samples/sec	 accuracy=43.322473	 loss=2.519205	 lr=0.032310
Epoch[133] Batch [2399]/[3759]	Speed: 117.969151 samples/sec	 accuracy=43.311849	 loss=2.520041	 lr=0.032297
Epoch[133] Batch [2449]/[3759]	Speed: 117.996108 samples/sec	 accuracy=43.329719	 loss=2.519679	 lr=0.032285
Epoch[133] Batch [2499]/[3759]	Speed: 117.174286 samples/sec	 accuracy=43.343750	 loss=2.519034	 lr=0.032273
Epoch[133] Batch [2549]/[3759]	Speed: 116.814100 samples/sec	 accuracy=43.348039	 loss=2.519575	 lr=0.032261
Epoch[133] Batch [2599]/[3759]	Speed: 117.559218 samples/sec	 accuracy=43.351562	 loss=2.519288	 lr=0.032249
Epoch[133] Batch [2649]/[3759]	Speed: 117.751325 samples/sec	 accuracy=43.333726	 loss=2.519761	 lr=0.032237
Epoch[133] Batch [2699]/[3759]	Speed: 118.046137 samples/sec	 accuracy=43.350116	 loss=2.519013	 lr=0.032225
Epoch[133] Batch [2749]/[3759]	Speed: 117.770886 samples/sec	 accuracy=43.349432	 loss=2.518739	 lr=0.032213
Epoch[133] Batch [2799]/[3759]	Speed: 117.681956 samples/sec	 accuracy=43.316964	 loss=2.519893	 lr=0.032201
Epoch[133] Batch [2849]/[3759]	Speed: 117.660013 samples/sec	 accuracy=43.307566	 loss=2.520477	 lr=0.032189
Epoch[133] Batch [2899]/[3759]	Speed: 117.390824 samples/sec	 accuracy=43.275862	 loss=2.522192	 lr=0.032177
Epoch[133] Batch [2949]/[3759]	Speed: 118.156747 samples/sec	 accuracy=43.274364	 loss=2.522023	 lr=0.032165
Epoch[133] Batch [2999]/[3759]	Speed: 116.961759 samples/sec	 accuracy=43.275000	 loss=2.521766	 lr=0.032153
Epoch[133] Batch [3049]/[3759]	Speed: 118.372862 samples/sec	 accuracy=43.289447	 loss=2.521297	 lr=0.032141
Epoch[133] Batch [3099]/[3759]	Speed: 117.245551 samples/sec	 accuracy=43.278226	 loss=2.521868	 lr=0.032129
Epoch[133] Batch [3149]/[3759]	Speed: 117.743368 samples/sec	 accuracy=43.246528	 loss=2.523585	 lr=0.032117
Epoch[133] Batch [3199]/[3759]	Speed: 117.923859 samples/sec	 accuracy=43.256348	 loss=2.523679	 lr=0.032105
Epoch[133] Batch [3249]/[3759]	Speed: 117.369329 samples/sec	 accuracy=43.256731	 loss=2.523835	 lr=0.032093
Epoch[133] Batch [3299]/[3759]	Speed: 117.161435 samples/sec	 accuracy=43.257576	 loss=2.523808	 lr=0.032081
Epoch[133] Batch [3349]/[3759]	Speed: 117.894650 samples/sec	 accuracy=43.253265	 loss=2.524255	 lr=0.032068
Epoch[133] Batch [3399]/[3759]	Speed: 117.023338 samples/sec	 accuracy=43.244945	 loss=2.524629	 lr=0.032056
Epoch[133] Batch [3449]/[3759]	Speed: 116.882772 samples/sec	 accuracy=43.235054	 loss=2.524383	 lr=0.032044
Epoch[133] Batch [3499]/[3759]	Speed: 117.375510 samples/sec	 accuracy=43.260714	 loss=2.523238	 lr=0.032032
Epoch[133] Batch [3549]/[3759]	Speed: 116.880079 samples/sec	 accuracy=43.270246	 loss=2.521912	 lr=0.032020
Epoch[133] Batch [3599]/[3759]	Speed: 117.333709 samples/sec	 accuracy=43.255208	 loss=2.522529	 lr=0.032008
Epoch[133] Batch [3649]/[3759]	Speed: 117.719011 samples/sec	 accuracy=43.230308	 loss=2.523812	 lr=0.031996
Epoch[133] Batch [3699]/[3759]	Speed: 117.444567 samples/sec	 accuracy=43.233108	 loss=2.522973	 lr=0.031984
Epoch[133] Batch [3749]/[3759]	Speed: 121.290156 samples/sec	 accuracy=43.219167	 loss=2.523589	 lr=0.031972
Batch [0049]/[0303]: acc-top1=43.968750 acc-top5=68.906250
Batch [0099]/[0303]: acc-top1=40.703125 acc-top5=67.015625
Batch [0149]/[0303]: acc-top1=42.395833 acc-top5=67.968750
Batch [0199]/[0303]: acc-top1=44.398438 acc-top5=69.601562
Batch [0249]/[0303]: acc-top1=43.768750 acc-top5=69.068750
Batch [0299]/[0303]: acc-top1=43.005208 acc-top5=68.973958
[Epoch 133] training: accuracy=43.222516	 loss=2.523503
[Epoch 133] speed: 116 samples/sec	time cost: 2174.408598
[Epoch 133] validation: acc-top1=42.961015 acc-top5=68.956271 loss=2.698692
Epoch[134] Batch [0049]/[3760]	Speed: 86.855909 samples/sec	 accuracy=43.625000	 loss=2.499408	 lr=0.031958
Epoch[134] Batch [0099]/[3760]	Speed: 117.924242 samples/sec	 accuracy=43.296875	 loss=2.522873	 lr=0.031946
Epoch[134] Batch [0149]/[3760]	Speed: 117.864832 samples/sec	 accuracy=43.614583	 loss=2.509140	 lr=0.031934
Epoch[134] Batch [0199]/[3760]	Speed: 116.868222 samples/sec	 accuracy=43.359375	 loss=2.523930	 lr=0.031922
Epoch[134] Batch [0249]/[3760]	Speed: 117.710437 samples/sec	 accuracy=43.387500	 loss=2.522301	 lr=0.031910
Epoch[134] Batch [0299]/[3760]	Speed: 117.763941 samples/sec	 accuracy=43.588542	 loss=2.524488	 lr=0.031898
Epoch[134] Batch [0349]/[3760]	Speed: 117.681528 samples/sec	 accuracy=43.500000	 loss=2.520858	 lr=0.031886
Epoch[134] Batch [0399]/[3760]	Speed: 117.177757 samples/sec	 accuracy=43.378906	 loss=2.519580	 lr=0.031874
Epoch[134] Batch [0449]/[3760]	Speed: 116.939116 samples/sec	 accuracy=43.645833	 loss=2.510717	 lr=0.031862
Epoch[134] Batch [0499]/[3760]	Speed: 117.520641 samples/sec	 accuracy=43.693750	 loss=2.510059	 lr=0.031850
Epoch[134] Batch [0549]/[3760]	Speed: 117.684074 samples/sec	 accuracy=43.710227	 loss=2.508917	 lr=0.031838
Epoch[134] Batch [0599]/[3760]	Speed: 117.333949 samples/sec	 accuracy=43.692708	 loss=2.506466	 lr=0.031826
Epoch[134] Batch [0649]/[3760]	Speed: 117.504887 samples/sec	 accuracy=43.701923	 loss=2.506805	 lr=0.031814
Epoch[134] Batch [0699]/[3760]	Speed: 117.448478 samples/sec	 accuracy=43.587054	 loss=2.509553	 lr=0.031802
Epoch[134] Batch [0749]/[3760]	Speed: 117.056698 samples/sec	 accuracy=43.641667	 loss=2.507321	 lr=0.031790
Epoch[134] Batch [0799]/[3760]	Speed: 117.471681 samples/sec	 accuracy=43.589844	 loss=2.510258	 lr=0.031778
Epoch[134] Batch [0849]/[3760]	Speed: 117.503877 samples/sec	 accuracy=43.505515	 loss=2.514761	 lr=0.031766
Epoch[134] Batch [0899]/[3760]	Speed: 115.991688 samples/sec	 accuracy=43.590278	 loss=2.511021	 lr=0.031754
Epoch[134] Batch [0949]/[3760]	Speed: 116.556298 samples/sec	 accuracy=43.629934	 loss=2.507256	 lr=0.031742
Epoch[134] Batch [0999]/[3760]	Speed: 117.447749 samples/sec	 accuracy=43.643750	 loss=2.507578	 lr=0.031730
Epoch[134] Batch [1049]/[3760]	Speed: 117.623401 samples/sec	 accuracy=43.616071	 loss=2.506412	 lr=0.031718
Epoch[134] Batch [1099]/[3760]	Speed: 118.465131 samples/sec	 accuracy=43.656250	 loss=2.505114	 lr=0.031706
Epoch[134] Batch [1149]/[3760]	Speed: 118.034914 samples/sec	 accuracy=43.663043	 loss=2.505005	 lr=0.031694
Epoch[134] Batch [1199]/[3760]	Speed: 117.020436 samples/sec	 accuracy=43.592448	 loss=2.509637	 lr=0.031682
Epoch[134] Batch [1249]/[3760]	Speed: 117.236756 samples/sec	 accuracy=43.608750	 loss=2.510596	 lr=0.031670
Epoch[134] Batch [1299]/[3760]	Speed: 117.816749 samples/sec	 accuracy=43.612981	 loss=2.509800	 lr=0.031658
Epoch[134] Batch [1349]/[3760]	Speed: 117.328722 samples/sec	 accuracy=43.622685	 loss=2.509071	 lr=0.031646
Epoch[134] Batch [1399]/[3760]	Speed: 117.441426 samples/sec	 accuracy=43.651786	 loss=2.506534	 lr=0.031634
Epoch[134] Batch [1449]/[3760]	Speed: 118.085728 samples/sec	 accuracy=43.691810	 loss=2.506152	 lr=0.031622
Epoch[134] Batch [1499]/[3760]	Speed: 116.978011 samples/sec	 accuracy=43.666667	 loss=2.504967	 lr=0.031610
Epoch[134] Batch [1549]/[3760]	Speed: 117.221493 samples/sec	 accuracy=43.636089	 loss=2.506498	 lr=0.031598
Epoch[134] Batch [1599]/[3760]	Speed: 117.341184 samples/sec	 accuracy=43.630859	 loss=2.506845	 lr=0.031586
Epoch[134] Batch [1649]/[3760]	Speed: 117.578731 samples/sec	 accuracy=43.602273	 loss=2.507174	 lr=0.031574
Epoch[134] Batch [1699]/[3760]	Speed: 117.437383 samples/sec	 accuracy=43.582721	 loss=2.507884	 lr=0.031562
Epoch[134] Batch [1749]/[3760]	Speed: 117.094232 samples/sec	 accuracy=43.566964	 loss=2.508851	 lr=0.031550
Epoch[134] Batch [1799]/[3760]	Speed: 118.085580 samples/sec	 accuracy=43.561632	 loss=2.509342	 lr=0.031538
Epoch[134] Batch [1849]/[3760]	Speed: 117.688474 samples/sec	 accuracy=43.554899	 loss=2.509716	 lr=0.031526
Epoch[134] Batch [1899]/[3760]	Speed: 117.231141 samples/sec	 accuracy=43.521382	 loss=2.511048	 lr=0.031514
Epoch[134] Batch [1949]/[3760]	Speed: 117.177331 samples/sec	 accuracy=43.521635	 loss=2.510726	 lr=0.031502
Epoch[134] Batch [1999]/[3760]	Speed: 117.502985 samples/sec	 accuracy=43.489844	 loss=2.512138	 lr=0.031490
Epoch[134] Batch [2049]/[3760]	Speed: 115.507429 samples/sec	 accuracy=43.487805	 loss=2.512130	 lr=0.031478
Epoch[134] Batch [2099]/[3760]	Speed: 117.294168 samples/sec	 accuracy=43.511905	 loss=2.511054	 lr=0.031466
Epoch[134] Batch [2149]/[3760]	Speed: 117.325425 samples/sec	 accuracy=43.502907	 loss=2.510663	 lr=0.031454
Epoch[134] Batch [2199]/[3760]	Speed: 117.292941 samples/sec	 accuracy=43.485085	 loss=2.510844	 lr=0.031442
Epoch[134] Batch [2249]/[3760]	Speed: 117.599303 samples/sec	 accuracy=43.490972	 loss=2.510624	 lr=0.031430
Epoch[134] Batch [2299]/[3760]	Speed: 116.647466 samples/sec	 accuracy=43.480978	 loss=2.511053	 lr=0.031418
Epoch[134] Batch [2349]/[3760]	Speed: 117.672394 samples/sec	 accuracy=43.502660	 loss=2.510737	 lr=0.031406
Epoch[134] Batch [2399]/[3760]	Speed: 117.244710 samples/sec	 accuracy=43.489583	 loss=2.511250	 lr=0.031394
Epoch[134] Batch [2449]/[3760]	Speed: 117.755463 samples/sec	 accuracy=43.501913	 loss=2.511915	 lr=0.031382
Epoch[134] Batch [2499]/[3760]	Speed: 117.698660 samples/sec	 accuracy=43.511250	 loss=2.510810	 lr=0.031370
Epoch[134] Batch [2549]/[3760]	Speed: 117.370491 samples/sec	 accuracy=43.517770	 loss=2.510913	 lr=0.031358
Epoch[134] Batch [2599]/[3760]	Speed: 117.682042 samples/sec	 accuracy=43.497596	 loss=2.512297	 lr=0.031346
Epoch[134] Batch [2649]/[3760]	Speed: 116.724428 samples/sec	 accuracy=43.471108	 loss=2.513583	 lr=0.031334
Epoch[134] Batch [2699]/[3760]	Speed: 117.384515 samples/sec	 accuracy=43.490741	 loss=2.512887	 lr=0.031322
Epoch[134] Batch [2749]/[3760]	Speed: 116.588712 samples/sec	 accuracy=43.469318	 loss=2.514053	 lr=0.031310
Epoch[134] Batch [2799]/[3760]	Speed: 117.983485 samples/sec	 accuracy=43.459821	 loss=2.514805	 lr=0.031298
Epoch[134] Batch [2849]/[3760]	Speed: 117.767804 samples/sec	 accuracy=43.436404	 loss=2.515568	 lr=0.031286
Epoch[134] Batch [2899]/[3760]	Speed: 117.377187 samples/sec	 accuracy=43.432651	 loss=2.515428	 lr=0.031274
Epoch[134] Batch [2949]/[3760]	Speed: 117.738287 samples/sec	 accuracy=43.427966	 loss=2.515539	 lr=0.031262
Epoch[134] Batch [2999]/[3760]	Speed: 117.594108 samples/sec	 accuracy=43.430208	 loss=2.515186	 lr=0.031250
Epoch[134] Batch [3049]/[3760]	Speed: 117.647763 samples/sec	 accuracy=43.430840	 loss=2.515011	 lr=0.031238
Epoch[134] Batch [3099]/[3760]	Speed: 117.743360 samples/sec	 accuracy=43.432460	 loss=2.514925	 lr=0.031227
Epoch[134] Batch [3149]/[3760]	Speed: 117.216501 samples/sec	 accuracy=43.462302	 loss=2.514307	 lr=0.031215
Epoch[134] Batch [3199]/[3760]	Speed: 118.379702 samples/sec	 accuracy=43.461914	 loss=2.514165	 lr=0.031203
Epoch[134] Batch [3249]/[3760]	Speed: 115.790066 samples/sec	 accuracy=43.461538	 loss=2.514058	 lr=0.031191
Epoch[134] Batch [3299]/[3760]	Speed: 118.340557 samples/sec	 accuracy=43.457386	 loss=2.513789	 lr=0.031179
Epoch[134] Batch [3349]/[3760]	Speed: 117.506194 samples/sec	 accuracy=43.454757	 loss=2.514383	 lr=0.031167
Epoch[134] Batch [3399]/[3760]	Speed: 117.892322 samples/sec	 accuracy=43.456801	 loss=2.514368	 lr=0.031155
Epoch[134] Batch [3449]/[3760]	Speed: 117.140301 samples/sec	 accuracy=43.456522	 loss=2.514672	 lr=0.031143
Epoch[134] Batch [3499]/[3760]	Speed: 117.067214 samples/sec	 accuracy=43.477679	 loss=2.513680	 lr=0.031131
Epoch[134] Batch [3549]/[3760]	Speed: 117.226811 samples/sec	 accuracy=43.470511	 loss=2.514031	 lr=0.031119
Epoch[134] Batch [3599]/[3760]	Speed: 117.550881 samples/sec	 accuracy=43.459201	 loss=2.514248	 lr=0.031107
Epoch[134] Batch [3649]/[3760]	Speed: 117.054847 samples/sec	 accuracy=43.479880	 loss=2.513650	 lr=0.031095
Epoch[134] Batch [3699]/[3760]	Speed: 117.413472 samples/sec	 accuracy=43.442990	 loss=2.514644	 lr=0.031083
Epoch[134] Batch [3749]/[3760]	Speed: 120.831390 samples/sec	 accuracy=43.445417	 loss=2.514472	 lr=0.031071
Batch [0049]/[0303]: acc-top1=43.906250 acc-top5=69.062500
Batch [0099]/[0303]: acc-top1=41.328125 acc-top5=68.375000
Batch [0149]/[0303]: acc-top1=43.229167 acc-top5=69.541667
Batch [0199]/[0303]: acc-top1=46.101562 acc-top5=71.570312
Batch [0249]/[0303]: acc-top1=45.387500 acc-top5=71.331250
Batch [0299]/[0303]: acc-top1=44.927083 acc-top5=71.369792
[Epoch 134] training: accuracy=43.442487	 loss=2.514268
[Epoch 134] speed: 116 samples/sec	time cost: 2175.409799
[Epoch 134] validation: acc-top1=44.884488 acc-top5=71.318069 loss=2.496016
Epoch[135] Batch [0049]/[3760]	Speed: 87.803374 samples/sec	 accuracy=44.000000	 loss=2.512243	 lr=0.031057
Epoch[135] Batch [0099]/[3760]	Speed: 117.215591 samples/sec	 accuracy=44.656250	 loss=2.468447	 lr=0.031045
Epoch[135] Batch [0149]/[3760]	Speed: 117.545586 samples/sec	 accuracy=44.645833	 loss=2.472719	 lr=0.031033
Epoch[135] Batch [0199]/[3760]	Speed: 117.425605 samples/sec	 accuracy=44.992188	 loss=2.461049	 lr=0.031021
Epoch[135] Batch [0249]/[3760]	Speed: 117.316808 samples/sec	 accuracy=44.737500	 loss=2.468690	 lr=0.031009
Epoch[135] Batch [0299]/[3760]	Speed: 117.702652 samples/sec	 accuracy=44.562500	 loss=2.473895	 lr=0.030997
Epoch[135] Batch [0349]/[3760]	Speed: 118.106735 samples/sec	 accuracy=44.410714	 loss=2.476921	 lr=0.030985
Epoch[135] Batch [0399]/[3760]	Speed: 116.682942 samples/sec	 accuracy=44.390625	 loss=2.477565	 lr=0.030973
Epoch[135] Batch [0449]/[3760]	Speed: 117.302456 samples/sec	 accuracy=44.454861	 loss=2.476426	 lr=0.030961
Epoch[135] Batch [0499]/[3760]	Speed: 117.297819 samples/sec	 accuracy=44.431250	 loss=2.478636	 lr=0.030950
Epoch[135] Batch [0549]/[3760]	Speed: 117.268379 samples/sec	 accuracy=44.636364	 loss=2.477200	 lr=0.030938
Epoch[135] Batch [0599]/[3760]	Speed: 118.207132 samples/sec	 accuracy=44.570312	 loss=2.477713	 lr=0.030926
Epoch[135] Batch [0649]/[3760]	Speed: 115.929989 samples/sec	 accuracy=44.480769	 loss=2.478491	 lr=0.030914
Epoch[135] Batch [0699]/[3760]	Speed: 117.787841 samples/sec	 accuracy=44.417411	 loss=2.482236	 lr=0.030902
Epoch[135] Batch [0749]/[3760]	Speed: 117.330221 samples/sec	 accuracy=44.370833	 loss=2.481284	 lr=0.030890
Epoch[135] Batch [0799]/[3760]	Speed: 117.254670 samples/sec	 accuracy=44.441406	 loss=2.480562	 lr=0.030878
Epoch[135] Batch [0849]/[3760]	Speed: 117.674407 samples/sec	 accuracy=44.415441	 loss=2.480246	 lr=0.030866
Epoch[135] Batch [0899]/[3760]	Speed: 117.527706 samples/sec	 accuracy=44.397569	 loss=2.479750	 lr=0.030854
Epoch[135] Batch [0949]/[3760]	Speed: 117.995382 samples/sec	 accuracy=44.414474	 loss=2.479169	 lr=0.030842
Epoch[135] Batch [0999]/[3760]	Speed: 116.882412 samples/sec	 accuracy=44.400000	 loss=2.477638	 lr=0.030830
Epoch[135] Batch [1049]/[3760]	Speed: 117.251365 samples/sec	 accuracy=44.327381	 loss=2.479059	 lr=0.030818
Epoch[135] Batch [1099]/[3760]	Speed: 117.369594 samples/sec	 accuracy=44.286932	 loss=2.478468	 lr=0.030807
Epoch[135] Batch [1149]/[3760]	Speed: 117.792215 samples/sec	 accuracy=44.293478	 loss=2.477877	 lr=0.030795
Epoch[135] Batch [1199]/[3760]	Speed: 118.015697 samples/sec	 accuracy=44.312500	 loss=2.476318	 lr=0.030783
Epoch[135] Batch [1249]/[3760]	Speed: 118.090409 samples/sec	 accuracy=44.356250	 loss=2.475142	 lr=0.030771
Epoch[135] Batch [1299]/[3760]	Speed: 117.802694 samples/sec	 accuracy=44.360577	 loss=2.476588	 lr=0.030759
Epoch[135] Batch [1349]/[3760]	Speed: 117.404968 samples/sec	 accuracy=44.384259	 loss=2.474740	 lr=0.030747
Epoch[135] Batch [1399]/[3760]	Speed: 117.143635 samples/sec	 accuracy=44.354911	 loss=2.476273	 lr=0.030735
Epoch[135] Batch [1449]/[3760]	Speed: 117.898164 samples/sec	 accuracy=44.407328	 loss=2.472624	 lr=0.030723
Epoch[135] Batch [1499]/[3760]	Speed: 117.319864 samples/sec	 accuracy=44.412500	 loss=2.472893	 lr=0.030711
Epoch[135] Batch [1549]/[3760]	Speed: 117.642426 samples/sec	 accuracy=44.329637	 loss=2.475730	 lr=0.030699
Epoch[135] Batch [1599]/[3760]	Speed: 117.700169 samples/sec	 accuracy=44.319336	 loss=2.477004	 lr=0.030688
Epoch[135] Batch [1649]/[3760]	Speed: 117.672576 samples/sec	 accuracy=44.338068	 loss=2.477599	 lr=0.030676
Epoch[135] Batch [1699]/[3760]	Speed: 117.912819 samples/sec	 accuracy=44.281250	 loss=2.479407	 lr=0.030664
Epoch[135] Batch [1749]/[3760]	Speed: 117.830998 samples/sec	 accuracy=44.233929	 loss=2.480169	 lr=0.030652
Epoch[135] Batch [1799]/[3760]	Speed: 118.134918 samples/sec	 accuracy=44.197049	 loss=2.481570	 lr=0.030640
Epoch[135] Batch [1849]/[3760]	Speed: 116.277961 samples/sec	 accuracy=44.216216	 loss=2.481168	 lr=0.030628
Epoch[135] Batch [1899]/[3760]	Speed: 117.499433 samples/sec	 accuracy=44.212171	 loss=2.480479	 lr=0.030616
Epoch[135] Batch [1949]/[3760]	Speed: 118.071372 samples/sec	 accuracy=44.157051	 loss=2.482004	 lr=0.030604
Epoch[135] Batch [1999]/[3760]	Speed: 117.039341 samples/sec	 accuracy=44.167187	 loss=2.481987	 lr=0.030592
Epoch[135] Batch [2049]/[3760]	Speed: 117.673982 samples/sec	 accuracy=44.158537	 loss=2.481997	 lr=0.030580
Epoch[135] Batch [2099]/[3760]	Speed: 117.740627 samples/sec	 accuracy=44.147321	 loss=2.482800	 lr=0.030569
Epoch[135] Batch [2149]/[3760]	Speed: 118.140565 samples/sec	 accuracy=44.125727	 loss=2.483417	 lr=0.030557
Epoch[135] Batch [2199]/[3760]	Speed: 117.033275 samples/sec	 accuracy=44.122869	 loss=2.483283	 lr=0.030545
Epoch[135] Batch [2249]/[3760]	Speed: 117.113928 samples/sec	 accuracy=44.129167	 loss=2.483826	 lr=0.030533
Epoch[135] Batch [2299]/[3760]	Speed: 117.831185 samples/sec	 accuracy=44.117527	 loss=2.484662	 lr=0.030521
Epoch[135] Batch [2349]/[3760]	Speed: 118.125926 samples/sec	 accuracy=44.102394	 loss=2.485880	 lr=0.030509
Epoch[135] Batch [2399]/[3760]	Speed: 118.005464 samples/sec	 accuracy=44.087891	 loss=2.486143	 lr=0.030497
Epoch[135] Batch [2449]/[3760]	Speed: 117.221003 samples/sec	 accuracy=44.105230	 loss=2.486608	 lr=0.030485
Epoch[135] Batch [2499]/[3760]	Speed: 117.066863 samples/sec	 accuracy=44.125000	 loss=2.486736	 lr=0.030474
Epoch[135] Batch [2549]/[3760]	Speed: 117.936122 samples/sec	 accuracy=44.138480	 loss=2.486231	 lr=0.030462
Epoch[135] Batch [2599]/[3760]	Speed: 117.550499 samples/sec	 accuracy=44.168269	 loss=2.485888	 lr=0.030450
Epoch[135] Batch [2649]/[3760]	Speed: 117.197646 samples/sec	 accuracy=44.157429	 loss=2.486263	 lr=0.030438
Epoch[135] Batch [2699]/[3760]	Speed: 116.752228 samples/sec	 accuracy=44.129051	 loss=2.488149	 lr=0.030426
Epoch[135] Batch [2749]/[3760]	Speed: 117.356765 samples/sec	 accuracy=44.149432	 loss=2.486722	 lr=0.030414
Epoch[135] Batch [2799]/[3760]	Speed: 117.731970 samples/sec	 accuracy=44.141741	 loss=2.486140	 lr=0.030402
Epoch[135] Batch [2849]/[3760]	Speed: 117.607733 samples/sec	 accuracy=44.160636	 loss=2.485785	 lr=0.030390
Epoch[135] Batch [2899]/[3760]	Speed: 117.254581 samples/sec	 accuracy=44.135776	 loss=2.486339	 lr=0.030379
Epoch[135] Batch [2949]/[3760]	Speed: 117.233280 samples/sec	 accuracy=44.121292	 loss=2.486441	 lr=0.030367
Epoch[135] Batch [2999]/[3760]	Speed: 117.726524 samples/sec	 accuracy=44.121354	 loss=2.486909	 lr=0.030355
Epoch[135] Batch [3049]/[3760]	Speed: 115.981378 samples/sec	 accuracy=44.133197	 loss=2.486358	 lr=0.030343
Epoch[135] Batch [3099]/[3760]	Speed: 117.059662 samples/sec	 accuracy=44.095766	 loss=2.488257	 lr=0.030331
Epoch[135] Batch [3149]/[3760]	Speed: 117.220281 samples/sec	 accuracy=44.083333	 loss=2.488830	 lr=0.030319
Epoch[135] Batch [3199]/[3760]	Speed: 117.375621 samples/sec	 accuracy=44.077637	 loss=2.489190	 lr=0.030307
Epoch[135] Batch [3249]/[3760]	Speed: 117.319192 samples/sec	 accuracy=44.075962	 loss=2.489297	 lr=0.030296
Epoch[135] Batch [3299]/[3760]	Speed: 117.851180 samples/sec	 accuracy=44.068655	 loss=2.489719	 lr=0.030284
Epoch[135] Batch [3349]/[3760]	Speed: 117.003712 samples/sec	 accuracy=44.055970	 loss=2.489952	 lr=0.030272
Epoch[135] Batch [3399]/[3760]	Speed: 117.429137 samples/sec	 accuracy=44.042279	 loss=2.490318	 lr=0.030260
Epoch[135] Batch [3449]/[3760]	Speed: 117.428005 samples/sec	 accuracy=44.023551	 loss=2.491446	 lr=0.030248
Epoch[135] Batch [3499]/[3760]	Speed: 117.309829 samples/sec	 accuracy=44.016518	 loss=2.491575	 lr=0.030236
Epoch[135] Batch [3549]/[3760]	Speed: 117.172607 samples/sec	 accuracy=44.028169	 loss=2.491314	 lr=0.030225
Epoch[135] Batch [3599]/[3760]	Speed: 117.324836 samples/sec	 accuracy=44.009549	 loss=2.492195	 lr=0.030213
Epoch[135] Batch [3649]/[3760]	Speed: 117.301737 samples/sec	 accuracy=44.016695	 loss=2.492566	 lr=0.030201
Epoch[135] Batch [3699]/[3760]	Speed: 117.747639 samples/sec	 accuracy=44.013514	 loss=2.492854	 lr=0.030189
Epoch[135] Batch [3749]/[3760]	Speed: 120.138650 samples/sec	 accuracy=44.002083	 loss=2.493101	 lr=0.030177
Batch [0049]/[0303]: acc-top1=46.500000 acc-top5=71.062500
Batch [0099]/[0303]: acc-top1=42.828125 acc-top5=69.031250
Batch [0149]/[0303]: acc-top1=44.135417 acc-top5=69.906250
Batch [0199]/[0303]: acc-top1=46.953125 acc-top5=71.460938
Batch [0249]/[0303]: acc-top1=46.087500 acc-top5=71.200000
Batch [0299]/[0303]: acc-top1=45.781250 acc-top5=71.260417
[Epoch 135] training: accuracy=44.000166	 loss=2.493088
[Epoch 135] speed: 116 samples/sec	time cost: 2173.638128
[Epoch 135] validation: acc-top1=45.812706 acc-top5=71.323226 loss=2.441470
Epoch[136] Batch [0049]/[3759]	Speed: 87.303092 samples/sec	 accuracy=45.843750	 loss=2.405571	 lr=0.030163
Epoch[136] Batch [0099]/[3759]	Speed: 117.082329 samples/sec	 accuracy=44.859375	 loss=2.461984	 lr=0.030151
Epoch[136] Batch [0149]/[3759]	Speed: 117.253856 samples/sec	 accuracy=45.093750	 loss=2.450917	 lr=0.030139
Epoch[136] Batch [0199]/[3759]	Speed: 117.566682 samples/sec	 accuracy=44.765625	 loss=2.447356	 lr=0.030127
Epoch[136] Batch [0249]/[3759]	Speed: 116.961249 samples/sec	 accuracy=44.512500	 loss=2.457475	 lr=0.030116
Epoch[136] Batch [0299]/[3759]	Speed: 116.500514 samples/sec	 accuracy=44.505208	 loss=2.449116	 lr=0.030104
Epoch[136] Batch [0349]/[3759]	Speed: 117.692110 samples/sec	 accuracy=44.625000	 loss=2.443839	 lr=0.030092
Epoch[136] Batch [0399]/[3759]	Speed: 116.948467 samples/sec	 accuracy=44.578125	 loss=2.444580	 lr=0.030080
Epoch[136] Batch [0449]/[3759]	Speed: 116.516874 samples/sec	 accuracy=44.729167	 loss=2.443634	 lr=0.030068
Epoch[136] Batch [0499]/[3759]	Speed: 116.938621 samples/sec	 accuracy=44.643750	 loss=2.449865	 lr=0.030056
Epoch[136] Batch [0549]/[3759]	Speed: 117.449646 samples/sec	 accuracy=44.613636	 loss=2.452324	 lr=0.030045
Epoch[136] Batch [0599]/[3759]	Speed: 117.016976 samples/sec	 accuracy=44.500000	 loss=2.454750	 lr=0.030033
Epoch[136] Batch [0649]/[3759]	Speed: 117.261902 samples/sec	 accuracy=44.459135	 loss=2.459354	 lr=0.030021
Epoch[136] Batch [0699]/[3759]	Speed: 117.259803 samples/sec	 accuracy=44.453125	 loss=2.461398	 lr=0.030009
Epoch[136] Batch [0749]/[3759]	Speed: 117.680851 samples/sec	 accuracy=44.333333	 loss=2.467094	 lr=0.029997
Epoch[136] Batch [0799]/[3759]	Speed: 117.664987 samples/sec	 accuracy=44.275391	 loss=2.468952	 lr=0.029986
Epoch[136] Batch [0849]/[3759]	Speed: 117.390428 samples/sec	 accuracy=44.253676	 loss=2.470134	 lr=0.029974
Epoch[136] Batch [0899]/[3759]	Speed: 116.780600 samples/sec	 accuracy=44.256944	 loss=2.471488	 lr=0.029962
Epoch[136] Batch [0949]/[3759]	Speed: 117.149475 samples/sec	 accuracy=44.241776	 loss=2.471305	 lr=0.029950
Epoch[136] Batch [0999]/[3759]	Speed: 118.160182 samples/sec	 accuracy=44.306250	 loss=2.469431	 lr=0.029938
Epoch[136] Batch [1049]/[3759]	Speed: 117.698329 samples/sec	 accuracy=44.215774	 loss=2.473494	 lr=0.029926
Epoch[136] Batch [1099]/[3759]	Speed: 116.977431 samples/sec	 accuracy=44.241477	 loss=2.472788	 lr=0.029915
Epoch[136] Batch [1149]/[3759]	Speed: 117.169872 samples/sec	 accuracy=44.190217	 loss=2.473290	 lr=0.029903
Epoch[136] Batch [1199]/[3759]	Speed: 117.848138 samples/sec	 accuracy=44.113281	 loss=2.478070	 lr=0.029891
Epoch[136] Batch [1249]/[3759]	Speed: 117.709273 samples/sec	 accuracy=44.126250	 loss=2.477151	 lr=0.029879
Epoch[136] Batch [1299]/[3759]	Speed: 117.202162 samples/sec	 accuracy=44.203125	 loss=2.475281	 lr=0.029867
Epoch[136] Batch [1349]/[3759]	Speed: 117.511663 samples/sec	 accuracy=44.215278	 loss=2.475393	 lr=0.029856
Epoch[136] Batch [1399]/[3759]	Speed: 117.556231 samples/sec	 accuracy=44.220982	 loss=2.474246	 lr=0.029844
Epoch[136] Batch [1449]/[3759]	Speed: 117.831900 samples/sec	 accuracy=44.201509	 loss=2.474220	 lr=0.029832
Epoch[136] Batch [1499]/[3759]	Speed: 117.833754 samples/sec	 accuracy=44.213542	 loss=2.474455	 lr=0.029820
Epoch[136] Batch [1549]/[3759]	Speed: 117.755105 samples/sec	 accuracy=44.195565	 loss=2.475522	 lr=0.029808
Epoch[136] Batch [1599]/[3759]	Speed: 116.564012 samples/sec	 accuracy=44.235352	 loss=2.474717	 lr=0.029797
Epoch[136] Batch [1649]/[3759]	Speed: 116.165647 samples/sec	 accuracy=44.208333	 loss=2.475448	 lr=0.029785
Epoch[136] Batch [1699]/[3759]	Speed: 117.362708 samples/sec	 accuracy=44.206801	 loss=2.476375	 lr=0.029773
Epoch[136] Batch [1749]/[3759]	Speed: 117.573065 samples/sec	 accuracy=44.212500	 loss=2.477038	 lr=0.029761
Epoch[136] Batch [1799]/[3759]	Speed: 116.651976 samples/sec	 accuracy=44.193576	 loss=2.477380	 lr=0.029749
Epoch[136] Batch [1849]/[3759]	Speed: 117.745647 samples/sec	 accuracy=44.223818	 loss=2.476886	 lr=0.029738
Epoch[136] Batch [1899]/[3759]	Speed: 118.171620 samples/sec	 accuracy=44.191612	 loss=2.477192	 lr=0.029726
Epoch[136] Batch [1949]/[3759]	Speed: 117.243970 samples/sec	 accuracy=44.161058	 loss=2.477940	 lr=0.029714
Epoch[136] Batch [1999]/[3759]	Speed: 117.169934 samples/sec	 accuracy=44.138281	 loss=2.479388	 lr=0.029702
Epoch[136] Batch [2049]/[3759]	Speed: 117.420604 samples/sec	 accuracy=44.162348	 loss=2.478642	 lr=0.029690
Epoch[136] Batch [2099]/[3759]	Speed: 117.371583 samples/sec	 accuracy=44.127232	 loss=2.479936	 lr=0.029679
Epoch[136] Batch [2149]/[3759]	Speed: 117.388662 samples/sec	 accuracy=44.139535	 loss=2.479321	 lr=0.029667
Epoch[136] Batch [2199]/[3759]	Speed: 117.185635 samples/sec	 accuracy=44.120028	 loss=2.480054	 lr=0.029655
Epoch[136] Batch [2249]/[3759]	Speed: 117.069602 samples/sec	 accuracy=44.111806	 loss=2.480917	 lr=0.029643
Epoch[136] Batch [2299]/[3759]	Speed: 116.767184 samples/sec	 accuracy=44.109375	 loss=2.481620	 lr=0.029632
Epoch[136] Batch [2349]/[3759]	Speed: 118.552723 samples/sec	 accuracy=44.103723	 loss=2.481742	 lr=0.029620
Epoch[136] Batch [2399]/[3759]	Speed: 117.819899 samples/sec	 accuracy=44.126302	 loss=2.481181	 lr=0.029608
Epoch[136] Batch [2449]/[3759]	Speed: 118.096299 samples/sec	 accuracy=44.126913	 loss=2.481313	 lr=0.029596
Epoch[136] Batch [2499]/[3759]	Speed: 117.001540 samples/sec	 accuracy=44.125000	 loss=2.481078	 lr=0.029584
Epoch[136] Batch [2549]/[3759]	Speed: 117.668822 samples/sec	 accuracy=44.096814	 loss=2.482304	 lr=0.029573
Epoch[136] Batch [2599]/[3759]	Speed: 117.583410 samples/sec	 accuracy=44.093149	 loss=2.482868	 lr=0.029561
Epoch[136] Batch [2649]/[3759]	Speed: 117.015455 samples/sec	 accuracy=44.097288	 loss=2.482330	 lr=0.029549
Epoch[136] Batch [2699]/[3759]	Speed: 117.320578 samples/sec	 accuracy=44.084491	 loss=2.481576	 lr=0.029537
Epoch[136] Batch [2749]/[3759]	Speed: 117.453041 samples/sec	 accuracy=44.065341	 loss=2.482637	 lr=0.029526
Epoch[136] Batch [2799]/[3759]	Speed: 115.716883 samples/sec	 accuracy=44.066964	 loss=2.482338	 lr=0.029514
Epoch[136] Batch [2849]/[3759]	Speed: 118.335348 samples/sec	 accuracy=44.071820	 loss=2.482567	 lr=0.029502
Epoch[136] Batch [2899]/[3759]	Speed: 117.122568 samples/sec	 accuracy=44.062500	 loss=2.482674	 lr=0.029490
Epoch[136] Batch [2949]/[3759]	Speed: 117.725795 samples/sec	 accuracy=44.055614	 loss=2.483211	 lr=0.029479
Epoch[136] Batch [2999]/[3759]	Speed: 116.884434 samples/sec	 accuracy=44.066667	 loss=2.483228	 lr=0.029467
Epoch[136] Batch [3049]/[3759]	Speed: 118.293005 samples/sec	 accuracy=44.031250	 loss=2.484714	 lr=0.029455
Epoch[136] Batch [3099]/[3759]	Speed: 117.644159 samples/sec	 accuracy=44.023185	 loss=2.485153	 lr=0.029443
Epoch[136] Batch [3149]/[3759]	Speed: 117.259413 samples/sec	 accuracy=44.037698	 loss=2.484327	 lr=0.029431
Epoch[136] Batch [3199]/[3759]	Speed: 118.634166 samples/sec	 accuracy=44.051270	 loss=2.483950	 lr=0.029420
Epoch[136] Batch [3249]/[3759]	Speed: 118.035420 samples/sec	 accuracy=44.039904	 loss=2.484659	 lr=0.029408
Epoch[136] Batch [3299]/[3759]	Speed: 117.200368 samples/sec	 accuracy=44.026989	 loss=2.485125	 lr=0.029396
Epoch[136] Batch [3349]/[3759]	Speed: 117.766721 samples/sec	 accuracy=44.029384	 loss=2.485370	 lr=0.029384
Epoch[136] Batch [3399]/[3759]	Speed: 117.616529 samples/sec	 accuracy=44.035846	 loss=2.485114	 lr=0.029373
Epoch[136] Batch [3449]/[3759]	Speed: 117.182546 samples/sec	 accuracy=44.021739	 loss=2.485162	 lr=0.029361
Epoch[136] Batch [3499]/[3759]	Speed: 117.817267 samples/sec	 accuracy=44.023661	 loss=2.484894	 lr=0.029349
Epoch[136] Batch [3549]/[3759]	Speed: 117.953266 samples/sec	 accuracy=44.001761	 loss=2.486221	 lr=0.029337
Epoch[136] Batch [3599]/[3759]	Speed: 117.683999 samples/sec	 accuracy=44.007812	 loss=2.486107	 lr=0.029326
Epoch[136] Batch [3649]/[3759]	Speed: 117.683368 samples/sec	 accuracy=44.000000	 loss=2.486141	 lr=0.029314
Epoch[136] Batch [3699]/[3759]	Speed: 117.341915 samples/sec	 accuracy=44.021115	 loss=2.485341	 lr=0.029302
Epoch[136] Batch [3749]/[3759]	Speed: 120.347254 samples/sec	 accuracy=44.022083	 loss=2.485175	 lr=0.029291
Batch [0049]/[0303]: acc-top1=42.562500 acc-top5=69.000000
Batch [0099]/[0303]: acc-top1=40.765625 acc-top5=67.609375
Batch [0149]/[0303]: acc-top1=43.302083 acc-top5=68.770833
Batch [0199]/[0303]: acc-top1=46.148438 acc-top5=70.578125
Batch [0249]/[0303]: acc-top1=45.900000 acc-top5=70.593750
Batch [0299]/[0303]: acc-top1=45.187500 acc-top5=70.661458
[Epoch 136] training: accuracy=44.018938	 loss=2.485234
[Epoch 136] speed: 116 samples/sec	time cost: 2174.724359
[Epoch 136] validation: acc-top1=45.183581 acc-top5=70.621906 loss=2.518225
Epoch[137] Batch [0049]/[3760]	Speed: 87.548431 samples/sec	 accuracy=44.000000	 loss=2.494285	 lr=0.029277
Epoch[137] Batch [0099]/[3760]	Speed: 116.936345 samples/sec	 accuracy=44.937500	 loss=2.448550	 lr=0.029265
Epoch[137] Batch [0149]/[3760]	Speed: 117.598891 samples/sec	 accuracy=45.583333	 loss=2.430589	 lr=0.029253
Epoch[137] Batch [0199]/[3760]	Speed: 117.280492 samples/sec	 accuracy=45.117188	 loss=2.448730	 lr=0.029241
Epoch[137] Batch [0249]/[3760]	Speed: 115.164492 samples/sec	 accuracy=44.881250	 loss=2.466260	 lr=0.029230
Epoch[137] Batch [0299]/[3760]	Speed: 117.488698 samples/sec	 accuracy=44.864583	 loss=2.455451	 lr=0.029218
Epoch[137] Batch [0349]/[3760]	Speed: 117.443024 samples/sec	 accuracy=44.991071	 loss=2.452242	 lr=0.029206
Epoch[137] Batch [0399]/[3760]	Speed: 117.405113 samples/sec	 accuracy=45.062500	 loss=2.444483	 lr=0.029195
Epoch[137] Batch [0449]/[3760]	Speed: 117.455229 samples/sec	 accuracy=44.996528	 loss=2.443764	 lr=0.029183
Epoch[137] Batch [0499]/[3760]	Speed: 117.461246 samples/sec	 accuracy=44.665625	 loss=2.456723	 lr=0.029171
Epoch[137] Batch [0549]/[3760]	Speed: 118.032131 samples/sec	 accuracy=44.602273	 loss=2.453480	 lr=0.029159
Epoch[137] Batch [0599]/[3760]	Speed: 116.856466 samples/sec	 accuracy=44.497396	 loss=2.459771	 lr=0.029148
Epoch[137] Batch [0649]/[3760]	Speed: 116.774374 samples/sec	 accuracy=44.495192	 loss=2.461563	 lr=0.029136
Epoch[137] Batch [0699]/[3760]	Speed: 117.217986 samples/sec	 accuracy=44.511161	 loss=2.458777	 lr=0.029124
Epoch[137] Batch [0749]/[3760]	Speed: 117.240183 samples/sec	 accuracy=44.614583	 loss=2.454525	 lr=0.029112
Epoch[137] Batch [0799]/[3760]	Speed: 117.093087 samples/sec	 accuracy=44.576172	 loss=2.454752	 lr=0.029101
Epoch[137] Batch [0849]/[3760]	Speed: 117.988124 samples/sec	 accuracy=44.621324	 loss=2.454010	 lr=0.029089
Epoch[137] Batch [0899]/[3760]	Speed: 117.573339 samples/sec	 accuracy=44.607639	 loss=2.454661	 lr=0.029077
Epoch[137] Batch [0949]/[3760]	Speed: 117.257023 samples/sec	 accuracy=44.610197	 loss=2.456463	 lr=0.029066
Epoch[137] Batch [0999]/[3760]	Speed: 116.383665 samples/sec	 accuracy=44.626562	 loss=2.455624	 lr=0.029054
Epoch[137] Batch [1049]/[3760]	Speed: 117.183056 samples/sec	 accuracy=44.648810	 loss=2.455433	 lr=0.029042
Epoch[137] Batch [1099]/[3760]	Speed: 117.106432 samples/sec	 accuracy=44.667614	 loss=2.455048	 lr=0.029030
Epoch[137] Batch [1149]/[3760]	Speed: 117.796496 samples/sec	 accuracy=44.710598	 loss=2.453892	 lr=0.029019
Epoch[137] Batch [1199]/[3760]	Speed: 117.152703 samples/sec	 accuracy=44.777344	 loss=2.450923	 lr=0.029007
Epoch[137] Batch [1249]/[3760]	Speed: 117.369126 samples/sec	 accuracy=44.782500	 loss=2.452065	 lr=0.028995
Epoch[137] Batch [1299]/[3760]	Speed: 117.897448 samples/sec	 accuracy=44.849760	 loss=2.448578	 lr=0.028984
Epoch[137] Batch [1349]/[3760]	Speed: 117.626350 samples/sec	 accuracy=44.820602	 loss=2.449248	 lr=0.028972
Epoch[137] Batch [1399]/[3760]	Speed: 117.935407 samples/sec	 accuracy=44.782366	 loss=2.451598	 lr=0.028960
Epoch[137] Batch [1449]/[3760]	Speed: 116.287673 samples/sec	 accuracy=44.806034	 loss=2.449851	 lr=0.028949
Epoch[137] Batch [1499]/[3760]	Speed: 117.513952 samples/sec	 accuracy=44.762500	 loss=2.451161	 lr=0.028937
Epoch[137] Batch [1549]/[3760]	Speed: 118.758410 samples/sec	 accuracy=44.782258	 loss=2.449408	 lr=0.028925
Epoch[137] Batch [1599]/[3760]	Speed: 117.445728 samples/sec	 accuracy=44.770508	 loss=2.451423	 lr=0.028913
Epoch[137] Batch [1649]/[3760]	Speed: 118.139009 samples/sec	 accuracy=44.765152	 loss=2.452882	 lr=0.028902
Epoch[137] Batch [1699]/[3760]	Speed: 116.908210 samples/sec	 accuracy=44.743566	 loss=2.454341	 lr=0.028890
Epoch[137] Batch [1749]/[3760]	Speed: 117.420463 samples/sec	 accuracy=44.717857	 loss=2.456286	 lr=0.028878
Epoch[137] Batch [1799]/[3760]	Speed: 117.098222 samples/sec	 accuracy=44.689236	 loss=2.457468	 lr=0.028867
Epoch[137] Batch [1849]/[3760]	Speed: 118.186471 samples/sec	 accuracy=44.692568	 loss=2.457354	 lr=0.028855
Epoch[137] Batch [1899]/[3760]	Speed: 117.355049 samples/sec	 accuracy=44.703125	 loss=2.455452	 lr=0.028843
Epoch[137] Batch [1949]/[3760]	Speed: 118.219951 samples/sec	 accuracy=44.685897	 loss=2.455682	 lr=0.028832
Epoch[137] Batch [1999]/[3760]	Speed: 117.851481 samples/sec	 accuracy=44.650000	 loss=2.456601	 lr=0.028820
Epoch[137] Batch [2049]/[3760]	Speed: 117.603571 samples/sec	 accuracy=44.653963	 loss=2.456577	 lr=0.028808
Epoch[137] Batch [2099]/[3760]	Speed: 117.374678 samples/sec	 accuracy=44.646577	 loss=2.456577	 lr=0.028797
Epoch[137] Batch [2149]/[3760]	Speed: 117.616894 samples/sec	 accuracy=44.638081	 loss=2.457365	 lr=0.028785
Epoch[137] Batch [2199]/[3760]	Speed: 117.999449 samples/sec	 accuracy=44.636364	 loss=2.456945	 lr=0.028773
Epoch[137] Batch [2249]/[3760]	Speed: 117.382377 samples/sec	 accuracy=44.602778	 loss=2.457772	 lr=0.028762
Epoch[137] Batch [2299]/[3760]	Speed: 117.752981 samples/sec	 accuracy=44.580842	 loss=2.459354	 lr=0.028750
Epoch[137] Batch [2349]/[3760]	Speed: 117.491655 samples/sec	 accuracy=44.586436	 loss=2.459136	 lr=0.028738
Epoch[137] Batch [2399]/[3760]	Speed: 116.970916 samples/sec	 accuracy=44.602214	 loss=2.458835	 lr=0.028727
Epoch[137] Batch [2449]/[3760]	Speed: 117.425437 samples/sec	 accuracy=44.555485	 loss=2.461121	 lr=0.028715
Epoch[137] Batch [2499]/[3760]	Speed: 117.616928 samples/sec	 accuracy=44.539375	 loss=2.462333	 lr=0.028703
Epoch[137] Batch [2549]/[3760]	Speed: 117.934081 samples/sec	 accuracy=44.553309	 loss=2.462154	 lr=0.028692
Epoch[137] Batch [2599]/[3760]	Speed: 117.465868 samples/sec	 accuracy=44.515625	 loss=2.464554	 lr=0.028680
Epoch[137] Batch [2649]/[3760]	Speed: 115.036863 samples/sec	 accuracy=44.495873	 loss=2.464748	 lr=0.028668
Epoch[137] Batch [2699]/[3760]	Speed: 117.153630 samples/sec	 accuracy=44.484375	 loss=2.465613	 lr=0.028657
Epoch[137] Batch [2749]/[3760]	Speed: 117.153602 samples/sec	 accuracy=44.451705	 loss=2.466739	 lr=0.028645
Epoch[137] Batch [2799]/[3760]	Speed: 117.722374 samples/sec	 accuracy=44.467634	 loss=2.465512	 lr=0.028633
Epoch[137] Batch [2849]/[3760]	Speed: 116.789438 samples/sec	 accuracy=44.475329	 loss=2.464996	 lr=0.028622
Epoch[137] Batch [2899]/[3760]	Speed: 116.913527 samples/sec	 accuracy=44.465517	 loss=2.465031	 lr=0.028610
Epoch[137] Batch [2949]/[3760]	Speed: 118.143465 samples/sec	 accuracy=44.478284	 loss=2.464795	 lr=0.028598
Epoch[137] Batch [2999]/[3760]	Speed: 118.077537 samples/sec	 accuracy=44.458333	 loss=2.465706	 lr=0.028587
Epoch[137] Batch [3049]/[3760]	Speed: 118.585609 samples/sec	 accuracy=44.481045	 loss=2.465091	 lr=0.028575
Epoch[137] Batch [3099]/[3760]	Speed: 117.378046 samples/sec	 accuracy=44.451109	 loss=2.466428	 lr=0.028563
Epoch[137] Batch [3149]/[3760]	Speed: 117.519963 samples/sec	 accuracy=44.462798	 loss=2.466941	 lr=0.028552
Epoch[137] Batch [3199]/[3760]	Speed: 117.520736 samples/sec	 accuracy=44.464355	 loss=2.467274	 lr=0.028540
Epoch[137] Batch [3249]/[3760]	Speed: 117.101062 samples/sec	 accuracy=44.481731	 loss=2.466348	 lr=0.028528
Epoch[137] Batch [3299]/[3760]	Speed: 117.173975 samples/sec	 accuracy=44.472064	 loss=2.467030	 lr=0.028517
Epoch[137] Batch [3349]/[3760]	Speed: 117.768582 samples/sec	 accuracy=44.478078	 loss=2.467099	 lr=0.028505
Epoch[137] Batch [3399]/[3760]	Speed: 117.639462 samples/sec	 accuracy=44.492188	 loss=2.466899	 lr=0.028493
Epoch[137] Batch [3449]/[3760]	Speed: 117.380356 samples/sec	 accuracy=44.495018	 loss=2.466352	 lr=0.028482
Epoch[137] Batch [3499]/[3760]	Speed: 117.675227 samples/sec	 accuracy=44.503571	 loss=2.465756	 lr=0.028470
Epoch[137] Batch [3549]/[3760]	Speed: 117.360522 samples/sec	 accuracy=44.489877	 loss=2.466474	 lr=0.028458
Epoch[137] Batch [3599]/[3760]	Speed: 117.392156 samples/sec	 accuracy=44.487847	 loss=2.466571	 lr=0.028447
Epoch[137] Batch [3649]/[3760]	Speed: 117.742999 samples/sec	 accuracy=44.500000	 loss=2.466580	 lr=0.028435
Epoch[137] Batch [3699]/[3760]	Speed: 117.875777 samples/sec	 accuracy=44.504645	 loss=2.466678	 lr=0.028424
Epoch[137] Batch [3749]/[3760]	Speed: 120.258581 samples/sec	 accuracy=44.490000	 loss=2.467243	 lr=0.028412
Batch [0049]/[0303]: acc-top1=46.125000 acc-top5=69.781250
Batch [0099]/[0303]: acc-top1=43.125000 acc-top5=68.984375
Batch [0149]/[0303]: acc-top1=44.781250 acc-top5=69.583333
Batch [0199]/[0303]: acc-top1=47.625000 acc-top5=71.585938
Batch [0249]/[0303]: acc-top1=46.693750 acc-top5=71.362500
Batch [0299]/[0303]: acc-top1=45.854167 acc-top5=71.515625
[Epoch 137] training: accuracy=44.497174	 loss=2.466977
[Epoch 137] speed: 116 samples/sec	time cost: 2174.794506
[Epoch 137] validation: acc-top1=45.776609 acc-top5=71.519183 loss=2.492379
Epoch[138] Batch [0049]/[3760]	Speed: 87.175704 samples/sec	 accuracy=45.593750	 loss=2.428923	 lr=0.028398
Epoch[138] Batch [0099]/[3760]	Speed: 116.248689 samples/sec	 accuracy=45.578125	 loss=2.420728	 lr=0.028386
Epoch[138] Batch [0149]/[3760]	Speed: 116.837630 samples/sec	 accuracy=45.312500	 loss=2.412772	 lr=0.028375
Epoch[138] Batch [0199]/[3760]	Speed: 116.399737 samples/sec	 accuracy=45.281250	 loss=2.424414	 lr=0.028363
Epoch[138] Batch [0249]/[3760]	Speed: 117.488319 samples/sec	 accuracy=45.237500	 loss=2.428561	 lr=0.028351
Epoch[138] Batch [0299]/[3760]	Speed: 117.660245 samples/sec	 accuracy=45.270833	 loss=2.429803	 lr=0.028340
Epoch[138] Batch [0349]/[3760]	Speed: 117.273498 samples/sec	 accuracy=45.258929	 loss=2.429529	 lr=0.028328
Epoch[138] Batch [0399]/[3760]	Speed: 117.322229 samples/sec	 accuracy=45.277344	 loss=2.425108	 lr=0.028317
Epoch[138] Batch [0449]/[3760]	Speed: 117.275651 samples/sec	 accuracy=45.225694	 loss=2.427982	 lr=0.028305
Epoch[138] Batch [0499]/[3760]	Speed: 117.005497 samples/sec	 accuracy=45.203125	 loss=2.431624	 lr=0.028293
Epoch[138] Batch [0549]/[3760]	Speed: 117.267853 samples/sec	 accuracy=45.056818	 loss=2.433253	 lr=0.028282
Epoch[138] Batch [0599]/[3760]	Speed: 117.550855 samples/sec	 accuracy=44.981771	 loss=2.433543	 lr=0.028270
Epoch[138] Batch [0649]/[3760]	Speed: 118.107609 samples/sec	 accuracy=44.987981	 loss=2.436026	 lr=0.028258
Epoch[138] Batch [0699]/[3760]	Speed: 117.999113 samples/sec	 accuracy=45.053571	 loss=2.435246	 lr=0.028247
Epoch[138] Batch [0749]/[3760]	Speed: 117.499893 samples/sec	 accuracy=45.037500	 loss=2.434588	 lr=0.028235
Epoch[138] Batch [0799]/[3760]	Speed: 117.460405 samples/sec	 accuracy=45.087891	 loss=2.432687	 lr=0.028224
Epoch[138] Batch [0849]/[3760]	Speed: 116.239552 samples/sec	 accuracy=45.095588	 loss=2.433884	 lr=0.028212
Epoch[138] Batch [0899]/[3760]	Speed: 117.082273 samples/sec	 accuracy=44.984375	 loss=2.438501	 lr=0.028200
Epoch[138] Batch [0949]/[3760]	Speed: 117.749051 samples/sec	 accuracy=44.955592	 loss=2.438223	 lr=0.028189
Epoch[138] Batch [0999]/[3760]	Speed: 117.568383 samples/sec	 accuracy=44.965625	 loss=2.439169	 lr=0.028177
Epoch[138] Batch [1049]/[3760]	Speed: 116.980731 samples/sec	 accuracy=44.909226	 loss=2.441935	 lr=0.028166
Epoch[138] Batch [1099]/[3760]	Speed: 117.397427 samples/sec	 accuracy=44.911932	 loss=2.442504	 lr=0.028154
Epoch[138] Batch [1149]/[3760]	Speed: 117.759843 samples/sec	 accuracy=44.975543	 loss=2.441616	 lr=0.028142
Epoch[138] Batch [1199]/[3760]	Speed: 117.242314 samples/sec	 accuracy=45.019531	 loss=2.440714	 lr=0.028131
Epoch[138] Batch [1249]/[3760]	Speed: 118.051480 samples/sec	 accuracy=45.012500	 loss=2.440632	 lr=0.028119
Epoch[138] Batch [1299]/[3760]	Speed: 115.485080 samples/sec	 accuracy=45.000000	 loss=2.442082	 lr=0.028108
Epoch[138] Batch [1349]/[3760]	Speed: 117.215903 samples/sec	 accuracy=44.918981	 loss=2.445103	 lr=0.028096
Epoch[138] Batch [1399]/[3760]	Speed: 117.671333 samples/sec	 accuracy=44.915179	 loss=2.444209	 lr=0.028084
Epoch[138] Batch [1449]/[3760]	Speed: 117.191135 samples/sec	 accuracy=44.884698	 loss=2.444852	 lr=0.028073
Epoch[138] Batch [1499]/[3760]	Speed: 117.338427 samples/sec	 accuracy=44.893750	 loss=2.445734	 lr=0.028061
Epoch[138] Batch [1549]/[3760]	Speed: 117.532818 samples/sec	 accuracy=44.921371	 loss=2.444172	 lr=0.028050
Epoch[138] Batch [1599]/[3760]	Speed: 117.162242 samples/sec	 accuracy=44.880859	 loss=2.444315	 lr=0.028038
Epoch[138] Batch [1649]/[3760]	Speed: 117.128354 samples/sec	 accuracy=44.863636	 loss=2.445537	 lr=0.028027
Epoch[138] Batch [1699]/[3760]	Speed: 117.925825 samples/sec	 accuracy=44.796875	 loss=2.448537	 lr=0.028015
Epoch[138] Batch [1749]/[3760]	Speed: 118.143681 samples/sec	 accuracy=44.781250	 loss=2.448705	 lr=0.028003
Epoch[138] Batch [1799]/[3760]	Speed: 118.096953 samples/sec	 accuracy=44.844618	 loss=2.447542	 lr=0.027992
Epoch[138] Batch [1849]/[3760]	Speed: 117.246863 samples/sec	 accuracy=44.826858	 loss=2.447636	 lr=0.027980
Epoch[138] Batch [1899]/[3760]	Speed: 117.969618 samples/sec	 accuracy=44.820724	 loss=2.447744	 lr=0.027969
Epoch[138] Batch [1949]/[3760]	Speed: 117.554116 samples/sec	 accuracy=44.816506	 loss=2.447813	 lr=0.027957
Epoch[138] Batch [1999]/[3760]	Speed: 117.253195 samples/sec	 accuracy=44.822656	 loss=2.446913	 lr=0.027945
Epoch[138] Batch [2049]/[3760]	Speed: 117.514560 samples/sec	 accuracy=44.833079	 loss=2.446935	 lr=0.027934
Epoch[138] Batch [2099]/[3760]	Speed: 117.715740 samples/sec	 accuracy=44.856399	 loss=2.446840	 lr=0.027922
Epoch[138] Batch [2149]/[3760]	Speed: 117.206157 samples/sec	 accuracy=44.887355	 loss=2.445997	 lr=0.027911
Epoch[138] Batch [2199]/[3760]	Speed: 117.443002 samples/sec	 accuracy=44.901278	 loss=2.446033	 lr=0.027899
Epoch[138] Batch [2249]/[3760]	Speed: 117.465705 samples/sec	 accuracy=44.902778	 loss=2.446269	 lr=0.027888
Epoch[138] Batch [2299]/[3760]	Speed: 117.502837 samples/sec	 accuracy=44.875000	 loss=2.447210	 lr=0.027876
Epoch[138] Batch [2349]/[3760]	Speed: 117.474833 samples/sec	 accuracy=44.861702	 loss=2.448396	 lr=0.027864
Epoch[138] Batch [2399]/[3760]	Speed: 117.677929 samples/sec	 accuracy=44.856120	 loss=2.448377	 lr=0.027853
Epoch[138] Batch [2449]/[3760]	Speed: 116.837864 samples/sec	 accuracy=44.865434	 loss=2.448513	 lr=0.027841
Epoch[138] Batch [2499]/[3760]	Speed: 116.009898 samples/sec	 accuracy=44.890000	 loss=2.447802	 lr=0.027830
Epoch[138] Batch [2549]/[3760]	Speed: 117.877093 samples/sec	 accuracy=44.897672	 loss=2.447295	 lr=0.027818
Epoch[138] Batch [2599]/[3760]	Speed: 117.927590 samples/sec	 accuracy=44.897837	 loss=2.447671	 lr=0.027807
Epoch[138] Batch [2649]/[3760]	Speed: 117.440395 samples/sec	 accuracy=44.899175	 loss=2.448366	 lr=0.027795
Epoch[138] Batch [2699]/[3760]	Speed: 117.665795 samples/sec	 accuracy=44.875000	 loss=2.449251	 lr=0.027784
Epoch[138] Batch [2749]/[3760]	Speed: 117.062486 samples/sec	 accuracy=44.832386	 loss=2.450671	 lr=0.027772
Epoch[138] Batch [2799]/[3760]	Speed: 117.857692 samples/sec	 accuracy=44.812500	 loss=2.451146	 lr=0.027760
Epoch[138] Batch [2849]/[3760]	Speed: 117.812109 samples/sec	 accuracy=44.797697	 loss=2.452134	 lr=0.027749
Epoch[138] Batch [2899]/[3760]	Speed: 117.620949 samples/sec	 accuracy=44.792565	 loss=2.452148	 lr=0.027737
Epoch[138] Batch [2949]/[3760]	Speed: 116.908023 samples/sec	 accuracy=44.779131	 loss=2.452829	 lr=0.027726
Epoch[138] Batch [2999]/[3760]	Speed: 117.307387 samples/sec	 accuracy=44.789062	 loss=2.452859	 lr=0.027714
Epoch[138] Batch [3049]/[3760]	Speed: 117.108359 samples/sec	 accuracy=44.783299	 loss=2.453668	 lr=0.027703
Epoch[138] Batch [3099]/[3760]	Speed: 117.639568 samples/sec	 accuracy=44.784274	 loss=2.453814	 lr=0.027691
Epoch[138] Batch [3149]/[3760]	Speed: 117.262610 samples/sec	 accuracy=44.820933	 loss=2.451978	 lr=0.027680
Epoch[138] Batch [3199]/[3760]	Speed: 117.782992 samples/sec	 accuracy=44.812988	 loss=2.452262	 lr=0.027668
Epoch[138] Batch [3249]/[3760]	Speed: 117.930181 samples/sec	 accuracy=44.807212	 loss=2.451918	 lr=0.027657
Epoch[138] Batch [3299]/[3760]	Speed: 117.606229 samples/sec	 accuracy=44.800189	 loss=2.452378	 lr=0.027645
Epoch[138] Batch [3349]/[3760]	Speed: 117.335680 samples/sec	 accuracy=44.781716	 loss=2.453535	 lr=0.027633
Epoch[138] Batch [3399]/[3760]	Speed: 117.509407 samples/sec	 accuracy=44.790901	 loss=2.453663	 lr=0.027622
Epoch[138] Batch [3449]/[3760]	Speed: 117.433071 samples/sec	 accuracy=44.793478	 loss=2.453570	 lr=0.027610
Epoch[138] Batch [3499]/[3760]	Speed: 117.540502 samples/sec	 accuracy=44.782589	 loss=2.453453	 lr=0.027599
Epoch[138] Batch [3549]/[3760]	Speed: 118.641016 samples/sec	 accuracy=44.760123	 loss=2.454049	 lr=0.027587
Epoch[138] Batch [3599]/[3760]	Speed: 117.756377 samples/sec	 accuracy=44.746528	 loss=2.454668	 lr=0.027576
Epoch[138] Batch [3649]/[3760]	Speed: 118.201879 samples/sec	 accuracy=44.779538	 loss=2.453767	 lr=0.027564
Epoch[138] Batch [3699]/[3760]	Speed: 115.407211 samples/sec	 accuracy=44.766047	 loss=2.454702	 lr=0.027553
Epoch[138] Batch [3749]/[3760]	Speed: 120.412640 samples/sec	 accuracy=44.762083	 loss=2.454479	 lr=0.027541
Batch [0049]/[0303]: acc-top1=44.812500 acc-top5=68.468750
Batch [0099]/[0303]: acc-top1=41.250000 acc-top5=66.843750
Batch [0149]/[0303]: acc-top1=43.531250 acc-top5=68.479167
Batch [0199]/[0303]: acc-top1=46.515625 acc-top5=70.593750
Batch [0249]/[0303]: acc-top1=45.893750 acc-top5=70.550000
Batch [0299]/[0303]: acc-top1=45.338542 acc-top5=70.583333
[Epoch 138] training: accuracy=44.759807	 loss=2.454676
[Epoch 138] speed: 116 samples/sec	time cost: 2175.762215
[Epoch 138] validation: acc-top1=45.271246 acc-top5=70.565182 loss=2.512373
Epoch[139] Batch [0049]/[3759]	Speed: 88.983366 samples/sec	 accuracy=46.500000	 loss=2.368841	 lr=0.027527
Epoch[139] Batch [0099]/[3759]	Speed: 118.299492 samples/sec	 accuracy=45.640625	 loss=2.391175	 lr=0.027516
Epoch[139] Batch [0149]/[3759]	Speed: 117.935961 samples/sec	 accuracy=45.989583	 loss=2.391678	 lr=0.027504
Epoch[139] Batch [0199]/[3759]	Speed: 117.545836 samples/sec	 accuracy=46.179688	 loss=2.396175	 lr=0.027493
Epoch[139] Batch [0249]/[3759]	Speed: 118.351703 samples/sec	 accuracy=46.312500	 loss=2.392904	 lr=0.027481
Epoch[139] Batch [0299]/[3759]	Speed: 117.236663 samples/sec	 accuracy=46.265625	 loss=2.398948	 lr=0.027470
Epoch[139] Batch [0349]/[3759]	Speed: 117.731262 samples/sec	 accuracy=46.049107	 loss=2.407846	 lr=0.027458
Epoch[139] Batch [0399]/[3759]	Speed: 117.918220 samples/sec	 accuracy=45.855469	 loss=2.410340	 lr=0.027447
Epoch[139] Batch [0449]/[3759]	Speed: 118.310393 samples/sec	 accuracy=45.739583	 loss=2.412381	 lr=0.027435
Epoch[139] Batch [0499]/[3759]	Speed: 117.780486 samples/sec	 accuracy=45.953125	 loss=2.403761	 lr=0.027424
Epoch[139] Batch [0549]/[3759]	Speed: 117.028742 samples/sec	 accuracy=46.017045	 loss=2.401656	 lr=0.027412
Epoch[139] Batch [0599]/[3759]	Speed: 117.729205 samples/sec	 accuracy=45.973958	 loss=2.401476	 lr=0.027401
Epoch[139] Batch [0649]/[3759]	Speed: 117.695092 samples/sec	 accuracy=45.860577	 loss=2.406130	 lr=0.027389
Epoch[139] Batch [0699]/[3759]	Speed: 117.192175 samples/sec	 accuracy=45.828125	 loss=2.405124	 lr=0.027378
Epoch[139] Batch [0749]/[3759]	Speed: 117.151902 samples/sec	 accuracy=45.777083	 loss=2.408957	 lr=0.027366
Epoch[139] Batch [0799]/[3759]	Speed: 117.895125 samples/sec	 accuracy=45.691406	 loss=2.413911	 lr=0.027355
Epoch[139] Batch [0849]/[3759]	Speed: 118.170681 samples/sec	 accuracy=45.641544	 loss=2.417488	 lr=0.027343
Epoch[139] Batch [0899]/[3759]	Speed: 117.599714 samples/sec	 accuracy=45.595486	 loss=2.418939	 lr=0.027332
Epoch[139] Batch [0949]/[3759]	Speed: 116.978658 samples/sec	 accuracy=45.565789	 loss=2.420550	 lr=0.027320
Epoch[139] Batch [0999]/[3759]	Speed: 116.489229 samples/sec	 accuracy=45.600000	 loss=2.418695	 lr=0.027309
Epoch[139] Batch [1049]/[3759]	Speed: 117.612921 samples/sec	 accuracy=45.593750	 loss=2.418389	 lr=0.027297
Epoch[139] Batch [1099]/[3759]	Speed: 117.152357 samples/sec	 accuracy=45.566761	 loss=2.418502	 lr=0.027286
Epoch[139] Batch [1149]/[3759]	Speed: 117.287458 samples/sec	 accuracy=45.471467	 loss=2.420760	 lr=0.027274
Epoch[139] Batch [1199]/[3759]	Speed: 117.841951 samples/sec	 accuracy=45.494792	 loss=2.420210	 lr=0.027263
Epoch[139] Batch [1249]/[3759]	Speed: 118.101426 samples/sec	 accuracy=45.511250	 loss=2.419213	 lr=0.027251
Epoch[139] Batch [1299]/[3759]	Speed: 117.019134 samples/sec	 accuracy=45.475962	 loss=2.420228	 lr=0.027240
Epoch[139] Batch [1349]/[3759]	Speed: 117.118772 samples/sec	 accuracy=45.467593	 loss=2.418913	 lr=0.027228
Epoch[139] Batch [1399]/[3759]	Speed: 117.480049 samples/sec	 accuracy=45.426339	 loss=2.419504	 lr=0.027217
Epoch[139] Batch [1449]/[3759]	Speed: 117.653983 samples/sec	 accuracy=45.357759	 loss=2.422648	 lr=0.027205
Epoch[139] Batch [1499]/[3759]	Speed: 117.771772 samples/sec	 accuracy=45.334375	 loss=2.423571	 lr=0.027194
Epoch[139] Batch [1549]/[3759]	Speed: 117.119229 samples/sec	 accuracy=45.332661	 loss=2.424875	 lr=0.027182
Epoch[139] Batch [1599]/[3759]	Speed: 117.385709 samples/sec	 accuracy=45.346680	 loss=2.423842	 lr=0.027171
Epoch[139] Batch [1649]/[3759]	Speed: 117.320419 samples/sec	 accuracy=45.324811	 loss=2.424827	 lr=0.027159
Epoch[139] Batch [1699]/[3759]	Speed: 116.601491 samples/sec	 accuracy=45.354779	 loss=2.424863	 lr=0.027148
Epoch[139] Batch [1749]/[3759]	Speed: 117.803139 samples/sec	 accuracy=45.333036	 loss=2.425867	 lr=0.027137
Epoch[139] Batch [1799]/[3759]	Speed: 117.497357 samples/sec	 accuracy=45.293403	 loss=2.428008	 lr=0.027125
Epoch[139] Batch [1849]/[3759]	Speed: 117.450015 samples/sec	 accuracy=45.276182	 loss=2.428505	 lr=0.027114
Epoch[139] Batch [1899]/[3759]	Speed: 118.064066 samples/sec	 accuracy=45.242599	 loss=2.427840	 lr=0.027102
Epoch[139] Batch [1949]/[3759]	Speed: 117.779683 samples/sec	 accuracy=45.258814	 loss=2.427976	 lr=0.027091
Epoch[139] Batch [1999]/[3759]	Speed: 117.751112 samples/sec	 accuracy=45.260156	 loss=2.427387	 lr=0.027079
Epoch[139] Batch [2049]/[3759]	Speed: 117.369458 samples/sec	 accuracy=45.263720	 loss=2.426468	 lr=0.027068
Epoch[139] Batch [2099]/[3759]	Speed: 118.050661 samples/sec	 accuracy=45.261905	 loss=2.427284	 lr=0.027056
Epoch[139] Batch [2149]/[3759]	Speed: 117.978511 samples/sec	 accuracy=45.218750	 loss=2.428090	 lr=0.027045
Epoch[139] Batch [2199]/[3759]	Speed: 116.149673 samples/sec	 accuracy=45.221591	 loss=2.428695	 lr=0.027033
Epoch[139] Batch [2249]/[3759]	Speed: 117.815578 samples/sec	 accuracy=45.194444	 loss=2.429689	 lr=0.027022
Epoch[139] Batch [2299]/[3759]	Speed: 117.788212 samples/sec	 accuracy=45.177989	 loss=2.430073	 lr=0.027010
Epoch[139] Batch [2349]/[3759]	Speed: 117.848050 samples/sec	 accuracy=45.178856	 loss=2.429774	 lr=0.026999
Epoch[139] Batch [2399]/[3759]	Speed: 117.185199 samples/sec	 accuracy=45.178385	 loss=2.429855	 lr=0.026988
Epoch[139] Batch [2449]/[3759]	Speed: 117.317984 samples/sec	 accuracy=45.198980	 loss=2.429573	 lr=0.026976
Epoch[139] Batch [2499]/[3759]	Speed: 117.272955 samples/sec	 accuracy=45.180625	 loss=2.429767	 lr=0.026965
Epoch[139] Batch [2549]/[3759]	Speed: 117.807221 samples/sec	 accuracy=45.170956	 loss=2.429873	 lr=0.026953
Epoch[139] Batch [2599]/[3759]	Speed: 118.364982 samples/sec	 accuracy=45.152644	 loss=2.430359	 lr=0.026942
Epoch[139] Batch [2649]/[3759]	Speed: 116.651847 samples/sec	 accuracy=45.168042	 loss=2.429428	 lr=0.026930
Epoch[139] Batch [2699]/[3759]	Speed: 118.458246 samples/sec	 accuracy=45.159144	 loss=2.429668	 lr=0.026919
Epoch[139] Batch [2749]/[3759]	Speed: 117.070588 samples/sec	 accuracy=45.138636	 loss=2.430743	 lr=0.026907
Epoch[139] Batch [2799]/[3759]	Speed: 118.329653 samples/sec	 accuracy=45.135603	 loss=2.431115	 lr=0.026896
Epoch[139] Batch [2849]/[3759]	Speed: 117.531489 samples/sec	 accuracy=45.139254	 loss=2.431851	 lr=0.026885
Epoch[139] Batch [2899]/[3759]	Speed: 117.524679 samples/sec	 accuracy=45.130927	 loss=2.432256	 lr=0.026873
Epoch[139] Batch [2949]/[3759]	Speed: 117.606756 samples/sec	 accuracy=45.128178	 loss=2.432297	 lr=0.026862
Epoch[139] Batch [2999]/[3759]	Speed: 117.321843 samples/sec	 accuracy=45.145313	 loss=2.431561	 lr=0.026850
Epoch[139] Batch [3049]/[3759]	Speed: 117.881075 samples/sec	 accuracy=45.121414	 loss=2.432645	 lr=0.026839
Epoch[139] Batch [3099]/[3759]	Speed: 116.518042 samples/sec	 accuracy=45.143649	 loss=2.431902	 lr=0.026827
Epoch[139] Batch [3149]/[3759]	Speed: 117.732418 samples/sec	 accuracy=45.117560	 loss=2.432406	 lr=0.026816
Epoch[139] Batch [3199]/[3759]	Speed: 117.964110 samples/sec	 accuracy=45.124512	 loss=2.432384	 lr=0.026805
Epoch[139] Batch [3249]/[3759]	Speed: 116.308539 samples/sec	 accuracy=45.131731	 loss=2.432683	 lr=0.026793
Epoch[139] Batch [3299]/[3759]	Speed: 117.409813 samples/sec	 accuracy=45.135890	 loss=2.432206	 lr=0.026782
Epoch[139] Batch [3349]/[3759]	Speed: 117.156769 samples/sec	 accuracy=45.141791	 loss=2.431851	 lr=0.026770
Epoch[139] Batch [3399]/[3759]	Speed: 115.440279 samples/sec	 accuracy=45.123162	 loss=2.432753	 lr=0.026759
Epoch[139] Batch [3449]/[3759]	Speed: 118.010984 samples/sec	 accuracy=45.132246	 loss=2.432082	 lr=0.026747
Epoch[139] Batch [3499]/[3759]	Speed: 116.982644 samples/sec	 accuracy=45.136161	 loss=2.431981	 lr=0.026736
Epoch[139] Batch [3549]/[3759]	Speed: 117.332621 samples/sec	 accuracy=45.141285	 loss=2.431521	 lr=0.026725
Epoch[139] Batch [3599]/[3759]	Speed: 117.757577 samples/sec	 accuracy=45.136719	 loss=2.432144	 lr=0.026713
Epoch[139] Batch [3649]/[3759]	Speed: 117.501043 samples/sec	 accuracy=45.147260	 loss=2.432015	 lr=0.026702
Epoch[139] Batch [3699]/[3759]	Speed: 117.732896 samples/sec	 accuracy=45.128801	 loss=2.432896	 lr=0.026690
Epoch[139] Batch [3749]/[3759]	Speed: 120.408788 samples/sec	 accuracy=45.111250	 loss=2.433676	 lr=0.026679
Batch [0049]/[0303]: acc-top1=45.656250 acc-top5=70.437500
Batch [0099]/[0303]: acc-top1=43.531250 acc-top5=69.375000
Batch [0149]/[0303]: acc-top1=44.906250 acc-top5=70.270833
Batch [0199]/[0303]: acc-top1=47.484375 acc-top5=71.929688
Batch [0249]/[0303]: acc-top1=47.006250 acc-top5=71.750000
Batch [0299]/[0303]: acc-top1=46.411458 acc-top5=71.713542
[Epoch 139] training: accuracy=45.112148	 loss=2.433632
[Epoch 139] speed: 117 samples/sec	time cost: 2172.679609
[Epoch 139] validation: acc-top1=46.416048 acc-top5=71.684200 loss=2.473795
Epoch[140] Batch [0049]/[3760]	Speed: 87.167664 samples/sec	 accuracy=45.750000	 loss=2.385765	 lr=0.026666
Epoch[140] Batch [0099]/[3760]	Speed: 117.267445 samples/sec	 accuracy=45.640625	 loss=2.389182	 lr=0.026654
Epoch[140] Batch [0149]/[3760]	Speed: 117.840831 samples/sec	 accuracy=45.416667	 loss=2.407494	 lr=0.026643
Epoch[140] Batch [0199]/[3760]	Speed: 117.817419 samples/sec	 accuracy=45.070312	 loss=2.416534	 lr=0.026631
Epoch[140] Batch [0249]/[3760]	Speed: 116.965546 samples/sec	 accuracy=45.406250	 loss=2.413709	 lr=0.026620
Epoch[140] Batch [0299]/[3760]	Speed: 117.467732 samples/sec	 accuracy=45.432292	 loss=2.413874	 lr=0.026608
Epoch[140] Batch [0349]/[3760]	Speed: 117.941416 samples/sec	 accuracy=45.477679	 loss=2.419124	 lr=0.026597
Epoch[140] Batch [0399]/[3760]	Speed: 117.052664 samples/sec	 accuracy=45.453125	 loss=2.416085	 lr=0.026586
Epoch[140] Batch [0449]/[3760]	Speed: 116.782331 samples/sec	 accuracy=45.458333	 loss=2.414871	 lr=0.026574
Epoch[140] Batch [0499]/[3760]	Speed: 117.013533 samples/sec	 accuracy=45.403125	 loss=2.415118	 lr=0.026563
Epoch[140] Batch [0549]/[3760]	Speed: 117.923750 samples/sec	 accuracy=45.346591	 loss=2.418611	 lr=0.026552
Epoch[140] Batch [0599]/[3760]	Speed: 117.566131 samples/sec	 accuracy=45.299479	 loss=2.419231	 lr=0.026540
Epoch[140] Batch [0649]/[3760]	Speed: 116.542958 samples/sec	 accuracy=45.271635	 loss=2.420600	 lr=0.026529
Epoch[140] Batch [0699]/[3760]	Speed: 117.571784 samples/sec	 accuracy=45.354911	 loss=2.414084	 lr=0.026517
Epoch[140] Batch [0749]/[3760]	Speed: 117.686286 samples/sec	 accuracy=45.350000	 loss=2.414918	 lr=0.026506
Epoch[140] Batch [0799]/[3760]	Speed: 117.439915 samples/sec	 accuracy=45.443359	 loss=2.412057	 lr=0.026495
Epoch[140] Batch [0849]/[3760]	Speed: 118.523403 samples/sec	 accuracy=45.349265	 loss=2.412907	 lr=0.026483
Epoch[140] Batch [0899]/[3760]	Speed: 117.069463 samples/sec	 accuracy=45.394097	 loss=2.410080	 lr=0.026472
Epoch[140] Batch [0949]/[3760]	Speed: 117.788888 samples/sec	 accuracy=45.391447	 loss=2.411571	 lr=0.026460
Epoch[140] Batch [0999]/[3760]	Speed: 117.197999 samples/sec	 accuracy=45.445312	 loss=2.407273	 lr=0.026449
Epoch[140] Batch [1049]/[3760]	Speed: 117.635025 samples/sec	 accuracy=45.444940	 loss=2.407613	 lr=0.026438
Epoch[140] Batch [1099]/[3760]	Speed: 117.914249 samples/sec	 accuracy=45.458807	 loss=2.408206	 lr=0.026426
Epoch[140] Batch [1149]/[3760]	Speed: 117.522699 samples/sec	 accuracy=45.438859	 loss=2.409083	 lr=0.026415
Epoch[140] Batch [1199]/[3760]	Speed: 117.653195 samples/sec	 accuracy=45.410156	 loss=2.410942	 lr=0.026404
Epoch[140] Batch [1249]/[3760]	Speed: 117.618759 samples/sec	 accuracy=45.412500	 loss=2.411336	 lr=0.026392
Epoch[140] Batch [1299]/[3760]	Speed: 117.966375 samples/sec	 accuracy=45.399038	 loss=2.412777	 lr=0.026381
Epoch[140] Batch [1349]/[3760]	Speed: 115.299559 samples/sec	 accuracy=45.384259	 loss=2.413035	 lr=0.026369
Epoch[140] Batch [1399]/[3760]	Speed: 117.581479 samples/sec	 accuracy=45.426339	 loss=2.411280	 lr=0.026358
Epoch[140] Batch [1449]/[3760]	Speed: 117.816367 samples/sec	 accuracy=45.464440	 loss=2.408860	 lr=0.026347
Epoch[140] Batch [1499]/[3760]	Speed: 117.859142 samples/sec	 accuracy=45.466667	 loss=2.410198	 lr=0.026335
Epoch[140] Batch [1549]/[3760]	Speed: 116.792086 samples/sec	 accuracy=45.435484	 loss=2.412386	 lr=0.026324
Epoch[140] Batch [1599]/[3760]	Speed: 117.508429 samples/sec	 accuracy=45.426758	 loss=2.412974	 lr=0.026313
Epoch[140] Batch [1649]/[3760]	Speed: 117.805397 samples/sec	 accuracy=45.417614	 loss=2.413022	 lr=0.026301
Epoch[140] Batch [1699]/[3760]	Speed: 117.350745 samples/sec	 accuracy=45.442096	 loss=2.411914	 lr=0.026290
Epoch[140] Batch [1749]/[3760]	Speed: 117.707892 samples/sec	 accuracy=45.461607	 loss=2.410666	 lr=0.026279
Epoch[140] Batch [1799]/[3760]	Speed: 117.515716 samples/sec	 accuracy=45.468750	 loss=2.411393	 lr=0.026267
Epoch[140] Batch [1849]/[3760]	Speed: 117.375560 samples/sec	 accuracy=45.451858	 loss=2.412885	 lr=0.026256
Epoch[140] Batch [1899]/[3760]	Speed: 117.789936 samples/sec	 accuracy=45.469572	 loss=2.413066	 lr=0.026245
Epoch[140] Batch [1949]/[3760]	Speed: 117.252679 samples/sec	 accuracy=45.451923	 loss=2.412984	 lr=0.026233
Epoch[140] Batch [1999]/[3760]	Speed: 117.923411 samples/sec	 accuracy=45.458594	 loss=2.412649	 lr=0.026222
Epoch[140] Batch [2049]/[3760]	Speed: 117.623383 samples/sec	 accuracy=45.442835	 loss=2.413237	 lr=0.026210
Epoch[140] Batch [2099]/[3760]	Speed: 117.239740 samples/sec	 accuracy=45.392113	 loss=2.415872	 lr=0.026199
Epoch[140] Batch [2149]/[3760]	Speed: 117.899247 samples/sec	 accuracy=45.372820	 loss=2.416680	 lr=0.026188
Epoch[140] Batch [2199]/[3760]	Speed: 117.579344 samples/sec	 accuracy=45.383523	 loss=2.416678	 lr=0.026176
Epoch[140] Batch [2249]/[3760]	Speed: 117.859404 samples/sec	 accuracy=45.375000	 loss=2.417734	 lr=0.026165
Epoch[140] Batch [2299]/[3760]	Speed: 117.658404 samples/sec	 accuracy=45.381114	 loss=2.418611	 lr=0.026154
Epoch[140] Batch [2349]/[3760]	Speed: 116.987186 samples/sec	 accuracy=45.403590	 loss=2.418551	 lr=0.026142
Epoch[140] Batch [2399]/[3760]	Speed: 116.752351 samples/sec	 accuracy=45.395182	 loss=2.419026	 lr=0.026131
Epoch[140] Batch [2449]/[3760]	Speed: 117.631861 samples/sec	 accuracy=45.382653	 loss=2.419752	 lr=0.026120
Epoch[140] Batch [2499]/[3760]	Speed: 117.005013 samples/sec	 accuracy=45.395000	 loss=2.419204	 lr=0.026108
Epoch[140] Batch [2549]/[3760]	Speed: 115.309177 samples/sec	 accuracy=45.407475	 loss=2.418688	 lr=0.026097
Epoch[140] Batch [2599]/[3760]	Speed: 117.069875 samples/sec	 accuracy=45.409255	 loss=2.418657	 lr=0.026086
Epoch[140] Batch [2649]/[3760]	Speed: 118.281287 samples/sec	 accuracy=45.380896	 loss=2.420085	 lr=0.026074
Epoch[140] Batch [2699]/[3760]	Speed: 117.827902 samples/sec	 accuracy=45.412037	 loss=2.419661	 lr=0.026063
Epoch[140] Batch [2749]/[3760]	Speed: 117.999852 samples/sec	 accuracy=45.415909	 loss=2.419193	 lr=0.026052
Epoch[140] Batch [2799]/[3760]	Speed: 117.886698 samples/sec	 accuracy=45.409040	 loss=2.418640	 lr=0.026040
Epoch[140] Batch [2849]/[3760]	Speed: 118.172798 samples/sec	 accuracy=45.394737	 loss=2.419583	 lr=0.026029
Epoch[140] Batch [2899]/[3760]	Speed: 117.777731 samples/sec	 accuracy=45.402478	 loss=2.418957	 lr=0.026018
Epoch[140] Batch [2949]/[3760]	Speed: 117.587027 samples/sec	 accuracy=45.381886	 loss=2.420139	 lr=0.026007
Epoch[140] Batch [2999]/[3760]	Speed: 117.823809 samples/sec	 accuracy=45.377604	 loss=2.420321	 lr=0.025995
Epoch[140] Batch [3049]/[3760]	Speed: 117.813811 samples/sec	 accuracy=45.370902	 loss=2.420078	 lr=0.025984
Epoch[140] Batch [3099]/[3760]	Speed: 118.455134 samples/sec	 accuracy=45.389617	 loss=2.420085	 lr=0.025973
Epoch[140] Batch [3149]/[3760]	Speed: 117.642144 samples/sec	 accuracy=45.413194	 loss=2.419801	 lr=0.025961
Epoch[140] Batch [3199]/[3760]	Speed: 117.887985 samples/sec	 accuracy=45.393066	 loss=2.420438	 lr=0.025950
Epoch[140] Batch [3249]/[3760]	Speed: 118.251873 samples/sec	 accuracy=45.402885	 loss=2.420173	 lr=0.025939
Epoch[140] Batch [3299]/[3760]	Speed: 116.962207 samples/sec	 accuracy=45.378788	 loss=2.420158	 lr=0.025927
Epoch[140] Batch [3349]/[3760]	Speed: 117.015552 samples/sec	 accuracy=45.401586	 loss=2.419494	 lr=0.025916
Epoch[140] Batch [3399]/[3760]	Speed: 117.418053 samples/sec	 accuracy=45.405790	 loss=2.419088	 lr=0.025905
Epoch[140] Batch [3449]/[3760]	Speed: 117.760934 samples/sec	 accuracy=45.398098	 loss=2.419731	 lr=0.025893
Epoch[140] Batch [3499]/[3760]	Speed: 117.640455 samples/sec	 accuracy=45.375000	 loss=2.420579	 lr=0.025882
Epoch[140] Batch [3549]/[3760]	Speed: 117.767746 samples/sec	 accuracy=45.388204	 loss=2.420299	 lr=0.025871
Epoch[140] Batch [3599]/[3760]	Speed: 118.182527 samples/sec	 accuracy=45.401910	 loss=2.419665	 lr=0.025860
Epoch[140] Batch [3649]/[3760]	Speed: 117.542726 samples/sec	 accuracy=45.390839	 loss=2.419632	 lr=0.025848
Epoch[140] Batch [3699]/[3760]	Speed: 117.771781 samples/sec	 accuracy=45.399916	 loss=2.419147	 lr=0.025837
Epoch[140] Batch [3749]/[3760]	Speed: 120.407550 samples/sec	 accuracy=45.380833	 loss=2.419885	 lr=0.025826
Batch [0049]/[0303]: acc-top1=46.593750 acc-top5=69.437500
Batch [0099]/[0303]: acc-top1=43.734375 acc-top5=69.343750
Batch [0149]/[0303]: acc-top1=45.375000 acc-top5=70.427083
Batch [0199]/[0303]: acc-top1=47.757812 acc-top5=72.015625
Batch [0249]/[0303]: acc-top1=47.125000 acc-top5=71.956250
Batch [0299]/[0303]: acc-top1=46.557292 acc-top5=72.020833
[Epoch 140] training: accuracy=45.386054	 loss=2.419442
[Epoch 140] speed: 117 samples/sec	time cost: 2173.786370
[Epoch 140] validation: acc-top1=46.606848 acc-top5=72.096741 loss=2.429477
Epoch[141] Batch [0049]/[3760]	Speed: 88.942230 samples/sec	 accuracy=45.750000	 loss=2.359158	 lr=0.025812
Epoch[141] Batch [0099]/[3760]	Speed: 116.828171 samples/sec	 accuracy=46.078125	 loss=2.341738	 lr=0.025801
Epoch[141] Batch [0149]/[3760]	Speed: 117.540379 samples/sec	 accuracy=45.885417	 loss=2.364736	 lr=0.025790
Epoch[141] Batch [0199]/[3760]	Speed: 117.898549 samples/sec	 accuracy=46.109375	 loss=2.351500	 lr=0.025778
Epoch[141] Batch [0249]/[3760]	Speed: 117.803874 samples/sec	 accuracy=46.325000	 loss=2.352767	 lr=0.025767
Epoch[141] Batch [0299]/[3760]	Speed: 116.419820 samples/sec	 accuracy=46.317708	 loss=2.353585	 lr=0.025756
Epoch[141] Batch [0349]/[3760]	Speed: 117.278801 samples/sec	 accuracy=46.361607	 loss=2.352470	 lr=0.025744
Epoch[141] Batch [0399]/[3760]	Speed: 116.649890 samples/sec	 accuracy=46.109375	 loss=2.363557	 lr=0.025733
Epoch[141] Batch [0449]/[3760]	Speed: 117.311790 samples/sec	 accuracy=46.013889	 loss=2.366755	 lr=0.025722
Epoch[141] Batch [0499]/[3760]	Speed: 117.618194 samples/sec	 accuracy=45.987500	 loss=2.372242	 lr=0.025711
Epoch[141] Batch [0549]/[3760]	Speed: 117.383157 samples/sec	 accuracy=45.886364	 loss=2.376338	 lr=0.025699
Epoch[141] Batch [0599]/[3760]	Speed: 117.363714 samples/sec	 accuracy=45.770833	 loss=2.383080	 lr=0.025688
Epoch[141] Batch [0649]/[3760]	Speed: 117.708714 samples/sec	 accuracy=45.721154	 loss=2.384342	 lr=0.025677
Epoch[141] Batch [0699]/[3760]	Speed: 118.377186 samples/sec	 accuracy=45.841518	 loss=2.380113	 lr=0.025666
Epoch[141] Batch [0749]/[3760]	Speed: 117.255850 samples/sec	 accuracy=45.816667	 loss=2.381058	 lr=0.025654
Epoch[141] Batch [0799]/[3760]	Speed: 117.649777 samples/sec	 accuracy=45.925781	 loss=2.378433	 lr=0.025643
Epoch[141] Batch [0849]/[3760]	Speed: 117.596780 samples/sec	 accuracy=45.961397	 loss=2.377285	 lr=0.025632
Epoch[141] Batch [0899]/[3760]	Speed: 117.293891 samples/sec	 accuracy=46.006944	 loss=2.375673	 lr=0.025620
Epoch[141] Batch [0949]/[3760]	Speed: 117.029200 samples/sec	 accuracy=45.985197	 loss=2.376734	 lr=0.025609
Epoch[141] Batch [0999]/[3760]	Speed: 116.893403 samples/sec	 accuracy=45.973437	 loss=2.376883	 lr=0.025598
Epoch[141] Batch [1049]/[3760]	Speed: 117.833893 samples/sec	 accuracy=45.949405	 loss=2.377023	 lr=0.025587
Epoch[141] Batch [1099]/[3760]	Speed: 117.682061 samples/sec	 accuracy=45.924716	 loss=2.378808	 lr=0.025575
Epoch[141] Batch [1149]/[3760]	Speed: 117.933760 samples/sec	 accuracy=45.945652	 loss=2.377053	 lr=0.025564
Epoch[141] Batch [1199]/[3760]	Speed: 116.919099 samples/sec	 accuracy=45.957031	 loss=2.378001	 lr=0.025553
Epoch[141] Batch [1249]/[3760]	Speed: 117.497995 samples/sec	 accuracy=45.887500	 loss=2.380079	 lr=0.025542
Epoch[141] Batch [1299]/[3760]	Speed: 115.136995 samples/sec	 accuracy=45.896635	 loss=2.380583	 lr=0.025530
Epoch[141] Batch [1349]/[3760]	Speed: 118.027759 samples/sec	 accuracy=45.968750	 loss=2.378672	 lr=0.025519
Epoch[141] Batch [1399]/[3760]	Speed: 117.219926 samples/sec	 accuracy=45.973214	 loss=2.377743	 lr=0.025508
Epoch[141] Batch [1449]/[3760]	Speed: 117.515944 samples/sec	 accuracy=45.935345	 loss=2.380077	 lr=0.025497
Epoch[141] Batch [1499]/[3760]	Speed: 116.472480 samples/sec	 accuracy=45.919792	 loss=2.381360	 lr=0.025485
Epoch[141] Batch [1549]/[3760]	Speed: 117.649280 samples/sec	 accuracy=45.872984	 loss=2.382840	 lr=0.025474
Epoch[141] Batch [1599]/[3760]	Speed: 117.972473 samples/sec	 accuracy=45.849609	 loss=2.383802	 lr=0.025463
Epoch[141] Batch [1649]/[3760]	Speed: 116.942736 samples/sec	 accuracy=45.843750	 loss=2.384617	 lr=0.025452
Epoch[141] Batch [1699]/[3760]	Speed: 117.843900 samples/sec	 accuracy=45.806985	 loss=2.387359	 lr=0.025441
Epoch[141] Batch [1749]/[3760]	Speed: 117.743589 samples/sec	 accuracy=45.780357	 loss=2.389461	 lr=0.025429
Epoch[141] Batch [1799]/[3760]	Speed: 117.476316 samples/sec	 accuracy=45.761285	 loss=2.390185	 lr=0.025418
Epoch[141] Batch [1849]/[3760]	Speed: 117.573659 samples/sec	 accuracy=45.727196	 loss=2.392261	 lr=0.025407
Epoch[141] Batch [1899]/[3760]	Speed: 117.831009 samples/sec	 accuracy=45.736842	 loss=2.392594	 lr=0.025396
Epoch[141] Batch [1949]/[3760]	Speed: 117.102794 samples/sec	 accuracy=45.724359	 loss=2.393905	 lr=0.025384
Epoch[141] Batch [1999]/[3760]	Speed: 117.020630 samples/sec	 accuracy=45.757812	 loss=2.392542	 lr=0.025373
Epoch[141] Batch [2049]/[3760]	Speed: 117.439853 samples/sec	 accuracy=45.744665	 loss=2.394204	 lr=0.025362
Epoch[141] Batch [2099]/[3760]	Speed: 118.108945 samples/sec	 accuracy=45.730655	 loss=2.394680	 lr=0.025351
Epoch[141] Batch [2149]/[3760]	Speed: 117.150441 samples/sec	 accuracy=45.721657	 loss=2.394862	 lr=0.025339
Epoch[141] Batch [2199]/[3760]	Speed: 117.237932 samples/sec	 accuracy=45.753551	 loss=2.393765	 lr=0.025328
Epoch[141] Batch [2249]/[3760]	Speed: 117.601759 samples/sec	 accuracy=45.752778	 loss=2.393663	 lr=0.025317
Epoch[141] Batch [2299]/[3760]	Speed: 117.334341 samples/sec	 accuracy=45.747962	 loss=2.393490	 lr=0.025306
Epoch[141] Batch [2349]/[3760]	Speed: 117.769716 samples/sec	 accuracy=45.762633	 loss=2.393144	 lr=0.025295
Epoch[141] Batch [2399]/[3760]	Speed: 117.918929 samples/sec	 accuracy=45.736979	 loss=2.394903	 lr=0.025283
Epoch[141] Batch [2449]/[3760]	Speed: 117.247106 samples/sec	 accuracy=45.752551	 loss=2.394327	 lr=0.025272
Epoch[141] Batch [2499]/[3760]	Speed: 118.353750 samples/sec	 accuracy=45.752500	 loss=2.394602	 lr=0.025261
Epoch[141] Batch [2549]/[3760]	Speed: 115.505086 samples/sec	 accuracy=45.755515	 loss=2.395956	 lr=0.025250
Epoch[141] Batch [2599]/[3760]	Speed: 117.831582 samples/sec	 accuracy=45.753606	 loss=2.396114	 lr=0.025239
Epoch[141] Batch [2649]/[3760]	Speed: 117.457198 samples/sec	 accuracy=45.715802	 loss=2.397898	 lr=0.025227
Epoch[141] Batch [2699]/[3760]	Speed: 117.842757 samples/sec	 accuracy=45.694444	 loss=2.398877	 lr=0.025216
Epoch[141] Batch [2749]/[3760]	Speed: 118.075689 samples/sec	 accuracy=45.717614	 loss=2.398845	 lr=0.025205
Epoch[141] Batch [2799]/[3760]	Speed: 117.148051 samples/sec	 accuracy=45.726562	 loss=2.398938	 lr=0.025194
Epoch[141] Batch [2849]/[3760]	Speed: 118.243495 samples/sec	 accuracy=45.706689	 loss=2.399753	 lr=0.025183
Epoch[141] Batch [2899]/[3760]	Speed: 117.712426 samples/sec	 accuracy=45.712823	 loss=2.400265	 lr=0.025171
Epoch[141] Batch [2949]/[3760]	Speed: 117.083065 samples/sec	 accuracy=45.718220	 loss=2.399859	 lr=0.025160
Epoch[141] Batch [2999]/[3760]	Speed: 116.979078 samples/sec	 accuracy=45.701042	 loss=2.400402	 lr=0.025149
Epoch[141] Batch [3049]/[3760]	Speed: 117.916507 samples/sec	 accuracy=45.679816	 loss=2.400901	 lr=0.025138
Epoch[141] Batch [3099]/[3760]	Speed: 117.316704 samples/sec	 accuracy=45.673387	 loss=2.401471	 lr=0.025127
Epoch[141] Batch [3149]/[3760]	Speed: 117.325095 samples/sec	 accuracy=45.639881	 loss=2.402657	 lr=0.025115
Epoch[141] Batch [3199]/[3760]	Speed: 117.026952 samples/sec	 accuracy=45.628418	 loss=2.402972	 lr=0.025104
Epoch[141] Batch [3249]/[3760]	Speed: 118.250478 samples/sec	 accuracy=45.637500	 loss=2.402851	 lr=0.025093
Epoch[141] Batch [3299]/[3760]	Speed: 117.778225 samples/sec	 accuracy=45.636837	 loss=2.402861	 lr=0.025082
Epoch[141] Batch [3349]/[3760]	Speed: 117.144866 samples/sec	 accuracy=45.649720	 loss=2.401699	 lr=0.025071
Epoch[141] Batch [3399]/[3760]	Speed: 117.290141 samples/sec	 accuracy=45.655331	 loss=2.401957	 lr=0.025060
Epoch[141] Batch [3449]/[3760]	Speed: 117.717364 samples/sec	 accuracy=45.656250	 loss=2.402090	 lr=0.025048
Epoch[141] Batch [3499]/[3760]	Speed: 117.377847 samples/sec	 accuracy=45.650446	 loss=2.402322	 lr=0.025037
Epoch[141] Batch [3549]/[3760]	Speed: 117.768830 samples/sec	 accuracy=45.643486	 loss=2.403078	 lr=0.025026
Epoch[141] Batch [3599]/[3760]	Speed: 117.492362 samples/sec	 accuracy=45.657986	 loss=2.402873	 lr=0.025015
Epoch[141] Batch [3649]/[3760]	Speed: 116.588281 samples/sec	 accuracy=45.666524	 loss=2.403362	 lr=0.025004
Epoch[141] Batch [3699]/[3760]	Speed: 116.776172 samples/sec	 accuracy=45.670608	 loss=2.403011	 lr=0.024992
Epoch[141] Batch [3749]/[3760]	Speed: 121.161062 samples/sec	 accuracy=45.694583	 loss=2.402030	 lr=0.024981
Batch [0049]/[0303]: acc-top1=46.250000 acc-top5=70.781250
Batch [0099]/[0303]: acc-top1=44.906250 acc-top5=70.937500
Batch [0149]/[0303]: acc-top1=46.218750 acc-top5=71.520833
Batch [0199]/[0303]: acc-top1=48.906250 acc-top5=73.101562
Batch [0249]/[0303]: acc-top1=48.243750 acc-top5=72.668750
Batch [0299]/[0303]: acc-top1=47.473958 acc-top5=72.479167
[Epoch 141] training: accuracy=45.693567	 loss=2.401954
[Epoch 141] speed: 116 samples/sec	time cost: 2175.077673
[Epoch 141] validation: acc-top1=47.406147 acc-top5=72.519596 loss=2.433384
Epoch[142] Batch [0049]/[3759]	Speed: 86.970302 samples/sec	 accuracy=47.125000	 loss=2.358990	 lr=0.024968
Epoch[142] Batch [0099]/[3759]	Speed: 117.822001 samples/sec	 accuracy=47.421875	 loss=2.340761	 lr=0.024957
Epoch[142] Batch [0149]/[3759]	Speed: 118.068968 samples/sec	 accuracy=46.791667	 loss=2.358315	 lr=0.024946
Epoch[142] Batch [0199]/[3759]	Speed: 117.334019 samples/sec	 accuracy=46.562500	 loss=2.369264	 lr=0.024934
Epoch[142] Batch [0249]/[3759]	Speed: 117.028202 samples/sec	 accuracy=46.537500	 loss=2.365215	 lr=0.024923
Epoch[142] Batch [0299]/[3759]	Speed: 117.411922 samples/sec	 accuracy=46.385417	 loss=2.364380	 lr=0.024912
Epoch[142] Batch [0349]/[3759]	Speed: 117.349084 samples/sec	 accuracy=46.241071	 loss=2.361123	 lr=0.024901
Epoch[142] Batch [0399]/[3759]	Speed: 118.112951 samples/sec	 accuracy=46.238281	 loss=2.358844	 lr=0.024890
Epoch[142] Batch [0449]/[3759]	Speed: 117.488077 samples/sec	 accuracy=46.256944	 loss=2.353789	 lr=0.024879
Epoch[142] Batch [0499]/[3759]	Speed: 117.807749 samples/sec	 accuracy=46.196875	 loss=2.357383	 lr=0.024867
Epoch[142] Batch [0549]/[3759]	Speed: 117.564769 samples/sec	 accuracy=46.190341	 loss=2.361271	 lr=0.024856
Epoch[142] Batch [0599]/[3759]	Speed: 117.224399 samples/sec	 accuracy=46.135417	 loss=2.369178	 lr=0.024845
Epoch[142] Batch [0649]/[3759]	Speed: 117.367827 samples/sec	 accuracy=46.197115	 loss=2.370737	 lr=0.024834
Epoch[142] Batch [0699]/[3759]	Speed: 117.808471 samples/sec	 accuracy=46.205357	 loss=2.372552	 lr=0.024823
Epoch[142] Batch [0749]/[3759]	Speed: 117.867770 samples/sec	 accuracy=46.152083	 loss=2.377518	 lr=0.024812
Epoch[142] Batch [0799]/[3759]	Speed: 117.181963 samples/sec	 accuracy=46.248047	 loss=2.372220	 lr=0.024801
Epoch[142] Batch [0849]/[3759]	Speed: 118.067550 samples/sec	 accuracy=46.244485	 loss=2.370833	 lr=0.024789
Epoch[142] Batch [0899]/[3759]	Speed: 117.986141 samples/sec	 accuracy=46.144097	 loss=2.374978	 lr=0.024778
Epoch[142] Batch [0949]/[3759]	Speed: 117.690354 samples/sec	 accuracy=46.069079	 loss=2.377446	 lr=0.024767
Epoch[142] Batch [0999]/[3759]	Speed: 117.485424 samples/sec	 accuracy=46.079687	 loss=2.376787	 lr=0.024756
Epoch[142] Batch [1049]/[3759]	Speed: 116.545412 samples/sec	 accuracy=46.111607	 loss=2.376443	 lr=0.024745
Epoch[142] Batch [1099]/[3759]	Speed: 118.196779 samples/sec	 accuracy=46.115057	 loss=2.376355	 lr=0.024734
Epoch[142] Batch [1149]/[3759]	Speed: 118.255959 samples/sec	 accuracy=46.158967	 loss=2.373037	 lr=0.024723
Epoch[142] Batch [1199]/[3759]	Speed: 117.837417 samples/sec	 accuracy=46.118490	 loss=2.373926	 lr=0.024712
Epoch[142] Batch [1249]/[3759]	Speed: 117.593882 samples/sec	 accuracy=46.130000	 loss=2.373046	 lr=0.024700
Epoch[142] Batch [1299]/[3759]	Speed: 116.168168 samples/sec	 accuracy=46.117788	 loss=2.372232	 lr=0.024689
Epoch[142] Batch [1349]/[3759]	Speed: 117.818574 samples/sec	 accuracy=46.106481	 loss=2.373865	 lr=0.024678
Epoch[142] Batch [1399]/[3759]	Speed: 117.269416 samples/sec	 accuracy=46.141741	 loss=2.373174	 lr=0.024667
Epoch[142] Batch [1449]/[3759]	Speed: 116.968896 samples/sec	 accuracy=46.161638	 loss=2.372842	 lr=0.024656
Epoch[142] Batch [1499]/[3759]	Speed: 117.018763 samples/sec	 accuracy=46.148958	 loss=2.373536	 lr=0.024645
Epoch[142] Batch [1549]/[3759]	Speed: 117.643578 samples/sec	 accuracy=46.152218	 loss=2.373703	 lr=0.024634
Epoch[142] Batch [1599]/[3759]	Speed: 117.372564 samples/sec	 accuracy=46.132812	 loss=2.372972	 lr=0.024623
Epoch[142] Batch [1649]/[3759]	Speed: 117.255081 samples/sec	 accuracy=46.147727	 loss=2.373236	 lr=0.024611
Epoch[142] Batch [1699]/[3759]	Speed: 117.643879 samples/sec	 accuracy=46.133272	 loss=2.373807	 lr=0.024600
Epoch[142] Batch [1749]/[3759]	Speed: 117.811225 samples/sec	 accuracy=46.108929	 loss=2.375882	 lr=0.024589
Epoch[142] Batch [1799]/[3759]	Speed: 117.296022 samples/sec	 accuracy=46.105903	 loss=2.375923	 lr=0.024578
Epoch[142] Batch [1849]/[3759]	Speed: 116.819963 samples/sec	 accuracy=46.091216	 loss=2.378123	 lr=0.024567
Epoch[142] Batch [1899]/[3759]	Speed: 117.128335 samples/sec	 accuracy=46.119243	 loss=2.378043	 lr=0.024556
Epoch[142] Batch [1949]/[3759]	Speed: 117.614074 samples/sec	 accuracy=46.136218	 loss=2.377239	 lr=0.024545
Epoch[142] Batch [1999]/[3759]	Speed: 117.990565 samples/sec	 accuracy=46.122656	 loss=2.377249	 lr=0.024534
Epoch[142] Batch [2049]/[3759]	Speed: 118.090317 samples/sec	 accuracy=46.101372	 loss=2.377494	 lr=0.024523
Epoch[142] Batch [2099]/[3759]	Speed: 117.477622 samples/sec	 accuracy=46.099702	 loss=2.378010	 lr=0.024512
Epoch[142] Batch [2149]/[3759]	Speed: 116.954685 samples/sec	 accuracy=46.097384	 loss=2.378577	 lr=0.024500
Epoch[142] Batch [2199]/[3759]	Speed: 117.498462 samples/sec	 accuracy=46.100852	 loss=2.378418	 lr=0.024489
Epoch[142] Batch [2249]/[3759]	Speed: 117.548462 samples/sec	 accuracy=46.095139	 loss=2.377093	 lr=0.024478
Epoch[142] Batch [2299]/[3759]	Speed: 118.282093 samples/sec	 accuracy=46.072690	 loss=2.378309	 lr=0.024467
Epoch[142] Batch [2349]/[3759]	Speed: 117.648550 samples/sec	 accuracy=46.081117	 loss=2.378560	 lr=0.024456
Epoch[142] Batch [2399]/[3759]	Speed: 118.135701 samples/sec	 accuracy=46.109375	 loss=2.376968	 lr=0.024445
Epoch[142] Batch [2449]/[3759]	Speed: 117.786327 samples/sec	 accuracy=46.099490	 loss=2.377663	 lr=0.024434
Epoch[142] Batch [2499]/[3759]	Speed: 117.740981 samples/sec	 accuracy=46.079375	 loss=2.378011	 lr=0.024423
Epoch[142] Batch [2549]/[3759]	Speed: 116.181438 samples/sec	 accuracy=46.101103	 loss=2.377650	 lr=0.024412
Epoch[142] Batch [2599]/[3759]	Speed: 117.666854 samples/sec	 accuracy=46.109976	 loss=2.377210	 lr=0.024401
Epoch[142] Batch [2649]/[3759]	Speed: 117.680960 samples/sec	 accuracy=46.097877	 loss=2.377896	 lr=0.024390
Epoch[142] Batch [2699]/[3759]	Speed: 117.943540 samples/sec	 accuracy=46.098958	 loss=2.378038	 lr=0.024379
Epoch[142] Batch [2749]/[3759]	Speed: 117.302530 samples/sec	 accuracy=46.098295	 loss=2.378631	 lr=0.024367
Epoch[142] Batch [2799]/[3759]	Speed: 118.012012 samples/sec	 accuracy=46.117746	 loss=2.377962	 lr=0.024356
Epoch[142] Batch [2849]/[3759]	Speed: 117.756300 samples/sec	 accuracy=46.092105	 loss=2.378859	 lr=0.024345
Epoch[142] Batch [2899]/[3759]	Speed: 117.431377 samples/sec	 accuracy=46.069504	 loss=2.379870	 lr=0.024334
Epoch[142] Batch [2949]/[3759]	Speed: 117.671233 samples/sec	 accuracy=46.030191	 loss=2.381605	 lr=0.024323
Epoch[142] Batch [2999]/[3759]	Speed: 117.937766 samples/sec	 accuracy=46.010417	 loss=2.382142	 lr=0.024312
Epoch[142] Batch [3049]/[3759]	Speed: 117.657611 samples/sec	 accuracy=46.021516	 loss=2.381836	 lr=0.024301
Epoch[142] Batch [3099]/[3759]	Speed: 117.030373 samples/sec	 accuracy=45.999496	 loss=2.383068	 lr=0.024290
Epoch[142] Batch [3149]/[3759]	Speed: 117.440550 samples/sec	 accuracy=46.000992	 loss=2.382954	 lr=0.024279
Epoch[142] Batch [3199]/[3759]	Speed: 117.045684 samples/sec	 accuracy=45.989746	 loss=2.383202	 lr=0.024268
Epoch[142] Batch [3249]/[3759]	Speed: 118.075595 samples/sec	 accuracy=45.971635	 loss=2.383804	 lr=0.024257
Epoch[142] Batch [3299]/[3759]	Speed: 117.290913 samples/sec	 accuracy=45.968277	 loss=2.384532	 lr=0.024246
Epoch[142] Batch [3349]/[3759]	Speed: 117.908263 samples/sec	 accuracy=45.959888	 loss=2.384708	 lr=0.024235
Epoch[142] Batch [3399]/[3759]	Speed: 117.878829 samples/sec	 accuracy=45.953125	 loss=2.384590	 lr=0.024224
Epoch[142] Batch [3449]/[3759]	Speed: 117.640491 samples/sec	 accuracy=45.937953	 loss=2.384526	 lr=0.024213
Epoch[142] Batch [3499]/[3759]	Speed: 117.114501 samples/sec	 accuracy=45.943750	 loss=2.384645	 lr=0.024202
Epoch[142] Batch [3549]/[3759]	Speed: 117.673797 samples/sec	 accuracy=45.935299	 loss=2.384378	 lr=0.024191
Epoch[142] Batch [3599]/[3759]	Speed: 117.090067 samples/sec	 accuracy=45.950955	 loss=2.384337	 lr=0.024179
Epoch[142] Batch [3649]/[3759]	Speed: 118.000531 samples/sec	 accuracy=45.948202	 loss=2.384549	 lr=0.024168
Epoch[142] Batch [3699]/[3759]	Speed: 117.037233 samples/sec	 accuracy=45.943412	 loss=2.384525	 lr=0.024157
Epoch[142] Batch [3749]/[3759]	Speed: 120.747341 samples/sec	 accuracy=45.947917	 loss=2.384361	 lr=0.024146
Batch [0049]/[0303]: acc-top1=48.031250 acc-top5=72.156250
Batch [0099]/[0303]: acc-top1=45.187500 acc-top5=70.828125
Batch [0149]/[0303]: acc-top1=46.958333 acc-top5=71.760417
Batch [0199]/[0303]: acc-top1=49.773438 acc-top5=73.601562
Batch [0249]/[0303]: acc-top1=48.806250 acc-top5=73.406250
Batch [0299]/[0303]: acc-top1=48.140625 acc-top5=73.369792
[Epoch 142] training: accuracy=45.937251	 loss=2.385014
[Epoch 142] speed: 117 samples/sec	time cost: 2173.245744
[Epoch 142] validation: acc-top1=48.040429 acc-top5=73.339521 loss=2.324256
Epoch[143] Batch [0049]/[3760]	Speed: 89.179719 samples/sec	 accuracy=47.843750	 loss=2.290724	 lr=0.024133
Epoch[143] Batch [0099]/[3760]	Speed: 117.864567 samples/sec	 accuracy=47.546875	 loss=2.311046	 lr=0.024122
Epoch[143] Batch [0149]/[3760]	Speed: 117.764501 samples/sec	 accuracy=47.270833	 loss=2.322431	 lr=0.024111
Epoch[143] Batch [0199]/[3760]	Speed: 117.734815 samples/sec	 accuracy=46.914062	 loss=2.335565	 lr=0.024100
Epoch[143] Batch [0249]/[3760]	Speed: 117.538210 samples/sec	 accuracy=47.237500	 loss=2.323416	 lr=0.024089
Epoch[143] Batch [0299]/[3760]	Speed: 117.885133 samples/sec	 accuracy=47.145833	 loss=2.321530	 lr=0.024078
Epoch[143] Batch [0349]/[3760]	Speed: 117.327195 samples/sec	 accuracy=47.098214	 loss=2.324434	 lr=0.024067
Epoch[143] Batch [0399]/[3760]	Speed: 118.314326 samples/sec	 accuracy=47.105469	 loss=2.328203	 lr=0.024056
Epoch[143] Batch [0449]/[3760]	Speed: 117.812462 samples/sec	 accuracy=46.989583	 loss=2.336102	 lr=0.024045
Epoch[143] Batch [0499]/[3760]	Speed: 116.978551 samples/sec	 accuracy=46.937500	 loss=2.339670	 lr=0.024034
Epoch[143] Batch [0549]/[3760]	Speed: 116.987149 samples/sec	 accuracy=47.096591	 loss=2.331853	 lr=0.024023
Epoch[143] Batch [0599]/[3760]	Speed: 117.538637 samples/sec	 accuracy=46.945312	 loss=2.335920	 lr=0.024012
Epoch[143] Batch [0649]/[3760]	Speed: 116.817857 samples/sec	 accuracy=46.927885	 loss=2.341347	 lr=0.024001
Epoch[143] Batch [0699]/[3760]	Speed: 118.147192 samples/sec	 accuracy=46.917411	 loss=2.343328	 lr=0.023990
Epoch[143] Batch [0749]/[3760]	Speed: 117.246668 samples/sec	 accuracy=46.837500	 loss=2.346760	 lr=0.023979
Epoch[143] Batch [0799]/[3760]	Speed: 117.656018 samples/sec	 accuracy=46.787109	 loss=2.347287	 lr=0.023968
Epoch[143] Batch [0849]/[3760]	Speed: 117.613795 samples/sec	 accuracy=46.748162	 loss=2.349589	 lr=0.023957
Epoch[143] Batch [0899]/[3760]	Speed: 118.201810 samples/sec	 accuracy=46.727431	 loss=2.349345	 lr=0.023946
Epoch[143] Batch [0949]/[3760]	Speed: 117.430892 samples/sec	 accuracy=46.745066	 loss=2.347404	 lr=0.023935
Epoch[143] Batch [0999]/[3760]	Speed: 117.286820 samples/sec	 accuracy=46.745313	 loss=2.348181	 lr=0.023924
Epoch[143] Batch [1049]/[3760]	Speed: 117.771280 samples/sec	 accuracy=46.711310	 loss=2.349933	 lr=0.023913
Epoch[143] Batch [1099]/[3760]	Speed: 115.853127 samples/sec	 accuracy=46.664773	 loss=2.351911	 lr=0.023902
Epoch[143] Batch [1149]/[3760]	Speed: 117.428889 samples/sec	 accuracy=46.653533	 loss=2.352715	 lr=0.023891
Epoch[143] Batch [1199]/[3760]	Speed: 117.382505 samples/sec	 accuracy=46.648438	 loss=2.353060	 lr=0.023880
Epoch[143] Batch [1249]/[3760]	Speed: 117.956760 samples/sec	 accuracy=46.678750	 loss=2.351870	 lr=0.023869
Epoch[143] Batch [1299]/[3760]	Speed: 117.890321 samples/sec	 accuracy=46.616587	 loss=2.353942	 lr=0.023858
Epoch[143] Batch [1349]/[3760]	Speed: 118.229025 samples/sec	 accuracy=46.586806	 loss=2.355646	 lr=0.023847
Epoch[143] Batch [1399]/[3760]	Speed: 117.514997 samples/sec	 accuracy=46.531250	 loss=2.357489	 lr=0.023836
Epoch[143] Batch [1449]/[3760]	Speed: 117.421302 samples/sec	 accuracy=46.534483	 loss=2.357757	 lr=0.023825
Epoch[143] Batch [1499]/[3760]	Speed: 116.907266 samples/sec	 accuracy=46.543750	 loss=2.357032	 lr=0.023814
Epoch[143] Batch [1549]/[3760]	Speed: 117.132105 samples/sec	 accuracy=46.562500	 loss=2.356354	 lr=0.023803
Epoch[143] Batch [1599]/[3760]	Speed: 118.055307 samples/sec	 accuracy=46.561523	 loss=2.356554	 lr=0.023792
Epoch[143] Batch [1649]/[3760]	Speed: 117.436653 samples/sec	 accuracy=46.539773	 loss=2.359325	 lr=0.023781
Epoch[143] Batch [1699]/[3760]	Speed: 117.511768 samples/sec	 accuracy=46.534926	 loss=2.359134	 lr=0.023770
Epoch[143] Batch [1749]/[3760]	Speed: 117.394828 samples/sec	 accuracy=46.533929	 loss=2.358306	 lr=0.023759
Epoch[143] Batch [1799]/[3760]	Speed: 117.041758 samples/sec	 accuracy=46.545139	 loss=2.357883	 lr=0.023748
Epoch[143] Batch [1849]/[3760]	Speed: 117.899239 samples/sec	 accuracy=46.514358	 loss=2.359509	 lr=0.023737
Epoch[143] Batch [1899]/[3760]	Speed: 117.273805 samples/sec	 accuracy=46.501645	 loss=2.359490	 lr=0.023726
Epoch[143] Batch [1949]/[3760]	Speed: 118.382152 samples/sec	 accuracy=46.484776	 loss=2.359304	 lr=0.023715
Epoch[143] Batch [1999]/[3760]	Speed: 116.772952 samples/sec	 accuracy=46.466406	 loss=2.358917	 lr=0.023704
Epoch[143] Batch [2049]/[3760]	Speed: 117.180595 samples/sec	 accuracy=46.502287	 loss=2.358804	 lr=0.023693
Epoch[143] Batch [2099]/[3760]	Speed: 117.334797 samples/sec	 accuracy=46.496280	 loss=2.358774	 lr=0.023682
Epoch[143] Batch [2149]/[3760]	Speed: 117.413170 samples/sec	 accuracy=46.505087	 loss=2.358263	 lr=0.023671
Epoch[143] Batch [2199]/[3760]	Speed: 117.615247 samples/sec	 accuracy=46.503551	 loss=2.358681	 lr=0.023660
Epoch[143] Batch [2249]/[3760]	Speed: 118.042557 samples/sec	 accuracy=46.519444	 loss=2.358162	 lr=0.023649
Epoch[143] Batch [2299]/[3760]	Speed: 117.464555 samples/sec	 accuracy=46.493886	 loss=2.359467	 lr=0.023638
Epoch[143] Batch [2349]/[3760]	Speed: 115.721709 samples/sec	 accuracy=46.484707	 loss=2.359727	 lr=0.023627
Epoch[143] Batch [2399]/[3760]	Speed: 116.770193 samples/sec	 accuracy=46.491536	 loss=2.359508	 lr=0.023616
Epoch[143] Batch [2449]/[3760]	Speed: 117.331080 samples/sec	 accuracy=46.456633	 loss=2.359861	 lr=0.023606
Epoch[143] Batch [2499]/[3760]	Speed: 117.152247 samples/sec	 accuracy=46.446250	 loss=2.360921	 lr=0.023595
Epoch[143] Batch [2549]/[3760]	Speed: 117.310122 samples/sec	 accuracy=46.448529	 loss=2.361195	 lr=0.023584
Epoch[143] Batch [2599]/[3760]	Speed: 117.395767 samples/sec	 accuracy=46.409255	 loss=2.362893	 lr=0.023573
Epoch[143] Batch [2649]/[3760]	Speed: 117.586296 samples/sec	 accuracy=46.419222	 loss=2.362818	 lr=0.023562
Epoch[143] Batch [2699]/[3760]	Speed: 117.789123 samples/sec	 accuracy=46.421875	 loss=2.362158	 lr=0.023551
Epoch[143] Batch [2749]/[3760]	Speed: 117.495599 samples/sec	 accuracy=46.415909	 loss=2.362767	 lr=0.023540
Epoch[143] Batch [2799]/[3760]	Speed: 117.233788 samples/sec	 accuracy=46.392857	 loss=2.363871	 lr=0.023529
Epoch[143] Batch [2849]/[3760]	Speed: 117.664954 samples/sec	 accuracy=46.361842	 loss=2.364889	 lr=0.023518
Epoch[143] Batch [2899]/[3760]	Speed: 117.473578 samples/sec	 accuracy=46.371767	 loss=2.364916	 lr=0.023507
Epoch[143] Batch [2949]/[3760]	Speed: 118.064489 samples/sec	 accuracy=46.359110	 loss=2.365396	 lr=0.023496
Epoch[143] Batch [2999]/[3760]	Speed: 117.296139 samples/sec	 accuracy=46.357292	 loss=2.365948	 lr=0.023485
Epoch[143] Batch [3049]/[3760]	Speed: 118.396177 samples/sec	 accuracy=46.373975	 loss=2.365261	 lr=0.023474
Epoch[143] Batch [3099]/[3760]	Speed: 118.089918 samples/sec	 accuracy=46.378024	 loss=2.365538	 lr=0.023463
Epoch[143] Batch [3149]/[3760]	Speed: 117.043108 samples/sec	 accuracy=46.367063	 loss=2.365904	 lr=0.023452
Epoch[143] Batch [3199]/[3760]	Speed: 116.910944 samples/sec	 accuracy=46.342773	 loss=2.366622	 lr=0.023441
Epoch[143] Batch [3249]/[3760]	Speed: 117.570299 samples/sec	 accuracy=46.351923	 loss=2.366484	 lr=0.023430
Epoch[143] Batch [3299]/[3760]	Speed: 116.751745 samples/sec	 accuracy=46.360795	 loss=2.366338	 lr=0.023420
Epoch[143] Batch [3349]/[3760]	Speed: 117.694568 samples/sec	 accuracy=46.335354	 loss=2.367395	 lr=0.023409
Epoch[143] Batch [3399]/[3760]	Speed: 117.501359 samples/sec	 accuracy=46.330882	 loss=2.367860	 lr=0.023398
Epoch[143] Batch [3449]/[3760]	Speed: 117.220700 samples/sec	 accuracy=46.322011	 loss=2.368575	 lr=0.023387
Epoch[143] Batch [3499]/[3760]	Speed: 118.283885 samples/sec	 accuracy=46.320089	 loss=2.368847	 lr=0.023376
Epoch[143] Batch [3549]/[3760]	Speed: 117.441852 samples/sec	 accuracy=46.319982	 loss=2.369278	 lr=0.023365
Epoch[143] Batch [3599]/[3760]	Speed: 115.983396 samples/sec	 accuracy=46.328993	 loss=2.368550	 lr=0.023354
Epoch[143] Batch [3649]/[3760]	Speed: 117.191233 samples/sec	 accuracy=46.336473	 loss=2.368365	 lr=0.023343
Epoch[143] Batch [3699]/[3760]	Speed: 117.406144 samples/sec	 accuracy=46.360220	 loss=2.367482	 lr=0.023332
Epoch[143] Batch [3749]/[3760]	Speed: 120.446733 samples/sec	 accuracy=46.343750	 loss=2.367815	 lr=0.023321
Batch [0049]/[0303]: acc-top1=47.906250 acc-top5=72.343750
Batch [0099]/[0303]: acc-top1=44.546875 acc-top5=71.046875
Batch [0149]/[0303]: acc-top1=46.364583 acc-top5=72.041667
Batch [0199]/[0303]: acc-top1=49.328125 acc-top5=73.914062
Batch [0249]/[0303]: acc-top1=48.912500 acc-top5=73.731250
Batch [0299]/[0303]: acc-top1=47.927083 acc-top5=73.416667
[Epoch 143] training: accuracy=46.342670	 loss=2.367929
[Epoch 143] speed: 116 samples/sec	time cost: 2172.786093
[Epoch 143] validation: acc-top1=47.865099 acc-top5=73.396246 loss=2.344995
Epoch[144] Batch [0049]/[3760]	Speed: 87.292062 samples/sec	 accuracy=46.875000	 loss=2.364786	 lr=0.023308
Epoch[144] Batch [0099]/[3760]	Speed: 117.876455 samples/sec	 accuracy=46.281250	 loss=2.346999	 lr=0.023297
Epoch[144] Batch [0149]/[3760]	Speed: 117.232532 samples/sec	 accuracy=46.760417	 loss=2.335138	 lr=0.023286
Epoch[144] Batch [0199]/[3760]	Speed: 117.714628 samples/sec	 accuracy=46.703125	 loss=2.348813	 lr=0.023276
Epoch[144] Batch [0249]/[3760]	Speed: 117.937124 samples/sec	 accuracy=46.612500	 loss=2.351295	 lr=0.023265
Epoch[144] Batch [0299]/[3760]	Speed: 116.915784 samples/sec	 accuracy=46.708333	 loss=2.344613	 lr=0.023254
Epoch[144] Batch [0349]/[3760]	Speed: 117.604278 samples/sec	 accuracy=46.763393	 loss=2.340098	 lr=0.023243
Epoch[144] Batch [0399]/[3760]	Speed: 117.334143 samples/sec	 accuracy=46.644531	 loss=2.339086	 lr=0.023232
Epoch[144] Batch [0449]/[3760]	Speed: 117.016873 samples/sec	 accuracy=46.638889	 loss=2.339698	 lr=0.023221
Epoch[144] Batch [0499]/[3760]	Speed: 117.649418 samples/sec	 accuracy=46.696875	 loss=2.334036	 lr=0.023210
Epoch[144] Batch [0549]/[3760]	Speed: 117.589950 samples/sec	 accuracy=46.536932	 loss=2.336399	 lr=0.023199
Epoch[144] Batch [0599]/[3760]	Speed: 116.648360 samples/sec	 accuracy=46.408854	 loss=2.337455	 lr=0.023188
Epoch[144] Batch [0649]/[3760]	Speed: 117.764159 samples/sec	 accuracy=46.461538	 loss=2.338327	 lr=0.023177
Epoch[144] Batch [0699]/[3760]	Speed: 117.307601 samples/sec	 accuracy=46.564732	 loss=2.335419	 lr=0.023167
Epoch[144] Batch [0749]/[3760]	Speed: 117.848371 samples/sec	 accuracy=46.716667	 loss=2.330685	 lr=0.023156
Epoch[144] Batch [0799]/[3760]	Speed: 117.566279 samples/sec	 accuracy=46.607422	 loss=2.333680	 lr=0.023145
Epoch[144] Batch [0849]/[3760]	Speed: 118.073236 samples/sec	 accuracy=46.485294	 loss=2.340496	 lr=0.023134
Epoch[144] Batch [0899]/[3760]	Speed: 117.104605 samples/sec	 accuracy=46.527778	 loss=2.338261	 lr=0.023123
Epoch[144] Batch [0949]/[3760]	Speed: 115.494960 samples/sec	 accuracy=46.565789	 loss=2.338133	 lr=0.023112
Epoch[144] Batch [0999]/[3760]	Speed: 118.022408 samples/sec	 accuracy=46.531250	 loss=2.341352	 lr=0.023101
Epoch[144] Batch [1049]/[3760]	Speed: 117.383156 samples/sec	 accuracy=46.555060	 loss=2.342363	 lr=0.023090
Epoch[144] Batch [1099]/[3760]	Speed: 117.236342 samples/sec	 accuracy=46.552557	 loss=2.342307	 lr=0.023080
Epoch[144] Batch [1149]/[3760]	Speed: 116.980196 samples/sec	 accuracy=46.593750	 loss=2.343544	 lr=0.023069
Epoch[144] Batch [1199]/[3760]	Speed: 117.823812 samples/sec	 accuracy=46.610677	 loss=2.343453	 lr=0.023058
Epoch[144] Batch [1249]/[3760]	Speed: 116.693334 samples/sec	 accuracy=46.582500	 loss=2.346397	 lr=0.023047
Epoch[144] Batch [1299]/[3760]	Speed: 117.231813 samples/sec	 accuracy=46.584135	 loss=2.346955	 lr=0.023036
Epoch[144] Batch [1349]/[3760]	Speed: 116.935592 samples/sec	 accuracy=46.636574	 loss=2.345591	 lr=0.023025
Epoch[144] Batch [1399]/[3760]	Speed: 116.829156 samples/sec	 accuracy=46.642857	 loss=2.343992	 lr=0.023014
Epoch[144] Batch [1449]/[3760]	Speed: 117.550103 samples/sec	 accuracy=46.600216	 loss=2.345823	 lr=0.023004
Epoch[144] Batch [1499]/[3760]	Speed: 117.163454 samples/sec	 accuracy=46.617708	 loss=2.346071	 lr=0.022993
Epoch[144] Batch [1549]/[3760]	Speed: 117.712790 samples/sec	 accuracy=46.610887	 loss=2.347714	 lr=0.022982
Epoch[144] Batch [1599]/[3760]	Speed: 118.033458 samples/sec	 accuracy=46.654297	 loss=2.346555	 lr=0.022971
Epoch[144] Batch [1649]/[3760]	Speed: 118.022698 samples/sec	 accuracy=46.675189	 loss=2.347007	 lr=0.022960
Epoch[144] Batch [1699]/[3760]	Speed: 117.140371 samples/sec	 accuracy=46.678309	 loss=2.347394	 lr=0.022949
Epoch[144] Batch [1749]/[3760]	Speed: 117.199959 samples/sec	 accuracy=46.636607	 loss=2.348515	 lr=0.022938
Epoch[144] Batch [1799]/[3760]	Speed: 117.092012 samples/sec	 accuracy=46.644097	 loss=2.348625	 lr=0.022928
Epoch[144] Batch [1849]/[3760]	Speed: 117.674988 samples/sec	 accuracy=46.658784	 loss=2.346878	 lr=0.022917
Epoch[144] Batch [1899]/[3760]	Speed: 117.483373 samples/sec	 accuracy=46.658717	 loss=2.346102	 lr=0.022906
Epoch[144] Batch [1949]/[3760]	Speed: 117.830703 samples/sec	 accuracy=46.642628	 loss=2.348327	 lr=0.022895
Epoch[144] Batch [1999]/[3760]	Speed: 117.425671 samples/sec	 accuracy=46.666406	 loss=2.348087	 lr=0.022884
Epoch[144] Batch [2049]/[3760]	Speed: 117.207276 samples/sec	 accuracy=46.660061	 loss=2.347216	 lr=0.022873
Epoch[144] Batch [2099]/[3760]	Speed: 117.801352 samples/sec	 accuracy=46.667411	 loss=2.346838	 lr=0.022863
Epoch[144] Batch [2149]/[3760]	Speed: 118.016453 samples/sec	 accuracy=46.696221	 loss=2.346446	 lr=0.022852
Epoch[144] Batch [2199]/[3760]	Speed: 115.478760 samples/sec	 accuracy=46.702415	 loss=2.345395	 lr=0.022841
Epoch[144] Batch [2249]/[3760]	Speed: 117.031553 samples/sec	 accuracy=46.713194	 loss=2.345326	 lr=0.022830
Epoch[144] Batch [2299]/[3760]	Speed: 116.265810 samples/sec	 accuracy=46.671196	 loss=2.346821	 lr=0.022819
Epoch[144] Batch [2349]/[3760]	Speed: 117.972558 samples/sec	 accuracy=46.667553	 loss=2.347574	 lr=0.022808
Epoch[144] Batch [2399]/[3760]	Speed: 116.931156 samples/sec	 accuracy=46.658203	 loss=2.346824	 lr=0.022798
Epoch[144] Batch [2449]/[3760]	Speed: 117.895792 samples/sec	 accuracy=46.663903	 loss=2.346758	 lr=0.022787
Epoch[144] Batch [2499]/[3760]	Speed: 116.963734 samples/sec	 accuracy=46.658125	 loss=2.346908	 lr=0.022776
Epoch[144] Batch [2549]/[3760]	Speed: 117.284166 samples/sec	 accuracy=46.645833	 loss=2.348141	 lr=0.022765
Epoch[144] Batch [2599]/[3760]	Speed: 116.675789 samples/sec	 accuracy=46.642428	 loss=2.348492	 lr=0.022754
Epoch[144] Batch [2649]/[3760]	Speed: 117.773409 samples/sec	 accuracy=46.613208	 loss=2.349082	 lr=0.022744
Epoch[144] Batch [2699]/[3760]	Speed: 117.261644 samples/sec	 accuracy=46.638310	 loss=2.347844	 lr=0.022733
Epoch[144] Batch [2749]/[3760]	Speed: 117.011745 samples/sec	 accuracy=46.641477	 loss=2.348466	 lr=0.022722
Epoch[144] Batch [2799]/[3760]	Speed: 117.348630 samples/sec	 accuracy=46.639509	 loss=2.347968	 lr=0.022711
Epoch[144] Batch [2849]/[3760]	Speed: 116.707032 samples/sec	 accuracy=46.632675	 loss=2.349232	 lr=0.022700
Epoch[144] Batch [2899]/[3760]	Speed: 117.977837 samples/sec	 accuracy=46.629849	 loss=2.349847	 lr=0.022689
Epoch[144] Batch [2949]/[3760]	Speed: 116.951639 samples/sec	 accuracy=46.627119	 loss=2.351033	 lr=0.022679
Epoch[144] Batch [2999]/[3760]	Speed: 117.806632 samples/sec	 accuracy=46.636458	 loss=2.351266	 lr=0.022668
Epoch[144] Batch [3049]/[3760]	Speed: 117.608654 samples/sec	 accuracy=46.631660	 loss=2.351983	 lr=0.022657
Epoch[144] Batch [3099]/[3760]	Speed: 117.214140 samples/sec	 accuracy=46.637097	 loss=2.351767	 lr=0.022646
Epoch[144] Batch [3149]/[3760]	Speed: 117.978736 samples/sec	 accuracy=46.632440	 loss=2.351882	 lr=0.022635
Epoch[144] Batch [3199]/[3760]	Speed: 117.609063 samples/sec	 accuracy=46.628906	 loss=2.352179	 lr=0.022625
Epoch[144] Batch [3249]/[3760]	Speed: 117.058697 samples/sec	 accuracy=46.633654	 loss=2.351789	 lr=0.022614
Epoch[144] Batch [3299]/[3760]	Speed: 117.721243 samples/sec	 accuracy=46.636837	 loss=2.351322	 lr=0.022603
Epoch[144] Batch [3349]/[3760]	Speed: 116.935945 samples/sec	 accuracy=46.645522	 loss=2.350554	 lr=0.022592
Epoch[144] Batch [3399]/[3760]	Speed: 117.079588 samples/sec	 accuracy=46.635110	 loss=2.350771	 lr=0.022582
Epoch[144] Batch [3449]/[3760]	Speed: 116.093221 samples/sec	 accuracy=46.637228	 loss=2.350574	 lr=0.022571
Epoch[144] Batch [3499]/[3760]	Speed: 117.554952 samples/sec	 accuracy=46.647768	 loss=2.350572	 lr=0.022560
Epoch[144] Batch [3549]/[3760]	Speed: 117.214679 samples/sec	 accuracy=46.673415	 loss=2.349045	 lr=0.022549
Epoch[144] Batch [3599]/[3760]	Speed: 116.717136 samples/sec	 accuracy=46.680556	 loss=2.348939	 lr=0.022538
Epoch[144] Batch [3649]/[3760]	Speed: 117.310468 samples/sec	 accuracy=46.670805	 loss=2.349846	 lr=0.022528
Epoch[144] Batch [3699]/[3760]	Speed: 117.233833 samples/sec	 accuracy=46.663851	 loss=2.349464	 lr=0.022517
Epoch[144] Batch [3749]/[3760]	Speed: 120.201724 samples/sec	 accuracy=46.662083	 loss=2.349576	 lr=0.022506
Batch [0049]/[0303]: acc-top1=45.656250 acc-top5=70.468750
Batch [0099]/[0303]: acc-top1=42.578125 acc-top5=68.453125
Batch [0149]/[0303]: acc-top1=43.718750 acc-top5=69.197917
Batch [0199]/[0303]: acc-top1=47.031250 acc-top5=71.570312
Batch [0249]/[0303]: acc-top1=46.231250 acc-top5=71.206250
Batch [0299]/[0303]: acc-top1=45.640625 acc-top5=71.114583
[Epoch 144] training: accuracy=46.658910	 loss=2.349590
[Epoch 144] speed: 116 samples/sec	time cost: 2176.327896
[Epoch 144] validation: acc-top1=45.554868 acc-top5=71.070545 loss=2.528845
Epoch[145] Batch [0049]/[3759]	Speed: 87.415667 samples/sec	 accuracy=48.312500	 loss=2.272730	 lr=0.022493
Epoch[145] Batch [0099]/[3759]	Speed: 118.326974 samples/sec	 accuracy=47.921875	 loss=2.299960	 lr=0.022482
Epoch[145] Batch [0149]/[3759]	Speed: 117.502299 samples/sec	 accuracy=47.520833	 loss=2.316947	 lr=0.022472
Epoch[145] Batch [0199]/[3759]	Speed: 117.439208 samples/sec	 accuracy=47.507812	 loss=2.319803	 lr=0.022461
Epoch[145] Batch [0249]/[3759]	Speed: 116.808230 samples/sec	 accuracy=47.418750	 loss=2.315344	 lr=0.022450
Epoch[145] Batch [0299]/[3759]	Speed: 118.257365 samples/sec	 accuracy=47.588542	 loss=2.303099	 lr=0.022439
Epoch[145] Batch [0349]/[3759]	Speed: 117.639303 samples/sec	 accuracy=47.468750	 loss=2.306772	 lr=0.022429
Epoch[145] Batch [0399]/[3759]	Speed: 117.439621 samples/sec	 accuracy=47.457031	 loss=2.306404	 lr=0.022418
Epoch[145] Batch [0449]/[3759]	Speed: 116.745598 samples/sec	 accuracy=47.503472	 loss=2.303465	 lr=0.022407
Epoch[145] Batch [0499]/[3759]	Speed: 117.037643 samples/sec	 accuracy=47.378125	 loss=2.312051	 lr=0.022396
Epoch[145] Batch [0549]/[3759]	Speed: 117.534157 samples/sec	 accuracy=47.312500	 loss=2.315495	 lr=0.022386
Epoch[145] Batch [0599]/[3759]	Speed: 117.635081 samples/sec	 accuracy=47.341146	 loss=2.314198	 lr=0.022375
Epoch[145] Batch [0649]/[3759]	Speed: 117.665813 samples/sec	 accuracy=47.283654	 loss=2.313473	 lr=0.022364
Epoch[145] Batch [0699]/[3759]	Speed: 118.143947 samples/sec	 accuracy=47.341518	 loss=2.308889	 lr=0.022353
Epoch[145] Batch [0749]/[3759]	Speed: 117.614782 samples/sec	 accuracy=47.297917	 loss=2.314481	 lr=0.022343
Epoch[145] Batch [0799]/[3759]	Speed: 115.574613 samples/sec	 accuracy=47.199219	 loss=2.315476	 lr=0.022332
Epoch[145] Batch [0849]/[3759]	Speed: 117.884268 samples/sec	 accuracy=47.264706	 loss=2.315301	 lr=0.022321
Epoch[145] Batch [0899]/[3759]	Speed: 118.050088 samples/sec	 accuracy=47.286458	 loss=2.315140	 lr=0.022310
Epoch[145] Batch [0949]/[3759]	Speed: 117.295599 samples/sec	 accuracy=47.378289	 loss=2.311061	 lr=0.022300
Epoch[145] Batch [0999]/[3759]	Speed: 118.007880 samples/sec	 accuracy=47.407812	 loss=2.310329	 lr=0.022289
Epoch[145] Batch [1049]/[3759]	Speed: 117.378968 samples/sec	 accuracy=47.489583	 loss=2.305358	 lr=0.022278
Epoch[145] Batch [1099]/[3759]	Speed: 117.532518 samples/sec	 accuracy=47.529830	 loss=2.305033	 lr=0.022267
Epoch[145] Batch [1149]/[3759]	Speed: 118.150974 samples/sec	 accuracy=47.486413	 loss=2.305807	 lr=0.022257
Epoch[145] Batch [1199]/[3759]	Speed: 117.260245 samples/sec	 accuracy=47.519531	 loss=2.304252	 lr=0.022246
Epoch[145] Batch [1249]/[3759]	Speed: 117.312494 samples/sec	 accuracy=47.551250	 loss=2.302493	 lr=0.022235
Epoch[145] Batch [1299]/[3759]	Speed: 117.797549 samples/sec	 accuracy=47.552885	 loss=2.302006	 lr=0.022224
Epoch[145] Batch [1349]/[3759]	Speed: 116.908018 samples/sec	 accuracy=47.510417	 loss=2.302480	 lr=0.022214
Epoch[145] Batch [1399]/[3759]	Speed: 116.985770 samples/sec	 accuracy=47.465402	 loss=2.303820	 lr=0.022203
Epoch[145] Batch [1449]/[3759]	Speed: 116.483577 samples/sec	 accuracy=47.475216	 loss=2.304113	 lr=0.022192
Epoch[145] Batch [1499]/[3759]	Speed: 117.772491 samples/sec	 accuracy=47.493750	 loss=2.303933	 lr=0.022182
Epoch[145] Batch [1549]/[3759]	Speed: 117.986612 samples/sec	 accuracy=47.470766	 loss=2.304706	 lr=0.022171
Epoch[145] Batch [1599]/[3759]	Speed: 116.800142 samples/sec	 accuracy=47.424805	 loss=2.307398	 lr=0.022160
Epoch[145] Batch [1649]/[3759]	Speed: 117.277311 samples/sec	 accuracy=47.392045	 loss=2.308934	 lr=0.022149
Epoch[145] Batch [1699]/[3759]	Speed: 117.112691 samples/sec	 accuracy=47.386949	 loss=2.309863	 lr=0.022139
Epoch[145] Batch [1749]/[3759]	Speed: 117.837931 samples/sec	 accuracy=47.325000	 loss=2.312326	 lr=0.022128
Epoch[145] Batch [1799]/[3759]	Speed: 116.434930 samples/sec	 accuracy=47.315104	 loss=2.313388	 lr=0.022117
Epoch[145] Batch [1849]/[3759]	Speed: 117.601139 samples/sec	 accuracy=47.288851	 loss=2.314430	 lr=0.022107
Epoch[145] Batch [1899]/[3759]	Speed: 117.459118 samples/sec	 accuracy=47.312500	 loss=2.313829	 lr=0.022096
Epoch[145] Batch [1949]/[3759]	Speed: 117.215964 samples/sec	 accuracy=47.298077	 loss=2.314744	 lr=0.022085
Epoch[145] Batch [1999]/[3759]	Speed: 117.891008 samples/sec	 accuracy=47.275000	 loss=2.315101	 lr=0.022075
Epoch[145] Batch [2049]/[3759]	Speed: 117.948785 samples/sec	 accuracy=47.237805	 loss=2.317056	 lr=0.022064
Epoch[145] Batch [2099]/[3759]	Speed: 115.540981 samples/sec	 accuracy=47.234375	 loss=2.317905	 lr=0.022053
Epoch[145] Batch [2149]/[3759]	Speed: 117.827247 samples/sec	 accuracy=47.233285	 loss=2.318124	 lr=0.022042
Epoch[145] Batch [2199]/[3759]	Speed: 117.426346 samples/sec	 accuracy=47.231534	 loss=2.318039	 lr=0.022032
Epoch[145] Batch [2249]/[3759]	Speed: 116.954745 samples/sec	 accuracy=47.247917	 loss=2.317323	 lr=0.022021
Epoch[145] Batch [2299]/[3759]	Speed: 117.482391 samples/sec	 accuracy=47.239810	 loss=2.316770	 lr=0.022010
Epoch[145] Batch [2349]/[3759]	Speed: 118.049230 samples/sec	 accuracy=47.231383	 loss=2.317776	 lr=0.022000
Epoch[145] Batch [2399]/[3759]	Speed: 117.966459 samples/sec	 accuracy=47.239583	 loss=2.318239	 lr=0.021989
Epoch[145] Batch [2449]/[3759]	Speed: 117.764335 samples/sec	 accuracy=47.210459	 loss=2.318672	 lr=0.021978
Epoch[145] Batch [2499]/[3759]	Speed: 116.958555 samples/sec	 accuracy=47.210000	 loss=2.317706	 lr=0.021968
Epoch[145] Batch [2549]/[3759]	Speed: 117.428645 samples/sec	 accuracy=47.185662	 loss=2.318509	 lr=0.021957
Epoch[145] Batch [2599]/[3759]	Speed: 116.966961 samples/sec	 accuracy=47.176683	 loss=2.318863	 lr=0.021946
Epoch[145] Batch [2649]/[3759]	Speed: 117.809474 samples/sec	 accuracy=47.159788	 loss=2.319822	 lr=0.021936
Epoch[145] Batch [2699]/[3759]	Speed: 117.727901 samples/sec	 accuracy=47.152778	 loss=2.320162	 lr=0.021925
Epoch[145] Batch [2749]/[3759]	Speed: 118.392898 samples/sec	 accuracy=47.146023	 loss=2.320978	 lr=0.021914
Epoch[145] Batch [2799]/[3759]	Speed: 117.485393 samples/sec	 accuracy=47.143973	 loss=2.321616	 lr=0.021904
Epoch[145] Batch [2849]/[3759]	Speed: 117.134564 samples/sec	 accuracy=47.140351	 loss=2.321639	 lr=0.021893
Epoch[145] Batch [2899]/[3759]	Speed: 117.965472 samples/sec	 accuracy=47.132004	 loss=2.321696	 lr=0.021882
Epoch[145] Batch [2949]/[3759]	Speed: 117.635075 samples/sec	 accuracy=47.146716	 loss=2.321460	 lr=0.021872
Epoch[145] Batch [2999]/[3759]	Speed: 117.686180 samples/sec	 accuracy=47.130729	 loss=2.322343	 lr=0.021861
Epoch[145] Batch [3049]/[3759]	Speed: 117.242745 samples/sec	 accuracy=47.130123	 loss=2.322485	 lr=0.021850
Epoch[145] Batch [3099]/[3759]	Speed: 117.204766 samples/sec	 accuracy=47.139113	 loss=2.322239	 lr=0.021840
Epoch[145] Batch [3149]/[3759]	Speed: 117.439500 samples/sec	 accuracy=47.133929	 loss=2.322558	 lr=0.021829
Epoch[145] Batch [3199]/[3759]	Speed: 117.365938 samples/sec	 accuracy=47.112305	 loss=2.324085	 lr=0.021818
Epoch[145] Batch [3249]/[3759]	Speed: 117.114112 samples/sec	 accuracy=47.138942	 loss=2.323479	 lr=0.021808
Epoch[145] Batch [3299]/[3759]	Speed: 117.899479 samples/sec	 accuracy=47.138731	 loss=2.323220	 lr=0.021797
Epoch[145] Batch [3349]/[3759]	Speed: 115.926803 samples/sec	 accuracy=47.148787	 loss=2.322767	 lr=0.021786
Epoch[145] Batch [3399]/[3759]	Speed: 117.353136 samples/sec	 accuracy=47.125460	 loss=2.323985	 lr=0.021776
Epoch[145] Batch [3449]/[3759]	Speed: 118.061040 samples/sec	 accuracy=47.105525	 loss=2.325073	 lr=0.021765
Epoch[145] Batch [3499]/[3759]	Speed: 117.434735 samples/sec	 accuracy=47.089732	 loss=2.325847	 lr=0.021754
Epoch[145] Batch [3549]/[3759]	Speed: 117.421099 samples/sec	 accuracy=47.064261	 loss=2.327604	 lr=0.021744
Epoch[145] Batch [3599]/[3759]	Speed: 117.454370 samples/sec	 accuracy=47.071181	 loss=2.327450	 lr=0.021733
Epoch[145] Batch [3649]/[3759]	Speed: 118.239237 samples/sec	 accuracy=47.067637	 loss=2.328206	 lr=0.021722
Epoch[145] Batch [3699]/[3759]	Speed: 117.699568 samples/sec	 accuracy=47.082348	 loss=2.327725	 lr=0.021712
Epoch[145] Batch [3749]/[3759]	Speed: 120.772691 samples/sec	 accuracy=47.083333	 loss=2.327710	 lr=0.021701
Batch [0049]/[0303]: acc-top1=47.000000 acc-top5=71.187500
Batch [0099]/[0303]: acc-top1=44.078125 acc-top5=70.500000
Batch [0149]/[0303]: acc-top1=46.468750 acc-top5=71.968750
Batch [0199]/[0303]: acc-top1=49.421875 acc-top5=73.703125
Batch [0249]/[0303]: acc-top1=48.837500 acc-top5=73.312500
Batch [0299]/[0303]: acc-top1=48.098958 acc-top5=73.343750
[Epoch 145] training: accuracy=47.078262	 loss=2.327998
[Epoch 145] speed: 116 samples/sec	time cost: 2173.401512
[Epoch 145] validation: acc-top1=48.097153 acc-top5=73.375619 loss=2.376777
Epoch[146] Batch [0049]/[3760]	Speed: 87.827102 samples/sec	 accuracy=48.343750	 loss=2.299939	 lr=0.021689
Epoch[146] Batch [0099]/[3760]	Speed: 117.794360 samples/sec	 accuracy=49.171875	 loss=2.266273	 lr=0.021678
Epoch[146] Batch [0149]/[3760]	Speed: 117.786379 samples/sec	 accuracy=48.125000	 loss=2.287780	 lr=0.021667
Epoch[146] Batch [0199]/[3760]	Speed: 117.338826 samples/sec	 accuracy=48.234375	 loss=2.282219	 lr=0.021657
Epoch[146] Batch [0249]/[3760]	Speed: 117.994048 samples/sec	 accuracy=48.200000	 loss=2.284049	 lr=0.021646
Epoch[146] Batch [0299]/[3760]	Speed: 117.782362 samples/sec	 accuracy=48.005208	 loss=2.288145	 lr=0.021636
Epoch[146] Batch [0349]/[3760]	Speed: 117.926391 samples/sec	 accuracy=48.080357	 loss=2.289131	 lr=0.021625
Epoch[146] Batch [0399]/[3760]	Speed: 116.989143 samples/sec	 accuracy=48.097656	 loss=2.289601	 lr=0.021614
Epoch[146] Batch [0449]/[3760]	Speed: 117.076682 samples/sec	 accuracy=48.045139	 loss=2.289296	 lr=0.021604
Epoch[146] Batch [0499]/[3760]	Speed: 117.209041 samples/sec	 accuracy=48.046875	 loss=2.288141	 lr=0.021593
Epoch[146] Batch [0549]/[3760]	Speed: 117.025861 samples/sec	 accuracy=47.917614	 loss=2.295919	 lr=0.021582
Epoch[146] Batch [0599]/[3760]	Speed: 116.771703 samples/sec	 accuracy=47.736979	 loss=2.302254	 lr=0.021572
Epoch[146] Batch [0649]/[3760]	Speed: 116.880135 samples/sec	 accuracy=47.747596	 loss=2.302357	 lr=0.021561
Epoch[146] Batch [0699]/[3760]	Speed: 117.817462 samples/sec	 accuracy=47.736607	 loss=2.305879	 lr=0.021551
Epoch[146] Batch [0749]/[3760]	Speed: 116.497728 samples/sec	 accuracy=47.737500	 loss=2.307564	 lr=0.021540
Epoch[146] Batch [0799]/[3760]	Speed: 116.818778 samples/sec	 accuracy=47.789062	 loss=2.304344	 lr=0.021529
Epoch[146] Batch [0849]/[3760]	Speed: 115.204716 samples/sec	 accuracy=47.744485	 loss=2.306430	 lr=0.021519
Epoch[146] Batch [0899]/[3760]	Speed: 118.102596 samples/sec	 accuracy=47.673611	 loss=2.306291	 lr=0.021508
Epoch[146] Batch [0949]/[3760]	Speed: 116.574558 samples/sec	 accuracy=47.722039	 loss=2.302556	 lr=0.021498
Epoch[146] Batch [0999]/[3760]	Speed: 117.791447 samples/sec	 accuracy=47.723437	 loss=2.302027	 lr=0.021487
Epoch[146] Batch [1049]/[3760]	Speed: 117.691247 samples/sec	 accuracy=47.668155	 loss=2.304150	 lr=0.021476
Epoch[146] Batch [1099]/[3760]	Speed: 117.540305 samples/sec	 accuracy=47.700284	 loss=2.302245	 lr=0.021466
Epoch[146] Batch [1149]/[3760]	Speed: 117.068733 samples/sec	 accuracy=47.679348	 loss=2.304376	 lr=0.021455
Epoch[146] Batch [1199]/[3760]	Speed: 116.966532 samples/sec	 accuracy=47.682292	 loss=2.304128	 lr=0.021445
Epoch[146] Batch [1249]/[3760]	Speed: 117.731793 samples/sec	 accuracy=47.685000	 loss=2.303764	 lr=0.021434
Epoch[146] Batch [1299]/[3760]	Speed: 117.514393 samples/sec	 accuracy=47.715144	 loss=2.303585	 lr=0.021423
Epoch[146] Batch [1349]/[3760]	Speed: 118.026675 samples/sec	 accuracy=47.641204	 loss=2.305451	 lr=0.021413
Epoch[146] Batch [1399]/[3760]	Speed: 116.565756 samples/sec	 accuracy=47.639509	 loss=2.306576	 lr=0.021402
Epoch[146] Batch [1449]/[3760]	Speed: 117.243968 samples/sec	 accuracy=47.646552	 loss=2.306818	 lr=0.021392
Epoch[146] Batch [1499]/[3760]	Speed: 117.659304 samples/sec	 accuracy=47.636458	 loss=2.306962	 lr=0.021381
Epoch[146] Batch [1549]/[3760]	Speed: 118.056859 samples/sec	 accuracy=47.605847	 loss=2.308262	 lr=0.021371
Epoch[146] Batch [1599]/[3760]	Speed: 117.912158 samples/sec	 accuracy=47.642578	 loss=2.308804	 lr=0.021360
Epoch[146] Batch [1649]/[3760]	Speed: 117.657491 samples/sec	 accuracy=47.681818	 loss=2.307306	 lr=0.021349
Epoch[146] Batch [1699]/[3760]	Speed: 117.298062 samples/sec	 accuracy=47.689338	 loss=2.306662	 lr=0.021339
Epoch[146] Batch [1749]/[3760]	Speed: 117.410213 samples/sec	 accuracy=47.649107	 loss=2.307961	 lr=0.021328
Epoch[146] Batch [1799]/[3760]	Speed: 116.075071 samples/sec	 accuracy=47.638889	 loss=2.308077	 lr=0.021318
Epoch[146] Batch [1849]/[3760]	Speed: 118.220314 samples/sec	 accuracy=47.657939	 loss=2.307396	 lr=0.021307
Epoch[146] Batch [1899]/[3760]	Speed: 117.052025 samples/sec	 accuracy=47.680099	 loss=2.305903	 lr=0.021297
Epoch[146] Batch [1949]/[3760]	Speed: 117.373049 samples/sec	 accuracy=47.683494	 loss=2.305544	 lr=0.021286
Epoch[146] Batch [1999]/[3760]	Speed: 117.051566 samples/sec	 accuracy=47.693750	 loss=2.304103	 lr=0.021276
Epoch[146] Batch [2049]/[3760]	Speed: 117.975261 samples/sec	 accuracy=47.673018	 loss=2.305538	 lr=0.021265
Epoch[146] Batch [2099]/[3760]	Speed: 116.879147 samples/sec	 accuracy=47.659226	 loss=2.305644	 lr=0.021254
Epoch[146] Batch [2149]/[3760]	Speed: 115.199620 samples/sec	 accuracy=47.657703	 loss=2.307284	 lr=0.021244
Epoch[146] Batch [2199]/[3760]	Speed: 116.746813 samples/sec	 accuracy=47.638494	 loss=2.308376	 lr=0.021233
Epoch[146] Batch [2249]/[3760]	Speed: 117.837131 samples/sec	 accuracy=47.614583	 loss=2.309385	 lr=0.021223
Epoch[146] Batch [2299]/[3760]	Speed: 118.043263 samples/sec	 accuracy=47.644022	 loss=2.308130	 lr=0.021212
Epoch[146] Batch [2349]/[3760]	Speed: 117.549922 samples/sec	 accuracy=47.657580	 loss=2.307960	 lr=0.021202
Epoch[146] Batch [2399]/[3760]	Speed: 117.810335 samples/sec	 accuracy=47.626953	 loss=2.309068	 lr=0.021191
Epoch[146] Batch [2449]/[3760]	Speed: 117.369012 samples/sec	 accuracy=47.611607	 loss=2.310360	 lr=0.021181
Epoch[146] Batch [2499]/[3760]	Speed: 117.385064 samples/sec	 accuracy=47.594375	 loss=2.311403	 lr=0.021170
Epoch[146] Batch [2549]/[3760]	Speed: 117.199564 samples/sec	 accuracy=47.561275	 loss=2.312941	 lr=0.021160
Epoch[146] Batch [2599]/[3760]	Speed: 117.087457 samples/sec	 accuracy=47.539663	 loss=2.313412	 lr=0.021149
Epoch[146] Batch [2649]/[3760]	Speed: 118.241793 samples/sec	 accuracy=47.543632	 loss=2.312830	 lr=0.021138
Epoch[146] Batch [2699]/[3760]	Speed: 117.135604 samples/sec	 accuracy=47.562500	 loss=2.311842	 lr=0.021128
Epoch[146] Batch [2749]/[3760]	Speed: 117.652961 samples/sec	 accuracy=47.552841	 loss=2.311855	 lr=0.021117
Epoch[146] Batch [2799]/[3760]	Speed: 116.600493 samples/sec	 accuracy=47.529576	 loss=2.313551	 lr=0.021107
Epoch[146] Batch [2849]/[3760]	Speed: 117.269278 samples/sec	 accuracy=47.528509	 loss=2.313177	 lr=0.021096
Epoch[146] Batch [2899]/[3760]	Speed: 117.040127 samples/sec	 accuracy=47.505927	 loss=2.314644	 lr=0.021086
Epoch[146] Batch [2949]/[3760]	Speed: 117.251614 samples/sec	 accuracy=47.497352	 loss=2.314114	 lr=0.021075
Epoch[146] Batch [2999]/[3760]	Speed: 117.051824 samples/sec	 accuracy=47.485417	 loss=2.314597	 lr=0.021065
Epoch[146] Batch [3049]/[3760]	Speed: 117.325621 samples/sec	 accuracy=47.474385	 loss=2.314812	 lr=0.021054
Epoch[146] Batch [3099]/[3760]	Speed: 117.652997 samples/sec	 accuracy=47.454133	 loss=2.314690	 lr=0.021044
Epoch[146] Batch [3149]/[3760]	Speed: 117.425571 samples/sec	 accuracy=47.480655	 loss=2.313386	 lr=0.021033
Epoch[146] Batch [3199]/[3760]	Speed: 117.269331 samples/sec	 accuracy=47.493164	 loss=2.313721	 lr=0.021023
Epoch[146] Batch [3249]/[3760]	Speed: 117.269543 samples/sec	 accuracy=47.497596	 loss=2.313146	 lr=0.021012
Epoch[146] Batch [3299]/[3760]	Speed: 117.051585 samples/sec	 accuracy=47.494792	 loss=2.313325	 lr=0.021002
Epoch[146] Batch [3349]/[3760]	Speed: 117.823309 samples/sec	 accuracy=47.494403	 loss=2.313487	 lr=0.020991
Epoch[146] Batch [3399]/[3760]	Speed: 117.481038 samples/sec	 accuracy=47.477941	 loss=2.313851	 lr=0.020981
Epoch[146] Batch [3449]/[3760]	Speed: 115.210009 samples/sec	 accuracy=47.472373	 loss=2.314389	 lr=0.020970
Epoch[146] Batch [3499]/[3760]	Speed: 117.689714 samples/sec	 accuracy=47.464732	 loss=2.314863	 lr=0.020960
Epoch[146] Batch [3549]/[3760]	Speed: 117.026539 samples/sec	 accuracy=47.452465	 loss=2.315452	 lr=0.020949
Epoch[146] Batch [3599]/[3760]	Speed: 118.215545 samples/sec	 accuracy=47.457899	 loss=2.315422	 lr=0.020939
Epoch[146] Batch [3649]/[3760]	Speed: 117.513429 samples/sec	 accuracy=47.467894	 loss=2.315425	 lr=0.020928
Epoch[146] Batch [3699]/[3760]	Speed: 117.632759 samples/sec	 accuracy=47.472551	 loss=2.315243	 lr=0.020918
Epoch[146] Batch [3749]/[3760]	Speed: 120.177927 samples/sec	 accuracy=47.475417	 loss=2.315773	 lr=0.020907
Batch [0049]/[0303]: acc-top1=49.125000 acc-top5=72.406250
Batch [0099]/[0303]: acc-top1=45.531250 acc-top5=70.140625
Batch [0149]/[0303]: acc-top1=47.604167 acc-top5=71.510417
Batch [0199]/[0303]: acc-top1=49.687500 acc-top5=73.375000
Batch [0249]/[0303]: acc-top1=48.693750 acc-top5=73.093750
Batch [0299]/[0303]: acc-top1=47.776042 acc-top5=72.843750
[Epoch 146] training: accuracy=47.470495	 loss=2.316125
[Epoch 146] speed: 116 samples/sec	time cost: 2176.165482
[Epoch 146] validation: acc-top1=47.715553 acc-top5=72.803218 loss=2.406818
Epoch[147] Batch [0049]/[3760]	Speed: 87.807411 samples/sec	 accuracy=47.843750	 loss=2.258411	 lr=0.020895
Epoch[147] Batch [0099]/[3760]	Speed: 117.689972 samples/sec	 accuracy=47.843750	 loss=2.258068	 lr=0.020884
Epoch[147] Batch [0149]/[3760]	Speed: 117.422196 samples/sec	 accuracy=47.583333	 loss=2.275325	 lr=0.020874
Epoch[147] Batch [0199]/[3760]	Speed: 117.165805 samples/sec	 accuracy=47.976562	 loss=2.265005	 lr=0.020863
Epoch[147] Batch [0249]/[3760]	Speed: 117.334047 samples/sec	 accuracy=48.000000	 loss=2.267718	 lr=0.020853
Epoch[147] Batch [0299]/[3760]	Speed: 117.937364 samples/sec	 accuracy=47.760417	 loss=2.280714	 lr=0.020842
Epoch[147] Batch [0349]/[3760]	Speed: 117.556285 samples/sec	 accuracy=47.995536	 loss=2.270710	 lr=0.020832
Epoch[147] Batch [0399]/[3760]	Speed: 116.585558 samples/sec	 accuracy=47.976562	 loss=2.268134	 lr=0.020821
Epoch[147] Batch [0449]/[3760]	Speed: 117.978840 samples/sec	 accuracy=48.083333	 loss=2.268537	 lr=0.020811
Epoch[147] Batch [0499]/[3760]	Speed: 117.608552 samples/sec	 accuracy=48.087500	 loss=2.267140	 lr=0.020800
Epoch[147] Batch [0549]/[3760]	Speed: 117.523279 samples/sec	 accuracy=47.954545	 loss=2.274492	 lr=0.020790
Epoch[147] Batch [0599]/[3760]	Speed: 116.944043 samples/sec	 accuracy=47.908854	 loss=2.278925	 lr=0.020779
Epoch[147] Batch [0649]/[3760]	Speed: 116.755679 samples/sec	 accuracy=47.867788	 loss=2.281999	 lr=0.020769
Epoch[147] Batch [0699]/[3760]	Speed: 117.428593 samples/sec	 accuracy=47.823661	 loss=2.285931	 lr=0.020758
Epoch[147] Batch [0749]/[3760]	Speed: 117.392774 samples/sec	 accuracy=47.793750	 loss=2.286291	 lr=0.020748
Epoch[147] Batch [0799]/[3760]	Speed: 118.042044 samples/sec	 accuracy=47.808594	 loss=2.287877	 lr=0.020738
Epoch[147] Batch [0849]/[3760]	Speed: 117.638635 samples/sec	 accuracy=47.779412	 loss=2.288292	 lr=0.020727
Epoch[147] Batch [0899]/[3760]	Speed: 116.972608 samples/sec	 accuracy=47.739583	 loss=2.289032	 lr=0.020717
Epoch[147] Batch [0949]/[3760]	Speed: 115.162265 samples/sec	 accuracy=47.781250	 loss=2.287688	 lr=0.020706
Epoch[147] Batch [0999]/[3760]	Speed: 117.899818 samples/sec	 accuracy=47.809375	 loss=2.286948	 lr=0.020696
Epoch[147] Batch [1049]/[3760]	Speed: 117.199469 samples/sec	 accuracy=47.840774	 loss=2.286024	 lr=0.020685
Epoch[147] Batch [1099]/[3760]	Speed: 117.570021 samples/sec	 accuracy=47.811080	 loss=2.287722	 lr=0.020675
Epoch[147] Batch [1149]/[3760]	Speed: 117.736412 samples/sec	 accuracy=47.820652	 loss=2.287062	 lr=0.020664
Epoch[147] Batch [1199]/[3760]	Speed: 116.456950 samples/sec	 accuracy=47.854167	 loss=2.285160	 lr=0.020654
Epoch[147] Batch [1249]/[3760]	Speed: 118.168232 samples/sec	 accuracy=47.902500	 loss=2.283342	 lr=0.020643
Epoch[147] Batch [1299]/[3760]	Speed: 117.030120 samples/sec	 accuracy=47.900240	 loss=2.283513	 lr=0.020633
Epoch[147] Batch [1349]/[3760]	Speed: 116.900132 samples/sec	 accuracy=47.857639	 loss=2.284289	 lr=0.020623
Epoch[147] Batch [1399]/[3760]	Speed: 116.894435 samples/sec	 accuracy=47.854911	 loss=2.285984	 lr=0.020612
Epoch[147] Batch [1449]/[3760]	Speed: 117.327450 samples/sec	 accuracy=47.810345	 loss=2.287880	 lr=0.020602
Epoch[147] Batch [1499]/[3760]	Speed: 117.329872 samples/sec	 accuracy=47.792708	 loss=2.289102	 lr=0.020591
Epoch[147] Batch [1549]/[3760]	Speed: 116.552747 samples/sec	 accuracy=47.757056	 loss=2.291968	 lr=0.020581
Epoch[147] Batch [1599]/[3760]	Speed: 117.507157 samples/sec	 accuracy=47.753906	 loss=2.293211	 lr=0.020570
Epoch[147] Batch [1649]/[3760]	Speed: 117.039247 samples/sec	 accuracy=47.777462	 loss=2.293256	 lr=0.020560
Epoch[147] Batch [1699]/[3760]	Speed: 117.458030 samples/sec	 accuracy=47.796875	 loss=2.292954	 lr=0.020550
Epoch[147] Batch [1749]/[3760]	Speed: 117.489165 samples/sec	 accuracy=47.794643	 loss=2.293567	 lr=0.020539
Epoch[147] Batch [1799]/[3760]	Speed: 117.962440 samples/sec	 accuracy=47.808160	 loss=2.294627	 lr=0.020529
Epoch[147] Batch [1849]/[3760]	Speed: 117.128376 samples/sec	 accuracy=47.799831	 loss=2.295068	 lr=0.020518
Epoch[147] Batch [1899]/[3760]	Speed: 117.841717 samples/sec	 accuracy=47.822368	 loss=2.294140	 lr=0.020508
Epoch[147] Batch [1949]/[3760]	Speed: 117.192455 samples/sec	 accuracy=47.815705	 loss=2.294617	 lr=0.020498
Epoch[147] Batch [1999]/[3760]	Speed: 117.398790 samples/sec	 accuracy=47.828125	 loss=2.295103	 lr=0.020487
Epoch[147] Batch [2049]/[3760]	Speed: 117.120802 samples/sec	 accuracy=47.853659	 loss=2.293905	 lr=0.020477
Epoch[147] Batch [2099]/[3760]	Speed: 117.410484 samples/sec	 accuracy=47.840774	 loss=2.295006	 lr=0.020466
Epoch[147] Batch [2149]/[3760]	Speed: 118.018591 samples/sec	 accuracy=47.858285	 loss=2.296458	 lr=0.020456
Epoch[147] Batch [2199]/[3760]	Speed: 117.226637 samples/sec	 accuracy=47.866477	 loss=2.296513	 lr=0.020445
Epoch[147] Batch [2249]/[3760]	Speed: 115.485868 samples/sec	 accuracy=47.854861	 loss=2.296619	 lr=0.020435
Epoch[147] Batch [2299]/[3760]	Speed: 117.353214 samples/sec	 accuracy=47.885190	 loss=2.294812	 lr=0.020425
Epoch[147] Batch [2349]/[3760]	Speed: 116.912367 samples/sec	 accuracy=47.867686	 loss=2.295042	 lr=0.020414
Epoch[147] Batch [2399]/[3760]	Speed: 117.340750 samples/sec	 accuracy=47.902344	 loss=2.294241	 lr=0.020404
Epoch[147] Batch [2449]/[3760]	Speed: 117.863007 samples/sec	 accuracy=47.902423	 loss=2.293546	 lr=0.020393
Epoch[147] Batch [2499]/[3760]	Speed: 118.006229 samples/sec	 accuracy=47.923750	 loss=2.292731	 lr=0.020383
Epoch[147] Batch [2549]/[3760]	Speed: 117.200030 samples/sec	 accuracy=47.911765	 loss=2.293002	 lr=0.020373
Epoch[147] Batch [2599]/[3760]	Speed: 117.354223 samples/sec	 accuracy=47.912260	 loss=2.293184	 lr=0.020362
Epoch[147] Batch [2649]/[3760]	Speed: 117.188389 samples/sec	 accuracy=47.922170	 loss=2.292092	 lr=0.020352
Epoch[147] Batch [2699]/[3760]	Speed: 117.690780 samples/sec	 accuracy=47.915509	 loss=2.291838	 lr=0.020342
Epoch[147] Batch [2749]/[3760]	Speed: 117.644243 samples/sec	 accuracy=47.920455	 loss=2.292386	 lr=0.020331
Epoch[147] Batch [2799]/[3760]	Speed: 117.301198 samples/sec	 accuracy=47.887835	 loss=2.294201	 lr=0.020321
Epoch[147] Batch [2849]/[3760]	Speed: 117.152646 samples/sec	 accuracy=47.894737	 loss=2.294241	 lr=0.020310
Epoch[147] Batch [2899]/[3760]	Speed: 117.144821 samples/sec	 accuracy=47.913793	 loss=2.293728	 lr=0.020300
Epoch[147] Batch [2949]/[3760]	Speed: 117.678584 samples/sec	 accuracy=47.901483	 loss=2.294380	 lr=0.020290
Epoch[147] Batch [2999]/[3760]	Speed: 117.265768 samples/sec	 accuracy=47.882812	 loss=2.295204	 lr=0.020279
Epoch[147] Batch [3049]/[3760]	Speed: 117.328064 samples/sec	 accuracy=47.843750	 loss=2.295832	 lr=0.020269
Epoch[147] Batch [3099]/[3760]	Speed: 117.488447 samples/sec	 accuracy=47.824597	 loss=2.296283	 lr=0.020259
Epoch[147] Batch [3149]/[3760]	Speed: 117.786637 samples/sec	 accuracy=47.817460	 loss=2.296590	 lr=0.020248
Epoch[147] Batch [3199]/[3760]	Speed: 118.029368 samples/sec	 accuracy=47.820312	 loss=2.296749	 lr=0.020238
Epoch[147] Batch [3249]/[3760]	Speed: 116.639147 samples/sec	 accuracy=47.798077	 loss=2.297374	 lr=0.020227
Epoch[147] Batch [3299]/[3760]	Speed: 117.623669 samples/sec	 accuracy=47.792614	 loss=2.297891	 lr=0.020217
Epoch[147] Batch [3349]/[3760]	Speed: 117.530694 samples/sec	 accuracy=47.801772	 loss=2.297921	 lr=0.020207
Epoch[147] Batch [3399]/[3760]	Speed: 117.626003 samples/sec	 accuracy=47.776654	 loss=2.298327	 lr=0.020196
Epoch[147] Batch [3449]/[3760]	Speed: 117.104394 samples/sec	 accuracy=47.771739	 loss=2.297763	 lr=0.020186
Epoch[147] Batch [3499]/[3760]	Speed: 115.413822 samples/sec	 accuracy=47.761607	 loss=2.298401	 lr=0.020176
Epoch[147] Batch [3549]/[3760]	Speed: 117.288973 samples/sec	 accuracy=47.761884	 loss=2.298803	 lr=0.020165
Epoch[147] Batch [3599]/[3760]	Speed: 116.973718 samples/sec	 accuracy=47.765625	 loss=2.298764	 lr=0.020155
Epoch[147] Batch [3649]/[3760]	Speed: 117.589472 samples/sec	 accuracy=47.779538	 loss=2.298237	 lr=0.020145
Epoch[147] Batch [3699]/[3760]	Speed: 117.317459 samples/sec	 accuracy=47.809122	 loss=2.297837	 lr=0.020134
Epoch[147] Batch [3749]/[3760]	Speed: 120.361261 samples/sec	 accuracy=47.785833	 loss=2.298617	 lr=0.020124
Batch [0049]/[0303]: acc-top1=48.937500 acc-top5=72.187500
Batch [0099]/[0303]: acc-top1=46.234375 acc-top5=71.406250
Batch [0149]/[0303]: acc-top1=47.604167 acc-top5=72.708333
Batch [0199]/[0303]: acc-top1=50.218750 acc-top5=74.664062
Batch [0249]/[0303]: acc-top1=49.312500 acc-top5=74.118750
Batch [0299]/[0303]: acc-top1=49.145833 acc-top5=74.062500
[Epoch 147] training: accuracy=47.793800	 loss=2.298203
[Epoch 147] speed: 116 samples/sec	time cost: 2177.161673
[Epoch 147] validation: acc-top1=49.076939 acc-top5=74.025371 loss=2.331671
Epoch[148] Batch [0049]/[3759]	Speed: 87.645105 samples/sec	 accuracy=48.750000	 loss=2.285290	 lr=0.020112
Epoch[148] Batch [0099]/[3759]	Speed: 117.403933 samples/sec	 accuracy=48.500000	 loss=2.260306	 lr=0.020101
Epoch[148] Batch [0149]/[3759]	Speed: 117.138178 samples/sec	 accuracy=48.052083	 loss=2.285170	 lr=0.020091
Epoch[148] Batch [0199]/[3759]	Speed: 117.898776 samples/sec	 accuracy=48.281250	 loss=2.262248	 lr=0.020081
Epoch[148] Batch [0249]/[3759]	Speed: 117.848572 samples/sec	 accuracy=48.537500	 loss=2.260873	 lr=0.020070
Epoch[148] Batch [0299]/[3759]	Speed: 117.371812 samples/sec	 accuracy=48.687500	 loss=2.250254	 lr=0.020060
Epoch[148] Batch [0349]/[3759]	Speed: 118.099853 samples/sec	 accuracy=48.691964	 loss=2.249381	 lr=0.020050
Epoch[148] Batch [0399]/[3759]	Speed: 117.235511 samples/sec	 accuracy=48.785156	 loss=2.249111	 lr=0.020039
Epoch[148] Batch [0449]/[3759]	Speed: 117.972503 samples/sec	 accuracy=48.795139	 loss=2.254225	 lr=0.020029
Epoch[148] Batch [0499]/[3759]	Speed: 117.649741 samples/sec	 accuracy=48.696875	 loss=2.255721	 lr=0.020019
Epoch[148] Batch [0549]/[3759]	Speed: 116.804880 samples/sec	 accuracy=48.707386	 loss=2.254774	 lr=0.020008
Epoch[148] Batch [0599]/[3759]	Speed: 117.048596 samples/sec	 accuracy=48.682292	 loss=2.254425	 lr=0.019998
Epoch[148] Batch [0649]/[3759]	Speed: 117.823977 samples/sec	 accuracy=48.682692	 loss=2.255078	 lr=0.019988
Epoch[148] Batch [0699]/[3759]	Speed: 117.786901 samples/sec	 accuracy=48.676339	 loss=2.254088	 lr=0.019977
Epoch[148] Batch [0749]/[3759]	Speed: 116.659674 samples/sec	 accuracy=48.608333	 loss=2.256269	 lr=0.019967
Epoch[148] Batch [0799]/[3759]	Speed: 117.622106 samples/sec	 accuracy=48.556641	 loss=2.256771	 lr=0.019957
Epoch[148] Batch [0849]/[3759]	Speed: 117.918705 samples/sec	 accuracy=48.597426	 loss=2.256005	 lr=0.019946
Epoch[148] Batch [0899]/[3759]	Speed: 115.824719 samples/sec	 accuracy=48.598958	 loss=2.255542	 lr=0.019936
Epoch[148] Batch [0949]/[3759]	Speed: 116.808460 samples/sec	 accuracy=48.557566	 loss=2.258326	 lr=0.019926
Epoch[148] Batch [0999]/[3759]	Speed: 116.811561 samples/sec	 accuracy=48.576562	 loss=2.258410	 lr=0.019915
Epoch[148] Batch [1049]/[3759]	Speed: 117.371668 samples/sec	 accuracy=48.526786	 loss=2.261893	 lr=0.019905
Epoch[148] Batch [1099]/[3759]	Speed: 117.059810 samples/sec	 accuracy=48.552557	 loss=2.261560	 lr=0.019895
Epoch[148] Batch [1149]/[3759]	Speed: 117.120845 samples/sec	 accuracy=48.536685	 loss=2.261087	 lr=0.019885
Epoch[148] Batch [1199]/[3759]	Speed: 117.474581 samples/sec	 accuracy=48.566406	 loss=2.259988	 lr=0.019874
Epoch[148] Batch [1249]/[3759]	Speed: 117.461733 samples/sec	 accuracy=48.573750	 loss=2.259968	 lr=0.019864
Epoch[148] Batch [1299]/[3759]	Speed: 117.579049 samples/sec	 accuracy=48.576923	 loss=2.258914	 lr=0.019854
Epoch[148] Batch [1349]/[3759]	Speed: 117.860625 samples/sec	 accuracy=48.521991	 loss=2.262087	 lr=0.019843
Epoch[148] Batch [1399]/[3759]	Speed: 117.238819 samples/sec	 accuracy=48.477679	 loss=2.263823	 lr=0.019833
Epoch[148] Batch [1449]/[3759]	Speed: 117.069828 samples/sec	 accuracy=48.487069	 loss=2.263918	 lr=0.019823
Epoch[148] Batch [1499]/[3759]	Speed: 117.182743 samples/sec	 accuracy=48.519792	 loss=2.262438	 lr=0.019813
Epoch[148] Batch [1549]/[3759]	Speed: 117.117740 samples/sec	 accuracy=48.522177	 loss=2.261961	 lr=0.019802
Epoch[148] Batch [1599]/[3759]	Speed: 117.571521 samples/sec	 accuracy=48.504883	 loss=2.263066	 lr=0.019792
Epoch[148] Batch [1649]/[3759]	Speed: 117.360863 samples/sec	 accuracy=48.498106	 loss=2.263796	 lr=0.019782
Epoch[148] Batch [1699]/[3759]	Speed: 118.280443 samples/sec	 accuracy=48.492647	 loss=2.263435	 lr=0.019771
Epoch[148] Batch [1749]/[3759]	Speed: 117.145329 samples/sec	 accuracy=48.508036	 loss=2.261936	 lr=0.019761
Epoch[148] Batch [1799]/[3759]	Speed: 117.394727 samples/sec	 accuracy=48.431424	 loss=2.264582	 lr=0.019751
Epoch[148] Batch [1849]/[3759]	Speed: 117.279640 samples/sec	 accuracy=48.415541	 loss=2.265298	 lr=0.019741
Epoch[148] Batch [1899]/[3759]	Speed: 117.786032 samples/sec	 accuracy=48.434211	 loss=2.264363	 lr=0.019730
Epoch[148] Batch [1949]/[3759]	Speed: 117.042577 samples/sec	 accuracy=48.475160	 loss=2.262075	 lr=0.019720
Epoch[148] Batch [1999]/[3759]	Speed: 117.956799 samples/sec	 accuracy=48.450781	 loss=2.262960	 lr=0.019710
Epoch[148] Batch [2049]/[3759]	Speed: 117.384969 samples/sec	 accuracy=48.453506	 loss=2.262668	 lr=0.019700
Epoch[148] Batch [2099]/[3759]	Speed: 117.316553 samples/sec	 accuracy=48.439732	 loss=2.263288	 lr=0.019689
Epoch[148] Batch [2149]/[3759]	Speed: 117.295083 samples/sec	 accuracy=48.437500	 loss=2.264236	 lr=0.019679
Epoch[148] Batch [2199]/[3759]	Speed: 115.563926 samples/sec	 accuracy=48.392756	 loss=2.265737	 lr=0.019669
Epoch[148] Batch [2249]/[3759]	Speed: 117.515680 samples/sec	 accuracy=48.391667	 loss=2.264680	 lr=0.019659
Epoch[148] Batch [2299]/[3759]	Speed: 117.023354 samples/sec	 accuracy=48.370924	 loss=2.265658	 lr=0.019648
Epoch[148] Batch [2349]/[3759]	Speed: 117.509217 samples/sec	 accuracy=48.402926	 loss=2.264278	 lr=0.019638
Epoch[148] Batch [2399]/[3759]	Speed: 117.870853 samples/sec	 accuracy=48.388021	 loss=2.265166	 lr=0.019628
Epoch[148] Batch [2449]/[3759]	Speed: 117.568972 samples/sec	 accuracy=48.377551	 loss=2.265429	 lr=0.019618
Epoch[148] Batch [2499]/[3759]	Speed: 117.716636 samples/sec	 accuracy=48.383750	 loss=2.264328	 lr=0.019607
Epoch[148] Batch [2549]/[3759]	Speed: 117.344714 samples/sec	 accuracy=48.399510	 loss=2.264211	 lr=0.019597
Epoch[148] Batch [2599]/[3759]	Speed: 117.768945 samples/sec	 accuracy=48.381010	 loss=2.264584	 lr=0.019587
Epoch[148] Batch [2649]/[3759]	Speed: 117.582707 samples/sec	 accuracy=48.390920	 loss=2.264611	 lr=0.019577
Epoch[148] Batch [2699]/[3759]	Speed: 117.353467 samples/sec	 accuracy=48.408565	 loss=2.264146	 lr=0.019566
Epoch[148] Batch [2749]/[3759]	Speed: 117.461029 samples/sec	 accuracy=48.414773	 loss=2.264145	 lr=0.019556
Epoch[148] Batch [2799]/[3759]	Speed: 117.850272 samples/sec	 accuracy=48.376674	 loss=2.266521	 lr=0.019546
Epoch[148] Batch [2849]/[3759]	Speed: 117.329471 samples/sec	 accuracy=48.369518	 loss=2.267049	 lr=0.019536
Epoch[148] Batch [2899]/[3759]	Speed: 117.069730 samples/sec	 accuracy=48.332974	 loss=2.269031	 lr=0.019525
Epoch[148] Batch [2949]/[3759]	Speed: 118.329471 samples/sec	 accuracy=48.318856	 loss=2.269008	 lr=0.019515
Epoch[148] Batch [2999]/[3759]	Speed: 117.156470 samples/sec	 accuracy=48.305208	 loss=2.269925	 lr=0.019505
Epoch[148] Batch [3049]/[3759]	Speed: 116.978703 samples/sec	 accuracy=48.307889	 loss=2.269519	 lr=0.019495
Epoch[148] Batch [3099]/[3759]	Speed: 117.325964 samples/sec	 accuracy=48.320060	 loss=2.269480	 lr=0.019485
Epoch[148] Batch [3149]/[3759]	Speed: 117.345342 samples/sec	 accuracy=48.327381	 loss=2.269816	 lr=0.019474
Epoch[148] Batch [3199]/[3759]	Speed: 117.808919 samples/sec	 accuracy=48.341797	 loss=2.269682	 lr=0.019464
Epoch[148] Batch [3249]/[3759]	Speed: 117.983917 samples/sec	 accuracy=48.342308	 loss=2.269844	 lr=0.019454
Epoch[148] Batch [3299]/[3759]	Speed: 117.412407 samples/sec	 accuracy=48.333807	 loss=2.269994	 lr=0.019444
Epoch[148] Batch [3349]/[3759]	Speed: 116.942580 samples/sec	 accuracy=48.338153	 loss=2.269745	 lr=0.019433
Epoch[148] Batch [3399]/[3759]	Speed: 117.733459 samples/sec	 accuracy=48.331801	 loss=2.269880	 lr=0.019423
Epoch[148] Batch [3449]/[3759]	Speed: 115.829032 samples/sec	 accuracy=48.310236	 loss=2.270523	 lr=0.019413
Epoch[148] Batch [3499]/[3759]	Speed: 118.072440 samples/sec	 accuracy=48.316964	 loss=2.270547	 lr=0.019403
Epoch[148] Batch [3549]/[3759]	Speed: 117.124668 samples/sec	 accuracy=48.326585	 loss=2.269990	 lr=0.019393
Epoch[148] Batch [3599]/[3759]	Speed: 117.125657 samples/sec	 accuracy=48.310764	 loss=2.270597	 lr=0.019382
Epoch[148] Batch [3649]/[3759]	Speed: 118.039709 samples/sec	 accuracy=48.305651	 loss=2.270731	 lr=0.019372
Epoch[148] Batch [3699]/[3759]	Speed: 117.495097 samples/sec	 accuracy=48.305743	 loss=2.270751	 lr=0.019362
Epoch[148] Batch [3749]/[3759]	Speed: 120.317520 samples/sec	 accuracy=48.330000	 loss=2.269813	 lr=0.019352
Batch [0049]/[0303]: acc-top1=47.156250 acc-top5=71.750000
Batch [0099]/[0303]: acc-top1=45.281250 acc-top5=71.656250
Batch [0149]/[0303]: acc-top1=46.854167 acc-top5=72.375000
Batch [0199]/[0303]: acc-top1=49.882812 acc-top5=74.320312
Batch [0249]/[0303]: acc-top1=49.575000 acc-top5=74.243750
Batch [0299]/[0303]: acc-top1=48.890625 acc-top5=74.218750
[Epoch 148] training: accuracy=48.323191	 loss=2.269907
[Epoch 148] speed: 116 samples/sec	time cost: 2174.416856
[Epoch 148] validation: acc-top1=48.973804 acc-top5=74.241955 loss=2.327983
Epoch[149] Batch [0049]/[3760]	Speed: 85.400264 samples/sec	 accuracy=48.718750	 loss=2.233452	 lr=0.019340
Epoch[149] Batch [0099]/[3760]	Speed: 117.384400 samples/sec	 accuracy=48.921875	 loss=2.232734	 lr=0.019330
Epoch[149] Batch [0149]/[3760]	Speed: 117.298622 samples/sec	 accuracy=48.885417	 loss=2.241225	 lr=0.019319
Epoch[149] Batch [0199]/[3760]	Speed: 117.675510 samples/sec	 accuracy=48.875000	 loss=2.232627	 lr=0.019309
Epoch[149] Batch [0249]/[3760]	Speed: 116.932613 samples/sec	 accuracy=48.768750	 loss=2.240157	 lr=0.019299
Epoch[149] Batch [0299]/[3760]	Speed: 116.666400 samples/sec	 accuracy=48.760417	 loss=2.229255	 lr=0.019289
Epoch[149] Batch [0349]/[3760]	Speed: 118.765459 samples/sec	 accuracy=48.839286	 loss=2.228163	 lr=0.019279
Epoch[149] Batch [0399]/[3760]	Speed: 116.767364 samples/sec	 accuracy=48.843750	 loss=2.233707	 lr=0.019269
Epoch[149] Batch [0449]/[3760]	Speed: 117.775177 samples/sec	 accuracy=48.909722	 loss=2.238259	 lr=0.019258
Epoch[149] Batch [0499]/[3760]	Speed: 116.783686 samples/sec	 accuracy=48.896875	 loss=2.239371	 lr=0.019248
Epoch[149] Batch [0549]/[3760]	Speed: 117.560687 samples/sec	 accuracy=48.931818	 loss=2.236496	 lr=0.019238
Epoch[149] Batch [0599]/[3760]	Speed: 117.524480 samples/sec	 accuracy=49.007812	 loss=2.236274	 lr=0.019228
Epoch[149] Batch [0649]/[3760]	Speed: 117.313572 samples/sec	 accuracy=48.927885	 loss=2.240600	 lr=0.019218
Epoch[149] Batch [0699]/[3760]	Speed: 117.506068 samples/sec	 accuracy=48.968750	 loss=2.237510	 lr=0.019208
Epoch[149] Batch [0749]/[3760]	Speed: 117.284790 samples/sec	 accuracy=48.966667	 loss=2.238092	 lr=0.019197
Epoch[149] Batch [0799]/[3760]	Speed: 117.166177 samples/sec	 accuracy=49.019531	 loss=2.237904	 lr=0.019187
Epoch[149] Batch [0849]/[3760]	Speed: 116.806232 samples/sec	 accuracy=48.948529	 loss=2.240667	 lr=0.019177
Epoch[149] Batch [0899]/[3760]	Speed: 117.445588 samples/sec	 accuracy=48.927083	 loss=2.241399	 lr=0.019167
Epoch[149] Batch [0949]/[3760]	Speed: 115.785176 samples/sec	 accuracy=48.930921	 loss=2.242660	 lr=0.019157
Epoch[149] Batch [0999]/[3760]	Speed: 116.834743 samples/sec	 accuracy=48.964062	 loss=2.239612	 lr=0.019147
Epoch[149] Batch [1049]/[3760]	Speed: 117.292101 samples/sec	 accuracy=48.950893	 loss=2.240137	 lr=0.019137
Epoch[149] Batch [1099]/[3760]	Speed: 116.517868 samples/sec	 accuracy=48.873580	 loss=2.243263	 lr=0.019126
Epoch[149] Batch [1149]/[3760]	Speed: 118.149008 samples/sec	 accuracy=48.860054	 loss=2.244882	 lr=0.019116
Epoch[149] Batch [1199]/[3760]	Speed: 116.986999 samples/sec	 accuracy=48.871094	 loss=2.243500	 lr=0.019106
Epoch[149] Batch [1249]/[3760]	Speed: 117.209756 samples/sec	 accuracy=48.836250	 loss=2.245902	 lr=0.019096
Epoch[149] Batch [1299]/[3760]	Speed: 118.103182 samples/sec	 accuracy=48.776442	 loss=2.249261	 lr=0.019086
Epoch[149] Batch [1349]/[3760]	Speed: 117.390370 samples/sec	 accuracy=48.744213	 loss=2.250726	 lr=0.019076
Epoch[149] Batch [1399]/[3760]	Speed: 117.053943 samples/sec	 accuracy=48.760045	 loss=2.250562	 lr=0.019066
Epoch[149] Batch [1449]/[3760]	Speed: 117.515341 samples/sec	 accuracy=48.775862	 loss=2.249563	 lr=0.019055
Epoch[149] Batch [1499]/[3760]	Speed: 116.941243 samples/sec	 accuracy=48.759375	 loss=2.249634	 lr=0.019045
Epoch[149] Batch [1549]/[3760]	Speed: 118.225116 samples/sec	 accuracy=48.792339	 loss=2.248769	 lr=0.019035
Epoch[149] Batch [1599]/[3760]	Speed: 117.465166 samples/sec	 accuracy=48.789062	 loss=2.247960	 lr=0.019025
Epoch[149] Batch [1649]/[3760]	Speed: 116.888627 samples/sec	 accuracy=48.747159	 loss=2.249848	 lr=0.019015
Epoch[149] Batch [1699]/[3760]	Speed: 118.060742 samples/sec	 accuracy=48.765625	 loss=2.248293	 lr=0.019005
Epoch[149] Batch [1749]/[3760]	Speed: 117.086800 samples/sec	 accuracy=48.775000	 loss=2.248647	 lr=0.018995
Epoch[149] Batch [1799]/[3760]	Speed: 116.872720 samples/sec	 accuracy=48.775174	 loss=2.247549	 lr=0.018985
Epoch[149] Batch [1849]/[3760]	Speed: 117.296378 samples/sec	 accuracy=48.756757	 loss=2.248516	 lr=0.018974
Epoch[149] Batch [1899]/[3760]	Speed: 117.018302 samples/sec	 accuracy=48.741776	 loss=2.249381	 lr=0.018964
Epoch[149] Batch [1949]/[3760]	Speed: 116.827833 samples/sec	 accuracy=48.692308	 loss=2.251460	 lr=0.018954
Epoch[149] Batch [1999]/[3760]	Speed: 116.860974 samples/sec	 accuracy=48.685938	 loss=2.251154	 lr=0.018944
Epoch[149] Batch [2049]/[3760]	Speed: 117.088369 samples/sec	 accuracy=48.641768	 loss=2.251403	 lr=0.018934
Epoch[149] Batch [2099]/[3760]	Speed: 116.665169 samples/sec	 accuracy=48.663690	 loss=2.250506	 lr=0.018924
Epoch[149] Batch [2149]/[3760]	Speed: 116.872221 samples/sec	 accuracy=48.664244	 loss=2.250888	 lr=0.018914
Epoch[149] Batch [2199]/[3760]	Speed: 114.654888 samples/sec	 accuracy=48.671875	 loss=2.250856	 lr=0.018904
Epoch[149] Batch [2249]/[3760]	Speed: 118.309200 samples/sec	 accuracy=48.665972	 loss=2.250931	 lr=0.018894
Epoch[149] Batch [2299]/[3760]	Speed: 116.910324 samples/sec	 accuracy=48.653533	 loss=2.251501	 lr=0.018883
Epoch[149] Batch [2349]/[3760]	Speed: 117.509460 samples/sec	 accuracy=48.686170	 loss=2.251583	 lr=0.018873
Epoch[149] Batch [2399]/[3760]	Speed: 118.035223 samples/sec	 accuracy=48.667318	 loss=2.252941	 lr=0.018863
Epoch[149] Batch [2449]/[3760]	Speed: 117.368670 samples/sec	 accuracy=48.690689	 loss=2.251772	 lr=0.018853
Epoch[149] Batch [2499]/[3760]	Speed: 118.161581 samples/sec	 accuracy=48.673125	 loss=2.252579	 lr=0.018843
Epoch[149] Batch [2549]/[3760]	Speed: 117.641541 samples/sec	 accuracy=48.656250	 loss=2.253350	 lr=0.018833
Epoch[149] Batch [2599]/[3760]	Speed: 117.088161 samples/sec	 accuracy=48.688101	 loss=2.252166	 lr=0.018823
Epoch[149] Batch [2649]/[3760]	Speed: 117.633020 samples/sec	 accuracy=48.711085	 loss=2.252500	 lr=0.018813
Epoch[149] Batch [2699]/[3760]	Speed: 117.334877 samples/sec	 accuracy=48.710648	 loss=2.251987	 lr=0.018803
Epoch[149] Batch [2749]/[3760]	Speed: 117.195253 samples/sec	 accuracy=48.715909	 loss=2.252082	 lr=0.018793
Epoch[149] Batch [2799]/[3760]	Speed: 117.862080 samples/sec	 accuracy=48.718750	 loss=2.251958	 lr=0.018783
Epoch[149] Batch [2849]/[3760]	Speed: 117.316553 samples/sec	 accuracy=48.715461	 loss=2.252495	 lr=0.018773
Epoch[149] Batch [2899]/[3760]	Speed: 117.635344 samples/sec	 accuracy=48.718211	 loss=2.252703	 lr=0.018762
Epoch[149] Batch [2949]/[3760]	Speed: 117.043569 samples/sec	 accuracy=48.699682	 loss=2.252393	 lr=0.018752
Epoch[149] Batch [2999]/[3760]	Speed: 116.634932 samples/sec	 accuracy=48.696875	 loss=2.252354	 lr=0.018742
Epoch[149] Batch [3049]/[3760]	Speed: 117.341294 samples/sec	 accuracy=48.681865	 loss=2.252125	 lr=0.018732
Epoch[149] Batch [3099]/[3760]	Speed: 118.184695 samples/sec	 accuracy=48.693044	 loss=2.251689	 lr=0.018722
Epoch[149] Batch [3149]/[3760]	Speed: 117.525400 samples/sec	 accuracy=48.688492	 loss=2.251547	 lr=0.018712
Epoch[149] Batch [3199]/[3760]	Speed: 117.330443 samples/sec	 accuracy=48.714355	 loss=2.251134	 lr=0.018702
Epoch[149] Batch [3249]/[3760]	Speed: 117.723800 samples/sec	 accuracy=48.731250	 loss=2.250432	 lr=0.018692
Epoch[149] Batch [3299]/[3760]	Speed: 117.244132 samples/sec	 accuracy=48.727746	 loss=2.250701	 lr=0.018682
Epoch[149] Batch [3349]/[3760]	Speed: 117.532169 samples/sec	 accuracy=48.699160	 loss=2.251882	 lr=0.018672
Epoch[149] Batch [3399]/[3760]	Speed: 117.993486 samples/sec	 accuracy=48.698070	 loss=2.251602	 lr=0.018662
Epoch[149] Batch [3449]/[3760]	Speed: 117.253198 samples/sec	 accuracy=48.673460	 loss=2.252524	 lr=0.018652
Epoch[149] Batch [3499]/[3760]	Speed: 115.476532 samples/sec	 accuracy=48.650446	 loss=2.253128	 lr=0.018642
Epoch[149] Batch [3549]/[3760]	Speed: 117.731812 samples/sec	 accuracy=48.632923	 loss=2.253630	 lr=0.018632
Epoch[149] Batch [3599]/[3760]	Speed: 117.386485 samples/sec	 accuracy=48.644531	 loss=2.253111	 lr=0.018622
Epoch[149] Batch [3649]/[3760]	Speed: 117.832975 samples/sec	 accuracy=48.630137	 loss=2.253923	 lr=0.018612
Epoch[149] Batch [3699]/[3760]	Speed: 117.958452 samples/sec	 accuracy=48.635980	 loss=2.254166	 lr=0.018602
Epoch[149] Batch [3749]/[3760]	Speed: 120.144219 samples/sec	 accuracy=48.625417	 loss=2.254670	 lr=0.018592
Batch [0049]/[0303]: acc-top1=49.531250 acc-top5=72.531250
Batch [0099]/[0303]: acc-top1=46.890625 acc-top5=71.593750
Batch [0149]/[0303]: acc-top1=48.406250 acc-top5=72.437500
Batch [0199]/[0303]: acc-top1=51.000000 acc-top5=74.195312
Batch [0249]/[0303]: acc-top1=50.187500 acc-top5=73.700000
Batch [0299]/[0303]: acc-top1=49.536458 acc-top5=73.828125
[Epoch 149] training: accuracy=48.614112	 loss=2.254824
[Epoch 149] speed: 116 samples/sec	time cost: 2177.089181
[Epoch 149] validation: acc-top1=49.546205 acc-top5=73.880982 loss=2.325133
Epoch[150] Batch [0049]/[3760]	Speed: 86.219719 samples/sec	 accuracy=51.468750	 loss=2.127971	 lr=0.018580
Epoch[150] Batch [0099]/[3760]	Speed: 118.122361 samples/sec	 accuracy=50.828125	 loss=2.150344	 lr=0.018570
Epoch[150] Batch [0149]/[3760]	Speed: 117.585542 samples/sec	 accuracy=51.020833	 loss=2.145385	 lr=0.018559
Epoch[150] Batch [0199]/[3760]	Speed: 118.054399 samples/sec	 accuracy=50.984375	 loss=2.154059	 lr=0.018549
Epoch[150] Batch [0249]/[3760]	Speed: 117.508405 samples/sec	 accuracy=50.437500	 loss=2.175425	 lr=0.018539
Epoch[150] Batch [0299]/[3760]	Speed: 117.480776 samples/sec	 accuracy=50.192708	 loss=2.187149	 lr=0.018529
Epoch[150] Batch [0349]/[3760]	Speed: 117.545992 samples/sec	 accuracy=49.866071	 loss=2.204014	 lr=0.018519
Epoch[150] Batch [0399]/[3760]	Speed: 116.362101 samples/sec	 accuracy=49.617188	 loss=2.216725	 lr=0.018509
Epoch[150] Batch [0449]/[3760]	Speed: 117.899071 samples/sec	 accuracy=49.697917	 loss=2.212105	 lr=0.018499
Epoch[150] Batch [0499]/[3760]	Speed: 117.640424 samples/sec	 accuracy=49.684375	 loss=2.213210	 lr=0.018489
Epoch[150] Batch [0549]/[3760]	Speed: 117.217037 samples/sec	 accuracy=49.573864	 loss=2.220355	 lr=0.018479
Epoch[150] Batch [0599]/[3760]	Speed: 117.127176 samples/sec	 accuracy=49.588542	 loss=2.217841	 lr=0.018469
Epoch[150] Batch [0649]/[3760]	Speed: 117.657447 samples/sec	 accuracy=49.586538	 loss=2.223239	 lr=0.018459
Epoch[150] Batch [0699]/[3760]	Speed: 117.868205 samples/sec	 accuracy=49.761161	 loss=2.215298	 lr=0.018449
Epoch[150] Batch [0749]/[3760]	Speed: 117.652310 samples/sec	 accuracy=49.754167	 loss=2.213648	 lr=0.018439
Epoch[150] Batch [0799]/[3760]	Speed: 118.463278 samples/sec	 accuracy=49.689453	 loss=2.216070	 lr=0.018429
Epoch[150] Batch [0849]/[3760]	Speed: 117.046546 samples/sec	 accuracy=49.669118	 loss=2.213951	 lr=0.018419
Epoch[150] Batch [0899]/[3760]	Speed: 117.297696 samples/sec	 accuracy=49.651042	 loss=2.213158	 lr=0.018409
Epoch[150] Batch [0949]/[3760]	Speed: 117.471443 samples/sec	 accuracy=49.613487	 loss=2.214367	 lr=0.018399
Epoch[150] Batch [0999]/[3760]	Speed: 117.592210 samples/sec	 accuracy=49.543750	 loss=2.218273	 lr=0.018389
Epoch[150] Batch [1049]/[3760]	Speed: 117.777223 samples/sec	 accuracy=49.438988	 loss=2.221291	 lr=0.018379
Epoch[150] Batch [1099]/[3760]	Speed: 115.235154 samples/sec	 accuracy=49.495739	 loss=2.218193	 lr=0.018369
Epoch[150] Batch [1149]/[3760]	Speed: 117.822189 samples/sec	 accuracy=49.392663	 loss=2.222700	 lr=0.018359
Epoch[150] Batch [1199]/[3760]	Speed: 117.787831 samples/sec	 accuracy=49.407552	 loss=2.222938	 lr=0.018349
Epoch[150] Batch [1249]/[3760]	Speed: 117.516165 samples/sec	 accuracy=49.390000	 loss=2.224123	 lr=0.018339
Epoch[150] Batch [1299]/[3760]	Speed: 117.207991 samples/sec	 accuracy=49.326923	 loss=2.227000	 lr=0.018329
Epoch[150] Batch [1349]/[3760]	Speed: 117.878159 samples/sec	 accuracy=49.353009	 loss=2.225623	 lr=0.018319
Epoch[150] Batch [1399]/[3760]	Speed: 117.365801 samples/sec	 accuracy=49.354911	 loss=2.227093	 lr=0.018309
Epoch[150] Batch [1449]/[3760]	Speed: 117.732829 samples/sec	 accuracy=49.360991	 loss=2.228642	 lr=0.018299
Epoch[150] Batch [1499]/[3760]	Speed: 117.535546 samples/sec	 accuracy=49.358333	 loss=2.228248	 lr=0.018289
Epoch[150] Batch [1549]/[3760]	Speed: 117.265961 samples/sec	 accuracy=49.344758	 loss=2.228635	 lr=0.018280
Epoch[150] Batch [1599]/[3760]	Speed: 117.378270 samples/sec	 accuracy=49.370117	 loss=2.226873	 lr=0.018270
Epoch[150] Batch [1649]/[3760]	Speed: 117.318864 samples/sec	 accuracy=49.339015	 loss=2.227762	 lr=0.018260
Epoch[150] Batch [1699]/[3760]	Speed: 117.980368 samples/sec	 accuracy=49.325368	 loss=2.226846	 lr=0.018250
Epoch[150] Batch [1749]/[3760]	Speed: 117.227427 samples/sec	 accuracy=49.293750	 loss=2.226903	 lr=0.018240
Epoch[150] Batch [1799]/[3760]	Speed: 117.336615 samples/sec	 accuracy=49.289931	 loss=2.227300	 lr=0.018230
Epoch[150] Batch [1849]/[3760]	Speed: 117.300585 samples/sec	 accuracy=49.293919	 loss=2.227730	 lr=0.018220
Epoch[150] Batch [1899]/[3760]	Speed: 117.803266 samples/sec	 accuracy=49.268092	 loss=2.227433	 lr=0.018210
Epoch[150] Batch [1949]/[3760]	Speed: 117.111715 samples/sec	 accuracy=49.247596	 loss=2.227932	 lr=0.018200
Epoch[150] Batch [1999]/[3760]	Speed: 117.892736 samples/sec	 accuracy=49.239844	 loss=2.228597	 lr=0.018190
Epoch[150] Batch [2049]/[3760]	Speed: 117.814877 samples/sec	 accuracy=49.236280	 loss=2.229094	 lr=0.018180
Epoch[150] Batch [2099]/[3760]	Speed: 117.636430 samples/sec	 accuracy=49.235119	 loss=2.229605	 lr=0.018170
Epoch[150] Batch [2149]/[3760]	Speed: 117.375039 samples/sec	 accuracy=49.235465	 loss=2.229479	 lr=0.018160
Epoch[150] Batch [2199]/[3760]	Speed: 117.578675 samples/sec	 accuracy=49.249290	 loss=2.228728	 lr=0.018150
Epoch[150] Batch [2249]/[3760]	Speed: 117.699770 samples/sec	 accuracy=49.257639	 loss=2.228877	 lr=0.018140
Epoch[150] Batch [2299]/[3760]	Speed: 116.611167 samples/sec	 accuracy=49.246603	 loss=2.230018	 lr=0.018130
Epoch[150] Batch [2349]/[3760]	Speed: 117.086354 samples/sec	 accuracy=49.265957	 loss=2.230402	 lr=0.018120
Epoch[150] Batch [2399]/[3760]	Speed: 116.112739 samples/sec	 accuracy=49.261719	 loss=2.231048	 lr=0.018110
Epoch[150] Batch [2449]/[3760]	Speed: 117.516534 samples/sec	 accuracy=49.258291	 loss=2.231265	 lr=0.018100
Epoch[150] Batch [2499]/[3760]	Speed: 118.202476 samples/sec	 accuracy=49.255625	 loss=2.231562	 lr=0.018090
Epoch[150] Batch [2549]/[3760]	Speed: 117.245826 samples/sec	 accuracy=49.251838	 loss=2.232037	 lr=0.018081
Epoch[150] Batch [2599]/[3760]	Speed: 117.274610 samples/sec	 accuracy=49.274639	 loss=2.231163	 lr=0.018071
Epoch[150] Batch [2649]/[3760]	Speed: 116.800010 samples/sec	 accuracy=49.232311	 loss=2.233316	 lr=0.018061
Epoch[150] Batch [2699]/[3760]	Speed: 117.435140 samples/sec	 accuracy=49.222801	 loss=2.233347	 lr=0.018051
Epoch[150] Batch [2749]/[3760]	Speed: 117.581149 samples/sec	 accuracy=49.238636	 loss=2.233133	 lr=0.018041
Epoch[150] Batch [2799]/[3760]	Speed: 117.305783 samples/sec	 accuracy=49.222656	 loss=2.233910	 lr=0.018031
Epoch[150] Batch [2849]/[3760]	Speed: 117.245214 samples/sec	 accuracy=49.232456	 loss=2.233105	 lr=0.018021
Epoch[150] Batch [2899]/[3760]	Speed: 117.562952 samples/sec	 accuracy=49.241918	 loss=2.232574	 lr=0.018011
Epoch[150] Batch [2949]/[3760]	Speed: 117.028636 samples/sec	 accuracy=49.240996	 loss=2.232274	 lr=0.018001
Epoch[150] Batch [2999]/[3760]	Speed: 116.692420 samples/sec	 accuracy=49.258333	 loss=2.232332	 lr=0.017991
Epoch[150] Batch [3049]/[3760]	Speed: 117.736421 samples/sec	 accuracy=49.256148	 loss=2.232698	 lr=0.017981
Epoch[150] Batch [3099]/[3760]	Speed: 117.978259 samples/sec	 accuracy=49.253024	 loss=2.233195	 lr=0.017971
Epoch[150] Batch [3149]/[3760]	Speed: 117.805489 samples/sec	 accuracy=49.251984	 loss=2.233153	 lr=0.017962
Epoch[150] Batch [3199]/[3760]	Speed: 117.952651 samples/sec	 accuracy=49.217285	 loss=2.235141	 lr=0.017952
Epoch[150] Batch [3249]/[3760]	Speed: 117.509663 samples/sec	 accuracy=49.201442	 loss=2.235479	 lr=0.017942
Epoch[150] Batch [3299]/[3760]	Speed: 116.398248 samples/sec	 accuracy=49.173769	 loss=2.237055	 lr=0.017932
Epoch[150] Batch [3349]/[3760]	Speed: 117.001016 samples/sec	 accuracy=49.170709	 loss=2.237161	 lr=0.017922
Epoch[150] Batch [3399]/[3760]	Speed: 117.561855 samples/sec	 accuracy=49.158548	 loss=2.237717	 lr=0.017912
Epoch[150] Batch [3449]/[3760]	Speed: 117.994036 samples/sec	 accuracy=49.153533	 loss=2.237603	 lr=0.017902
Epoch[150] Batch [3499]/[3760]	Speed: 117.626813 samples/sec	 accuracy=49.149107	 loss=2.238033	 lr=0.017892
Epoch[150] Batch [3549]/[3760]	Speed: 117.493456 samples/sec	 accuracy=49.150968	 loss=2.237627	 lr=0.017882
Epoch[150] Batch [3599]/[3760]	Speed: 117.328918 samples/sec	 accuracy=49.160156	 loss=2.236835	 lr=0.017873
Epoch[150] Batch [3649]/[3760]	Speed: 117.946253 samples/sec	 accuracy=49.173801	 loss=2.235939	 lr=0.017863
Epoch[150] Batch [3699]/[3760]	Speed: 117.534911 samples/sec	 accuracy=49.165118	 loss=2.236456	 lr=0.017853
Epoch[150] Batch [3749]/[3760]	Speed: 118.192674 samples/sec	 accuracy=49.163333	 loss=2.236301	 lr=0.017843
Batch [0049]/[0303]: acc-top1=49.906250 acc-top5=72.875000
Batch [0099]/[0303]: acc-top1=46.046875 acc-top5=71.828125
Batch [0149]/[0303]: acc-top1=47.833333 acc-top5=73.010417
Batch [0199]/[0303]: acc-top1=50.750000 acc-top5=74.875000
Batch [0249]/[0303]: acc-top1=50.075000 acc-top5=74.462500
Batch [0299]/[0303]: acc-top1=49.750000 acc-top5=74.609375
[Epoch 150] training: accuracy=49.163481	 loss=2.236280
[Epoch 150] speed: 116 samples/sec	time cost: 2175.776002
[Epoch 150] validation: acc-top1=49.649340 acc-top5=74.582302 loss=2.269983
Epoch[151] Batch [0049]/[3759]	Speed: 87.898441 samples/sec	 accuracy=49.968750	 loss=2.182908	 lr=0.017831
Epoch[151] Batch [0099]/[3759]	Speed: 117.652843 samples/sec	 accuracy=49.625000	 loss=2.206917	 lr=0.017821
Epoch[151] Batch [0149]/[3759]	Speed: 117.295477 samples/sec	 accuracy=49.531250	 loss=2.198448	 lr=0.017811
Epoch[151] Batch [0199]/[3759]	Speed: 116.742853 samples/sec	 accuracy=49.843750	 loss=2.191153	 lr=0.017801
Epoch[151] Batch [0249]/[3759]	Speed: 117.778837 samples/sec	 accuracy=49.843750	 loss=2.191205	 lr=0.017792
Epoch[151] Batch [0299]/[3759]	Speed: 116.987355 samples/sec	 accuracy=49.750000	 loss=2.188046	 lr=0.017782
Epoch[151] Batch [0349]/[3759]	Speed: 117.131813 samples/sec	 accuracy=50.040179	 loss=2.179626	 lr=0.017772
Epoch[151] Batch [0399]/[3759]	Speed: 118.434638 samples/sec	 accuracy=49.773438	 loss=2.193017	 lr=0.017762
Epoch[151] Batch [0449]/[3759]	Speed: 117.005755 samples/sec	 accuracy=49.777778	 loss=2.195584	 lr=0.017752
Epoch[151] Batch [0499]/[3759]	Speed: 116.833825 samples/sec	 accuracy=49.815625	 loss=2.191621	 lr=0.017742
Epoch[151] Batch [0549]/[3759]	Speed: 117.241361 samples/sec	 accuracy=49.738636	 loss=2.194885	 lr=0.017732
Epoch[151] Batch [0599]/[3759]	Speed: 117.161927 samples/sec	 accuracy=49.645833	 loss=2.197270	 lr=0.017723
Epoch[151] Batch [0649]/[3759]	Speed: 117.456549 samples/sec	 accuracy=49.634615	 loss=2.198742	 lr=0.017713
Epoch[151] Batch [0699]/[3759]	Speed: 117.692666 samples/sec	 accuracy=49.676339	 loss=2.199379	 lr=0.017703
Epoch[151] Batch [0749]/[3759]	Speed: 118.431619 samples/sec	 accuracy=49.758333	 loss=2.195051	 lr=0.017693
Epoch[151] Batch [0799]/[3759]	Speed: 117.682503 samples/sec	 accuracy=49.675781	 loss=2.198919	 lr=0.017683
Epoch[151] Batch [0849]/[3759]	Speed: 116.692988 samples/sec	 accuracy=49.715074	 loss=2.198153	 lr=0.017673
Epoch[151] Batch [0899]/[3759]	Speed: 116.969609 samples/sec	 accuracy=49.779514	 loss=2.196787	 lr=0.017664
Epoch[151] Batch [0949]/[3759]	Speed: 116.966528 samples/sec	 accuracy=49.751645	 loss=2.198100	 lr=0.017654
Epoch[151] Batch [0999]/[3759]	Speed: 117.293151 samples/sec	 accuracy=49.760938	 loss=2.198583	 lr=0.017644
Epoch[151] Batch [1049]/[3759]	Speed: 117.016358 samples/sec	 accuracy=49.796131	 loss=2.197536	 lr=0.017634
Epoch[151] Batch [1099]/[3759]	Speed: 117.169834 samples/sec	 accuracy=49.740057	 loss=2.199494	 lr=0.017624
Epoch[151] Batch [1149]/[3759]	Speed: 118.416977 samples/sec	 accuracy=49.698370	 loss=2.202088	 lr=0.017614
Epoch[151] Batch [1199]/[3759]	Speed: 114.535277 samples/sec	 accuracy=49.710938	 loss=2.201246	 lr=0.017605
Epoch[151] Batch [1249]/[3759]	Speed: 117.691052 samples/sec	 accuracy=49.662500	 loss=2.204435	 lr=0.017595
Epoch[151] Batch [1299]/[3759]	Speed: 117.909402 samples/sec	 accuracy=49.615385	 loss=2.205975	 lr=0.017585
Epoch[151] Batch [1349]/[3759]	Speed: 118.216773 samples/sec	 accuracy=49.635417	 loss=2.206811	 lr=0.017575
Epoch[151] Batch [1399]/[3759]	Speed: 117.378813 samples/sec	 accuracy=49.639509	 loss=2.206673	 lr=0.017565
Epoch[151] Batch [1449]/[3759]	Speed: 117.468560 samples/sec	 accuracy=49.639009	 loss=2.207297	 lr=0.017555
Epoch[151] Batch [1499]/[3759]	Speed: 117.811656 samples/sec	 accuracy=49.602083	 loss=2.209238	 lr=0.017546
Epoch[151] Batch [1549]/[3759]	Speed: 117.478512 samples/sec	 accuracy=49.658266	 loss=2.210402	 lr=0.017536
Epoch[151] Batch [1599]/[3759]	Speed: 117.511621 samples/sec	 accuracy=49.624023	 loss=2.211941	 lr=0.017526
Epoch[151] Batch [1649]/[3759]	Speed: 117.174104 samples/sec	 accuracy=49.635417	 loss=2.212371	 lr=0.017516
Epoch[151] Batch [1699]/[3759]	Speed: 117.652205 samples/sec	 accuracy=49.622243	 loss=2.212696	 lr=0.017506
Epoch[151] Batch [1749]/[3759]	Speed: 117.231129 samples/sec	 accuracy=49.615179	 loss=2.212945	 lr=0.017497
Epoch[151] Batch [1799]/[3759]	Speed: 116.775388 samples/sec	 accuracy=49.617188	 loss=2.212100	 lr=0.017487
Epoch[151] Batch [1849]/[3759]	Speed: 117.241848 samples/sec	 accuracy=49.605574	 loss=2.213632	 lr=0.017477
Epoch[151] Batch [1899]/[3759]	Speed: 117.131529 samples/sec	 accuracy=49.631579	 loss=2.212925	 lr=0.017467
Epoch[151] Batch [1949]/[3759]	Speed: 117.737817 samples/sec	 accuracy=49.616987	 loss=2.213173	 lr=0.017457
Epoch[151] Batch [1999]/[3759]	Speed: 116.701774 samples/sec	 accuracy=49.600781	 loss=2.214183	 lr=0.017448
Epoch[151] Batch [2049]/[3759]	Speed: 117.164277 samples/sec	 accuracy=49.575457	 loss=2.214315	 lr=0.017438
Epoch[151] Batch [2099]/[3759]	Speed: 117.161141 samples/sec	 accuracy=49.551339	 loss=2.215329	 lr=0.017428
Epoch[151] Batch [2149]/[3759]	Speed: 117.404010 samples/sec	 accuracy=49.571221	 loss=2.215096	 lr=0.017418
Epoch[151] Batch [2199]/[3759]	Speed: 117.789821 samples/sec	 accuracy=49.553977	 loss=2.215386	 lr=0.017408
Epoch[151] Batch [2249]/[3759]	Speed: 117.011606 samples/sec	 accuracy=49.529861	 loss=2.215913	 lr=0.017399
Epoch[151] Batch [2299]/[3759]	Speed: 117.356172 samples/sec	 accuracy=49.502717	 loss=2.215891	 lr=0.017389
Epoch[151] Batch [2349]/[3759]	Speed: 117.930679 samples/sec	 accuracy=49.507979	 loss=2.214862	 lr=0.017379
Epoch[151] Batch [2399]/[3759]	Speed: 117.617956 samples/sec	 accuracy=49.463542	 loss=2.216851	 lr=0.017369
Epoch[151] Batch [2449]/[3759]	Speed: 117.283071 samples/sec	 accuracy=49.464923	 loss=2.216871	 lr=0.017360
Epoch[151] Batch [2499]/[3759]	Speed: 117.872894 samples/sec	 accuracy=49.449375	 loss=2.217342	 lr=0.017350
Epoch[151] Batch [2549]/[3759]	Speed: 115.991385 samples/sec	 accuracy=49.465686	 loss=2.216878	 lr=0.017340
Epoch[151] Batch [2599]/[3759]	Speed: 117.938270 samples/sec	 accuracy=49.440505	 loss=2.217603	 lr=0.017330
Epoch[151] Batch [2649]/[3759]	Speed: 118.200559 samples/sec	 accuracy=49.416863	 loss=2.217810	 lr=0.017321
Epoch[151] Batch [2699]/[3759]	Speed: 117.572472 samples/sec	 accuracy=49.394676	 loss=2.218714	 lr=0.017311
Epoch[151] Batch [2749]/[3759]	Speed: 117.310994 samples/sec	 accuracy=49.384659	 loss=2.219692	 lr=0.017301
Epoch[151] Batch [2799]/[3759]	Speed: 117.005977 samples/sec	 accuracy=49.420759	 loss=2.218658	 lr=0.017291
Epoch[151] Batch [2849]/[3759]	Speed: 117.601126 samples/sec	 accuracy=49.402412	 loss=2.219284	 lr=0.017281
Epoch[151] Batch [2899]/[3759]	Speed: 117.459789 samples/sec	 accuracy=49.407328	 loss=2.219053	 lr=0.017272
Epoch[151] Batch [2949]/[3759]	Speed: 117.269007 samples/sec	 accuracy=49.390360	 loss=2.219531	 lr=0.017262
Epoch[151] Batch [2999]/[3759]	Speed: 118.055819 samples/sec	 accuracy=49.413021	 loss=2.219379	 lr=0.017252
Epoch[151] Batch [3049]/[3759]	Speed: 118.035793 samples/sec	 accuracy=49.411885	 loss=2.218952	 lr=0.017243
Epoch[151] Batch [3099]/[3759]	Speed: 117.337148 samples/sec	 accuracy=49.398690	 loss=2.218931	 lr=0.017233
Epoch[151] Batch [3149]/[3759]	Speed: 117.013217 samples/sec	 accuracy=49.384425	 loss=2.219960	 lr=0.017223
Epoch[151] Batch [3199]/[3759]	Speed: 118.358543 samples/sec	 accuracy=49.384766	 loss=2.219642	 lr=0.017213
Epoch[151] Batch [3249]/[3759]	Speed: 117.371947 samples/sec	 accuracy=49.388942	 loss=2.219823	 lr=0.017204
Epoch[151] Batch [3299]/[3759]	Speed: 118.616359 samples/sec	 accuracy=49.396307	 loss=2.220515	 lr=0.017194
Epoch[151] Batch [3349]/[3759]	Speed: 117.724965 samples/sec	 accuracy=49.429571	 loss=2.218891	 lr=0.017184
Epoch[151] Batch [3399]/[3759]	Speed: 116.700929 samples/sec	 accuracy=49.414062	 loss=2.219171	 lr=0.017174
Epoch[151] Batch [3449]/[3759]	Speed: 118.413036 samples/sec	 accuracy=49.400362	 loss=2.219534	 lr=0.017165
Epoch[151] Batch [3499]/[3759]	Speed: 117.257375 samples/sec	 accuracy=49.391071	 loss=2.219700	 lr=0.017155
Epoch[151] Batch [3549]/[3759]	Speed: 117.728155 samples/sec	 accuracy=49.405370	 loss=2.218424	 lr=0.017145
Epoch[151] Batch [3599]/[3759]	Speed: 117.249110 samples/sec	 accuracy=49.400608	 loss=2.219219	 lr=0.017135
Epoch[151] Batch [3649]/[3759]	Speed: 117.147669 samples/sec	 accuracy=49.424229	 loss=2.218933	 lr=0.017126
Epoch[151] Batch [3699]/[3759]	Speed: 118.229323 samples/sec	 accuracy=49.402027	 loss=2.219480	 lr=0.017116
Epoch[151] Batch [3749]/[3759]	Speed: 120.930063 samples/sec	 accuracy=49.407083	 loss=2.219506	 lr=0.017106
Batch [0049]/[0303]: acc-top1=50.343750 acc-top5=73.312500
Batch [0099]/[0303]: acc-top1=47.015625 acc-top5=71.859375
Batch [0149]/[0303]: acc-top1=48.927083 acc-top5=72.989583
Batch [0199]/[0303]: acc-top1=51.679688 acc-top5=74.671875
Batch [0249]/[0303]: acc-top1=50.762500 acc-top5=74.575000
Batch [0299]/[0303]: acc-top1=49.932292 acc-top5=74.562500
[Epoch 151] training: accuracy=49.406425	 loss=2.219754
[Epoch 151] speed: 116 samples/sec	time cost: 2174.887120
[Epoch 151] validation: acc-top1=49.943276 acc-top5=74.587459 loss=2.294312
Epoch[152] Batch [0049]/[3760]	Speed: 87.330003 samples/sec	 accuracy=50.406250	 loss=2.149132	 lr=0.017095
Epoch[152] Batch [0099]/[3760]	Speed: 114.632603 samples/sec	 accuracy=50.015625	 loss=2.160105	 lr=0.017085
Epoch[152] Batch [0149]/[3760]	Speed: 116.956328 samples/sec	 accuracy=50.197917	 loss=2.161250	 lr=0.017075
Epoch[152] Batch [0199]/[3760]	Speed: 117.077338 samples/sec	 accuracy=50.054687	 loss=2.168359	 lr=0.017066
Epoch[152] Batch [0249]/[3760]	Speed: 117.513626 samples/sec	 accuracy=50.393750	 loss=2.165231	 lr=0.017056
Epoch[152] Batch [0299]/[3760]	Speed: 117.066017 samples/sec	 accuracy=50.526042	 loss=2.163381	 lr=0.017046
Epoch[152] Batch [0349]/[3760]	Speed: 118.089504 samples/sec	 accuracy=50.473214	 loss=2.168528	 lr=0.017037
Epoch[152] Batch [0399]/[3760]	Speed: 116.898139 samples/sec	 accuracy=50.433594	 loss=2.172551	 lr=0.017027
Epoch[152] Batch [0449]/[3760]	Speed: 117.574083 samples/sec	 accuracy=50.312500	 loss=2.176788	 lr=0.017017
Epoch[152] Batch [0499]/[3760]	Speed: 117.226651 samples/sec	 accuracy=50.237500	 loss=2.179024	 lr=0.017008
Epoch[152] Batch [0549]/[3760]	Speed: 117.190967 samples/sec	 accuracy=50.238636	 loss=2.177660	 lr=0.016998
Epoch[152] Batch [0599]/[3760]	Speed: 118.126722 samples/sec	 accuracy=50.164062	 loss=2.178973	 lr=0.016988
Epoch[152] Batch [0649]/[3760]	Speed: 117.559698 samples/sec	 accuracy=50.170673	 loss=2.179986	 lr=0.016978
Epoch[152] Batch [0699]/[3760]	Speed: 117.886382 samples/sec	 accuracy=50.171875	 loss=2.178083	 lr=0.016969
Epoch[152] Batch [0749]/[3760]	Speed: 117.222839 samples/sec	 accuracy=50.164583	 loss=2.179579	 lr=0.016959
Epoch[152] Batch [0799]/[3760]	Speed: 117.926853 samples/sec	 accuracy=50.179688	 loss=2.180606	 lr=0.016949
Epoch[152] Batch [0849]/[3760]	Speed: 117.516506 samples/sec	 accuracy=50.053309	 loss=2.184265	 lr=0.016940
Epoch[152] Batch [0899]/[3760]	Speed: 117.976755 samples/sec	 accuracy=50.017361	 loss=2.186392	 lr=0.016930
Epoch[152] Batch [0949]/[3760]	Speed: 117.366757 samples/sec	 accuracy=49.916118	 loss=2.189598	 lr=0.016920
Epoch[152] Batch [0999]/[3760]	Speed: 117.759035 samples/sec	 accuracy=49.915625	 loss=2.189570	 lr=0.016911
Epoch[152] Batch [1049]/[3760]	Speed: 116.924525 samples/sec	 accuracy=49.933036	 loss=2.190842	 lr=0.016901
Epoch[152] Batch [1099]/[3760]	Speed: 117.735895 samples/sec	 accuracy=49.923295	 loss=2.191190	 lr=0.016891
Epoch[152] Batch [1149]/[3760]	Speed: 117.931780 samples/sec	 accuracy=49.925272	 loss=2.189304	 lr=0.016882
Epoch[152] Batch [1199]/[3760]	Speed: 116.624258 samples/sec	 accuracy=49.941406	 loss=2.188218	 lr=0.016872
Epoch[152] Batch [1249]/[3760]	Speed: 117.296704 samples/sec	 accuracy=49.936250	 loss=2.188644	 lr=0.016862
Epoch[152] Batch [1299]/[3760]	Speed: 117.100028 samples/sec	 accuracy=49.950721	 loss=2.188969	 lr=0.016853
Epoch[152] Batch [1349]/[3760]	Speed: 118.039208 samples/sec	 accuracy=49.900463	 loss=2.191524	 lr=0.016843
Epoch[152] Batch [1399]/[3760]	Speed: 115.512678 samples/sec	 accuracy=49.832589	 loss=2.193881	 lr=0.016833
Epoch[152] Batch [1449]/[3760]	Speed: 117.920955 samples/sec	 accuracy=49.815733	 loss=2.195293	 lr=0.016824
Epoch[152] Batch [1499]/[3760]	Speed: 118.110239 samples/sec	 accuracy=49.803125	 loss=2.195765	 lr=0.016814
Epoch[152] Batch [1549]/[3760]	Speed: 117.808762 samples/sec	 accuracy=49.762097	 loss=2.197759	 lr=0.016805
Epoch[152] Batch [1599]/[3760]	Speed: 116.612678 samples/sec	 accuracy=49.776367	 loss=2.197809	 lr=0.016795
Epoch[152] Batch [1649]/[3760]	Speed: 117.543199 samples/sec	 accuracy=49.759470	 loss=2.197405	 lr=0.016785
Epoch[152] Batch [1699]/[3760]	Speed: 116.724550 samples/sec	 accuracy=49.762868	 loss=2.197822	 lr=0.016776
Epoch[152] Batch [1749]/[3760]	Speed: 117.079119 samples/sec	 accuracy=49.810714	 loss=2.197765	 lr=0.016766
Epoch[152] Batch [1799]/[3760]	Speed: 117.854125 samples/sec	 accuracy=49.784722	 loss=2.199396	 lr=0.016756
Epoch[152] Batch [1849]/[3760]	Speed: 117.764833 samples/sec	 accuracy=49.818412	 loss=2.199717	 lr=0.016747
Epoch[152] Batch [1899]/[3760]	Speed: 117.542804 samples/sec	 accuracy=49.792763	 loss=2.200342	 lr=0.016737
Epoch[152] Batch [1949]/[3760]	Speed: 117.790922 samples/sec	 accuracy=49.810897	 loss=2.200719	 lr=0.016727
Epoch[152] Batch [1999]/[3760]	Speed: 117.115720 samples/sec	 accuracy=49.805469	 loss=2.199701	 lr=0.016718
Epoch[152] Batch [2049]/[3760]	Speed: 118.101916 samples/sec	 accuracy=49.889482	 loss=2.196540	 lr=0.016708
Epoch[152] Batch [2099]/[3760]	Speed: 117.523159 samples/sec	 accuracy=49.853423	 loss=2.197480	 lr=0.016699
Epoch[152] Batch [2149]/[3760]	Speed: 117.607569 samples/sec	 accuracy=49.835029	 loss=2.197415	 lr=0.016689
Epoch[152] Batch [2199]/[3760]	Speed: 117.091970 samples/sec	 accuracy=49.830966	 loss=2.197784	 lr=0.016679
Epoch[152] Batch [2249]/[3760]	Speed: 117.410732 samples/sec	 accuracy=49.823611	 loss=2.199039	 lr=0.016670
Epoch[152] Batch [2299]/[3760]	Speed: 117.531149 samples/sec	 accuracy=49.853940	 loss=2.198306	 lr=0.016660
Epoch[152] Batch [2349]/[3760]	Speed: 117.137885 samples/sec	 accuracy=49.872340	 loss=2.197069	 lr=0.016650
Epoch[152] Batch [2399]/[3760]	Speed: 118.100634 samples/sec	 accuracy=49.873047	 loss=2.196496	 lr=0.016641
Epoch[152] Batch [2449]/[3760]	Speed: 117.728823 samples/sec	 accuracy=49.890306	 loss=2.197005	 lr=0.016631
Epoch[152] Batch [2499]/[3760]	Speed: 117.642582 samples/sec	 accuracy=49.900625	 loss=2.196156	 lr=0.016622
Epoch[152] Batch [2549]/[3760]	Speed: 118.248148 samples/sec	 accuracy=49.862132	 loss=2.197581	 lr=0.016612
Epoch[152] Batch [2599]/[3760]	Speed: 118.064833 samples/sec	 accuracy=49.846755	 loss=2.197972	 lr=0.016602
Epoch[152] Batch [2649]/[3760]	Speed: 117.684429 samples/sec	 accuracy=49.872642	 loss=2.197669	 lr=0.016593
Epoch[152] Batch [2699]/[3760]	Speed: 114.985989 samples/sec	 accuracy=49.850694	 loss=2.197804	 lr=0.016583
Epoch[152] Batch [2749]/[3760]	Speed: 118.601606 samples/sec	 accuracy=49.860227	 loss=2.197762	 lr=0.016574
Epoch[152] Batch [2799]/[3760]	Speed: 117.337508 samples/sec	 accuracy=49.847656	 loss=2.198542	 lr=0.016564
Epoch[152] Batch [2849]/[3760]	Speed: 117.457678 samples/sec	 accuracy=49.844298	 loss=2.199001	 lr=0.016554
Epoch[152] Batch [2899]/[3760]	Speed: 117.256013 samples/sec	 accuracy=49.834052	 loss=2.198867	 lr=0.016545
Epoch[152] Batch [2949]/[3760]	Speed: 117.598366 samples/sec	 accuracy=49.861229	 loss=2.198054	 lr=0.016535
Epoch[152] Batch [2999]/[3760]	Speed: 117.521851 samples/sec	 accuracy=49.872396	 loss=2.197456	 lr=0.016526
Epoch[152] Batch [3049]/[3760]	Speed: 117.424281 samples/sec	 accuracy=49.883709	 loss=2.197220	 lr=0.016516
Epoch[152] Batch [3099]/[3760]	Speed: 116.842192 samples/sec	 accuracy=49.877520	 loss=2.197445	 lr=0.016507
Epoch[152] Batch [3149]/[3760]	Speed: 117.947759 samples/sec	 accuracy=49.883433	 loss=2.197040	 lr=0.016497
Epoch[152] Batch [3199]/[3760]	Speed: 117.534566 samples/sec	 accuracy=49.863281	 loss=2.197444	 lr=0.016487
Epoch[152] Batch [3249]/[3760]	Speed: 117.365614 samples/sec	 accuracy=49.829808	 loss=2.198257	 lr=0.016478
Epoch[152] Batch [3299]/[3760]	Speed: 117.763836 samples/sec	 accuracy=49.851326	 loss=2.197319	 lr=0.016468
Epoch[152] Batch [3349]/[3760]	Speed: 117.176455 samples/sec	 accuracy=49.845149	 loss=2.196935	 lr=0.016459
Epoch[152] Batch [3399]/[3760]	Speed: 117.341036 samples/sec	 accuracy=49.855699	 loss=2.197055	 lr=0.016449
Epoch[152] Batch [3449]/[3760]	Speed: 116.230614 samples/sec	 accuracy=49.839221	 loss=2.196904	 lr=0.016440
Epoch[152] Batch [3499]/[3760]	Speed: 117.429440 samples/sec	 accuracy=49.808036	 loss=2.198273	 lr=0.016430
Epoch[152] Batch [3549]/[3760]	Speed: 117.509120 samples/sec	 accuracy=49.801056	 loss=2.198890	 lr=0.016420
Epoch[152] Batch [3599]/[3760]	Speed: 117.171703 samples/sec	 accuracy=49.816840	 loss=2.198655	 lr=0.016411
Epoch[152] Batch [3649]/[3760]	Speed: 117.673891 samples/sec	 accuracy=49.805223	 loss=2.198479	 lr=0.016401
Epoch[152] Batch [3699]/[3760]	Speed: 118.153943 samples/sec	 accuracy=49.784206	 loss=2.199152	 lr=0.016392
Epoch[152] Batch [3749]/[3760]	Speed: 120.631368 samples/sec	 accuracy=49.798333	 loss=2.198830	 lr=0.016382
Batch [0049]/[0303]: acc-top1=50.468750 acc-top5=75.218750
Batch [0099]/[0303]: acc-top1=48.203125 acc-top5=73.312500
Batch [0149]/[0303]: acc-top1=49.541667 acc-top5=73.947917
Batch [0199]/[0303]: acc-top1=51.843750 acc-top5=75.093750
Batch [0249]/[0303]: acc-top1=51.050000 acc-top5=74.750000
Batch [0299]/[0303]: acc-top1=50.307292 acc-top5=74.489583
[Epoch 152] training: accuracy=49.794714	 loss=2.198830
[Epoch 152] speed: 116 samples/sec	time cost: 2175.525828
[Epoch 152] validation: acc-top1=50.324876 acc-top5=74.520421 loss=2.309442
Epoch[153] Batch [0049]/[3760]	Speed: 87.281543 samples/sec	 accuracy=49.843750	 loss=2.181835	 lr=0.016371
Epoch[153] Batch [0099]/[3760]	Speed: 117.499632 samples/sec	 accuracy=49.812500	 loss=2.161890	 lr=0.016361
Epoch[153] Batch [0149]/[3760]	Speed: 116.486691 samples/sec	 accuracy=49.687500	 loss=2.172028	 lr=0.016352
Epoch[153] Batch [0199]/[3760]	Speed: 118.307644 samples/sec	 accuracy=49.757812	 loss=2.172090	 lr=0.016342
Epoch[153] Batch [0249]/[3760]	Speed: 117.722875 samples/sec	 accuracy=49.606250	 loss=2.187613	 lr=0.016333
Epoch[153] Batch [0299]/[3760]	Speed: 117.810082 samples/sec	 accuracy=49.843750	 loss=2.185479	 lr=0.016323
Epoch[153] Batch [0349]/[3760]	Speed: 117.588513 samples/sec	 accuracy=49.946429	 loss=2.180580	 lr=0.016314
Epoch[153] Batch [0399]/[3760]	Speed: 115.395060 samples/sec	 accuracy=49.992188	 loss=2.180052	 lr=0.016304
Epoch[153] Batch [0449]/[3760]	Speed: 117.553298 samples/sec	 accuracy=49.947917	 loss=2.184314	 lr=0.016295
Epoch[153] Batch [0499]/[3760]	Speed: 117.571283 samples/sec	 accuracy=49.756250	 loss=2.187857	 lr=0.016285
Epoch[153] Batch [0549]/[3760]	Speed: 117.320086 samples/sec	 accuracy=49.778409	 loss=2.189338	 lr=0.016275
Epoch[153] Batch [0599]/[3760]	Speed: 117.562934 samples/sec	 accuracy=49.932292	 loss=2.186094	 lr=0.016266
Epoch[153] Batch [0649]/[3760]	Speed: 117.259819 samples/sec	 accuracy=49.918269	 loss=2.186723	 lr=0.016256
Epoch[153] Batch [0699]/[3760]	Speed: 116.967500 samples/sec	 accuracy=50.004464	 loss=2.181827	 lr=0.016247
Epoch[153] Batch [0749]/[3760]	Speed: 117.735859 samples/sec	 accuracy=50.025000	 loss=2.179682	 lr=0.016237
Epoch[153] Batch [0799]/[3760]	Speed: 117.386577 samples/sec	 accuracy=50.083984	 loss=2.176576	 lr=0.016228
Epoch[153] Batch [0849]/[3760]	Speed: 116.806429 samples/sec	 accuracy=50.064338	 loss=2.177243	 lr=0.016218
Epoch[153] Batch [0899]/[3760]	Speed: 117.748064 samples/sec	 accuracy=50.093750	 loss=2.177147	 lr=0.016209
Epoch[153] Batch [0949]/[3760]	Speed: 117.684385 samples/sec	 accuracy=49.986842	 loss=2.179124	 lr=0.016199
Epoch[153] Batch [0999]/[3760]	Speed: 117.070297 samples/sec	 accuracy=49.990625	 loss=2.177207	 lr=0.016190
Epoch[153] Batch [1049]/[3760]	Speed: 117.987121 samples/sec	 accuracy=50.066964	 loss=2.176064	 lr=0.016180
Epoch[153] Batch [1099]/[3760]	Speed: 117.085361 samples/sec	 accuracy=50.065341	 loss=2.176606	 lr=0.016171
Epoch[153] Batch [1149]/[3760]	Speed: 117.667395 samples/sec	 accuracy=50.040761	 loss=2.177528	 lr=0.016161
Epoch[153] Batch [1199]/[3760]	Speed: 117.633069 samples/sec	 accuracy=50.070312	 loss=2.176453	 lr=0.016152
Epoch[153] Batch [1249]/[3760]	Speed: 117.367348 samples/sec	 accuracy=50.042500	 loss=2.177630	 lr=0.016142
Epoch[153] Batch [1299]/[3760]	Speed: 117.079572 samples/sec	 accuracy=50.016827	 loss=2.177843	 lr=0.016133
Epoch[153] Batch [1349]/[3760]	Speed: 117.743418 samples/sec	 accuracy=50.053241	 loss=2.177619	 lr=0.016123
Epoch[153] Batch [1399]/[3760]	Speed: 117.908314 samples/sec	 accuracy=50.045759	 loss=2.177736	 lr=0.016114
Epoch[153] Batch [1449]/[3760]	Speed: 117.484561 samples/sec	 accuracy=50.087284	 loss=2.176219	 lr=0.016104
Epoch[153] Batch [1499]/[3760]	Speed: 117.163432 samples/sec	 accuracy=50.152083	 loss=2.175399	 lr=0.016095
Epoch[153] Batch [1549]/[3760]	Speed: 117.695225 samples/sec	 accuracy=50.192540	 loss=2.173889	 lr=0.016086
Epoch[153] Batch [1599]/[3760]	Speed: 117.652027 samples/sec	 accuracy=50.195312	 loss=2.172984	 lr=0.016076
Epoch[153] Batch [1649]/[3760]	Speed: 117.765415 samples/sec	 accuracy=50.168561	 loss=2.174522	 lr=0.016067
Epoch[153] Batch [1699]/[3760]	Speed: 117.383256 samples/sec	 accuracy=50.162684	 loss=2.174827	 lr=0.016057
Epoch[153] Batch [1749]/[3760]	Speed: 115.520184 samples/sec	 accuracy=50.146429	 loss=2.174442	 lr=0.016048
Epoch[153] Batch [1799]/[3760]	Speed: 117.401352 samples/sec	 accuracy=50.146701	 loss=2.175123	 lr=0.016038
Epoch[153] Batch [1849]/[3760]	Speed: 117.234203 samples/sec	 accuracy=50.150338	 loss=2.175057	 lr=0.016029
Epoch[153] Batch [1899]/[3760]	Speed: 117.573757 samples/sec	 accuracy=50.144737	 loss=2.174634	 lr=0.016019
Epoch[153] Batch [1949]/[3760]	Speed: 117.578465 samples/sec	 accuracy=50.138622	 loss=2.175422	 lr=0.016010
Epoch[153] Batch [1999]/[3760]	Speed: 117.310183 samples/sec	 accuracy=50.152344	 loss=2.175250	 lr=0.016000
Epoch[153] Batch [2049]/[3760]	Speed: 117.797327 samples/sec	 accuracy=50.150915	 loss=2.176215	 lr=0.015991
Epoch[153] Batch [2099]/[3760]	Speed: 116.325712 samples/sec	 accuracy=50.132440	 loss=2.176617	 lr=0.015981
Epoch[153] Batch [2149]/[3760]	Speed: 117.483482 samples/sec	 accuracy=50.105378	 loss=2.177631	 lr=0.015972
Epoch[153] Batch [2199]/[3760]	Speed: 117.915046 samples/sec	 accuracy=50.142045	 loss=2.176308	 lr=0.015962
Epoch[153] Batch [2249]/[3760]	Speed: 117.503631 samples/sec	 accuracy=50.173611	 loss=2.175032	 lr=0.015953
Epoch[153] Batch [2299]/[3760]	Speed: 117.296782 samples/sec	 accuracy=50.161005	 loss=2.175261	 lr=0.015944
Epoch[153] Batch [2349]/[3760]	Speed: 117.720413 samples/sec	 accuracy=50.178191	 loss=2.174862	 lr=0.015934
Epoch[153] Batch [2399]/[3760]	Speed: 117.031328 samples/sec	 accuracy=50.154948	 loss=2.175861	 lr=0.015925
Epoch[153] Batch [2449]/[3760]	Speed: 117.754986 samples/sec	 accuracy=50.155612	 loss=2.175628	 lr=0.015915
Epoch[153] Batch [2499]/[3760]	Speed: 117.241429 samples/sec	 accuracy=50.146250	 loss=2.176269	 lr=0.015906
Epoch[153] Batch [2549]/[3760]	Speed: 117.595483 samples/sec	 accuracy=50.137255	 loss=2.176258	 lr=0.015896
Epoch[153] Batch [2599]/[3760]	Speed: 117.838611 samples/sec	 accuracy=50.140625	 loss=2.176265	 lr=0.015887
Epoch[153] Batch [2649]/[3760]	Speed: 117.329914 samples/sec	 accuracy=50.129127	 loss=2.176686	 lr=0.015878
Epoch[153] Batch [2699]/[3760]	Speed: 117.417603 samples/sec	 accuracy=50.152199	 loss=2.175917	 lr=0.015868
Epoch[153] Batch [2749]/[3760]	Speed: 117.902128 samples/sec	 accuracy=50.181250	 loss=2.174535	 lr=0.015859
Epoch[153] Batch [2799]/[3760]	Speed: 117.571148 samples/sec	 accuracy=50.169643	 loss=2.175092	 lr=0.015849
Epoch[153] Batch [2849]/[3760]	Speed: 117.594155 samples/sec	 accuracy=50.174342	 loss=2.174775	 lr=0.015840
Epoch[153] Batch [2899]/[3760]	Speed: 116.715297 samples/sec	 accuracy=50.166487	 loss=2.174616	 lr=0.015830
Epoch[153] Batch [2949]/[3760]	Speed: 117.472608 samples/sec	 accuracy=50.184852	 loss=2.174407	 lr=0.015821
Epoch[153] Batch [2999]/[3760]	Speed: 117.693082 samples/sec	 accuracy=50.188021	 loss=2.174787	 lr=0.015812
Epoch[153] Batch [3049]/[3760]	Speed: 118.287306 samples/sec	 accuracy=50.186475	 loss=2.174736	 lr=0.015802
Epoch[153] Batch [3099]/[3760]	Speed: 115.292290 samples/sec	 accuracy=50.201613	 loss=2.175125	 lr=0.015793
Epoch[153] Batch [3149]/[3760]	Speed: 117.565180 samples/sec	 accuracy=50.200893	 loss=2.176007	 lr=0.015783
Epoch[153] Batch [3199]/[3760]	Speed: 117.837058 samples/sec	 accuracy=50.208984	 loss=2.175743	 lr=0.015774
Epoch[153] Batch [3249]/[3760]	Speed: 117.499540 samples/sec	 accuracy=50.202404	 loss=2.176049	 lr=0.015765
Epoch[153] Batch [3299]/[3760]	Speed: 117.362304 samples/sec	 accuracy=50.205966	 loss=2.175744	 lr=0.015755
Epoch[153] Batch [3349]/[3760]	Speed: 117.665280 samples/sec	 accuracy=50.210821	 loss=2.175526	 lr=0.015746
Epoch[153] Batch [3399]/[3760]	Speed: 117.475266 samples/sec	 accuracy=50.204504	 loss=2.175691	 lr=0.015736
Epoch[153] Batch [3449]/[3760]	Speed: 116.987405 samples/sec	 accuracy=50.178442	 loss=2.176814	 lr=0.015727
Epoch[153] Batch [3499]/[3760]	Speed: 117.799222 samples/sec	 accuracy=50.171875	 loss=2.177362	 lr=0.015718
Epoch[153] Batch [3549]/[3760]	Speed: 117.792800 samples/sec	 accuracy=50.167694	 loss=2.178048	 lr=0.015708
Epoch[153] Batch [3599]/[3760]	Speed: 117.103512 samples/sec	 accuracy=50.171875	 loss=2.177726	 lr=0.015699
Epoch[153] Batch [3649]/[3760]	Speed: 117.222183 samples/sec	 accuracy=50.172089	 loss=2.178204	 lr=0.015689
Epoch[153] Batch [3699]/[3760]	Speed: 117.299961 samples/sec	 accuracy=50.157517	 loss=2.178235	 lr=0.015680
Epoch[153] Batch [3749]/[3760]	Speed: 119.978644 samples/sec	 accuracy=50.155833	 loss=2.178190	 lr=0.015671
Batch [0049]/[0303]: acc-top1=48.781250 acc-top5=72.531250
Batch [0099]/[0303]: acc-top1=46.734375 acc-top5=71.718750
Batch [0149]/[0303]: acc-top1=48.510417 acc-top5=72.968750
Batch [0199]/[0303]: acc-top1=51.585938 acc-top5=75.109375
Batch [0249]/[0303]: acc-top1=50.943750 acc-top5=74.962500
Batch [0299]/[0303]: acc-top1=50.390625 acc-top5=74.875000
[Epoch 153] training: accuracy=50.155834	 loss=2.178260
[Epoch 153] speed: 116 samples/sec	time cost: 2175.294101
[Epoch 153] validation: acc-top1=50.262995 acc-top5=74.804043 loss=2.244161
Epoch[154] Batch [0049]/[3759]	Speed: 86.231565 samples/sec	 accuracy=50.906250	 loss=2.110173	 lr=0.015659
Epoch[154] Batch [0099]/[3759]	Speed: 116.854364 samples/sec	 accuracy=50.406250	 loss=2.130114	 lr=0.015650
Epoch[154] Batch [0149]/[3759]	Speed: 117.621817 samples/sec	 accuracy=50.635417	 loss=2.129311	 lr=0.015641
Epoch[154] Batch [0199]/[3759]	Speed: 117.421409 samples/sec	 accuracy=50.195312	 loss=2.131723	 lr=0.015631
Epoch[154] Batch [0249]/[3759]	Speed: 117.078398 samples/sec	 accuracy=50.250000	 loss=2.132989	 lr=0.015622
Epoch[154] Batch [0299]/[3759]	Speed: 117.069340 samples/sec	 accuracy=50.651042	 loss=2.120397	 lr=0.015613
Epoch[154] Batch [0349]/[3759]	Speed: 117.372839 samples/sec	 accuracy=50.678571	 loss=2.127342	 lr=0.015603
Epoch[154] Batch [0399]/[3759]	Speed: 116.804980 samples/sec	 accuracy=50.886719	 loss=2.126098	 lr=0.015594
Epoch[154] Batch [0449]/[3759]	Speed: 117.474932 samples/sec	 accuracy=50.746528	 loss=2.132078	 lr=0.015585
Epoch[154] Batch [0499]/[3759]	Speed: 117.477399 samples/sec	 accuracy=50.759375	 loss=2.130434	 lr=0.015575
Epoch[154] Batch [0549]/[3759]	Speed: 117.156314 samples/sec	 accuracy=50.752841	 loss=2.132040	 lr=0.015566
Epoch[154] Batch [0599]/[3759]	Speed: 116.988559 samples/sec	 accuracy=50.747396	 loss=2.132530	 lr=0.015556
Epoch[154] Batch [0649]/[3759]	Speed: 116.985599 samples/sec	 accuracy=50.930288	 loss=2.124205	 lr=0.015547
Epoch[154] Batch [0699]/[3759]	Speed: 115.924481 samples/sec	 accuracy=50.823661	 loss=2.127706	 lr=0.015538
Epoch[154] Batch [0749]/[3759]	Speed: 117.999381 samples/sec	 accuracy=50.752083	 loss=2.132662	 lr=0.015528
Epoch[154] Batch [0799]/[3759]	Speed: 117.125977 samples/sec	 accuracy=50.775391	 loss=2.134129	 lr=0.015519
Epoch[154] Batch [0849]/[3759]	Speed: 117.227751 samples/sec	 accuracy=50.733456	 loss=2.136642	 lr=0.015510
Epoch[154] Batch [0899]/[3759]	Speed: 117.214554 samples/sec	 accuracy=50.756944	 loss=2.135823	 lr=0.015500
Epoch[154] Batch [0949]/[3759]	Speed: 117.865577 samples/sec	 accuracy=50.753289	 loss=2.137697	 lr=0.015491
Epoch[154] Batch [0999]/[3759]	Speed: 117.652403 samples/sec	 accuracy=50.720312	 loss=2.140713	 lr=0.015482
Epoch[154] Batch [1049]/[3759]	Speed: 117.731685 samples/sec	 accuracy=50.705357	 loss=2.141736	 lr=0.015472
Epoch[154] Batch [1099]/[3759]	Speed: 116.969783 samples/sec	 accuracy=50.671875	 loss=2.143214	 lr=0.015463
Epoch[154] Batch [1149]/[3759]	Speed: 117.639935 samples/sec	 accuracy=50.633152	 loss=2.144951	 lr=0.015454
Epoch[154] Batch [1199]/[3759]	Speed: 118.194495 samples/sec	 accuracy=50.606771	 loss=2.146396	 lr=0.015444
Epoch[154] Batch [1249]/[3759]	Speed: 117.273974 samples/sec	 accuracy=50.580000	 loss=2.148531	 lr=0.015435
Epoch[154] Batch [1299]/[3759]	Speed: 118.036963 samples/sec	 accuracy=50.573317	 loss=2.149435	 lr=0.015426
Epoch[154] Batch [1349]/[3759]	Speed: 116.958908 samples/sec	 accuracy=50.585648	 loss=2.149213	 lr=0.015417
Epoch[154] Batch [1399]/[3759]	Speed: 117.525811 samples/sec	 accuracy=50.631696	 loss=2.147280	 lr=0.015407
Epoch[154] Batch [1449]/[3759]	Speed: 117.235779 samples/sec	 accuracy=50.663793	 loss=2.146417	 lr=0.015398
Epoch[154] Batch [1499]/[3759]	Speed: 117.508266 samples/sec	 accuracy=50.673958	 loss=2.145696	 lr=0.015389
Epoch[154] Batch [1549]/[3759]	Speed: 117.208122 samples/sec	 accuracy=50.632056	 loss=2.147327	 lr=0.015379
Epoch[154] Batch [1599]/[3759]	Speed: 116.484070 samples/sec	 accuracy=50.640625	 loss=2.148562	 lr=0.015370
Epoch[154] Batch [1649]/[3759]	Speed: 117.223892 samples/sec	 accuracy=50.656250	 loss=2.148272	 lr=0.015361
Epoch[154] Batch [1699]/[3759]	Speed: 117.459805 samples/sec	 accuracy=50.659007	 loss=2.149587	 lr=0.015351
Epoch[154] Batch [1749]/[3759]	Speed: 117.164548 samples/sec	 accuracy=50.682143	 loss=2.150234	 lr=0.015342
Epoch[154] Batch [1799]/[3759]	Speed: 116.883950 samples/sec	 accuracy=50.671007	 loss=2.150758	 lr=0.015333
Epoch[154] Batch [1849]/[3759]	Speed: 116.966465 samples/sec	 accuracy=50.654561	 loss=2.151688	 lr=0.015323
Epoch[154] Batch [1899]/[3759]	Speed: 117.610567 samples/sec	 accuracy=50.651316	 loss=2.151750	 lr=0.015314
Epoch[154] Batch [1949]/[3759]	Speed: 116.920862 samples/sec	 accuracy=50.630609	 loss=2.152759	 lr=0.015305
Epoch[154] Batch [1999]/[3759]	Speed: 116.884335 samples/sec	 accuracy=50.621875	 loss=2.153371	 lr=0.015296
Epoch[154] Batch [2049]/[3759]	Speed: 114.866326 samples/sec	 accuracy=50.625762	 loss=2.152215	 lr=0.015286
Epoch[154] Batch [2099]/[3759]	Speed: 116.955116 samples/sec	 accuracy=50.662202	 loss=2.151758	 lr=0.015277
Epoch[154] Batch [2149]/[3759]	Speed: 117.193909 samples/sec	 accuracy=50.661337	 loss=2.152622	 lr=0.015268
Epoch[154] Batch [2199]/[3759]	Speed: 117.475808 samples/sec	 accuracy=50.635653	 loss=2.153755	 lr=0.015258
Epoch[154] Batch [2249]/[3759]	Speed: 118.314708 samples/sec	 accuracy=50.642361	 loss=2.153613	 lr=0.015249
Epoch[154] Batch [2299]/[3759]	Speed: 117.597760 samples/sec	 accuracy=50.650136	 loss=2.154436	 lr=0.015240
Epoch[154] Batch [2349]/[3759]	Speed: 117.006871 samples/sec	 accuracy=50.648936	 loss=2.153922	 lr=0.015231
Epoch[154] Batch [2399]/[3759]	Speed: 117.481228 samples/sec	 accuracy=50.638021	 loss=2.154572	 lr=0.015221
Epoch[154] Batch [2449]/[3759]	Speed: 117.079883 samples/sec	 accuracy=50.633291	 loss=2.154731	 lr=0.015212
Epoch[154] Batch [2499]/[3759]	Speed: 117.275121 samples/sec	 accuracy=50.596875	 loss=2.155952	 lr=0.015203
Epoch[154] Batch [2549]/[3759]	Speed: 117.099226 samples/sec	 accuracy=50.597426	 loss=2.156388	 lr=0.015194
Epoch[154] Batch [2599]/[3759]	Speed: 117.662563 samples/sec	 accuracy=50.581731	 loss=2.157373	 lr=0.015184
Epoch[154] Batch [2649]/[3759]	Speed: 117.807284 samples/sec	 accuracy=50.573703	 loss=2.157477	 lr=0.015175
Epoch[154] Batch [2699]/[3759]	Speed: 117.877623 samples/sec	 accuracy=50.592593	 loss=2.156922	 lr=0.015166
Epoch[154] Batch [2749]/[3759]	Speed: 117.560422 samples/sec	 accuracy=50.572159	 loss=2.157253	 lr=0.015157
Epoch[154] Batch [2799]/[3759]	Speed: 118.267730 samples/sec	 accuracy=50.524554	 loss=2.158964	 lr=0.015147
Epoch[154] Batch [2849]/[3759]	Speed: 118.047534 samples/sec	 accuracy=50.560855	 loss=2.156759	 lr=0.015138
Epoch[154] Batch [2899]/[3759]	Speed: 117.843983 samples/sec	 accuracy=50.551185	 loss=2.157449	 lr=0.015129
Epoch[154] Batch [2949]/[3759]	Speed: 117.820733 samples/sec	 accuracy=50.566208	 loss=2.157356	 lr=0.015120
Epoch[154] Batch [2999]/[3759]	Speed: 117.844510 samples/sec	 accuracy=50.568229	 loss=2.157311	 lr=0.015110
Epoch[154] Batch [3049]/[3759]	Speed: 117.347666 samples/sec	 accuracy=50.557377	 loss=2.157441	 lr=0.015101
Epoch[154] Batch [3099]/[3759]	Speed: 117.378402 samples/sec	 accuracy=50.552923	 loss=2.157237	 lr=0.015092
Epoch[154] Batch [3149]/[3759]	Speed: 117.792650 samples/sec	 accuracy=50.564484	 loss=2.156439	 lr=0.015083
Epoch[154] Batch [3199]/[3759]	Speed: 116.998546 samples/sec	 accuracy=50.542480	 loss=2.156961	 lr=0.015073
Epoch[154] Batch [3249]/[3759]	Speed: 117.659732 samples/sec	 accuracy=50.548558	 loss=2.157228	 lr=0.015064
Epoch[154] Batch [3299]/[3759]	Speed: 117.549955 samples/sec	 accuracy=50.535511	 loss=2.158551	 lr=0.015055
Epoch[154] Batch [3349]/[3759]	Speed: 117.745456 samples/sec	 accuracy=50.520056	 loss=2.159139	 lr=0.015046
Epoch[154] Batch [3399]/[3759]	Speed: 115.564919 samples/sec	 accuracy=50.534467	 loss=2.158948	 lr=0.015037
Epoch[154] Batch [3449]/[3759]	Speed: 117.613309 samples/sec	 accuracy=50.519928	 loss=2.160024	 lr=0.015027
Epoch[154] Batch [3499]/[3759]	Speed: 117.142557 samples/sec	 accuracy=50.509375	 loss=2.160422	 lr=0.015018
Epoch[154] Batch [3549]/[3759]	Speed: 117.023058 samples/sec	 accuracy=50.525088	 loss=2.160155	 lr=0.015009
Epoch[154] Batch [3599]/[3759]	Speed: 118.237714 samples/sec	 accuracy=50.520833	 loss=2.160085	 lr=0.015000
Epoch[154] Batch [3649]/[3759]	Speed: 117.808862 samples/sec	 accuracy=50.509418	 loss=2.159935	 lr=0.014990
Epoch[154] Batch [3699]/[3759]	Speed: 117.779107 samples/sec	 accuracy=50.515625	 loss=2.160047	 lr=0.014981
Epoch[154] Batch [3749]/[3759]	Speed: 120.706522 samples/sec	 accuracy=50.498333	 loss=2.161039	 lr=0.014972
Batch [0049]/[0303]: acc-top1=52.093750 acc-top5=75.718750
Batch [0099]/[0303]: acc-top1=48.593750 acc-top5=73.453125
Batch [0149]/[0303]: acc-top1=49.979167 acc-top5=74.166667
Batch [0199]/[0303]: acc-top1=52.515625 acc-top5=76.000000
Batch [0249]/[0303]: acc-top1=51.718750 acc-top5=75.656250
Batch [0299]/[0303]: acc-top1=51.192708 acc-top5=75.786458
[Epoch 154] training: accuracy=50.497972	 loss=2.161160
[Epoch 154] speed: 116 samples/sec	time cost: 2176.686934
[Epoch 154] validation: acc-top1=51.149959 acc-top5=75.783828 loss=2.196522
Epoch[155] Batch [0049]/[3760]	Speed: 87.568151 samples/sec	 accuracy=51.906250	 loss=2.129714	 lr=0.014961
Epoch[155] Batch [0099]/[3760]	Speed: 117.309269 samples/sec	 accuracy=51.234375	 loss=2.139921	 lr=0.014952
Epoch[155] Batch [0149]/[3760]	Speed: 117.728738 samples/sec	 accuracy=52.062500	 loss=2.104381	 lr=0.014943
Epoch[155] Batch [0199]/[3760]	Speed: 116.877712 samples/sec	 accuracy=52.156250	 loss=2.096067	 lr=0.014934
Epoch[155] Batch [0249]/[3760]	Speed: 117.329104 samples/sec	 accuracy=51.831250	 loss=2.100678	 lr=0.014924
Epoch[155] Batch [0299]/[3760]	Speed: 117.535546 samples/sec	 accuracy=51.635417	 loss=2.107508	 lr=0.014915
Epoch[155] Batch [0349]/[3760]	Speed: 116.874875 samples/sec	 accuracy=51.691964	 loss=2.105747	 lr=0.014906
Epoch[155] Batch [0399]/[3760]	Speed: 117.518293 samples/sec	 accuracy=51.730469	 loss=2.106119	 lr=0.014897
Epoch[155] Batch [0449]/[3760]	Speed: 117.702076 samples/sec	 accuracy=51.708333	 loss=2.109722	 lr=0.014888
Epoch[155] Batch [0499]/[3760]	Speed: 118.158495 samples/sec	 accuracy=51.793750	 loss=2.106486	 lr=0.014878
Epoch[155] Batch [0549]/[3760]	Speed: 117.182611 samples/sec	 accuracy=51.863636	 loss=2.107270	 lr=0.014869
Epoch[155] Batch [0599]/[3760]	Speed: 118.122551 samples/sec	 accuracy=51.833333	 loss=2.106311	 lr=0.014860
Epoch[155] Batch [0649]/[3760]	Speed: 117.722322 samples/sec	 accuracy=51.747596	 loss=2.111898	 lr=0.014851
Epoch[155] Batch [0699]/[3760]	Speed: 117.379919 samples/sec	 accuracy=51.747768	 loss=2.109507	 lr=0.014842
Epoch[155] Batch [0749]/[3760]	Speed: 117.644458 samples/sec	 accuracy=51.714583	 loss=2.111177	 lr=0.014833
Epoch[155] Batch [0799]/[3760]	Speed: 116.992580 samples/sec	 accuracy=51.675781	 loss=2.114825	 lr=0.014823
Epoch[155] Batch [0849]/[3760]	Speed: 117.142895 samples/sec	 accuracy=51.602941	 loss=2.113229	 lr=0.014814
Epoch[155] Batch [0899]/[3760]	Speed: 116.686930 samples/sec	 accuracy=51.600694	 loss=2.113388	 lr=0.014805
Epoch[155] Batch [0949]/[3760]	Speed: 114.891908 samples/sec	 accuracy=51.534539	 loss=2.114930	 lr=0.014796
Epoch[155] Batch [0999]/[3760]	Speed: 117.563325 samples/sec	 accuracy=51.492188	 loss=2.117050	 lr=0.014787
Epoch[155] Batch [1049]/[3760]	Speed: 117.330215 samples/sec	 accuracy=51.483631	 loss=2.116956	 lr=0.014778
Epoch[155] Batch [1099]/[3760]	Speed: 117.310710 samples/sec	 accuracy=51.528409	 loss=2.114647	 lr=0.014768
Epoch[155] Batch [1149]/[3760]	Speed: 117.558389 samples/sec	 accuracy=51.547554	 loss=2.114172	 lr=0.014759
Epoch[155] Batch [1199]/[3760]	Speed: 116.856748 samples/sec	 accuracy=51.492188	 loss=2.116221	 lr=0.014750
Epoch[155] Batch [1249]/[3760]	Speed: 117.710824 samples/sec	 accuracy=51.517500	 loss=2.114938	 lr=0.014741
Epoch[155] Batch [1299]/[3760]	Speed: 117.583122 samples/sec	 accuracy=51.495192	 loss=2.116866	 lr=0.014732
Epoch[155] Batch [1349]/[3760]	Speed: 117.271625 samples/sec	 accuracy=51.409722	 loss=2.118851	 lr=0.014723
Epoch[155] Batch [1399]/[3760]	Speed: 119.019636 samples/sec	 accuracy=51.373884	 loss=2.120611	 lr=0.014714
Epoch[155] Batch [1449]/[3760]	Speed: 117.633297 samples/sec	 accuracy=51.342672	 loss=2.121751	 lr=0.014704
Epoch[155] Batch [1499]/[3760]	Speed: 117.504544 samples/sec	 accuracy=51.416667	 loss=2.119216	 lr=0.014695
Epoch[155] Batch [1549]/[3760]	Speed: 117.155841 samples/sec	 accuracy=51.412298	 loss=2.119229	 lr=0.014686
Epoch[155] Batch [1599]/[3760]	Speed: 117.774718 samples/sec	 accuracy=51.389648	 loss=2.119581	 lr=0.014677
Epoch[155] Batch [1649]/[3760]	Speed: 117.790187 samples/sec	 accuracy=51.388258	 loss=2.120114	 lr=0.014668
Epoch[155] Batch [1699]/[3760]	Speed: 117.116604 samples/sec	 accuracy=51.368566	 loss=2.120332	 lr=0.014659
Epoch[155] Batch [1749]/[3760]	Speed: 118.019425 samples/sec	 accuracy=51.338393	 loss=2.120171	 lr=0.014650
Epoch[155] Batch [1799]/[3760]	Speed: 117.387166 samples/sec	 accuracy=51.328125	 loss=2.121513	 lr=0.014641
Epoch[155] Batch [1849]/[3760]	Speed: 117.470176 samples/sec	 accuracy=51.352196	 loss=2.121283	 lr=0.014631
Epoch[155] Batch [1899]/[3760]	Speed: 117.598675 samples/sec	 accuracy=51.353618	 loss=2.120994	 lr=0.014622
Epoch[155] Batch [1949]/[3760]	Speed: 117.447461 samples/sec	 accuracy=51.347756	 loss=2.121671	 lr=0.014613
Epoch[155] Batch [1999]/[3760]	Speed: 117.706283 samples/sec	 accuracy=51.357813	 loss=2.121186	 lr=0.014604
Epoch[155] Batch [2049]/[3760]	Speed: 117.736401 samples/sec	 accuracy=51.307927	 loss=2.122756	 lr=0.014595
Epoch[155] Batch [2099]/[3760]	Speed: 117.614710 samples/sec	 accuracy=51.301339	 loss=2.123270	 lr=0.014586
Epoch[155] Batch [2149]/[3760]	Speed: 116.831528 samples/sec	 accuracy=51.239099	 loss=2.124738	 lr=0.014577
Epoch[155] Batch [2199]/[3760]	Speed: 117.473414 samples/sec	 accuracy=51.191761	 loss=2.128092	 lr=0.014568
Epoch[155] Batch [2249]/[3760]	Speed: 117.227227 samples/sec	 accuracy=51.218056	 loss=2.127597	 lr=0.014559
Epoch[155] Batch [2299]/[3760]	Speed: 115.555733 samples/sec	 accuracy=51.201766	 loss=2.128359	 lr=0.014550
Epoch[155] Batch [2349]/[3760]	Speed: 117.932885 samples/sec	 accuracy=51.178856	 loss=2.128900	 lr=0.014540
Epoch[155] Batch [2399]/[3760]	Speed: 116.817075 samples/sec	 accuracy=51.171224	 loss=2.129760	 lr=0.014531
Epoch[155] Batch [2449]/[3760]	Speed: 117.274198 samples/sec	 accuracy=51.168367	 loss=2.129971	 lr=0.014522
Epoch[155] Batch [2499]/[3760]	Speed: 117.419356 samples/sec	 accuracy=51.177500	 loss=2.130392	 lr=0.014513
Epoch[155] Batch [2549]/[3760]	Speed: 117.025851 samples/sec	 accuracy=51.191176	 loss=2.130312	 lr=0.014504
Epoch[155] Batch [2599]/[3760]	Speed: 117.867569 samples/sec	 accuracy=51.166466	 loss=2.131151	 lr=0.014495
Epoch[155] Batch [2649]/[3760]	Speed: 117.408305 samples/sec	 accuracy=51.170401	 loss=2.131710	 lr=0.014486
Epoch[155] Batch [2699]/[3760]	Speed: 116.903337 samples/sec	 accuracy=51.168403	 loss=2.132009	 lr=0.014477
Epoch[155] Batch [2749]/[3760]	Speed: 117.498608 samples/sec	 accuracy=51.181250	 loss=2.132394	 lr=0.014468
Epoch[155] Batch [2799]/[3760]	Speed: 117.544145 samples/sec	 accuracy=51.198661	 loss=2.130731	 lr=0.014459
Epoch[155] Batch [2849]/[3760]	Speed: 117.358408 samples/sec	 accuracy=51.213816	 loss=2.130430	 lr=0.014450
Epoch[155] Batch [2899]/[3760]	Speed: 117.077347 samples/sec	 accuracy=51.203664	 loss=2.130344	 lr=0.014441
Epoch[155] Batch [2949]/[3760]	Speed: 117.393117 samples/sec	 accuracy=51.175318	 loss=2.131889	 lr=0.014431
Epoch[155] Batch [2999]/[3760]	Speed: 118.253359 samples/sec	 accuracy=51.158333	 loss=2.131887	 lr=0.014422
Epoch[155] Batch [3049]/[3760]	Speed: 117.417630 samples/sec	 accuracy=51.165984	 loss=2.131203	 lr=0.014413
Epoch[155] Batch [3099]/[3760]	Speed: 117.387869 samples/sec	 accuracy=51.193548	 loss=2.130196	 lr=0.014404
Epoch[155] Batch [3149]/[3760]	Speed: 116.986762 samples/sec	 accuracy=51.167163	 loss=2.131615	 lr=0.014395
Epoch[155] Batch [3199]/[3760]	Speed: 117.719574 samples/sec	 accuracy=51.185547	 loss=2.130579	 lr=0.014386
Epoch[155] Batch [3249]/[3760]	Speed: 117.694494 samples/sec	 accuracy=51.151442	 loss=2.131370	 lr=0.014377
Epoch[155] Batch [3299]/[3760]	Speed: 117.520431 samples/sec	 accuracy=51.142519	 loss=2.131855	 lr=0.014368
Epoch[155] Batch [3349]/[3760]	Speed: 117.159892 samples/sec	 accuracy=51.168377	 loss=2.130467	 lr=0.014359
Epoch[155] Batch [3399]/[3760]	Speed: 117.511547 samples/sec	 accuracy=51.187960	 loss=2.130615	 lr=0.014350
Epoch[155] Batch [3449]/[3760]	Speed: 117.764285 samples/sec	 accuracy=51.185236	 loss=2.130547	 lr=0.014341
Epoch[155] Batch [3499]/[3760]	Speed: 117.243679 samples/sec	 accuracy=51.197321	 loss=2.130702	 lr=0.014332
Epoch[155] Batch [3549]/[3760]	Speed: 117.533480 samples/sec	 accuracy=51.188820	 loss=2.131265	 lr=0.014323
Epoch[155] Batch [3599]/[3760]	Speed: 115.121681 samples/sec	 accuracy=51.171875	 loss=2.132319	 lr=0.014314
Epoch[155] Batch [3649]/[3760]	Speed: 117.864526 samples/sec	 accuracy=51.184075	 loss=2.132677	 lr=0.014305
Epoch[155] Batch [3699]/[3760]	Speed: 117.822438 samples/sec	 accuracy=51.167230	 loss=2.133407	 lr=0.014296
Epoch[155] Batch [3749]/[3760]	Speed: 119.920381 samples/sec	 accuracy=51.161667	 loss=2.133547	 lr=0.014287
Batch [0049]/[0303]: acc-top1=51.531250 acc-top5=73.500000
Batch [0099]/[0303]: acc-top1=48.781250 acc-top5=72.718750
Batch [0149]/[0303]: acc-top1=50.062500 acc-top5=73.562500
Batch [0199]/[0303]: acc-top1=52.914062 acc-top5=75.320312
Batch [0249]/[0303]: acc-top1=52.112500 acc-top5=74.975000
Batch [0299]/[0303]: acc-top1=51.817708 acc-top5=75.213542
[Epoch 155] training: accuracy=51.164395	 loss=2.133418
[Epoch 155] speed: 116 samples/sec	time cost: 2175.815996
[Epoch 155] validation: acc-top1=51.742987 acc-top5=75.242368 loss=2.216556
Epoch[156] Batch [0049]/[3760]	Speed: 87.610866 samples/sec	 accuracy=52.875000	 loss=2.073863	 lr=0.014276
Epoch[156] Batch [0099]/[3760]	Speed: 117.352948 samples/sec	 accuracy=52.500000	 loss=2.063570	 lr=0.014267
Epoch[156] Batch [0149]/[3760]	Speed: 117.570211 samples/sec	 accuracy=52.354167	 loss=2.085503	 lr=0.014258
Epoch[156] Batch [0199]/[3760]	Speed: 117.725667 samples/sec	 accuracy=51.835937	 loss=2.100597	 lr=0.014249
Epoch[156] Batch [0249]/[3760]	Speed: 117.761025 samples/sec	 accuracy=51.650000	 loss=2.105735	 lr=0.014240
Epoch[156] Batch [0299]/[3760]	Speed: 117.726378 samples/sec	 accuracy=51.661458	 loss=2.105411	 lr=0.014231
Epoch[156] Batch [0349]/[3760]	Speed: 117.476334 samples/sec	 accuracy=51.745536	 loss=2.100625	 lr=0.014222
Epoch[156] Batch [0399]/[3760]	Speed: 117.432966 samples/sec	 accuracy=51.800781	 loss=2.095727	 lr=0.014213
Epoch[156] Batch [0449]/[3760]	Speed: 117.741926 samples/sec	 accuracy=51.843750	 loss=2.093985	 lr=0.014204
Epoch[156] Batch [0499]/[3760]	Speed: 117.633413 samples/sec	 accuracy=51.821875	 loss=2.093213	 lr=0.014195
Epoch[156] Batch [0549]/[3760]	Speed: 117.591267 samples/sec	 accuracy=51.786932	 loss=2.098884	 lr=0.014186
Epoch[156] Batch [0599]/[3760]	Speed: 117.890343 samples/sec	 accuracy=51.687500	 loss=2.104721	 lr=0.014177
Epoch[156] Batch [0649]/[3760]	Speed: 117.935844 samples/sec	 accuracy=51.649038	 loss=2.103692	 lr=0.014168
Epoch[156] Batch [0699]/[3760]	Speed: 117.473446 samples/sec	 accuracy=51.720982	 loss=2.103822	 lr=0.014159
Epoch[156] Batch [0749]/[3760]	Speed: 118.397741 samples/sec	 accuracy=51.739583	 loss=2.102856	 lr=0.014150
Epoch[156] Batch [0799]/[3760]	Speed: 117.259183 samples/sec	 accuracy=51.796875	 loss=2.101180	 lr=0.014141
Epoch[156] Batch [0849]/[3760]	Speed: 117.897938 samples/sec	 accuracy=51.851103	 loss=2.099660	 lr=0.014132
Epoch[156] Batch [0899]/[3760]	Speed: 118.195094 samples/sec	 accuracy=51.840278	 loss=2.100823	 lr=0.014123
Epoch[156] Batch [0949]/[3760]	Speed: 117.506582 samples/sec	 accuracy=51.777961	 loss=2.103488	 lr=0.014114
Epoch[156] Batch [0999]/[3760]	Speed: 117.287998 samples/sec	 accuracy=51.709375	 loss=2.108053	 lr=0.014105
Epoch[156] Batch [1049]/[3760]	Speed: 115.500970 samples/sec	 accuracy=51.717262	 loss=2.108184	 lr=0.014096
Epoch[156] Batch [1099]/[3760]	Speed: 117.998332 samples/sec	 accuracy=51.701705	 loss=2.109902	 lr=0.014087
Epoch[156] Batch [1149]/[3760]	Speed: 117.658293 samples/sec	 accuracy=51.686141	 loss=2.112530	 lr=0.014078
Epoch[156] Batch [1199]/[3760]	Speed: 117.244276 samples/sec	 accuracy=51.652344	 loss=2.113683	 lr=0.014069
Epoch[156] Batch [1249]/[3760]	Speed: 117.693347 samples/sec	 accuracy=51.590000	 loss=2.114788	 lr=0.014060
Epoch[156] Batch [1299]/[3760]	Speed: 117.368746 samples/sec	 accuracy=51.572115	 loss=2.115737	 lr=0.014051
Epoch[156] Batch [1349]/[3760]	Speed: 117.686149 samples/sec	 accuracy=51.579861	 loss=2.115882	 lr=0.014042
Epoch[156] Batch [1399]/[3760]	Speed: 117.208213 samples/sec	 accuracy=51.581473	 loss=2.116393	 lr=0.014033
Epoch[156] Batch [1449]/[3760]	Speed: 117.491924 samples/sec	 accuracy=51.525862	 loss=2.117519	 lr=0.014024
Epoch[156] Batch [1499]/[3760]	Speed: 117.107128 samples/sec	 accuracy=51.528125	 loss=2.116238	 lr=0.014015
Epoch[156] Batch [1549]/[3760]	Speed: 117.579693 samples/sec	 accuracy=51.540323	 loss=2.116157	 lr=0.014006
Epoch[156] Batch [1599]/[3760]	Speed: 118.038870 samples/sec	 accuracy=51.536133	 loss=2.116185	 lr=0.013997
Epoch[156] Batch [1649]/[3760]	Speed: 116.683032 samples/sec	 accuracy=51.567235	 loss=2.114378	 lr=0.013988
Epoch[156] Batch [1699]/[3760]	Speed: 117.449396 samples/sec	 accuracy=51.574449	 loss=2.113224	 lr=0.013979
Epoch[156] Batch [1749]/[3760]	Speed: 116.825958 samples/sec	 accuracy=51.545536	 loss=2.113795	 lr=0.013971
Epoch[156] Batch [1799]/[3760]	Speed: 117.714531 samples/sec	 accuracy=51.559028	 loss=2.113293	 lr=0.013962
Epoch[156] Batch [1849]/[3760]	Speed: 118.675789 samples/sec	 accuracy=51.543919	 loss=2.114829	 lr=0.013953
Epoch[156] Batch [1899]/[3760]	Speed: 117.999757 samples/sec	 accuracy=51.559211	 loss=2.114611	 lr=0.013944
Epoch[156] Batch [1949]/[3760]	Speed: 117.776003 samples/sec	 accuracy=51.568910	 loss=2.114138	 lr=0.013935
Epoch[156] Batch [1999]/[3760]	Speed: 117.439704 samples/sec	 accuracy=51.574219	 loss=2.114102	 lr=0.013926
Epoch[156] Batch [2049]/[3760]	Speed: 117.563503 samples/sec	 accuracy=51.544207	 loss=2.114866	 lr=0.013917
Epoch[156] Batch [2099]/[3760]	Speed: 118.335873 samples/sec	 accuracy=51.523810	 loss=2.116464	 lr=0.013908
Epoch[156] Batch [2149]/[3760]	Speed: 117.700761 samples/sec	 accuracy=51.535610	 loss=2.115993	 lr=0.013899
Epoch[156] Batch [2199]/[3760]	Speed: 118.002319 samples/sec	 accuracy=51.519176	 loss=2.116958	 lr=0.013890
Epoch[156] Batch [2249]/[3760]	Speed: 117.967879 samples/sec	 accuracy=51.537500	 loss=2.116638	 lr=0.013881
Epoch[156] Batch [2299]/[3760]	Speed: 117.575274 samples/sec	 accuracy=51.559103	 loss=2.116461	 lr=0.013872
Epoch[156] Batch [2349]/[3760]	Speed: 118.098942 samples/sec	 accuracy=51.533910	 loss=2.117900	 lr=0.013863
Epoch[156] Batch [2399]/[3760]	Speed: 115.771864 samples/sec	 accuracy=51.545573	 loss=2.118101	 lr=0.013854
Epoch[156] Batch [2449]/[3760]	Speed: 117.265580 samples/sec	 accuracy=51.552934	 loss=2.117775	 lr=0.013846
Epoch[156] Batch [2499]/[3760]	Speed: 117.573779 samples/sec	 accuracy=51.574375	 loss=2.116232	 lr=0.013837
Epoch[156] Batch [2549]/[3760]	Speed: 117.586311 samples/sec	 accuracy=51.560049	 loss=2.116068	 lr=0.013828
Epoch[156] Batch [2599]/[3760]	Speed: 116.465244 samples/sec	 accuracy=51.556490	 loss=2.116666	 lr=0.013819
Epoch[156] Batch [2649]/[3760]	Speed: 117.750064 samples/sec	 accuracy=51.579599	 loss=2.115546	 lr=0.013810
Epoch[156] Batch [2699]/[3760]	Speed: 118.016776 samples/sec	 accuracy=51.560185	 loss=2.116035	 lr=0.013801
Epoch[156] Batch [2749]/[3760]	Speed: 117.213752 samples/sec	 accuracy=51.537500	 loss=2.117237	 lr=0.013792
Epoch[156] Batch [2799]/[3760]	Speed: 117.293755 samples/sec	 accuracy=51.544643	 loss=2.116546	 lr=0.013783
Epoch[156] Batch [2849]/[3760]	Speed: 117.709265 samples/sec	 accuracy=51.537281	 loss=2.116588	 lr=0.013774
Epoch[156] Batch [2899]/[3760]	Speed: 117.435052 samples/sec	 accuracy=51.514009	 loss=2.117861	 lr=0.013765
Epoch[156] Batch [2949]/[3760]	Speed: 118.346933 samples/sec	 accuracy=51.517479	 loss=2.117809	 lr=0.013757
Epoch[156] Batch [2999]/[3760]	Speed: 117.369713 samples/sec	 accuracy=51.529688	 loss=2.117574	 lr=0.013748
Epoch[156] Batch [3049]/[3760]	Speed: 117.747217 samples/sec	 accuracy=51.562500	 loss=2.115882	 lr=0.013739
Epoch[156] Batch [3099]/[3760]	Speed: 117.482990 samples/sec	 accuracy=51.531754	 loss=2.117280	 lr=0.013730
Epoch[156] Batch [3149]/[3760]	Speed: 117.736419 samples/sec	 accuracy=51.502976	 loss=2.118399	 lr=0.013721
Epoch[156] Batch [3199]/[3760]	Speed: 117.439188 samples/sec	 accuracy=51.509766	 loss=2.118024	 lr=0.013712
Epoch[156] Batch [3249]/[3760]	Speed: 117.791976 samples/sec	 accuracy=51.505769	 loss=2.117832	 lr=0.013703
Epoch[156] Batch [3299]/[3760]	Speed: 117.870351 samples/sec	 accuracy=51.517992	 loss=2.117584	 lr=0.013694
Epoch[156] Batch [3349]/[3760]	Speed: 117.322797 samples/sec	 accuracy=51.506530	 loss=2.117245	 lr=0.013686
Epoch[156] Batch [3399]/[3760]	Speed: 117.267044 samples/sec	 accuracy=51.523438	 loss=2.116090	 lr=0.013677
Epoch[156] Batch [3449]/[3760]	Speed: 118.161044 samples/sec	 accuracy=51.524457	 loss=2.116072	 lr=0.013668
Epoch[156] Batch [3499]/[3760]	Speed: 117.784210 samples/sec	 accuracy=51.499554	 loss=2.116161	 lr=0.013659
Epoch[156] Batch [3549]/[3760]	Speed: 116.755605 samples/sec	 accuracy=51.514085	 loss=2.115118	 lr=0.013650
Epoch[156] Batch [3599]/[3760]	Speed: 117.999228 samples/sec	 accuracy=51.509549	 loss=2.115010	 lr=0.013641
Epoch[156] Batch [3649]/[3760]	Speed: 117.028389 samples/sec	 accuracy=51.527397	 loss=2.114548	 lr=0.013632
Epoch[156] Batch [3699]/[3760]	Speed: 117.359843 samples/sec	 accuracy=51.495777	 loss=2.115552	 lr=0.013624
Epoch[156] Batch [3749]/[3760]	Speed: 118.348485 samples/sec	 accuracy=51.506667	 loss=2.114885	 lr=0.013615
Batch [0049]/[0303]: acc-top1=49.875000 acc-top5=72.906250
Batch [0099]/[0303]: acc-top1=48.000000 acc-top5=72.718750
Batch [0149]/[0303]: acc-top1=49.958333 acc-top5=73.760417
Batch [0199]/[0303]: acc-top1=53.054688 acc-top5=75.773438
Batch [0249]/[0303]: acc-top1=51.806250 acc-top5=75.187500
Batch [0299]/[0303]: acc-top1=51.026042 acc-top5=75.052083
[Epoch 156] training: accuracy=51.497257	 loss=2.115337
[Epoch 156] speed: 117 samples/sec	time cost: 2172.584457
[Epoch 156] validation: acc-top1=50.943688 acc-top5=75.030941 loss=2.247844
Epoch[157] Batch [0049]/[3759]	Speed: 87.913801 samples/sec	 accuracy=50.968750	 loss=2.123201	 lr=0.013604
Epoch[157] Batch [0099]/[3759]	Speed: 118.186775 samples/sec	 accuracy=51.250000	 loss=2.102196	 lr=0.013595
Epoch[157] Batch [0149]/[3759]	Speed: 117.383446 samples/sec	 accuracy=51.812500	 loss=2.093595	 lr=0.013586
Epoch[157] Batch [0199]/[3759]	Speed: 117.211185 samples/sec	 accuracy=52.125000	 loss=2.082592	 lr=0.013578
Epoch[157] Batch [0249]/[3759]	Speed: 116.834612 samples/sec	 accuracy=52.350000	 loss=2.073389	 lr=0.013569
Epoch[157] Batch [0299]/[3759]	Speed: 117.799070 samples/sec	 accuracy=52.421875	 loss=2.073113	 lr=0.013560
Epoch[157] Batch [0349]/[3759]	Speed: 117.742596 samples/sec	 accuracy=52.406250	 loss=2.067670	 lr=0.013551
Epoch[157] Batch [0399]/[3759]	Speed: 117.586845 samples/sec	 accuracy=52.425781	 loss=2.067179	 lr=0.013542
Epoch[157] Batch [0449]/[3759]	Speed: 117.823759 samples/sec	 accuracy=52.392361	 loss=2.071270	 lr=0.013533
Epoch[157] Batch [0499]/[3759]	Speed: 117.576189 samples/sec	 accuracy=52.300000	 loss=2.073820	 lr=0.013525
Epoch[157] Batch [0549]/[3759]	Speed: 118.375804 samples/sec	 accuracy=52.281250	 loss=2.078474	 lr=0.013516
Epoch[157] Batch [0599]/[3759]	Speed: 116.528889 samples/sec	 accuracy=52.286458	 loss=2.079815	 lr=0.013507
Epoch[157] Batch [0649]/[3759]	Speed: 117.457920 samples/sec	 accuracy=52.379808	 loss=2.076056	 lr=0.013498
Epoch[157] Batch [0699]/[3759]	Speed: 117.370145 samples/sec	 accuracy=52.424107	 loss=2.074611	 lr=0.013489
Epoch[157] Batch [0749]/[3759]	Speed: 117.564135 samples/sec	 accuracy=52.389583	 loss=2.076936	 lr=0.013481
Epoch[157] Batch [0799]/[3759]	Speed: 117.784614 samples/sec	 accuracy=52.390625	 loss=2.078114	 lr=0.013472
Epoch[157] Batch [0849]/[3759]	Speed: 117.086690 samples/sec	 accuracy=52.363971	 loss=2.078507	 lr=0.013463
Epoch[157] Batch [0899]/[3759]	Speed: 118.018239 samples/sec	 accuracy=52.428819	 loss=2.075857	 lr=0.013454
Epoch[157] Batch [0949]/[3759]	Speed: 117.227621 samples/sec	 accuracy=52.463816	 loss=2.075821	 lr=0.013445
Epoch[157] Batch [0999]/[3759]	Speed: 117.336291 samples/sec	 accuracy=52.492188	 loss=2.073628	 lr=0.013437
Epoch[157] Batch [1049]/[3759]	Speed: 116.900065 samples/sec	 accuracy=52.468750	 loss=2.076899	 lr=0.013428
Epoch[157] Batch [1099]/[3759]	Speed: 116.853222 samples/sec	 accuracy=52.458807	 loss=2.076414	 lr=0.013419
Epoch[157] Batch [1149]/[3759]	Speed: 115.748956 samples/sec	 accuracy=52.396739	 loss=2.078643	 lr=0.013410
Epoch[157] Batch [1199]/[3759]	Speed: 116.563658 samples/sec	 accuracy=52.369792	 loss=2.079647	 lr=0.013401
Epoch[157] Batch [1249]/[3759]	Speed: 117.666990 samples/sec	 accuracy=52.258750	 loss=2.082075	 lr=0.013393
Epoch[157] Batch [1299]/[3759]	Speed: 117.380829 samples/sec	 accuracy=52.287260	 loss=2.080535	 lr=0.013384
Epoch[157] Batch [1349]/[3759]	Speed: 116.358993 samples/sec	 accuracy=52.310185	 loss=2.080104	 lr=0.013375
Epoch[157] Batch [1399]/[3759]	Speed: 117.619796 samples/sec	 accuracy=52.302455	 loss=2.079930	 lr=0.013366
Epoch[157] Batch [1449]/[3759]	Speed: 117.333284 samples/sec	 accuracy=52.290948	 loss=2.080179	 lr=0.013357
Epoch[157] Batch [1499]/[3759]	Speed: 116.770018 samples/sec	 accuracy=52.247917	 loss=2.082388	 lr=0.013349
Epoch[157] Batch [1549]/[3759]	Speed: 117.194454 samples/sec	 accuracy=52.218750	 loss=2.082364	 lr=0.013340
Epoch[157] Batch [1599]/[3759]	Speed: 117.450304 samples/sec	 accuracy=52.223633	 loss=2.082039	 lr=0.013331
Epoch[157] Batch [1649]/[3759]	Speed: 117.928232 samples/sec	 accuracy=52.187500	 loss=2.081385	 lr=0.013322
Epoch[157] Batch [1699]/[3759]	Speed: 117.221723 samples/sec	 accuracy=52.187500	 loss=2.081083	 lr=0.013314
Epoch[157] Batch [1749]/[3759]	Speed: 117.923284 samples/sec	 accuracy=52.167857	 loss=2.081573	 lr=0.013305
Epoch[157] Batch [1799]/[3759]	Speed: 117.315078 samples/sec	 accuracy=52.144965	 loss=2.081568	 lr=0.013296
Epoch[157] Batch [1849]/[3759]	Speed: 117.496879 samples/sec	 accuracy=52.119932	 loss=2.083145	 lr=0.013287
Epoch[157] Batch [1899]/[3759]	Speed: 117.666834 samples/sec	 accuracy=52.111842	 loss=2.082621	 lr=0.013279
Epoch[157] Batch [1949]/[3759]	Speed: 116.534318 samples/sec	 accuracy=52.125000	 loss=2.082668	 lr=0.013270
Epoch[157] Batch [1999]/[3759]	Speed: 117.223291 samples/sec	 accuracy=52.145312	 loss=2.081843	 lr=0.013261
Epoch[157] Batch [2049]/[3759]	Speed: 117.586693 samples/sec	 accuracy=52.096799	 loss=2.082940	 lr=0.013252
Epoch[157] Batch [2099]/[3759]	Speed: 118.008013 samples/sec	 accuracy=52.092262	 loss=2.084003	 lr=0.013244
Epoch[157] Batch [2149]/[3759]	Speed: 118.244212 samples/sec	 accuracy=52.119913	 loss=2.083725	 lr=0.013235
Epoch[157] Batch [2199]/[3759]	Speed: 117.965533 samples/sec	 accuracy=52.091619	 loss=2.084278	 lr=0.013226
Epoch[157] Batch [2249]/[3759]	Speed: 117.481682 samples/sec	 accuracy=52.079167	 loss=2.085082	 lr=0.013217
Epoch[157] Batch [2299]/[3759]	Speed: 116.942262 samples/sec	 accuracy=52.096467	 loss=2.085206	 lr=0.013209
Epoch[157] Batch [2349]/[3759]	Speed: 117.990780 samples/sec	 accuracy=52.119681	 loss=2.084057	 lr=0.013200
Epoch[157] Batch [2399]/[3759]	Speed: 118.320678 samples/sec	 accuracy=52.083333	 loss=2.085588	 lr=0.013191
Epoch[157] Batch [2449]/[3759]	Speed: 117.081346 samples/sec	 accuracy=52.049745	 loss=2.086617	 lr=0.013182
Epoch[157] Batch [2499]/[3759]	Speed: 114.960807 samples/sec	 accuracy=52.039375	 loss=2.086419	 lr=0.013174
Epoch[157] Batch [2549]/[3759]	Speed: 117.278541 samples/sec	 accuracy=52.030637	 loss=2.087088	 lr=0.013165
Epoch[157] Batch [2599]/[3759]	Speed: 118.244516 samples/sec	 accuracy=52.020433	 loss=2.087683	 lr=0.013156
Epoch[157] Batch [2649]/[3759]	Speed: 118.079756 samples/sec	 accuracy=52.008255	 loss=2.088135	 lr=0.013148
Epoch[157] Batch [2699]/[3759]	Speed: 117.542449 samples/sec	 accuracy=52.016204	 loss=2.087824	 lr=0.013139
Epoch[157] Batch [2749]/[3759]	Speed: 118.096337 samples/sec	 accuracy=52.006818	 loss=2.088135	 lr=0.013130
Epoch[157] Batch [2799]/[3759]	Speed: 118.647259 samples/sec	 accuracy=52.003906	 loss=2.088001	 lr=0.013121
Epoch[157] Batch [2849]/[3759]	Speed: 117.259091 samples/sec	 accuracy=52.000548	 loss=2.087980	 lr=0.013113
Epoch[157] Batch [2899]/[3759]	Speed: 116.854327 samples/sec	 accuracy=52.007543	 loss=2.087112	 lr=0.013104
Epoch[157] Batch [2949]/[3759]	Speed: 117.979173 samples/sec	 accuracy=52.012712	 loss=2.087392	 lr=0.013095
Epoch[157] Batch [2999]/[3759]	Speed: 117.522803 samples/sec	 accuracy=51.982292	 loss=2.088293	 lr=0.013087
Epoch[157] Batch [3049]/[3759]	Speed: 118.197018 samples/sec	 accuracy=51.975410	 loss=2.088430	 lr=0.013078
Epoch[157] Batch [3099]/[3759]	Speed: 117.085475 samples/sec	 accuracy=51.967742	 loss=2.088255	 lr=0.013069
Epoch[157] Batch [3149]/[3759]	Speed: 118.493118 samples/sec	 accuracy=51.956845	 loss=2.088519	 lr=0.013061
Epoch[157] Batch [3199]/[3759]	Speed: 117.428170 samples/sec	 accuracy=51.952148	 loss=2.089336	 lr=0.013052
Epoch[157] Batch [3249]/[3759]	Speed: 117.524666 samples/sec	 accuracy=51.964423	 loss=2.089715	 lr=0.013043
Epoch[157] Batch [3299]/[3759]	Speed: 117.702150 samples/sec	 accuracy=51.950284	 loss=2.090823	 lr=0.013034
Epoch[157] Batch [3349]/[3759]	Speed: 117.356963 samples/sec	 accuracy=51.948228	 loss=2.090841	 lr=0.013026
Epoch[157] Batch [3399]/[3759]	Speed: 117.612352 samples/sec	 accuracy=51.934283	 loss=2.090674	 lr=0.013017
Epoch[157] Batch [3449]/[3759]	Speed: 118.095340 samples/sec	 accuracy=51.918931	 loss=2.091354	 lr=0.013008
Epoch[157] Batch [3499]/[3759]	Speed: 117.443932 samples/sec	 accuracy=51.907143	 loss=2.092128	 lr=0.013000
Epoch[157] Batch [3549]/[3759]	Speed: 118.325606 samples/sec	 accuracy=51.909331	 loss=2.092105	 lr=0.012991
Epoch[157] Batch [3599]/[3759]	Speed: 117.432735 samples/sec	 accuracy=51.931424	 loss=2.091163	 lr=0.012982
Epoch[157] Batch [3649]/[3759]	Speed: 118.140385 samples/sec	 accuracy=51.936216	 loss=2.090959	 lr=0.012974
Epoch[157] Batch [3699]/[3759]	Speed: 117.555293 samples/sec	 accuracy=51.948480	 loss=2.090122	 lr=0.012965
Epoch[157] Batch [3749]/[3759]	Speed: 120.776483 samples/sec	 accuracy=51.948333	 loss=2.090460	 lr=0.012956
Batch [0049]/[0303]: acc-top1=51.718750 acc-top5=74.125000
Batch [0099]/[0303]: acc-top1=48.843750 acc-top5=72.828125
Batch [0149]/[0303]: acc-top1=49.968750 acc-top5=73.437500
Batch [0199]/[0303]: acc-top1=52.867188 acc-top5=75.484375
Batch [0249]/[0303]: acc-top1=52.375000 acc-top5=75.362500
Batch [0299]/[0303]: acc-top1=51.947917 acc-top5=75.546875
[Epoch 157] training: accuracy=51.954060	 loss=2.090545
[Epoch 157] speed: 116 samples/sec	time cost: 2172.758605
[Epoch 157] validation: acc-top1=51.918317 acc-top5=75.577558 loss=2.204536
Epoch[158] Batch [0049]/[3760]	Speed: 87.719835 samples/sec	 accuracy=52.125000	 loss=2.083519	 lr=0.012946
Epoch[158] Batch [0099]/[3760]	Speed: 116.017342 samples/sec	 accuracy=51.921875	 loss=2.095363	 lr=0.012937
Epoch[158] Batch [0149]/[3760]	Speed: 117.445430 samples/sec	 accuracy=52.520833	 loss=2.067792	 lr=0.012929
Epoch[158] Batch [0199]/[3760]	Speed: 117.587032 samples/sec	 accuracy=52.671875	 loss=2.068877	 lr=0.012920
Epoch[158] Batch [0249]/[3760]	Speed: 118.334863 samples/sec	 accuracy=52.600000	 loss=2.063006	 lr=0.012912
Epoch[158] Batch [0299]/[3760]	Speed: 118.010646 samples/sec	 accuracy=53.197917	 loss=2.039514	 lr=0.012903
Epoch[158] Batch [0349]/[3760]	Speed: 117.988706 samples/sec	 accuracy=53.089286	 loss=2.043447	 lr=0.012894
Epoch[158] Batch [0399]/[3760]	Speed: 116.343270 samples/sec	 accuracy=53.089844	 loss=2.042701	 lr=0.012886
Epoch[158] Batch [0449]/[3760]	Speed: 117.035273 samples/sec	 accuracy=53.013889	 loss=2.042321	 lr=0.012877
Epoch[158] Batch [0499]/[3760]	Speed: 117.205186 samples/sec	 accuracy=53.093750	 loss=2.039056	 lr=0.012868
Epoch[158] Batch [0549]/[3760]	Speed: 117.566902 samples/sec	 accuracy=53.076705	 loss=2.037474	 lr=0.012860
Epoch[158] Batch [0599]/[3760]	Speed: 117.447320 samples/sec	 accuracy=53.070313	 loss=2.040907	 lr=0.012851
Epoch[158] Batch [0649]/[3760]	Speed: 116.881410 samples/sec	 accuracy=52.983173	 loss=2.044543	 lr=0.012842
Epoch[158] Batch [0699]/[3760]	Speed: 117.373063 samples/sec	 accuracy=52.939732	 loss=2.047215	 lr=0.012834
Epoch[158] Batch [0749]/[3760]	Speed: 117.237126 samples/sec	 accuracy=52.925000	 loss=2.049158	 lr=0.012825
Epoch[158] Batch [0799]/[3760]	Speed: 117.400525 samples/sec	 accuracy=52.914062	 loss=2.046312	 lr=0.012817
Epoch[158] Batch [0849]/[3760]	Speed: 117.425464 samples/sec	 accuracy=52.926471	 loss=2.044119	 lr=0.012808
Epoch[158] Batch [0899]/[3760]	Speed: 118.000364 samples/sec	 accuracy=52.944444	 loss=2.042392	 lr=0.012799
Epoch[158] Batch [0949]/[3760]	Speed: 117.449274 samples/sec	 accuracy=52.876645	 loss=2.043201	 lr=0.012791
Epoch[158] Batch [0999]/[3760]	Speed: 117.395677 samples/sec	 accuracy=52.826562	 loss=2.044531	 lr=0.012782
Epoch[158] Batch [1049]/[3760]	Speed: 117.580592 samples/sec	 accuracy=52.742560	 loss=2.047904	 lr=0.012773
Epoch[158] Batch [1099]/[3760]	Speed: 116.973016 samples/sec	 accuracy=52.723011	 loss=2.049783	 lr=0.012765
Epoch[158] Batch [1149]/[3760]	Speed: 117.379089 samples/sec	 accuracy=52.698370	 loss=2.051722	 lr=0.012756
Epoch[158] Batch [1199]/[3760]	Speed: 117.209666 samples/sec	 accuracy=52.742188	 loss=2.050361	 lr=0.012748
Epoch[158] Batch [1249]/[3760]	Speed: 117.806538 samples/sec	 accuracy=52.728750	 loss=2.051825	 lr=0.012739
Epoch[158] Batch [1299]/[3760]	Speed: 117.741510 samples/sec	 accuracy=52.719952	 loss=2.052942	 lr=0.012730
Epoch[158] Batch [1349]/[3760]	Speed: 118.327832 samples/sec	 accuracy=52.711806	 loss=2.053278	 lr=0.012722
Epoch[158] Batch [1399]/[3760]	Speed: 117.589098 samples/sec	 accuracy=52.680804	 loss=2.054398	 lr=0.012713
Epoch[158] Batch [1449]/[3760]	Speed: 117.923785 samples/sec	 accuracy=52.664871	 loss=2.054932	 lr=0.012705
Epoch[158] Batch [1499]/[3760]	Speed: 115.716802 samples/sec	 accuracy=52.661458	 loss=2.054400	 lr=0.012696
Epoch[158] Batch [1549]/[3760]	Speed: 117.399769 samples/sec	 accuracy=52.635081	 loss=2.056439	 lr=0.012687
Epoch[158] Batch [1599]/[3760]	Speed: 117.245644 samples/sec	 accuracy=52.629883	 loss=2.057421	 lr=0.012679
Epoch[158] Batch [1649]/[3760]	Speed: 118.118288 samples/sec	 accuracy=52.690341	 loss=2.056783	 lr=0.012670
Epoch[158] Batch [1699]/[3760]	Speed: 117.516035 samples/sec	 accuracy=52.695772	 loss=2.055905	 lr=0.012662
Epoch[158] Batch [1749]/[3760]	Speed: 117.347104 samples/sec	 accuracy=52.668750	 loss=2.057366	 lr=0.012653
Epoch[158] Batch [1799]/[3760]	Speed: 116.971404 samples/sec	 accuracy=52.636285	 loss=2.059152	 lr=0.012645
Epoch[158] Batch [1849]/[3760]	Speed: 117.681745 samples/sec	 accuracy=52.635135	 loss=2.058940	 lr=0.012636
Epoch[158] Batch [1899]/[3760]	Speed: 117.343891 samples/sec	 accuracy=52.622533	 loss=2.059091	 lr=0.012627
Epoch[158] Batch [1949]/[3760]	Speed: 117.820074 samples/sec	 accuracy=52.660256	 loss=2.057622	 lr=0.012619
Epoch[158] Batch [1999]/[3760]	Speed: 117.402721 samples/sec	 accuracy=52.659375	 loss=2.057145	 lr=0.012610
Epoch[158] Batch [2049]/[3760]	Speed: 117.901515 samples/sec	 accuracy=52.676829	 loss=2.056902	 lr=0.012602
Epoch[158] Batch [2099]/[3760]	Speed: 117.778434 samples/sec	 accuracy=52.687500	 loss=2.056948	 lr=0.012593
Epoch[158] Batch [2149]/[3760]	Speed: 117.703240 samples/sec	 accuracy=52.678052	 loss=2.057483	 lr=0.012585
Epoch[158] Batch [2199]/[3760]	Speed: 117.958058 samples/sec	 accuracy=52.649858	 loss=2.057902	 lr=0.012576
Epoch[158] Batch [2249]/[3760]	Speed: 116.886185 samples/sec	 accuracy=52.649306	 loss=2.057818	 lr=0.012568
Epoch[158] Batch [2299]/[3760]	Speed: 117.542057 samples/sec	 accuracy=52.604620	 loss=2.059630	 lr=0.012559
Epoch[158] Batch [2349]/[3760]	Speed: 117.435688 samples/sec	 accuracy=52.580452	 loss=2.059765	 lr=0.012550
Epoch[158] Batch [2399]/[3760]	Speed: 118.186419 samples/sec	 accuracy=52.551432	 loss=2.060702	 lr=0.012542
Epoch[158] Batch [2449]/[3760]	Speed: 117.484774 samples/sec	 accuracy=52.551658	 loss=2.061225	 lr=0.012533
Epoch[158] Batch [2499]/[3760]	Speed: 117.488794 samples/sec	 accuracy=52.559375	 loss=2.061347	 lr=0.012525
Epoch[158] Batch [2549]/[3760]	Speed: 117.179439 samples/sec	 accuracy=52.549632	 loss=2.061915	 lr=0.012516
Epoch[158] Batch [2599]/[3760]	Speed: 117.392967 samples/sec	 accuracy=52.554087	 loss=2.061761	 lr=0.012508
Epoch[158] Batch [2649]/[3760]	Speed: 117.359478 samples/sec	 accuracy=52.524764	 loss=2.063102	 lr=0.012499
Epoch[158] Batch [2699]/[3760]	Speed: 117.196233 samples/sec	 accuracy=52.504051	 loss=2.063820	 lr=0.012491
Epoch[158] Batch [2749]/[3760]	Speed: 117.806174 samples/sec	 accuracy=52.509091	 loss=2.063928	 lr=0.012482
Epoch[158] Batch [2799]/[3760]	Speed: 117.640012 samples/sec	 accuracy=52.516183	 loss=2.064004	 lr=0.012474
Epoch[158] Batch [2849]/[3760]	Speed: 117.808518 samples/sec	 accuracy=52.493969	 loss=2.064778	 lr=0.012465
Epoch[158] Batch [2899]/[3760]	Speed: 115.798919 samples/sec	 accuracy=52.491918	 loss=2.065180	 lr=0.012457
Epoch[158] Batch [2949]/[3760]	Speed: 117.448394 samples/sec	 accuracy=52.507945	 loss=2.065317	 lr=0.012448
Epoch[158] Batch [2999]/[3760]	Speed: 117.644682 samples/sec	 accuracy=52.504167	 loss=2.065790	 lr=0.012440
Epoch[158] Batch [3049]/[3760]	Speed: 117.502216 samples/sec	 accuracy=52.516906	 loss=2.065227	 lr=0.012431
Epoch[158] Batch [3099]/[3760]	Speed: 116.806674 samples/sec	 accuracy=52.482359	 loss=2.066247	 lr=0.012423
Epoch[158] Batch [3149]/[3760]	Speed: 117.512258 samples/sec	 accuracy=52.482639	 loss=2.066923	 lr=0.012414
Epoch[158] Batch [3199]/[3760]	Speed: 117.829401 samples/sec	 accuracy=52.497070	 loss=2.066755	 lr=0.012406
Epoch[158] Batch [3249]/[3760]	Speed: 116.974627 samples/sec	 accuracy=52.510096	 loss=2.066241	 lr=0.012397
Epoch[158] Batch [3299]/[3760]	Speed: 117.310526 samples/sec	 accuracy=52.489110	 loss=2.066119	 lr=0.012389
Epoch[158] Batch [3349]/[3760]	Speed: 117.883277 samples/sec	 accuracy=52.480410	 loss=2.066515	 lr=0.012380
Epoch[158] Batch [3399]/[3760]	Speed: 117.553791 samples/sec	 accuracy=52.499081	 loss=2.065668	 lr=0.012372
Epoch[158] Batch [3449]/[3760]	Speed: 117.284276 samples/sec	 accuracy=52.492301	 loss=2.065742	 lr=0.012363
Epoch[158] Batch [3499]/[3760]	Speed: 117.261622 samples/sec	 accuracy=52.462500	 loss=2.067027	 lr=0.012355
Epoch[158] Batch [3549]/[3760]	Speed: 116.469726 samples/sec	 accuracy=52.432658	 loss=2.067842	 lr=0.012346
Epoch[158] Batch [3599]/[3760]	Speed: 117.802997 samples/sec	 accuracy=52.425781	 loss=2.068208	 lr=0.012338
Epoch[158] Batch [3649]/[3760]	Speed: 118.002376 samples/sec	 accuracy=52.442637	 loss=2.068017	 lr=0.012329
Epoch[158] Batch [3699]/[3760]	Speed: 117.935271 samples/sec	 accuracy=52.435389	 loss=2.068486	 lr=0.012321
Epoch[158] Batch [3749]/[3760]	Speed: 120.830019 samples/sec	 accuracy=52.445000	 loss=2.068338	 lr=0.012312
Batch [0049]/[0303]: acc-top1=53.687500 acc-top5=75.125000
Batch [0099]/[0303]: acc-top1=49.515625 acc-top5=73.687500
Batch [0149]/[0303]: acc-top1=50.583333 acc-top5=74.416667
Batch [0199]/[0303]: acc-top1=53.414062 acc-top5=76.539062
Batch [0249]/[0303]: acc-top1=52.637500 acc-top5=76.206250
Batch [0299]/[0303]: acc-top1=52.187500 acc-top5=76.197917
[Epoch 158] training: accuracy=52.447640	 loss=2.068178
[Epoch 158] speed: 116 samples/sec	time cost: 2173.347190
[Epoch 158] validation: acc-top1=52.140058 acc-top5=76.139645 loss=2.220627
Epoch[159] Batch [0049]/[3760]	Speed: 87.269676 samples/sec	 accuracy=53.687500	 loss=2.016698	 lr=0.012302
Epoch[159] Batch [0099]/[3760]	Speed: 118.169261 samples/sec	 accuracy=52.875000	 loss=2.041492	 lr=0.012293
Epoch[159] Batch [0149]/[3760]	Speed: 117.095859 samples/sec	 accuracy=52.906250	 loss=2.028191	 lr=0.012285
Epoch[159] Batch [0199]/[3760]	Speed: 117.047188 samples/sec	 accuracy=53.492188	 loss=2.010561	 lr=0.012277
Epoch[159] Batch [0249]/[3760]	Speed: 117.437577 samples/sec	 accuracy=53.087500	 loss=2.026704	 lr=0.012268
Epoch[159] Batch [0299]/[3760]	Speed: 117.233816 samples/sec	 accuracy=53.223958	 loss=2.027839	 lr=0.012260
Epoch[159] Batch [0349]/[3760]	Speed: 117.653127 samples/sec	 accuracy=53.000000	 loss=2.037574	 lr=0.012251
Epoch[159] Batch [0399]/[3760]	Speed: 117.218591 samples/sec	 accuracy=52.949219	 loss=2.035694	 lr=0.012243
Epoch[159] Batch [0449]/[3760]	Speed: 115.520543 samples/sec	 accuracy=52.930556	 loss=2.037282	 lr=0.012234
Epoch[159] Batch [0499]/[3760]	Speed: 117.451050 samples/sec	 accuracy=52.921875	 loss=2.037266	 lr=0.012226
Epoch[159] Batch [0549]/[3760]	Speed: 117.503314 samples/sec	 accuracy=52.877841	 loss=2.037902	 lr=0.012217
Epoch[159] Batch [0599]/[3760]	Speed: 117.238380 samples/sec	 accuracy=52.885417	 loss=2.040466	 lr=0.012209
Epoch[159] Batch [0649]/[3760]	Speed: 116.579570 samples/sec	 accuracy=52.923077	 loss=2.040410	 lr=0.012200
Epoch[159] Batch [0699]/[3760]	Speed: 117.440392 samples/sec	 accuracy=53.000000	 loss=2.037806	 lr=0.012192
Epoch[159] Batch [0749]/[3760]	Speed: 117.235615 samples/sec	 accuracy=52.933333	 loss=2.042022	 lr=0.012184
Epoch[159] Batch [0799]/[3760]	Speed: 117.619546 samples/sec	 accuracy=53.001953	 loss=2.040278	 lr=0.012175
Epoch[159] Batch [0849]/[3760]	Speed: 116.686086 samples/sec	 accuracy=53.023897	 loss=2.038860	 lr=0.012167
Epoch[159] Batch [0899]/[3760]	Speed: 116.924506 samples/sec	 accuracy=53.013889	 loss=2.039810	 lr=0.012158
Epoch[159] Batch [0949]/[3760]	Speed: 117.683906 samples/sec	 accuracy=53.003289	 loss=2.039107	 lr=0.012150
Epoch[159] Batch [0999]/[3760]	Speed: 117.261315 samples/sec	 accuracy=52.998437	 loss=2.038036	 lr=0.012141
Epoch[159] Batch [1049]/[3760]	Speed: 117.235942 samples/sec	 accuracy=53.055060	 loss=2.036384	 lr=0.012133
Epoch[159] Batch [1099]/[3760]	Speed: 117.175958 samples/sec	 accuracy=53.063920	 loss=2.036463	 lr=0.012125
Epoch[159] Batch [1149]/[3760]	Speed: 117.227173 samples/sec	 accuracy=53.043478	 loss=2.037688	 lr=0.012116
Epoch[159] Batch [1199]/[3760]	Speed: 116.465264 samples/sec	 accuracy=52.994792	 loss=2.040459	 lr=0.012108
Epoch[159] Batch [1249]/[3760]	Speed: 117.613710 samples/sec	 accuracy=52.987500	 loss=2.039747	 lr=0.012099
Epoch[159] Batch [1299]/[3760]	Speed: 117.398566 samples/sec	 accuracy=53.024038	 loss=2.038679	 lr=0.012091
Epoch[159] Batch [1349]/[3760]	Speed: 118.448266 samples/sec	 accuracy=53.030093	 loss=2.038918	 lr=0.012083
Epoch[159] Batch [1399]/[3760]	Speed: 118.020311 samples/sec	 accuracy=52.997768	 loss=2.041222	 lr=0.012074
Epoch[159] Batch [1449]/[3760]	Speed: 117.338017 samples/sec	 accuracy=52.989224	 loss=2.041312	 lr=0.012066
Epoch[159] Batch [1499]/[3760]	Speed: 116.306832 samples/sec	 accuracy=52.973958	 loss=2.042660	 lr=0.012057
Epoch[159] Batch [1549]/[3760]	Speed: 117.625030 samples/sec	 accuracy=52.942540	 loss=2.045061	 lr=0.012049
Epoch[159] Batch [1599]/[3760]	Speed: 117.777017 samples/sec	 accuracy=52.901367	 loss=2.046376	 lr=0.012041
Epoch[159] Batch [1649]/[3760]	Speed: 117.813662 samples/sec	 accuracy=52.910985	 loss=2.046782	 lr=0.012032
Epoch[159] Batch [1699]/[3760]	Speed: 117.431642 samples/sec	 accuracy=52.897059	 loss=2.047701	 lr=0.012024
Epoch[159] Batch [1749]/[3760]	Speed: 117.775475 samples/sec	 accuracy=52.849107	 loss=2.049181	 lr=0.012015
Epoch[159] Batch [1799]/[3760]	Speed: 117.235124 samples/sec	 accuracy=52.891493	 loss=2.048382	 lr=0.012007
Epoch[159] Batch [1849]/[3760]	Speed: 115.809169 samples/sec	 accuracy=52.849662	 loss=2.048575	 lr=0.011999
Epoch[159] Batch [1899]/[3760]	Speed: 117.514092 samples/sec	 accuracy=52.838816	 loss=2.048696	 lr=0.011990
Epoch[159] Batch [1949]/[3760]	Speed: 117.457570 samples/sec	 accuracy=52.835737	 loss=2.048453	 lr=0.011982
Epoch[159] Batch [1999]/[3760]	Speed: 117.157494 samples/sec	 accuracy=52.811719	 loss=2.050116	 lr=0.011973
Epoch[159] Batch [2049]/[3760]	Speed: 117.546720 samples/sec	 accuracy=52.849848	 loss=2.048331	 lr=0.011965
Epoch[159] Batch [2099]/[3760]	Speed: 117.688246 samples/sec	 accuracy=52.849702	 loss=2.048801	 lr=0.011957
Epoch[159] Batch [2149]/[3760]	Speed: 118.112022 samples/sec	 accuracy=52.864099	 loss=2.049084	 lr=0.011948
Epoch[159] Batch [2199]/[3760]	Speed: 115.427917 samples/sec	 accuracy=52.870739	 loss=2.048598	 lr=0.011940
Epoch[159] Batch [2249]/[3760]	Speed: 117.463687 samples/sec	 accuracy=52.868056	 loss=2.048753	 lr=0.011932
Epoch[159] Batch [2299]/[3760]	Speed: 117.548669 samples/sec	 accuracy=52.875679	 loss=2.047897	 lr=0.011923
Epoch[159] Batch [2349]/[3760]	Speed: 117.913181 samples/sec	 accuracy=52.890293	 loss=2.046748	 lr=0.011915
Epoch[159] Batch [2399]/[3760]	Speed: 118.324364 samples/sec	 accuracy=52.906901	 loss=2.046292	 lr=0.011907
Epoch[159] Batch [2449]/[3760]	Speed: 117.745829 samples/sec	 accuracy=52.901148	 loss=2.045983	 lr=0.011898
Epoch[159] Batch [2499]/[3760]	Speed: 117.600228 samples/sec	 accuracy=52.891875	 loss=2.046316	 lr=0.011890
Epoch[159] Batch [2549]/[3760]	Speed: 118.159253 samples/sec	 accuracy=52.848039	 loss=2.048761	 lr=0.011881
Epoch[159] Batch [2599]/[3760]	Speed: 117.332615 samples/sec	 accuracy=52.838942	 loss=2.048982	 lr=0.011873
Epoch[159] Batch [2649]/[3760]	Speed: 118.403038 samples/sec	 accuracy=52.855542	 loss=2.047819	 lr=0.011865
Epoch[159] Batch [2699]/[3760]	Speed: 117.795458 samples/sec	 accuracy=52.861690	 loss=2.047646	 lr=0.011856
Epoch[159] Batch [2749]/[3760]	Speed: 117.772584 samples/sec	 accuracy=52.850000	 loss=2.048567	 lr=0.011848
Epoch[159] Batch [2799]/[3760]	Speed: 118.241684 samples/sec	 accuracy=52.848772	 loss=2.048937	 lr=0.011840
Epoch[159] Batch [2849]/[3760]	Speed: 117.511728 samples/sec	 accuracy=52.838816	 loss=2.049398	 lr=0.011831
Epoch[159] Batch [2899]/[3760]	Speed: 117.099027 samples/sec	 accuracy=52.802263	 loss=2.051318	 lr=0.011823
Epoch[159] Batch [2949]/[3760]	Speed: 117.723782 samples/sec	 accuracy=52.791843	 loss=2.051292	 lr=0.011815
Epoch[159] Batch [2999]/[3760]	Speed: 117.766986 samples/sec	 accuracy=52.804688	 loss=2.050278	 lr=0.011806
Epoch[159] Batch [3049]/[3760]	Speed: 117.603712 samples/sec	 accuracy=52.793545	 loss=2.051258	 lr=0.011798
Epoch[159] Batch [3099]/[3760]	Speed: 117.257645 samples/sec	 accuracy=52.783770	 loss=2.051318	 lr=0.011790
Epoch[159] Batch [3149]/[3760]	Speed: 117.449054 samples/sec	 accuracy=52.775298	 loss=2.051899	 lr=0.011782
Epoch[159] Batch [3199]/[3760]	Speed: 115.284532 samples/sec	 accuracy=52.776855	 loss=2.051946	 lr=0.011773
Epoch[159] Batch [3249]/[3760]	Speed: 118.484921 samples/sec	 accuracy=52.798558	 loss=2.051741	 lr=0.011765
Epoch[159] Batch [3299]/[3760]	Speed: 117.838854 samples/sec	 accuracy=52.797822	 loss=2.052223	 lr=0.011757
Epoch[159] Batch [3349]/[3760]	Speed: 117.746968 samples/sec	 accuracy=52.784981	 loss=2.052150	 lr=0.011748
Epoch[159] Batch [3399]/[3760]	Speed: 117.939352 samples/sec	 accuracy=52.791360	 loss=2.052658	 lr=0.011740
Epoch[159] Batch [3449]/[3760]	Speed: 118.233233 samples/sec	 accuracy=52.795743	 loss=2.052301	 lr=0.011732
Epoch[159] Batch [3499]/[3760]	Speed: 117.464157 samples/sec	 accuracy=52.791071	 loss=2.052578	 lr=0.011723
Epoch[159] Batch [3549]/[3760]	Speed: 117.410432 samples/sec	 accuracy=52.813820	 loss=2.051888	 lr=0.011715
Epoch[159] Batch [3599]/[3760]	Speed: 118.155186 samples/sec	 accuracy=52.796875	 loss=2.052735	 lr=0.011707
Epoch[159] Batch [3649]/[3760]	Speed: 117.672343 samples/sec	 accuracy=52.793664	 loss=2.052914	 lr=0.011698
Epoch[159] Batch [3699]/[3760]	Speed: 117.456513 samples/sec	 accuracy=52.792652	 loss=2.053278	 lr=0.011690
Epoch[159] Batch [3749]/[3760]	Speed: 120.715148 samples/sec	 accuracy=52.813333	 loss=2.052791	 lr=0.011682
Batch [0049]/[0303]: acc-top1=52.531250 acc-top5=75.000000
Batch [0099]/[0303]: acc-top1=49.406250 acc-top5=74.109375
Batch [0149]/[0303]: acc-top1=50.760417 acc-top5=74.645833
Batch [0199]/[0303]: acc-top1=53.515625 acc-top5=76.476562
Batch [0249]/[0303]: acc-top1=52.881250 acc-top5=76.225000
Batch [0299]/[0303]: acc-top1=52.390625 acc-top5=76.057292
[Epoch 159] training: accuracy=52.819564	 loss=2.052533
[Epoch 159] speed: 116 samples/sec	time cost: 2174.697634
[Epoch 159] validation: acc-top1=52.310231 acc-top5=76.036510 loss=2.213319
Epoch[160] Batch [0049]/[3759]	Speed: 86.863827 samples/sec	 accuracy=53.531250	 loss=2.034313	 lr=0.011672
Epoch[160] Batch [0099]/[3759]	Speed: 116.954554 samples/sec	 accuracy=54.046875	 loss=1.999459	 lr=0.011664
Epoch[160] Batch [0149]/[3759]	Speed: 117.548843 samples/sec	 accuracy=53.843750	 loss=1.999180	 lr=0.011655
Epoch[160] Batch [0199]/[3759]	Speed: 116.803450 samples/sec	 accuracy=54.070312	 loss=1.993704	 lr=0.011647
Epoch[160] Batch [0249]/[3759]	Speed: 117.951504 samples/sec	 accuracy=54.150000	 loss=1.997503	 lr=0.011639
Epoch[160] Batch [0299]/[3759]	Speed: 117.114007 samples/sec	 accuracy=53.937500	 loss=2.004409	 lr=0.011631
Epoch[160] Batch [0349]/[3759]	Speed: 117.997428 samples/sec	 accuracy=53.977679	 loss=2.005759	 lr=0.011622
Epoch[160] Batch [0399]/[3759]	Speed: 117.376897 samples/sec	 accuracy=54.257812	 loss=1.998046	 lr=0.011614
Epoch[160] Batch [0449]/[3759]	Speed: 117.282815 samples/sec	 accuracy=54.125000	 loss=2.000837	 lr=0.011606
Epoch[160] Batch [0499]/[3759]	Speed: 117.285326 samples/sec	 accuracy=53.887500	 loss=2.012066	 lr=0.011598
Epoch[160] Batch [0549]/[3759]	Speed: 117.327481 samples/sec	 accuracy=53.821023	 loss=2.014867	 lr=0.011589
Epoch[160] Batch [0599]/[3759]	Speed: 117.656959 samples/sec	 accuracy=53.809896	 loss=2.010639	 lr=0.011581
Epoch[160] Batch [0649]/[3759]	Speed: 117.112802 samples/sec	 accuracy=53.829327	 loss=2.008399	 lr=0.011573
Epoch[160] Batch [0699]/[3759]	Speed: 118.272798 samples/sec	 accuracy=53.785714	 loss=2.008774	 lr=0.011565
Epoch[160] Batch [0749]/[3759]	Speed: 115.566444 samples/sec	 accuracy=53.852083	 loss=2.005750	 lr=0.011556
Epoch[160] Batch [0799]/[3759]	Speed: 118.147984 samples/sec	 accuracy=53.855469	 loss=2.003537	 lr=0.011548
Epoch[160] Batch [0849]/[3759]	Speed: 117.796285 samples/sec	 accuracy=53.803309	 loss=2.004263	 lr=0.011540
Epoch[160] Batch [0899]/[3759]	Speed: 117.559916 samples/sec	 accuracy=53.857639	 loss=2.002597	 lr=0.011532
Epoch[160] Batch [0949]/[3759]	Speed: 116.996107 samples/sec	 accuracy=53.748355	 loss=2.006567	 lr=0.011523
Epoch[160] Batch [0999]/[3759]	Speed: 117.568914 samples/sec	 accuracy=53.748438	 loss=2.008672	 lr=0.011515
Epoch[160] Batch [1049]/[3759]	Speed: 117.739073 samples/sec	 accuracy=53.747024	 loss=2.007868	 lr=0.011507
Epoch[160] Batch [1099]/[3759]	Speed: 116.849842 samples/sec	 accuracy=53.710227	 loss=2.007689	 lr=0.011499
Epoch[160] Batch [1149]/[3759]	Speed: 118.375736 samples/sec	 accuracy=53.672554	 loss=2.009010	 lr=0.011490
Epoch[160] Batch [1199]/[3759]	Speed: 116.714470 samples/sec	 accuracy=53.669271	 loss=2.010637	 lr=0.011482
Epoch[160] Batch [1249]/[3759]	Speed: 117.569079 samples/sec	 accuracy=53.668750	 loss=2.011507	 lr=0.011474
Epoch[160] Batch [1299]/[3759]	Speed: 117.711956 samples/sec	 accuracy=53.703125	 loss=2.011269	 lr=0.011466
Epoch[160] Batch [1349]/[3759]	Speed: 117.182820 samples/sec	 accuracy=53.686343	 loss=2.011283	 lr=0.011457
Epoch[160] Batch [1399]/[3759]	Speed: 117.318639 samples/sec	 accuracy=53.700893	 loss=2.010967	 lr=0.011449
Epoch[160] Batch [1449]/[3759]	Speed: 117.233898 samples/sec	 accuracy=53.704741	 loss=2.011682	 lr=0.011441
Epoch[160] Batch [1499]/[3759]	Speed: 117.539954 samples/sec	 accuracy=53.713542	 loss=2.012305	 lr=0.011433
Epoch[160] Batch [1549]/[3759]	Speed: 117.271579 samples/sec	 accuracy=53.723790	 loss=2.010917	 lr=0.011425
Epoch[160] Batch [1599]/[3759]	Speed: 118.236088 samples/sec	 accuracy=53.681641	 loss=2.012420	 lr=0.011416
Epoch[160] Batch [1649]/[3759]	Speed: 117.926663 samples/sec	 accuracy=53.669508	 loss=2.013601	 lr=0.011408
Epoch[160] Batch [1699]/[3759]	Speed: 117.454312 samples/sec	 accuracy=53.663603	 loss=2.014035	 lr=0.011400
Epoch[160] Batch [1749]/[3759]	Speed: 117.474563 samples/sec	 accuracy=53.688393	 loss=2.013486	 lr=0.011392
Epoch[160] Batch [1799]/[3759]	Speed: 117.380769 samples/sec	 accuracy=53.717882	 loss=2.012223	 lr=0.011384
Epoch[160] Batch [1849]/[3759]	Speed: 117.273307 samples/sec	 accuracy=53.713682	 loss=2.012965	 lr=0.011375
Epoch[160] Batch [1899]/[3759]	Speed: 117.427313 samples/sec	 accuracy=53.717928	 loss=2.012888	 lr=0.011367
Epoch[160] Batch [1949]/[3759]	Speed: 117.750179 samples/sec	 accuracy=53.673077	 loss=2.014537	 lr=0.011359
Epoch[160] Batch [1999]/[3759]	Speed: 117.718999 samples/sec	 accuracy=53.675781	 loss=2.015157	 lr=0.011351
Epoch[160] Batch [2049]/[3759]	Speed: 117.633915 samples/sec	 accuracy=53.658537	 loss=2.015597	 lr=0.011343
Epoch[160] Batch [2099]/[3759]	Speed: 115.620541 samples/sec	 accuracy=53.640625	 loss=2.016273	 lr=0.011335
Epoch[160] Batch [2149]/[3759]	Speed: 117.747862 samples/sec	 accuracy=53.634448	 loss=2.016338	 lr=0.011326
Epoch[160] Batch [2199]/[3759]	Speed: 117.433448 samples/sec	 accuracy=53.699574	 loss=2.014027	 lr=0.011318
Epoch[160] Batch [2249]/[3759]	Speed: 118.044372 samples/sec	 accuracy=53.692361	 loss=2.014202	 lr=0.011310
Epoch[160] Batch [2299]/[3759]	Speed: 117.792016 samples/sec	 accuracy=53.713995	 loss=2.013356	 lr=0.011302
Epoch[160] Batch [2349]/[3759]	Speed: 118.105410 samples/sec	 accuracy=53.672207	 loss=2.014395	 lr=0.011294
Epoch[160] Batch [2399]/[3759]	Speed: 117.922718 samples/sec	 accuracy=53.655599	 loss=2.016184	 lr=0.011286
Epoch[160] Batch [2449]/[3759]	Speed: 117.754951 samples/sec	 accuracy=53.648597	 loss=2.017059	 lr=0.011277
Epoch[160] Batch [2499]/[3759]	Speed: 117.483529 samples/sec	 accuracy=53.655000	 loss=2.017337	 lr=0.011269
Epoch[160] Batch [2549]/[3759]	Speed: 117.023217 samples/sec	 accuracy=53.688725	 loss=2.015592	 lr=0.011261
Epoch[160] Batch [2599]/[3759]	Speed: 118.090585 samples/sec	 accuracy=53.685096	 loss=2.015356	 lr=0.011253
Epoch[160] Batch [2649]/[3759]	Speed: 117.502798 samples/sec	 accuracy=53.664505	 loss=2.015696	 lr=0.011245
Epoch[160] Batch [2699]/[3759]	Speed: 117.634268 samples/sec	 accuracy=53.650463	 loss=2.016542	 lr=0.011237
Epoch[160] Batch [2749]/[3759]	Speed: 117.375777 samples/sec	 accuracy=53.617614	 loss=2.017784	 lr=0.011228
Epoch[160] Batch [2799]/[3759]	Speed: 117.515549 samples/sec	 accuracy=53.609375	 loss=2.018398	 lr=0.011220
Epoch[160] Batch [2849]/[3759]	Speed: 117.625628 samples/sec	 accuracy=53.589364	 loss=2.018531	 lr=0.011212
Epoch[160] Batch [2899]/[3759]	Speed: 117.636328 samples/sec	 accuracy=53.598599	 loss=2.017928	 lr=0.011204
Epoch[160] Batch [2949]/[3759]	Speed: 117.619445 samples/sec	 accuracy=53.571504	 loss=2.018888	 lr=0.011196
Epoch[160] Batch [2999]/[3759]	Speed: 117.146223 samples/sec	 accuracy=53.581250	 loss=2.018730	 lr=0.011188
Epoch[160] Batch [3049]/[3759]	Speed: 117.472199 samples/sec	 accuracy=53.588627	 loss=2.018435	 lr=0.011180
Epoch[160] Batch [3099]/[3759]	Speed: 117.677011 samples/sec	 accuracy=53.572581	 loss=2.018876	 lr=0.011172
Epoch[160] Batch [3149]/[3759]	Speed: 116.699155 samples/sec	 accuracy=53.573413	 loss=2.019385	 lr=0.011163
Epoch[160] Batch [3199]/[3759]	Speed: 117.706245 samples/sec	 accuracy=53.542969	 loss=2.021041	 lr=0.011155
Epoch[160] Batch [3249]/[3759]	Speed: 117.375038 samples/sec	 accuracy=53.518269	 loss=2.022136	 lr=0.011147
Epoch[160] Batch [3299]/[3759]	Speed: 117.885654 samples/sec	 accuracy=53.536458	 loss=2.021510	 lr=0.011139
Epoch[160] Batch [3349]/[3759]	Speed: 117.587747 samples/sec	 accuracy=53.536847	 loss=2.021979	 lr=0.011131
Epoch[160] Batch [3399]/[3759]	Speed: 117.743110 samples/sec	 accuracy=53.528033	 loss=2.021773	 lr=0.011123
Epoch[160] Batch [3449]/[3759]	Speed: 117.304646 samples/sec	 accuracy=53.527174	 loss=2.022054	 lr=0.011115
Epoch[160] Batch [3499]/[3759]	Speed: 115.851430 samples/sec	 accuracy=53.521875	 loss=2.021612	 lr=0.011107
Epoch[160] Batch [3549]/[3759]	Speed: 117.393165 samples/sec	 accuracy=53.522447	 loss=2.022207	 lr=0.011098
Epoch[160] Batch [3599]/[3759]	Speed: 117.101708 samples/sec	 accuracy=53.502604	 loss=2.023492	 lr=0.011090
Epoch[160] Batch [3649]/[3759]	Speed: 118.398971 samples/sec	 accuracy=53.492723	 loss=2.023497	 lr=0.011082
Epoch[160] Batch [3699]/[3759]	Speed: 117.575664 samples/sec	 accuracy=53.484797	 loss=2.023919	 lr=0.011074
Epoch[160] Batch [3749]/[3759]	Speed: 121.014141 samples/sec	 accuracy=53.485000	 loss=2.024241	 lr=0.011066
Batch [0049]/[0303]: acc-top1=52.937500 acc-top5=75.531250
Batch [0099]/[0303]: acc-top1=50.765625 acc-top5=74.656250
Batch [0149]/[0303]: acc-top1=52.072917 acc-top5=75.552083
Batch [0199]/[0303]: acc-top1=54.343750 acc-top5=77.304688
Batch [0249]/[0303]: acc-top1=53.475000 acc-top5=77.143750
Batch [0299]/[0303]: acc-top1=53.109375 acc-top5=77.088542
[Epoch 160] training: accuracy=53.477903	 loss=2.024371
[Epoch 160] speed: 116 samples/sec	time cost: 2173.306725
[Epoch 160] validation: acc-top1=53.063119 acc-top5=77.098804 loss=2.096457
Epoch[161] Batch [0049]/[3760]	Speed: 88.404642 samples/sec	 accuracy=55.468750	 loss=1.959799	 lr=0.011057
Epoch[161] Batch [0099]/[3760]	Speed: 117.643135 samples/sec	 accuracy=53.750000	 loss=1.997870	 lr=0.011048
Epoch[161] Batch [0149]/[3760]	Speed: 117.900411 samples/sec	 accuracy=53.937500	 loss=2.000034	 lr=0.011040
Epoch[161] Batch [0199]/[3760]	Speed: 117.392971 samples/sec	 accuracy=53.796875	 loss=2.008049	 lr=0.011032
Epoch[161] Batch [0249]/[3760]	Speed: 117.971438 samples/sec	 accuracy=53.725000	 loss=2.012168	 lr=0.011024
Epoch[161] Batch [0299]/[3760]	Speed: 118.406328 samples/sec	 accuracy=53.770833	 loss=2.007725	 lr=0.011016
Epoch[161] Batch [0349]/[3760]	Speed: 116.710795 samples/sec	 accuracy=53.665179	 loss=2.006873	 lr=0.011008
Epoch[161] Batch [0399]/[3760]	Speed: 117.863306 samples/sec	 accuracy=53.839844	 loss=2.003135	 lr=0.011000
Epoch[161] Batch [0449]/[3760]	Speed: 117.651516 samples/sec	 accuracy=53.843750	 loss=2.002497	 lr=0.010992
Epoch[161] Batch [0499]/[3760]	Speed: 117.236774 samples/sec	 accuracy=53.812500	 loss=2.005637	 lr=0.010984
Epoch[161] Batch [0549]/[3760]	Speed: 117.121683 samples/sec	 accuracy=53.863636	 loss=2.002377	 lr=0.010976
Epoch[161] Batch [0599]/[3760]	Speed: 117.372402 samples/sec	 accuracy=53.940104	 loss=1.996949	 lr=0.010968
Epoch[161] Batch [0649]/[3760]	Speed: 117.327879 samples/sec	 accuracy=54.036058	 loss=1.995506	 lr=0.010960
Epoch[161] Batch [0699]/[3760]	Speed: 117.081161 samples/sec	 accuracy=53.979911	 loss=1.999386	 lr=0.010952
Epoch[161] Batch [0749]/[3760]	Speed: 117.537675 samples/sec	 accuracy=53.987500	 loss=1.999385	 lr=0.010944
Epoch[161] Batch [0799]/[3760]	Speed: 117.984167 samples/sec	 accuracy=54.068359	 loss=1.998516	 lr=0.010935
Epoch[161] Batch [0849]/[3760]	Speed: 117.851438 samples/sec	 accuracy=54.112132	 loss=1.995362	 lr=0.010927
Epoch[161] Batch [0899]/[3760]	Speed: 117.830414 samples/sec	 accuracy=54.163194	 loss=1.994365	 lr=0.010919
Epoch[161] Batch [0949]/[3760]	Speed: 117.098357 samples/sec	 accuracy=54.146382	 loss=1.996382	 lr=0.010911
Epoch[161] Batch [0999]/[3760]	Speed: 117.297650 samples/sec	 accuracy=54.190625	 loss=1.994767	 lr=0.010903
Epoch[161] Batch [1049]/[3760]	Speed: 117.701803 samples/sec	 accuracy=54.212798	 loss=1.995739	 lr=0.010895
Epoch[161] Batch [1099]/[3760]	Speed: 117.622250 samples/sec	 accuracy=54.198864	 loss=1.994787	 lr=0.010887
Epoch[161] Batch [1149]/[3760]	Speed: 115.360546 samples/sec	 accuracy=54.201087	 loss=1.995455	 lr=0.010879
Epoch[161] Batch [1199]/[3760]	Speed: 117.135780 samples/sec	 accuracy=54.248698	 loss=1.993278	 lr=0.010871
Epoch[161] Batch [1249]/[3760]	Speed: 116.737945 samples/sec	 accuracy=54.267500	 loss=1.993819	 lr=0.010863
Epoch[161] Batch [1299]/[3760]	Speed: 117.601272 samples/sec	 accuracy=54.274038	 loss=1.992229	 lr=0.010855
Epoch[161] Batch [1349]/[3760]	Speed: 117.429266 samples/sec	 accuracy=54.297454	 loss=1.991270	 lr=0.010847
Epoch[161] Batch [1399]/[3760]	Speed: 117.382229 samples/sec	 accuracy=54.299107	 loss=1.991383	 lr=0.010839
Epoch[161] Batch [1449]/[3760]	Speed: 117.769560 samples/sec	 accuracy=54.273707	 loss=1.992062	 lr=0.010831
Epoch[161] Batch [1499]/[3760]	Speed: 117.483942 samples/sec	 accuracy=54.251042	 loss=1.991954	 lr=0.010823
Epoch[161] Batch [1549]/[3760]	Speed: 117.877083 samples/sec	 accuracy=54.263105	 loss=1.991987	 lr=0.010815
Epoch[161] Batch [1599]/[3760]	Speed: 118.140184 samples/sec	 accuracy=54.282227	 loss=1.993061	 lr=0.010807
Epoch[161] Batch [1649]/[3760]	Speed: 117.054578 samples/sec	 accuracy=54.307765	 loss=1.991555	 lr=0.010799
Epoch[161] Batch [1699]/[3760]	Speed: 117.795320 samples/sec	 accuracy=54.264706	 loss=1.993332	 lr=0.010791
Epoch[161] Batch [1749]/[3760]	Speed: 117.375221 samples/sec	 accuracy=54.259821	 loss=1.994588	 lr=0.010783
Epoch[161] Batch [1799]/[3760]	Speed: 117.414124 samples/sec	 accuracy=54.315104	 loss=1.992994	 lr=0.010775
Epoch[161] Batch [1849]/[3760]	Speed: 117.154969 samples/sec	 accuracy=54.311655	 loss=1.992586	 lr=0.010767
Epoch[161] Batch [1899]/[3760]	Speed: 117.360976 samples/sec	 accuracy=54.279605	 loss=1.993741	 lr=0.010759
Epoch[161] Batch [1949]/[3760]	Speed: 117.634961 samples/sec	 accuracy=54.254006	 loss=1.993896	 lr=0.010751
Epoch[161] Batch [1999]/[3760]	Speed: 116.272491 samples/sec	 accuracy=54.238281	 loss=1.995214	 lr=0.010743
Epoch[161] Batch [2049]/[3760]	Speed: 117.448871 samples/sec	 accuracy=54.231707	 loss=1.994961	 lr=0.010735
Epoch[161] Batch [2099]/[3760]	Speed: 117.926570 samples/sec	 accuracy=54.242560	 loss=1.995058	 lr=0.010727
Epoch[161] Batch [2149]/[3760]	Speed: 117.090378 samples/sec	 accuracy=54.241279	 loss=1.995164	 lr=0.010719
Epoch[161] Batch [2199]/[3760]	Speed: 117.291150 samples/sec	 accuracy=54.218750	 loss=1.995779	 lr=0.010711
Epoch[161] Batch [2249]/[3760]	Speed: 117.698747 samples/sec	 accuracy=54.196528	 loss=1.996630	 lr=0.010703
Epoch[161] Batch [2299]/[3760]	Speed: 117.311549 samples/sec	 accuracy=54.180707	 loss=1.997450	 lr=0.010695
Epoch[161] Batch [2349]/[3760]	Speed: 117.447577 samples/sec	 accuracy=54.168218	 loss=1.997795	 lr=0.010687
Epoch[161] Batch [2399]/[3760]	Speed: 116.818883 samples/sec	 accuracy=54.172526	 loss=1.997501	 lr=0.010679
Epoch[161] Batch [2449]/[3760]	Speed: 117.607957 samples/sec	 accuracy=54.156888	 loss=1.997110	 lr=0.010671
Epoch[161] Batch [2499]/[3760]	Speed: 117.294903 samples/sec	 accuracy=54.122500	 loss=1.997728	 lr=0.010663
Epoch[161] Batch [2549]/[3760]	Speed: 115.737147 samples/sec	 accuracy=54.098039	 loss=1.998169	 lr=0.010655
Epoch[161] Batch [2599]/[3760]	Speed: 117.160838 samples/sec	 accuracy=54.116587	 loss=1.997220	 lr=0.010647
Epoch[161] Batch [2649]/[3760]	Speed: 117.026876 samples/sec	 accuracy=54.093160	 loss=1.998341	 lr=0.010639
Epoch[161] Batch [2699]/[3760]	Speed: 117.192549 samples/sec	 accuracy=54.077546	 loss=1.998774	 lr=0.010631
Epoch[161] Batch [2749]/[3760]	Speed: 118.056619 samples/sec	 accuracy=54.069886	 loss=1.999121	 lr=0.010624
Epoch[161] Batch [2799]/[3760]	Speed: 117.590753 samples/sec	 accuracy=54.055246	 loss=1.999431	 lr=0.010616
Epoch[161] Batch [2849]/[3760]	Speed: 117.870210 samples/sec	 accuracy=54.037829	 loss=2.000576	 lr=0.010608
Epoch[161] Batch [2899]/[3760]	Speed: 117.377649 samples/sec	 accuracy=54.043642	 loss=2.000402	 lr=0.010600
Epoch[161] Batch [2949]/[3760]	Speed: 116.890882 samples/sec	 accuracy=54.040784	 loss=2.000804	 lr=0.010592
Epoch[161] Batch [2999]/[3760]	Speed: 117.727791 samples/sec	 accuracy=54.042708	 loss=2.000221	 lr=0.010584
Epoch[161] Batch [3049]/[3760]	Speed: 117.820519 samples/sec	 accuracy=54.046619	 loss=2.000306	 lr=0.010576
Epoch[161] Batch [3099]/[3760]	Speed: 116.842248 samples/sec	 accuracy=54.045867	 loss=2.000521	 lr=0.010568
Epoch[161] Batch [3149]/[3760]	Speed: 117.552255 samples/sec	 accuracy=54.018849	 loss=2.001542	 lr=0.010560
Epoch[161] Batch [3199]/[3760]	Speed: 117.918398 samples/sec	 accuracy=53.983398	 loss=2.002651	 lr=0.010552
Epoch[161] Batch [3249]/[3760]	Speed: 117.398541 samples/sec	 accuracy=53.964423	 loss=2.003791	 lr=0.010544
Epoch[161] Batch [3299]/[3760]	Speed: 118.198985 samples/sec	 accuracy=53.959280	 loss=2.004107	 lr=0.010536
Epoch[161] Batch [3349]/[3760]	Speed: 117.987099 samples/sec	 accuracy=53.982743	 loss=2.003559	 lr=0.010528
Epoch[161] Batch [3399]/[3760]	Speed: 117.072895 samples/sec	 accuracy=53.998162	 loss=2.003107	 lr=0.010520
Epoch[161] Batch [3449]/[3760]	Speed: 118.053261 samples/sec	 accuracy=54.013587	 loss=2.002622	 lr=0.010512
Epoch[161] Batch [3499]/[3760]	Speed: 117.700127 samples/sec	 accuracy=54.025893	 loss=2.001987	 lr=0.010505
Epoch[161] Batch [3549]/[3760]	Speed: 117.466765 samples/sec	 accuracy=54.022447	 loss=2.001793	 lr=0.010497
Epoch[161] Batch [3599]/[3760]	Speed: 117.784385 samples/sec	 accuracy=54.008247	 loss=2.002589	 lr=0.010489
Epoch[161] Batch [3649]/[3760]	Speed: 117.933429 samples/sec	 accuracy=54.010274	 loss=2.003052	 lr=0.010481
Epoch[161] Batch [3699]/[3760]	Speed: 117.838778 samples/sec	 accuracy=54.025338	 loss=2.002841	 lr=0.010473
Epoch[161] Batch [3749]/[3760]	Speed: 120.414675 samples/sec	 accuracy=54.027083	 loss=2.003038	 lr=0.010465
Batch [0049]/[0303]: acc-top1=52.906250 acc-top5=76.312500
Batch [0099]/[0303]: acc-top1=50.031250 acc-top5=74.406250
Batch [0149]/[0303]: acc-top1=51.927083 acc-top5=75.468750
Batch [0199]/[0303]: acc-top1=54.523438 acc-top5=77.046875
Batch [0249]/[0303]: acc-top1=54.100000 acc-top5=77.000000
Batch [0299]/[0303]: acc-top1=53.562500 acc-top5=77.109375
[Epoch 161] training: accuracy=54.032995	 loss=2.002853
[Epoch 161] speed: 116 samples/sec	time cost: 2173.345336
[Epoch 161] validation: acc-top1=53.511757 acc-top5=77.109117 loss=2.143722
Epoch[162] Batch [0049]/[3759]	Speed: 86.443286 samples/sec	 accuracy=54.093750	 loss=2.020745	 lr=0.010456
Epoch[162] Batch [0099]/[3759]	Speed: 117.122529 samples/sec	 accuracy=54.078125	 loss=1.999025	 lr=0.010448
Epoch[162] Batch [0149]/[3759]	Speed: 117.792524 samples/sec	 accuracy=54.052083	 loss=2.002526	 lr=0.010440
Epoch[162] Batch [0199]/[3759]	Speed: 115.325483 samples/sec	 accuracy=54.539062	 loss=1.988901	 lr=0.010432
Epoch[162] Batch [0249]/[3759]	Speed: 117.512940 samples/sec	 accuracy=54.537500	 loss=1.988092	 lr=0.010424
Epoch[162] Batch [0299]/[3759]	Speed: 117.250889 samples/sec	 accuracy=54.536458	 loss=1.983559	 lr=0.010416
Epoch[162] Batch [0349]/[3759]	Speed: 117.787354 samples/sec	 accuracy=54.401786	 loss=1.985066	 lr=0.010408
Epoch[162] Batch [0399]/[3759]	Speed: 117.589660 samples/sec	 accuracy=54.566406	 loss=1.979388	 lr=0.010400
Epoch[162] Batch [0449]/[3759]	Speed: 118.146338 samples/sec	 accuracy=54.541667	 loss=1.980775	 lr=0.010393
Epoch[162] Batch [0499]/[3759]	Speed: 116.629936 samples/sec	 accuracy=54.450000	 loss=1.982353	 lr=0.010385
Epoch[162] Batch [0549]/[3759]	Speed: 117.640000 samples/sec	 accuracy=54.326705	 loss=1.987902	 lr=0.010377
Epoch[162] Batch [0599]/[3759]	Speed: 117.258291 samples/sec	 accuracy=54.320312	 loss=1.986301	 lr=0.010369
Epoch[162] Batch [0649]/[3759]	Speed: 117.954089 samples/sec	 accuracy=54.302885	 loss=1.986119	 lr=0.010361
Epoch[162] Batch [0699]/[3759]	Speed: 117.559778 samples/sec	 accuracy=54.310268	 loss=1.984942	 lr=0.010353
Epoch[162] Batch [0749]/[3759]	Speed: 116.497058 samples/sec	 accuracy=54.331250	 loss=1.983663	 lr=0.010345
Epoch[162] Batch [0799]/[3759]	Speed: 118.035000 samples/sec	 accuracy=54.285156	 loss=1.985834	 lr=0.010338
Epoch[162] Batch [0849]/[3759]	Speed: 116.946431 samples/sec	 accuracy=54.283088	 loss=1.986290	 lr=0.010330
Epoch[162] Batch [0899]/[3759]	Speed: 118.119192 samples/sec	 accuracy=54.296875	 loss=1.984327	 lr=0.010322
Epoch[162] Batch [0949]/[3759]	Speed: 117.231631 samples/sec	 accuracy=54.384868	 loss=1.982249	 lr=0.010314
Epoch[162] Batch [0999]/[3759]	Speed: 116.956028 samples/sec	 accuracy=54.382812	 loss=1.982037	 lr=0.010306
Epoch[162] Batch [1049]/[3759]	Speed: 117.109795 samples/sec	 accuracy=54.379464	 loss=1.982561	 lr=0.010298
Epoch[162] Batch [1099]/[3759]	Speed: 116.986686 samples/sec	 accuracy=54.329545	 loss=1.984590	 lr=0.010290
Epoch[162] Batch [1149]/[3759]	Speed: 117.324576 samples/sec	 accuracy=54.369565	 loss=1.982303	 lr=0.010283
Epoch[162] Batch [1199]/[3759]	Speed: 116.430559 samples/sec	 accuracy=54.384115	 loss=1.983265	 lr=0.010275
Epoch[162] Batch [1249]/[3759]	Speed: 117.663211 samples/sec	 accuracy=54.432500	 loss=1.980504	 lr=0.010267
Epoch[162] Batch [1299]/[3759]	Speed: 118.081035 samples/sec	 accuracy=54.456731	 loss=1.979198	 lr=0.010259
Epoch[162] Batch [1349]/[3759]	Speed: 117.213980 samples/sec	 accuracy=54.454861	 loss=1.979444	 lr=0.010251
Epoch[162] Batch [1399]/[3759]	Speed: 117.764199 samples/sec	 accuracy=54.468750	 loss=1.980122	 lr=0.010243
Epoch[162] Batch [1449]/[3759]	Speed: 116.972626 samples/sec	 accuracy=54.441810	 loss=1.981859	 lr=0.010236
Epoch[162] Batch [1499]/[3759]	Speed: 116.963286 samples/sec	 accuracy=54.466667	 loss=1.981715	 lr=0.010228
Epoch[162] Batch [1549]/[3759]	Speed: 118.294507 samples/sec	 accuracy=54.413306	 loss=1.984602	 lr=0.010220
Epoch[162] Batch [1599]/[3759]	Speed: 114.910897 samples/sec	 accuracy=54.397461	 loss=1.984355	 lr=0.010212
Epoch[162] Batch [1649]/[3759]	Speed: 117.723315 samples/sec	 accuracy=54.392992	 loss=1.983570	 lr=0.010204
Epoch[162] Batch [1699]/[3759]	Speed: 118.161862 samples/sec	 accuracy=54.341912	 loss=1.986408	 lr=0.010197
Epoch[162] Batch [1749]/[3759]	Speed: 117.404959 samples/sec	 accuracy=54.352679	 loss=1.986577	 lr=0.010189
Epoch[162] Batch [1799]/[3759]	Speed: 117.487478 samples/sec	 accuracy=54.297743	 loss=1.987866	 lr=0.010181
Epoch[162] Batch [1849]/[3759]	Speed: 117.553251 samples/sec	 accuracy=54.328547	 loss=1.987620	 lr=0.010173
Epoch[162] Batch [1899]/[3759]	Speed: 117.806442 samples/sec	 accuracy=54.361842	 loss=1.986846	 lr=0.010165
Epoch[162] Batch [1949]/[3759]	Speed: 117.199578 samples/sec	 accuracy=54.370192	 loss=1.985782	 lr=0.010158
Epoch[162] Batch [1999]/[3759]	Speed: 117.198019 samples/sec	 accuracy=54.374219	 loss=1.986020	 lr=0.010150
Epoch[162] Batch [2049]/[3759]	Speed: 117.581243 samples/sec	 accuracy=54.387195	 loss=1.986654	 lr=0.010142
Epoch[162] Batch [2099]/[3759]	Speed: 116.900147 samples/sec	 accuracy=54.377976	 loss=1.986395	 lr=0.010134
Epoch[162] Batch [2149]/[3759]	Speed: 118.174368 samples/sec	 accuracy=54.365552	 loss=1.987380	 lr=0.010126
Epoch[162] Batch [2199]/[3759]	Speed: 117.257927 samples/sec	 accuracy=54.364347	 loss=1.987920	 lr=0.010119
Epoch[162] Batch [2249]/[3759]	Speed: 117.099580 samples/sec	 accuracy=54.350694	 loss=1.987836	 lr=0.010111
Epoch[162] Batch [2299]/[3759]	Speed: 117.750419 samples/sec	 accuracy=54.375000	 loss=1.987661	 lr=0.010103
Epoch[162] Batch [2349]/[3759]	Speed: 117.703504 samples/sec	 accuracy=54.380319	 loss=1.988198	 lr=0.010095
Epoch[162] Batch [2399]/[3759]	Speed: 116.620865 samples/sec	 accuracy=54.362630	 loss=1.988085	 lr=0.010088
Epoch[162] Batch [2449]/[3759]	Speed: 117.095156 samples/sec	 accuracy=54.348852	 loss=1.988310	 lr=0.010080
Epoch[162] Batch [2499]/[3759]	Speed: 116.759390 samples/sec	 accuracy=54.366250	 loss=1.987019	 lr=0.010072
Epoch[162] Batch [2549]/[3759]	Speed: 116.980961 samples/sec	 accuracy=54.375000	 loss=1.987399	 lr=0.010064
Epoch[162] Batch [2599]/[3759]	Speed: 117.389853 samples/sec	 accuracy=54.368389	 loss=1.987260	 lr=0.010057
Epoch[162] Batch [2649]/[3759]	Speed: 116.798488 samples/sec	 accuracy=54.356132	 loss=1.987318	 lr=0.010049
Epoch[162] Batch [2699]/[3759]	Speed: 116.857561 samples/sec	 accuracy=54.358218	 loss=1.987472	 lr=0.010041
Epoch[162] Batch [2749]/[3759]	Speed: 116.988524 samples/sec	 accuracy=54.364773	 loss=1.987098	 lr=0.010033
Epoch[162] Batch [2799]/[3759]	Speed: 116.982592 samples/sec	 accuracy=54.372210	 loss=1.985804	 lr=0.010025
Epoch[162] Batch [2849]/[3759]	Speed: 116.889266 samples/sec	 accuracy=54.385417	 loss=1.985654	 lr=0.010018
Epoch[162] Batch [2899]/[3759]	Speed: 117.750797 samples/sec	 accuracy=54.331358	 loss=1.986931	 lr=0.010010
Epoch[162] Batch [2949]/[3759]	Speed: 116.688180 samples/sec	 accuracy=54.351165	 loss=1.986016	 lr=0.010002
Epoch[162] Batch [2999]/[3759]	Speed: 116.016957 samples/sec	 accuracy=54.330729	 loss=1.986265	 lr=0.009995
Epoch[162] Batch [3049]/[3759]	Speed: 117.172191 samples/sec	 accuracy=54.312500	 loss=1.986694	 lr=0.009987
Epoch[162] Batch [3099]/[3759]	Speed: 116.447852 samples/sec	 accuracy=54.312500	 loss=1.986264	 lr=0.009979
Epoch[162] Batch [3149]/[3759]	Speed: 116.962259 samples/sec	 accuracy=54.318452	 loss=1.986339	 lr=0.009971
Epoch[162] Batch [3199]/[3759]	Speed: 117.132089 samples/sec	 accuracy=54.315430	 loss=1.986433	 lr=0.009964
Epoch[162] Batch [3249]/[3759]	Speed: 117.163289 samples/sec	 accuracy=54.349038	 loss=1.984951	 lr=0.009956
Epoch[162] Batch [3299]/[3759]	Speed: 117.923276 samples/sec	 accuracy=54.381155	 loss=1.984470	 lr=0.009948
Epoch[162] Batch [3349]/[3759]	Speed: 117.191876 samples/sec	 accuracy=54.386194	 loss=1.984152	 lr=0.009940
Epoch[162] Batch [3399]/[3759]	Speed: 117.739782 samples/sec	 accuracy=54.388327	 loss=1.983358	 lr=0.009933
Epoch[162] Batch [3449]/[3759]	Speed: 117.814106 samples/sec	 accuracy=54.356884	 loss=1.984585	 lr=0.009925
Epoch[162] Batch [3499]/[3759]	Speed: 117.138037 samples/sec	 accuracy=54.366964	 loss=1.984445	 lr=0.009917
Epoch[162] Batch [3549]/[3759]	Speed: 117.247061 samples/sec	 accuracy=54.363996	 loss=1.984034	 lr=0.009910
Epoch[162] Batch [3599]/[3759]	Speed: 116.742052 samples/sec	 accuracy=54.355903	 loss=1.984103	 lr=0.009902
Epoch[162] Batch [3649]/[3759]	Speed: 117.123560 samples/sec	 accuracy=54.377140	 loss=1.983309	 lr=0.009894
Epoch[162] Batch [3699]/[3759]	Speed: 117.566104 samples/sec	 accuracy=54.377111	 loss=1.982952	 lr=0.009886
Epoch[162] Batch [3749]/[3759]	Speed: 120.832877 samples/sec	 accuracy=54.385417	 loss=1.982548	 lr=0.009879
Batch [0049]/[0303]: acc-top1=54.312500 acc-top5=76.156250
Batch [0099]/[0303]: acc-top1=51.125000 acc-top5=75.406250
Batch [0149]/[0303]: acc-top1=52.885417 acc-top5=76.010417
Batch [0199]/[0303]: acc-top1=55.203125 acc-top5=77.554688
Batch [0249]/[0303]: acc-top1=54.437500 acc-top5=77.387500
Batch [0299]/[0303]: acc-top1=53.880208 acc-top5=77.302083
[Epoch 162] training: accuracy=54.383230	 loss=1.982487
[Epoch 162] speed: 116 samples/sec	time cost: 2177.325732
[Epoch 162] validation: acc-top1=53.764439 acc-top5=77.268977 loss=2.133919
Epoch[163] Batch [0049]/[3760]	Speed: 87.315381 samples/sec	 accuracy=54.781250	 loss=1.926534	 lr=0.009870
Epoch[163] Batch [0099]/[3760]	Speed: 117.771506 samples/sec	 accuracy=55.703125	 loss=1.908287	 lr=0.009862
Epoch[163] Batch [0149]/[3760]	Speed: 117.434151 samples/sec	 accuracy=55.833333	 loss=1.903327	 lr=0.009854
Epoch[163] Batch [0199]/[3760]	Speed: 116.999557 samples/sec	 accuracy=55.820312	 loss=1.904153	 lr=0.009847
Epoch[163] Batch [0249]/[3760]	Speed: 117.930931 samples/sec	 accuracy=55.856250	 loss=1.904347	 lr=0.009839
Epoch[163] Batch [0299]/[3760]	Speed: 116.529346 samples/sec	 accuracy=55.843750	 loss=1.909869	 lr=0.009831
Epoch[163] Batch [0349]/[3760]	Speed: 117.227652 samples/sec	 accuracy=55.625000	 loss=1.914510	 lr=0.009824
Epoch[163] Batch [0399]/[3760]	Speed: 117.359841 samples/sec	 accuracy=55.511719	 loss=1.917991	 lr=0.009816
Epoch[163] Batch [0449]/[3760]	Speed: 117.546131 samples/sec	 accuracy=55.538194	 loss=1.923817	 lr=0.009808
Epoch[163] Batch [0499]/[3760]	Speed: 117.490442 samples/sec	 accuracy=55.387500	 loss=1.929857	 lr=0.009801
Epoch[163] Batch [0549]/[3760]	Speed: 118.204048 samples/sec	 accuracy=55.360795	 loss=1.934616	 lr=0.009793
Epoch[163] Batch [0599]/[3760]	Speed: 117.877534 samples/sec	 accuracy=55.367188	 loss=1.934177	 lr=0.009785
Epoch[163] Batch [0649]/[3760]	Speed: 117.260664 samples/sec	 accuracy=55.437500	 loss=1.932962	 lr=0.009778
Epoch[163] Batch [0699]/[3760]	Speed: 117.800305 samples/sec	 accuracy=55.453125	 loss=1.928142	 lr=0.009770
Epoch[163] Batch [0749]/[3760]	Speed: 117.164109 samples/sec	 accuracy=55.412500	 loss=1.930133	 lr=0.009762
Epoch[163] Batch [0799]/[3760]	Speed: 118.299924 samples/sec	 accuracy=55.394531	 loss=1.930497	 lr=0.009755
Epoch[163] Batch [0849]/[3760]	Speed: 117.330066 samples/sec	 accuracy=55.316176	 loss=1.933128	 lr=0.009747
Epoch[163] Batch [0899]/[3760]	Speed: 116.947573 samples/sec	 accuracy=55.237847	 loss=1.937608	 lr=0.009739
Epoch[163] Batch [0949]/[3760]	Speed: 115.552372 samples/sec	 accuracy=55.253289	 loss=1.937020	 lr=0.009732
Epoch[163] Batch [0999]/[3760]	Speed: 117.326040 samples/sec	 accuracy=55.234375	 loss=1.938303	 lr=0.009724
Epoch[163] Batch [1049]/[3760]	Speed: 117.796336 samples/sec	 accuracy=55.230655	 loss=1.935638	 lr=0.009716
Epoch[163] Batch [1099]/[3760]	Speed: 117.273156 samples/sec	 accuracy=55.178977	 loss=1.936344	 lr=0.009709
Epoch[163] Batch [1149]/[3760]	Speed: 117.820747 samples/sec	 accuracy=55.198370	 loss=1.936124	 lr=0.009701
Epoch[163] Batch [1199]/[3760]	Speed: 117.519755 samples/sec	 accuracy=55.157552	 loss=1.938997	 lr=0.009693
Epoch[163] Batch [1249]/[3760]	Speed: 117.994681 samples/sec	 accuracy=55.138750	 loss=1.940312	 lr=0.009686
Epoch[163] Batch [1299]/[3760]	Speed: 116.773556 samples/sec	 accuracy=55.140625	 loss=1.940443	 lr=0.009678
Epoch[163] Batch [1349]/[3760]	Speed: 117.156432 samples/sec	 accuracy=55.202546	 loss=1.939349	 lr=0.009671
Epoch[163] Batch [1399]/[3760]	Speed: 117.696754 samples/sec	 accuracy=55.220982	 loss=1.938896	 lr=0.009663
Epoch[163] Batch [1449]/[3760]	Speed: 117.837101 samples/sec	 accuracy=55.132543	 loss=1.941346	 lr=0.009655
Epoch[163] Batch [1499]/[3760]	Speed: 117.942660 samples/sec	 accuracy=55.159375	 loss=1.939908	 lr=0.009648
Epoch[163] Batch [1549]/[3760]	Speed: 117.664092 samples/sec	 accuracy=55.156250	 loss=1.941102	 lr=0.009640
Epoch[163] Batch [1599]/[3760]	Speed: 118.177516 samples/sec	 accuracy=55.130859	 loss=1.942145	 lr=0.009632
Epoch[163] Batch [1649]/[3760]	Speed: 117.016231 samples/sec	 accuracy=55.136364	 loss=1.941318	 lr=0.009625
Epoch[163] Batch [1699]/[3760]	Speed: 117.103022 samples/sec	 accuracy=55.084559	 loss=1.943827	 lr=0.009617
Epoch[163] Batch [1749]/[3760]	Speed: 117.375193 samples/sec	 accuracy=55.100000	 loss=1.944111	 lr=0.009610
Epoch[163] Batch [1799]/[3760]	Speed: 116.554432 samples/sec	 accuracy=55.078125	 loss=1.944313	 lr=0.009602
Epoch[163] Batch [1849]/[3760]	Speed: 117.849826 samples/sec	 accuracy=55.093750	 loss=1.944029	 lr=0.009594
Epoch[163] Batch [1899]/[3760]	Speed: 117.434929 samples/sec	 accuracy=55.121711	 loss=1.944183	 lr=0.009587
Epoch[163] Batch [1949]/[3760]	Speed: 117.123045 samples/sec	 accuracy=55.154647	 loss=1.943486	 lr=0.009579
Epoch[163] Batch [1999]/[3760]	Speed: 117.196442 samples/sec	 accuracy=55.128125	 loss=1.944239	 lr=0.009572
Epoch[163] Batch [2049]/[3760]	Speed: 118.139320 samples/sec	 accuracy=55.089177	 loss=1.946188	 lr=0.009564
Epoch[163] Batch [2099]/[3760]	Speed: 117.725460 samples/sec	 accuracy=55.101935	 loss=1.944693	 lr=0.009557
Epoch[163] Batch [2149]/[3760]	Speed: 117.976809 samples/sec	 accuracy=55.073401	 loss=1.945675	 lr=0.009549
Epoch[163] Batch [2199]/[3760]	Speed: 117.546010 samples/sec	 accuracy=55.060369	 loss=1.946155	 lr=0.009541
Epoch[163] Batch [2249]/[3760]	Speed: 118.436932 samples/sec	 accuracy=55.064583	 loss=1.945517	 lr=0.009534
Epoch[163] Batch [2299]/[3760]	Speed: 117.952192 samples/sec	 accuracy=55.024457	 loss=1.946584	 lr=0.009526
Epoch[163] Batch [2349]/[3760]	Speed: 115.058475 samples/sec	 accuracy=55.013298	 loss=1.946888	 lr=0.009519
Epoch[163] Batch [2399]/[3760]	Speed: 117.863987 samples/sec	 accuracy=54.992188	 loss=1.947473	 lr=0.009511
Epoch[163] Batch [2449]/[3760]	Speed: 117.305403 samples/sec	 accuracy=54.978954	 loss=1.948842	 lr=0.009504
Epoch[163] Batch [2499]/[3760]	Speed: 117.969104 samples/sec	 accuracy=54.986250	 loss=1.948667	 lr=0.009496
Epoch[163] Batch [2549]/[3760]	Speed: 117.765898 samples/sec	 accuracy=54.999387	 loss=1.948792	 lr=0.009488
Epoch[163] Batch [2599]/[3760]	Speed: 116.994729 samples/sec	 accuracy=54.986779	 loss=1.949626	 lr=0.009481
Epoch[163] Batch [2649]/[3760]	Speed: 116.800401 samples/sec	 accuracy=55.001769	 loss=1.949673	 lr=0.009473
Epoch[163] Batch [2699]/[3760]	Speed: 117.701869 samples/sec	 accuracy=54.983796	 loss=1.950557	 lr=0.009466
Epoch[163] Batch [2749]/[3760]	Speed: 118.131409 samples/sec	 accuracy=54.972727	 loss=1.950905	 lr=0.009458
Epoch[163] Batch [2799]/[3760]	Speed: 118.536017 samples/sec	 accuracy=54.970424	 loss=1.950402	 lr=0.009451
Epoch[163] Batch [2849]/[3760]	Speed: 118.719059 samples/sec	 accuracy=54.967654	 loss=1.950365	 lr=0.009443
Epoch[163] Batch [2899]/[3760]	Speed: 117.037895 samples/sec	 accuracy=54.980603	 loss=1.950659	 lr=0.009436
Epoch[163] Batch [2949]/[3760]	Speed: 117.703235 samples/sec	 accuracy=54.966631	 loss=1.951036	 lr=0.009428
Epoch[163] Batch [2999]/[3760]	Speed: 118.255829 samples/sec	 accuracy=54.955208	 loss=1.952033	 lr=0.009420
Epoch[163] Batch [3049]/[3760]	Speed: 117.471822 samples/sec	 accuracy=54.961578	 loss=1.952107	 lr=0.009413
Epoch[163] Batch [3099]/[3760]	Speed: 116.663177 samples/sec	 accuracy=54.954133	 loss=1.952655	 lr=0.009405
Epoch[163] Batch [3149]/[3760]	Speed: 118.116613 samples/sec	 accuracy=54.924107	 loss=1.953818	 lr=0.009398
Epoch[163] Batch [3199]/[3760]	Speed: 117.602120 samples/sec	 accuracy=54.936035	 loss=1.954062	 lr=0.009390
Epoch[163] Batch [3249]/[3760]	Speed: 118.149748 samples/sec	 accuracy=54.930769	 loss=1.954258	 lr=0.009383
Epoch[163] Batch [3299]/[3760]	Speed: 118.697473 samples/sec	 accuracy=54.930871	 loss=1.953657	 lr=0.009375
Epoch[163] Batch [3349]/[3760]	Speed: 118.042461 samples/sec	 accuracy=54.927705	 loss=1.953871	 lr=0.009368
Epoch[163] Batch [3399]/[3760]	Speed: 117.952214 samples/sec	 accuracy=54.932445	 loss=1.953811	 lr=0.009360
Epoch[163] Batch [3449]/[3760]	Speed: 118.728962 samples/sec	 accuracy=54.938859	 loss=1.952940	 lr=0.009353
Epoch[163] Batch [3499]/[3760]	Speed: 117.780873 samples/sec	 accuracy=54.933036	 loss=1.952933	 lr=0.009345
Epoch[163] Batch [3549]/[3760]	Speed: 117.814420 samples/sec	 accuracy=54.911532	 loss=1.953086	 lr=0.009338
Epoch[163] Batch [3599]/[3760]	Speed: 117.512665 samples/sec	 accuracy=54.895399	 loss=1.953461	 lr=0.009330
Epoch[163] Batch [3649]/[3760]	Speed: 117.888414 samples/sec	 accuracy=54.887414	 loss=1.953818	 lr=0.009323
Epoch[163] Batch [3699]/[3760]	Speed: 117.688999 samples/sec	 accuracy=54.865709	 loss=1.954685	 lr=0.009315
Epoch[163] Batch [3749]/[3760]	Speed: 120.464255 samples/sec	 accuracy=54.852500	 loss=1.955121	 lr=0.009308
Batch [0049]/[0303]: acc-top1=53.093750 acc-top5=75.875000
Batch [0099]/[0303]: acc-top1=51.359375 acc-top5=75.375000
Batch [0149]/[0303]: acc-top1=53.197917 acc-top5=76.218750
Batch [0199]/[0303]: acc-top1=55.570312 acc-top5=77.875000
Batch [0249]/[0303]: acc-top1=54.700000 acc-top5=77.650000
Batch [0299]/[0303]: acc-top1=54.302083 acc-top5=77.598958
[Epoch 163] training: accuracy=54.843750	 loss=1.955405
[Epoch 163] speed: 117 samples/sec	time cost: 2172.749846
[Epoch 163] validation: acc-top1=54.254332 acc-top5=77.599010 loss=2.089704
Epoch[164] Batch [0049]/[3760]	Speed: 87.445647 samples/sec	 accuracy=56.562500	 loss=1.864887	 lr=0.009299
Epoch[164] Batch [0099]/[3760]	Speed: 117.649569 samples/sec	 accuracy=56.578125	 loss=1.868322	 lr=0.009291
Epoch[164] Batch [0149]/[3760]	Speed: 117.603982 samples/sec	 accuracy=56.781250	 loss=1.880717	 lr=0.009284
Epoch[164] Batch [0199]/[3760]	Speed: 116.425795 samples/sec	 accuracy=56.914062	 loss=1.878969	 lr=0.009276
Epoch[164] Batch [0249]/[3760]	Speed: 116.970008 samples/sec	 accuracy=56.643750	 loss=1.890225	 lr=0.009269
Epoch[164] Batch [0299]/[3760]	Speed: 117.102141 samples/sec	 accuracy=56.406250	 loss=1.895042	 lr=0.009261
Epoch[164] Batch [0349]/[3760]	Speed: 117.802078 samples/sec	 accuracy=56.459821	 loss=1.894509	 lr=0.009254
Epoch[164] Batch [0399]/[3760]	Speed: 116.971927 samples/sec	 accuracy=56.289062	 loss=1.900110	 lr=0.009246
Epoch[164] Batch [0449]/[3760]	Speed: 117.850442 samples/sec	 accuracy=56.302083	 loss=1.900262	 lr=0.009239
Epoch[164] Batch [0499]/[3760]	Speed: 117.417746 samples/sec	 accuracy=56.165625	 loss=1.904911	 lr=0.009231
Epoch[164] Batch [0549]/[3760]	Speed: 117.546618 samples/sec	 accuracy=56.048295	 loss=1.912743	 lr=0.009224
Epoch[164] Batch [0599]/[3760]	Speed: 117.806108 samples/sec	 accuracy=56.054688	 loss=1.914903	 lr=0.009217
Epoch[164] Batch [0649]/[3760]	Speed: 117.525280 samples/sec	 accuracy=56.060096	 loss=1.912680	 lr=0.009209
Epoch[164] Batch [0699]/[3760]	Speed: 118.095509 samples/sec	 accuracy=56.031250	 loss=1.914166	 lr=0.009202
Epoch[164] Batch [0749]/[3760]	Speed: 117.344424 samples/sec	 accuracy=56.010417	 loss=1.913280	 lr=0.009194
Epoch[164] Batch [0799]/[3760]	Speed: 117.680348 samples/sec	 accuracy=55.939453	 loss=1.916889	 lr=0.009187
Epoch[164] Batch [0849]/[3760]	Speed: 117.253347 samples/sec	 accuracy=55.886029	 loss=1.919399	 lr=0.009179
Epoch[164] Batch [0899]/[3760]	Speed: 117.508021 samples/sec	 accuracy=55.876736	 loss=1.918229	 lr=0.009172
Epoch[164] Batch [0949]/[3760]	Speed: 116.796160 samples/sec	 accuracy=55.807566	 loss=1.920203	 lr=0.009164
Epoch[164] Batch [0999]/[3760]	Speed: 117.567269 samples/sec	 accuracy=55.826563	 loss=1.917517	 lr=0.009157
Epoch[164] Batch [1049]/[3760]	Speed: 117.259448 samples/sec	 accuracy=55.855655	 loss=1.915397	 lr=0.009149
Epoch[164] Batch [1099]/[3760]	Speed: 117.479488 samples/sec	 accuracy=55.744318	 loss=1.917649	 lr=0.009142
Epoch[164] Batch [1149]/[3760]	Speed: 118.870462 samples/sec	 accuracy=55.717391	 loss=1.918543	 lr=0.009135
Epoch[164] Batch [1199]/[3760]	Speed: 117.609415 samples/sec	 accuracy=55.697917	 loss=1.919377	 lr=0.009127
Epoch[164] Batch [1249]/[3760]	Speed: 117.497690 samples/sec	 accuracy=55.768750	 loss=1.917460	 lr=0.009120
Epoch[164] Batch [1299]/[3760]	Speed: 117.458084 samples/sec	 accuracy=55.728365	 loss=1.918264	 lr=0.009112
Epoch[164] Batch [1349]/[3760]	Speed: 114.532257 samples/sec	 accuracy=55.773148	 loss=1.916192	 lr=0.009105
Epoch[164] Batch [1399]/[3760]	Speed: 117.047127 samples/sec	 accuracy=55.774554	 loss=1.916792	 lr=0.009097
Epoch[164] Batch [1449]/[3760]	Speed: 118.170408 samples/sec	 accuracy=55.775862	 loss=1.916869	 lr=0.009090
Epoch[164] Batch [1499]/[3760]	Speed: 116.898821 samples/sec	 accuracy=55.803125	 loss=1.915679	 lr=0.009083
Epoch[164] Batch [1549]/[3760]	Speed: 116.819227 samples/sec	 accuracy=55.817540	 loss=1.915545	 lr=0.009075
Epoch[164] Batch [1599]/[3760]	Speed: 116.880673 samples/sec	 accuracy=55.781250	 loss=1.916316	 lr=0.009068
Epoch[164] Batch [1649]/[3760]	Speed: 117.670485 samples/sec	 accuracy=55.806818	 loss=1.916335	 lr=0.009060
Epoch[164] Batch [1699]/[3760]	Speed: 117.261045 samples/sec	 accuracy=55.802390	 loss=1.916282	 lr=0.009053
Epoch[164] Batch [1749]/[3760]	Speed: 117.372197 samples/sec	 accuracy=55.791964	 loss=1.915987	 lr=0.009046
Epoch[164] Batch [1799]/[3760]	Speed: 117.093980 samples/sec	 accuracy=55.769965	 loss=1.916669	 lr=0.009038
Epoch[164] Batch [1849]/[3760]	Speed: 117.747715 samples/sec	 accuracy=55.774493	 loss=1.917452	 lr=0.009031
Epoch[164] Batch [1899]/[3760]	Speed: 116.957087 samples/sec	 accuracy=55.731086	 loss=1.918856	 lr=0.009023
Epoch[164] Batch [1949]/[3760]	Speed: 117.248281 samples/sec	 accuracy=55.681090	 loss=1.919715	 lr=0.009016
Epoch[164] Batch [1999]/[3760]	Speed: 116.882029 samples/sec	 accuracy=55.684375	 loss=1.919818	 lr=0.009009
Epoch[164] Batch [2049]/[3760]	Speed: 118.148759 samples/sec	 accuracy=55.663872	 loss=1.921089	 lr=0.009001
Epoch[164] Batch [2099]/[3760]	Speed: 117.019307 samples/sec	 accuracy=55.662946	 loss=1.920386	 lr=0.008994
Epoch[164] Batch [2149]/[3760]	Speed: 117.893757 samples/sec	 accuracy=55.658430	 loss=1.921338	 lr=0.008987
Epoch[164] Batch [2199]/[3760]	Speed: 117.076268 samples/sec	 accuracy=55.673295	 loss=1.921108	 lr=0.008979
Epoch[164] Batch [2249]/[3760]	Speed: 117.275299 samples/sec	 accuracy=55.681944	 loss=1.920532	 lr=0.008972
Epoch[164] Batch [2299]/[3760]	Speed: 116.767961 samples/sec	 accuracy=55.672554	 loss=1.920802	 lr=0.008964
Epoch[164] Batch [2349]/[3760]	Speed: 117.062977 samples/sec	 accuracy=55.683511	 loss=1.920368	 lr=0.008957
Epoch[164] Batch [2399]/[3760]	Speed: 117.984796 samples/sec	 accuracy=55.647786	 loss=1.922049	 lr=0.008950
Epoch[164] Batch [2449]/[3760]	Speed: 117.393464 samples/sec	 accuracy=55.618622	 loss=1.923293	 lr=0.008942
Epoch[164] Batch [2499]/[3760]	Speed: 117.639910 samples/sec	 accuracy=55.591875	 loss=1.923908	 lr=0.008935
Epoch[164] Batch [2549]/[3760]	Speed: 117.141961 samples/sec	 accuracy=55.558824	 loss=1.925031	 lr=0.008928
Epoch[164] Batch [2599]/[3760]	Speed: 117.620047 samples/sec	 accuracy=55.529447	 loss=1.925885	 lr=0.008920
Epoch[164] Batch [2649]/[3760]	Speed: 117.275359 samples/sec	 accuracy=55.500590	 loss=1.927102	 lr=0.008913
Epoch[164] Batch [2699]/[3760]	Speed: 117.551646 samples/sec	 accuracy=55.479745	 loss=1.927376	 lr=0.008906
Epoch[164] Batch [2749]/[3760]	Speed: 118.060997 samples/sec	 accuracy=55.451136	 loss=1.928599	 lr=0.008898
Epoch[164] Batch [2799]/[3760]	Speed: 116.082510 samples/sec	 accuracy=55.421317	 loss=1.930050	 lr=0.008891
Epoch[164] Batch [2849]/[3760]	Speed: 116.010313 samples/sec	 accuracy=55.417763	 loss=1.930120	 lr=0.008883
Epoch[164] Batch [2899]/[3760]	Speed: 117.874492 samples/sec	 accuracy=55.432112	 loss=1.929842	 lr=0.008876
Epoch[164] Batch [2949]/[3760]	Speed: 118.117645 samples/sec	 accuracy=55.452331	 loss=1.928773	 lr=0.008869
Epoch[164] Batch [2999]/[3760]	Speed: 116.732360 samples/sec	 accuracy=55.455729	 loss=1.928352	 lr=0.008861
Epoch[164] Batch [3049]/[3760]	Speed: 117.440177 samples/sec	 accuracy=55.460041	 loss=1.928251	 lr=0.008854
Epoch[164] Batch [3099]/[3760]	Speed: 117.852664 samples/sec	 accuracy=55.440524	 loss=1.928524	 lr=0.008847
Epoch[164] Batch [3149]/[3760]	Speed: 117.084388 samples/sec	 accuracy=55.434524	 loss=1.928863	 lr=0.008840
Epoch[164] Batch [3199]/[3760]	Speed: 117.046394 samples/sec	 accuracy=55.437012	 loss=1.929045	 lr=0.008832
Epoch[164] Batch [3249]/[3760]	Speed: 118.269983 samples/sec	 accuracy=55.439904	 loss=1.929065	 lr=0.008825
Epoch[164] Batch [3299]/[3760]	Speed: 117.821075 samples/sec	 accuracy=55.456439	 loss=1.927835	 lr=0.008818
Epoch[164] Batch [3349]/[3760]	Speed: 117.644501 samples/sec	 accuracy=55.455224	 loss=1.928504	 lr=0.008810
Epoch[164] Batch [3399]/[3760]	Speed: 118.273065 samples/sec	 accuracy=55.440257	 loss=1.929684	 lr=0.008803
Epoch[164] Batch [3449]/[3760]	Speed: 117.068791 samples/sec	 accuracy=55.425725	 loss=1.930322	 lr=0.008796
Epoch[164] Batch [3499]/[3760]	Speed: 117.707483 samples/sec	 accuracy=55.404464	 loss=1.931285	 lr=0.008788
Epoch[164] Batch [3549]/[3760]	Speed: 117.567943 samples/sec	 accuracy=55.393046	 loss=1.931744	 lr=0.008781
Epoch[164] Batch [3599]/[3760]	Speed: 117.951138 samples/sec	 accuracy=55.390191	 loss=1.932043	 lr=0.008774
Epoch[164] Batch [3649]/[3760]	Speed: 118.133739 samples/sec	 accuracy=55.377140	 loss=1.932559	 lr=0.008766
Epoch[164] Batch [3699]/[3760]	Speed: 117.029746 samples/sec	 accuracy=55.377956	 loss=1.932661	 lr=0.008759
Epoch[164] Batch [3749]/[3760]	Speed: 120.485719 samples/sec	 accuracy=55.381250	 loss=1.933118	 lr=0.008752
Batch [0049]/[0303]: acc-top1=54.250000 acc-top5=76.281250
Batch [0099]/[0303]: acc-top1=51.406250 acc-top5=75.109375
Batch [0149]/[0303]: acc-top1=52.479167 acc-top5=75.552083
Batch [0199]/[0303]: acc-top1=55.398438 acc-top5=77.367188
Batch [0249]/[0303]: acc-top1=54.831250 acc-top5=77.237500
Batch [0299]/[0303]: acc-top1=54.437500 acc-top5=77.364583
[Epoch 164] training: accuracy=55.387716	 loss=1.932930
[Epoch 164] speed: 116 samples/sec	time cost: 2175.758063
[Epoch 164] validation: acc-top1=54.393564 acc-top5=77.341172 loss=2.097981
Epoch[165] Batch [0049]/[3759]	Speed: 87.396990 samples/sec	 accuracy=56.000000	 loss=1.934685	 lr=0.008743
Epoch[165] Batch [0099]/[3759]	Speed: 116.823896 samples/sec	 accuracy=56.453125	 loss=1.895264	 lr=0.008736
Epoch[165] Batch [0149]/[3759]	Speed: 117.346394 samples/sec	 accuracy=56.156250	 loss=1.909558	 lr=0.008729
Epoch[165] Batch [0199]/[3759]	Speed: 117.706578 samples/sec	 accuracy=56.234375	 loss=1.901517	 lr=0.008721
Epoch[165] Batch [0249]/[3759]	Speed: 117.642459 samples/sec	 accuracy=56.631250	 loss=1.889897	 lr=0.008714
Epoch[165] Batch [0299]/[3759]	Speed: 116.668174 samples/sec	 accuracy=56.552083	 loss=1.887893	 lr=0.008707
Epoch[165] Batch [0349]/[3759]	Speed: 117.561688 samples/sec	 accuracy=56.441964	 loss=1.890266	 lr=0.008699
Epoch[165] Batch [0399]/[3759]	Speed: 115.224171 samples/sec	 accuracy=56.402344	 loss=1.890230	 lr=0.008692
Epoch[165] Batch [0449]/[3759]	Speed: 117.229498 samples/sec	 accuracy=56.440972	 loss=1.889518	 lr=0.008685
Epoch[165] Batch [0499]/[3759]	Speed: 117.799606 samples/sec	 accuracy=56.406250	 loss=1.887360	 lr=0.008678
Epoch[165] Batch [0549]/[3759]	Speed: 116.848830 samples/sec	 accuracy=56.184659	 loss=1.896975	 lr=0.008670
Epoch[165] Batch [0599]/[3759]	Speed: 118.106227 samples/sec	 accuracy=56.111979	 loss=1.901145	 lr=0.008663
Epoch[165] Batch [0649]/[3759]	Speed: 118.008616 samples/sec	 accuracy=56.129808	 loss=1.901779	 lr=0.008656
Epoch[165] Batch [0699]/[3759]	Speed: 117.816611 samples/sec	 accuracy=56.263393	 loss=1.898058	 lr=0.008649
Epoch[165] Batch [0749]/[3759]	Speed: 116.946069 samples/sec	 accuracy=56.183333	 loss=1.899562	 lr=0.008641
Epoch[165] Batch [0799]/[3759]	Speed: 117.437526 samples/sec	 accuracy=56.113281	 loss=1.901842	 lr=0.008634
Epoch[165] Batch [0849]/[3759]	Speed: 117.478681 samples/sec	 accuracy=56.128676	 loss=1.901882	 lr=0.008627
Epoch[165] Batch [0899]/[3759]	Speed: 117.393791 samples/sec	 accuracy=56.142361	 loss=1.897975	 lr=0.008620
Epoch[165] Batch [0949]/[3759]	Speed: 116.992424 samples/sec	 accuracy=56.054276	 loss=1.901635	 lr=0.008612
Epoch[165] Batch [0999]/[3759]	Speed: 117.271135 samples/sec	 accuracy=56.006250	 loss=1.903694	 lr=0.008605
Epoch[165] Batch [1049]/[3759]	Speed: 117.340410 samples/sec	 accuracy=56.046131	 loss=1.900734	 lr=0.008598
Epoch[165] Batch [1099]/[3759]	Speed: 117.026312 samples/sec	 accuracy=56.071023	 loss=1.899447	 lr=0.008591
Epoch[165] Batch [1149]/[3759]	Speed: 117.747025 samples/sec	 accuracy=56.065217	 loss=1.900922	 lr=0.008583
Epoch[165] Batch [1199]/[3759]	Speed: 117.478498 samples/sec	 accuracy=56.063802	 loss=1.900112	 lr=0.008576
Epoch[165] Batch [1249]/[3759]	Speed: 117.398179 samples/sec	 accuracy=56.061250	 loss=1.900195	 lr=0.008569
Epoch[165] Batch [1299]/[3759]	Speed: 117.506336 samples/sec	 accuracy=56.137019	 loss=1.898737	 lr=0.008562
Epoch[165] Batch [1349]/[3759]	Speed: 117.998748 samples/sec	 accuracy=56.122685	 loss=1.899474	 lr=0.008555
Epoch[165] Batch [1399]/[3759]	Speed: 117.424264 samples/sec	 accuracy=56.117188	 loss=1.900086	 lr=0.008547
Epoch[165] Batch [1449]/[3759]	Speed: 117.859830 samples/sec	 accuracy=56.109914	 loss=1.900549	 lr=0.008540
Epoch[165] Batch [1499]/[3759]	Speed: 116.792960 samples/sec	 accuracy=56.093750	 loss=1.900959	 lr=0.008533
Epoch[165] Batch [1549]/[3759]	Speed: 118.109652 samples/sec	 accuracy=56.046371	 loss=1.901809	 lr=0.008526
Epoch[165] Batch [1599]/[3759]	Speed: 117.321754 samples/sec	 accuracy=56.050781	 loss=1.900942	 lr=0.008519
Epoch[165] Batch [1649]/[3759]	Speed: 118.002581 samples/sec	 accuracy=56.062500	 loss=1.899497	 lr=0.008511
Epoch[165] Batch [1699]/[3759]	Speed: 117.566473 samples/sec	 accuracy=56.032169	 loss=1.900119	 lr=0.008504
Epoch[165] Batch [1749]/[3759]	Speed: 117.047773 samples/sec	 accuracy=55.997321	 loss=1.900823	 lr=0.008497
Epoch[165] Batch [1799]/[3759]	Speed: 114.723315 samples/sec	 accuracy=55.975694	 loss=1.901696	 lr=0.008490
Epoch[165] Batch [1849]/[3759]	Speed: 117.396089 samples/sec	 accuracy=55.943412	 loss=1.903329	 lr=0.008483
Epoch[165] Batch [1899]/[3759]	Speed: 117.814575 samples/sec	 accuracy=55.981908	 loss=1.901995	 lr=0.008475
Epoch[165] Batch [1949]/[3759]	Speed: 116.847450 samples/sec	 accuracy=55.968750	 loss=1.901743	 lr=0.008468
Epoch[165] Batch [1999]/[3759]	Speed: 117.156115 samples/sec	 accuracy=55.935156	 loss=1.902571	 lr=0.008461
Epoch[165] Batch [2049]/[3759]	Speed: 117.917893 samples/sec	 accuracy=55.918445	 loss=1.904044	 lr=0.008454
Epoch[165] Batch [2099]/[3759]	Speed: 118.090211 samples/sec	 accuracy=55.875744	 loss=1.906165	 lr=0.008447
Epoch[165] Batch [2149]/[3759]	Speed: 117.235619 samples/sec	 accuracy=55.850291	 loss=1.907358	 lr=0.008440
Epoch[165] Batch [2199]/[3759]	Speed: 117.963372 samples/sec	 accuracy=55.833807	 loss=1.907276	 lr=0.008432
Epoch[165] Batch [2249]/[3759]	Speed: 117.960395 samples/sec	 accuracy=55.853472	 loss=1.906767	 lr=0.008425
Epoch[165] Batch [2299]/[3759]	Speed: 117.828279 samples/sec	 accuracy=55.845109	 loss=1.906434	 lr=0.008418
Epoch[165] Batch [2349]/[3759]	Speed: 117.302250 samples/sec	 accuracy=55.823138	 loss=1.907086	 lr=0.008411
Epoch[165] Batch [2399]/[3759]	Speed: 117.098889 samples/sec	 accuracy=55.810547	 loss=1.906423	 lr=0.008404
Epoch[165] Batch [2449]/[3759]	Speed: 117.332184 samples/sec	 accuracy=55.774872	 loss=1.907833	 lr=0.008397
Epoch[165] Batch [2499]/[3759]	Speed: 118.241727 samples/sec	 accuracy=55.801875	 loss=1.907175	 lr=0.008389
Epoch[165] Batch [2549]/[3759]	Speed: 116.648371 samples/sec	 accuracy=55.784314	 loss=1.907954	 lr=0.008382
Epoch[165] Batch [2599]/[3759]	Speed: 117.301302 samples/sec	 accuracy=55.756010	 loss=1.908809	 lr=0.008375
Epoch[165] Batch [2649]/[3759]	Speed: 117.343582 samples/sec	 accuracy=55.762972	 loss=1.908591	 lr=0.008368
Epoch[165] Batch [2699]/[3759]	Speed: 117.109163 samples/sec	 accuracy=55.774884	 loss=1.908031	 lr=0.008361
Epoch[165] Batch [2749]/[3759]	Speed: 117.124481 samples/sec	 accuracy=55.772159	 loss=1.907544	 lr=0.008354
Epoch[165] Batch [2799]/[3759]	Speed: 117.780397 samples/sec	 accuracy=55.780692	 loss=1.907756	 lr=0.008347
Epoch[165] Batch [2849]/[3759]	Speed: 118.391468 samples/sec	 accuracy=55.812500	 loss=1.907347	 lr=0.008339
Epoch[165] Batch [2899]/[3759]	Speed: 117.146140 samples/sec	 accuracy=55.780172	 loss=1.907677	 lr=0.008332
Epoch[165] Batch [2949]/[3759]	Speed: 116.804916 samples/sec	 accuracy=55.785487	 loss=1.907302	 lr=0.008325
Epoch[165] Batch [2999]/[3759]	Speed: 117.304864 samples/sec	 accuracy=55.782812	 loss=1.907106	 lr=0.008318
Epoch[165] Batch [3049]/[3759]	Speed: 117.365827 samples/sec	 accuracy=55.790984	 loss=1.907073	 lr=0.008311
Epoch[165] Batch [3099]/[3759]	Speed: 118.002540 samples/sec	 accuracy=55.806452	 loss=1.906705	 lr=0.008304
Epoch[165] Batch [3149]/[3759]	Speed: 116.851080 samples/sec	 accuracy=55.807044	 loss=1.906893	 lr=0.008297
Epoch[165] Batch [3199]/[3759]	Speed: 118.526694 samples/sec	 accuracy=55.807617	 loss=1.907080	 lr=0.008290
Epoch[165] Batch [3249]/[3759]	Speed: 114.835765 samples/sec	 accuracy=55.792308	 loss=1.908127	 lr=0.008282
Epoch[165] Batch [3299]/[3759]	Speed: 117.751807 samples/sec	 accuracy=55.772254	 loss=1.908721	 lr=0.008275
Epoch[165] Batch [3349]/[3759]	Speed: 117.644164 samples/sec	 accuracy=55.772854	 loss=1.909049	 lr=0.008268
Epoch[165] Batch [3399]/[3759]	Speed: 117.256701 samples/sec	 accuracy=55.761949	 loss=1.909710	 lr=0.008261
Epoch[165] Batch [3449]/[3759]	Speed: 117.395618 samples/sec	 accuracy=55.764040	 loss=1.909176	 lr=0.008254
Epoch[165] Batch [3499]/[3759]	Speed: 117.415760 samples/sec	 accuracy=55.781696	 loss=1.908072	 lr=0.008247
Epoch[165] Batch [3549]/[3759]	Speed: 118.260388 samples/sec	 accuracy=55.797975	 loss=1.907072	 lr=0.008240
Epoch[165] Batch [3599]/[3759]	Speed: 117.788598 samples/sec	 accuracy=55.795139	 loss=1.906426	 lr=0.008233
Epoch[165] Batch [3649]/[3759]	Speed: 117.899342 samples/sec	 accuracy=55.811216	 loss=1.905947	 lr=0.008226
Epoch[165] Batch [3699]/[3759]	Speed: 118.132352 samples/sec	 accuracy=55.801943	 loss=1.906667	 lr=0.008219
Epoch[165] Batch [3749]/[3759]	Speed: 120.718883 samples/sec	 accuracy=55.785000	 loss=1.906917	 lr=0.008211
Batch [0049]/[0303]: acc-top1=54.750000 acc-top5=77.218750
Batch [0099]/[0303]: acc-top1=52.171875 acc-top5=75.375000
Batch [0149]/[0303]: acc-top1=53.052083 acc-top5=76.125000
Batch [0199]/[0303]: acc-top1=55.718750 acc-top5=77.726562
Batch [0249]/[0303]: acc-top1=54.743750 acc-top5=77.581250
Batch [0299]/[0303]: acc-top1=54.447917 acc-top5=77.739583
[Epoch 165] training: accuracy=55.790270	 loss=1.906664
[Epoch 165] speed: 116 samples/sec	time cost: 2175.169499
[Epoch 165] validation: acc-top1=54.409035 acc-top5=77.743399 loss=2.103142
Epoch[166] Batch [0049]/[3760]	Speed: 86.792911 samples/sec	 accuracy=57.812500	 loss=1.860271	 lr=0.008203
Epoch[166] Batch [0099]/[3760]	Speed: 116.622043 samples/sec	 accuracy=56.718750	 loss=1.871606	 lr=0.008196
Epoch[166] Batch [0149]/[3760]	Speed: 116.886490 samples/sec	 accuracy=57.083333	 loss=1.862392	 lr=0.008189
Epoch[166] Batch [0199]/[3760]	Speed: 117.299009 samples/sec	 accuracy=57.328125	 loss=1.848495	 lr=0.008182
Epoch[166] Batch [0249]/[3760]	Speed: 118.102804 samples/sec	 accuracy=57.156250	 loss=1.856050	 lr=0.008175
Epoch[166] Batch [0299]/[3760]	Speed: 118.389103 samples/sec	 accuracy=57.010417	 loss=1.861100	 lr=0.008168
Epoch[166] Batch [0349]/[3760]	Speed: 117.007595 samples/sec	 accuracy=56.834821	 loss=1.866642	 lr=0.008161
Epoch[166] Batch [0399]/[3760]	Speed: 117.266341 samples/sec	 accuracy=56.664062	 loss=1.873702	 lr=0.008154
Epoch[166] Batch [0449]/[3760]	Speed: 118.375463 samples/sec	 accuracy=56.715278	 loss=1.872058	 lr=0.008147
Epoch[166] Batch [0499]/[3760]	Speed: 117.498497 samples/sec	 accuracy=56.871875	 loss=1.871019	 lr=0.008140
Epoch[166] Batch [0549]/[3760]	Speed: 117.986506 samples/sec	 accuracy=56.852273	 loss=1.868880	 lr=0.008132
Epoch[166] Batch [0599]/[3760]	Speed: 117.778726 samples/sec	 accuracy=56.869792	 loss=1.870519	 lr=0.008125
Epoch[166] Batch [0649]/[3760]	Speed: 116.708303 samples/sec	 accuracy=56.896635	 loss=1.868442	 lr=0.008118
Epoch[166] Batch [0699]/[3760]	Speed: 117.645118 samples/sec	 accuracy=56.848214	 loss=1.872619	 lr=0.008111
Epoch[166] Batch [0749]/[3760]	Speed: 117.452897 samples/sec	 accuracy=56.756250	 loss=1.875476	 lr=0.008104
Epoch[166] Batch [0799]/[3760]	Speed: 117.682825 samples/sec	 accuracy=56.707031	 loss=1.877581	 lr=0.008097
Epoch[166] Batch [0849]/[3760]	Speed: 117.769619 samples/sec	 accuracy=56.742647	 loss=1.876454	 lr=0.008090
Epoch[166] Batch [0899]/[3760]	Speed: 114.634470 samples/sec	 accuracy=56.760417	 loss=1.876286	 lr=0.008083
Epoch[166] Batch [0949]/[3760]	Speed: 117.577494 samples/sec	 accuracy=56.651316	 loss=1.879948	 lr=0.008076
Epoch[166] Batch [0999]/[3760]	Speed: 117.720142 samples/sec	 accuracy=56.634375	 loss=1.877835	 lr=0.008069
Epoch[166] Batch [1049]/[3760]	Speed: 117.620278 samples/sec	 accuracy=56.635417	 loss=1.878365	 lr=0.008062
Epoch[166] Batch [1099]/[3760]	Speed: 117.214371 samples/sec	 accuracy=56.616477	 loss=1.878406	 lr=0.008055
Epoch[166] Batch [1149]/[3760]	Speed: 117.456099 samples/sec	 accuracy=56.618207	 loss=1.879452	 lr=0.008048
Epoch[166] Batch [1199]/[3760]	Speed: 117.480507 samples/sec	 accuracy=56.553385	 loss=1.881129	 lr=0.008041
Epoch[166] Batch [1249]/[3760]	Speed: 117.098609 samples/sec	 accuracy=56.588750	 loss=1.879437	 lr=0.008034
Epoch[166] Batch [1299]/[3760]	Speed: 117.029455 samples/sec	 accuracy=56.608173	 loss=1.880254	 lr=0.008027
Epoch[166] Batch [1349]/[3760]	Speed: 117.045191 samples/sec	 accuracy=56.609954	 loss=1.879774	 lr=0.008020
Epoch[166] Batch [1399]/[3760]	Speed: 117.122587 samples/sec	 accuracy=56.621652	 loss=1.879641	 lr=0.008013
Epoch[166] Batch [1449]/[3760]	Speed: 117.119509 samples/sec	 accuracy=56.584052	 loss=1.880506	 lr=0.008006
Epoch[166] Batch [1499]/[3760]	Speed: 117.685024 samples/sec	 accuracy=56.507292	 loss=1.883789	 lr=0.007999
Epoch[166] Batch [1549]/[3760]	Speed: 116.953606 samples/sec	 accuracy=56.503024	 loss=1.883297	 lr=0.007992
Epoch[166] Batch [1599]/[3760]	Speed: 117.444557 samples/sec	 accuracy=56.532227	 loss=1.881623	 lr=0.007985
Epoch[166] Batch [1649]/[3760]	Speed: 117.291436 samples/sec	 accuracy=56.534091	 loss=1.881415	 lr=0.007978
Epoch[166] Batch [1699]/[3760]	Speed: 118.027568 samples/sec	 accuracy=56.564338	 loss=1.879683	 lr=0.007971
Epoch[166] Batch [1749]/[3760]	Speed: 117.278371 samples/sec	 accuracy=56.541964	 loss=1.880545	 lr=0.007964
Epoch[166] Batch [1799]/[3760]	Speed: 117.935360 samples/sec	 accuracy=56.568576	 loss=1.878810	 lr=0.007957
Epoch[166] Batch [1849]/[3760]	Speed: 117.572535 samples/sec	 accuracy=56.532095	 loss=1.879766	 lr=0.007950
Epoch[166] Batch [1899]/[3760]	Speed: 117.677725 samples/sec	 accuracy=56.530428	 loss=1.880414	 lr=0.007943
Epoch[166] Batch [1949]/[3760]	Speed: 118.186892 samples/sec	 accuracy=56.548077	 loss=1.879524	 lr=0.007936
Epoch[166] Batch [1999]/[3760]	Speed: 116.866208 samples/sec	 accuracy=56.574219	 loss=1.878385	 lr=0.007929
Epoch[166] Batch [2049]/[3760]	Speed: 117.738667 samples/sec	 accuracy=56.532012	 loss=1.880373	 lr=0.007922
Epoch[166] Batch [2099]/[3760]	Speed: 116.887501 samples/sec	 accuracy=56.583333	 loss=1.877373	 lr=0.007915
Epoch[166] Batch [2149]/[3760]	Speed: 117.616437 samples/sec	 accuracy=56.603924	 loss=1.877172	 lr=0.007908
Epoch[166] Batch [2199]/[3760]	Speed: 117.775185 samples/sec	 accuracy=56.625710	 loss=1.876620	 lr=0.007901
Epoch[166] Batch [2249]/[3760]	Speed: 117.742828 samples/sec	 accuracy=56.624306	 loss=1.875782	 lr=0.007894
Epoch[166] Batch [2299]/[3760]	Speed: 117.786226 samples/sec	 accuracy=56.633152	 loss=1.874294	 lr=0.007887
Epoch[166] Batch [2349]/[3760]	Speed: 118.002581 samples/sec	 accuracy=56.613032	 loss=1.874588	 lr=0.007880
Epoch[166] Batch [2399]/[3760]	Speed: 114.878884 samples/sec	 accuracy=56.609375	 loss=1.874632	 lr=0.007874
Epoch[166] Batch [2449]/[3760]	Speed: 117.581270 samples/sec	 accuracy=56.621173	 loss=1.873990	 lr=0.007867
Epoch[166] Batch [2499]/[3760]	Speed: 118.131330 samples/sec	 accuracy=56.600000	 loss=1.874389	 lr=0.007860
Epoch[166] Batch [2549]/[3760]	Speed: 116.718270 samples/sec	 accuracy=56.594363	 loss=1.874706	 lr=0.007853
Epoch[166] Batch [2599]/[3760]	Speed: 118.075909 samples/sec	 accuracy=56.565505	 loss=1.875599	 lr=0.007846
Epoch[166] Batch [2649]/[3760]	Speed: 117.174544 samples/sec	 accuracy=56.557193	 loss=1.875343	 lr=0.007839
Epoch[166] Batch [2699]/[3760]	Speed: 117.652220 samples/sec	 accuracy=56.552662	 loss=1.875196	 lr=0.007832
Epoch[166] Batch [2749]/[3760]	Speed: 117.831195 samples/sec	 accuracy=56.557386	 loss=1.874562	 lr=0.007825
Epoch[166] Batch [2799]/[3760]	Speed: 117.650647 samples/sec	 accuracy=56.559152	 loss=1.875218	 lr=0.007818
Epoch[166] Batch [2849]/[3760]	Speed: 117.380427 samples/sec	 accuracy=56.563596	 loss=1.875086	 lr=0.007811
Epoch[166] Batch [2899]/[3760]	Speed: 117.858039 samples/sec	 accuracy=56.580819	 loss=1.874632	 lr=0.007804
Epoch[166] Batch [2949]/[3760]	Speed: 117.834357 samples/sec	 accuracy=56.570975	 loss=1.875193	 lr=0.007797
Epoch[166] Batch [2999]/[3760]	Speed: 116.695126 samples/sec	 accuracy=56.571354	 loss=1.875106	 lr=0.007790
Epoch[166] Batch [3049]/[3760]	Speed: 117.515170 samples/sec	 accuracy=56.582992	 loss=1.874528	 lr=0.007783
Epoch[166] Batch [3099]/[3760]	Speed: 117.914191 samples/sec	 accuracy=56.603327	 loss=1.874171	 lr=0.007777
Epoch[166] Batch [3149]/[3760]	Speed: 117.287333 samples/sec	 accuracy=56.613095	 loss=1.874205	 lr=0.007770
Epoch[166] Batch [3199]/[3760]	Speed: 117.321162 samples/sec	 accuracy=56.606445	 loss=1.874735	 lr=0.007763
Epoch[166] Batch [3249]/[3760]	Speed: 117.584201 samples/sec	 accuracy=56.589904	 loss=1.874971	 lr=0.007756
Epoch[166] Batch [3299]/[3760]	Speed: 116.881683 samples/sec	 accuracy=56.584754	 loss=1.874578	 lr=0.007749
Epoch[166] Batch [3349]/[3760]	Speed: 117.439853 samples/sec	 accuracy=56.590019	 loss=1.874460	 lr=0.007742
Epoch[166] Batch [3399]/[3760]	Speed: 117.915273 samples/sec	 accuracy=56.601103	 loss=1.873875	 lr=0.007735
Epoch[166] Batch [3449]/[3760]	Speed: 116.845007 samples/sec	 accuracy=56.616395	 loss=1.872963	 lr=0.007728
Epoch[166] Batch [3499]/[3760]	Speed: 117.616741 samples/sec	 accuracy=56.611607	 loss=1.872832	 lr=0.007721
Epoch[166] Batch [3549]/[3760]	Speed: 116.635501 samples/sec	 accuracy=56.624120	 loss=1.872135	 lr=0.007714
Epoch[166] Batch [3599]/[3760]	Speed: 117.753397 samples/sec	 accuracy=56.631510	 loss=1.871023	 lr=0.007708
Epoch[166] Batch [3649]/[3760]	Speed: 118.156934 samples/sec	 accuracy=56.624144	 loss=1.871860	 lr=0.007701
Epoch[166] Batch [3699]/[3760]	Speed: 116.814200 samples/sec	 accuracy=56.624155	 loss=1.871897	 lr=0.007694
Epoch[166] Batch [3749]/[3760]	Speed: 120.533408 samples/sec	 accuracy=56.625000	 loss=1.871251	 lr=0.007687
Batch [0049]/[0303]: acc-top1=55.937500 acc-top5=76.250000
Batch [0099]/[0303]: acc-top1=52.640625 acc-top5=75.625000
Batch [0149]/[0303]: acc-top1=53.750000 acc-top5=76.437500
Batch [0199]/[0303]: acc-top1=56.289062 acc-top5=78.046875
Batch [0249]/[0303]: acc-top1=55.593750 acc-top5=77.850000
Batch [0299]/[0303]: acc-top1=55.000000 acc-top5=77.770833
[Epoch 166] training: accuracy=56.628158	 loss=1.871144
[Epoch 166] speed: 116 samples/sec	time cost: 2174.777872
[Epoch 166] validation: acc-top1=54.991749 acc-top5=77.800124 loss=2.115971
Epoch[167] Batch [0049]/[3760]	Speed: 85.312890 samples/sec	 accuracy=58.000000	 loss=1.793766	 lr=0.007679
Epoch[167] Batch [0099]/[3760]	Speed: 117.660920 samples/sec	 accuracy=57.750000	 loss=1.807856	 lr=0.007672
Epoch[167] Batch [0149]/[3760]	Speed: 117.591160 samples/sec	 accuracy=57.541667	 loss=1.820917	 lr=0.007665
Epoch[167] Batch [0199]/[3760]	Speed: 117.205805 samples/sec	 accuracy=57.453125	 loss=1.832438	 lr=0.007658
Epoch[167] Batch [0249]/[3760]	Speed: 118.324610 samples/sec	 accuracy=57.287500	 loss=1.838604	 lr=0.007651
Epoch[167] Batch [0299]/[3760]	Speed: 117.566010 samples/sec	 accuracy=57.166667	 loss=1.841507	 lr=0.007644
Epoch[167] Batch [0349]/[3760]	Speed: 117.988262 samples/sec	 accuracy=56.941964	 loss=1.848101	 lr=0.007638
Epoch[167] Batch [0399]/[3760]	Speed: 117.387498 samples/sec	 accuracy=57.203125	 loss=1.839153	 lr=0.007631
Epoch[167] Batch [0449]/[3760]	Speed: 117.706932 samples/sec	 accuracy=57.225694	 loss=1.839071	 lr=0.007624
Epoch[167] Batch [0499]/[3760]	Speed: 116.697054 samples/sec	 accuracy=57.331250	 loss=1.833096	 lr=0.007617
Epoch[167] Batch [0549]/[3760]	Speed: 117.737141 samples/sec	 accuracy=57.295455	 loss=1.831240	 lr=0.007610
Epoch[167] Batch [0599]/[3760]	Speed: 117.408286 samples/sec	 accuracy=57.190104	 loss=1.832375	 lr=0.007603
Epoch[167] Batch [0649]/[3760]	Speed: 116.935443 samples/sec	 accuracy=57.076923	 loss=1.834113	 lr=0.007596
Epoch[167] Batch [0699]/[3760]	Speed: 118.232638 samples/sec	 accuracy=57.183036	 loss=1.830033	 lr=0.007590
Epoch[167] Batch [0749]/[3760]	Speed: 117.529605 samples/sec	 accuracy=57.222917	 loss=1.832102	 lr=0.007583
Epoch[167] Batch [0799]/[3760]	Speed: 117.949752 samples/sec	 accuracy=57.195312	 loss=1.831591	 lr=0.007576
Epoch[167] Batch [0849]/[3760]	Speed: 117.192278 samples/sec	 accuracy=57.226103	 loss=1.832605	 lr=0.007569
Epoch[167] Batch [0899]/[3760]	Speed: 117.442301 samples/sec	 accuracy=57.194444	 loss=1.837035	 lr=0.007562
Epoch[167] Batch [0949]/[3760]	Speed: 117.100410 samples/sec	 accuracy=57.113487	 loss=1.841194	 lr=0.007556
Epoch[167] Batch [0999]/[3760]	Speed: 117.228356 samples/sec	 accuracy=57.100000	 loss=1.843368	 lr=0.007549
Epoch[167] Batch [1049]/[3760]	Speed: 117.496176 samples/sec	 accuracy=57.148810	 loss=1.842448	 lr=0.007542
Epoch[167] Batch [1099]/[3760]	Speed: 116.740457 samples/sec	 accuracy=57.147727	 loss=1.842081	 lr=0.007535
Epoch[167] Batch [1149]/[3760]	Speed: 117.860242 samples/sec	 accuracy=57.220109	 loss=1.840873	 lr=0.007528
Epoch[167] Batch [1199]/[3760]	Speed: 116.514724 samples/sec	 accuracy=57.227865	 loss=1.842602	 lr=0.007521
Epoch[167] Batch [1249]/[3760]	Speed: 117.292546 samples/sec	 accuracy=57.248750	 loss=1.843031	 lr=0.007515
Epoch[167] Batch [1299]/[3760]	Speed: 116.526742 samples/sec	 accuracy=57.272837	 loss=1.843530	 lr=0.007508
Epoch[167] Batch [1349]/[3760]	Speed: 117.066483 samples/sec	 accuracy=57.258102	 loss=1.842482	 lr=0.007501
Epoch[167] Batch [1399]/[3760]	Speed: 117.367575 samples/sec	 accuracy=57.236607	 loss=1.843158	 lr=0.007494
Epoch[167] Batch [1449]/[3760]	Speed: 117.655382 samples/sec	 accuracy=57.214440	 loss=1.844504	 lr=0.007488
Epoch[167] Batch [1499]/[3760]	Speed: 114.780044 samples/sec	 accuracy=57.245833	 loss=1.843602	 lr=0.007481
Epoch[167] Batch [1549]/[3760]	Speed: 117.392721 samples/sec	 accuracy=57.166331	 loss=1.847266	 lr=0.007474
Epoch[167] Batch [1599]/[3760]	Speed: 117.817257 samples/sec	 accuracy=57.080078	 loss=1.850578	 lr=0.007467
Epoch[167] Batch [1649]/[3760]	Speed: 117.494046 samples/sec	 accuracy=57.076705	 loss=1.850263	 lr=0.007460
Epoch[167] Batch [1699]/[3760]	Speed: 117.528421 samples/sec	 accuracy=57.068015	 loss=1.850375	 lr=0.007454
Epoch[167] Batch [1749]/[3760]	Speed: 117.604319 samples/sec	 accuracy=57.063393	 loss=1.850235	 lr=0.007447
Epoch[167] Batch [1799]/[3760]	Speed: 117.214709 samples/sec	 accuracy=57.070312	 loss=1.849125	 lr=0.007440
Epoch[167] Batch [1849]/[3760]	Speed: 117.023604 samples/sec	 accuracy=57.079392	 loss=1.847948	 lr=0.007433
Epoch[167] Batch [1899]/[3760]	Speed: 117.187150 samples/sec	 accuracy=57.111020	 loss=1.847211	 lr=0.007427
Epoch[167] Batch [1949]/[3760]	Speed: 116.708132 samples/sec	 accuracy=57.141026	 loss=1.846104	 lr=0.007420
Epoch[167] Batch [1999]/[3760]	Speed: 117.585550 samples/sec	 accuracy=57.116406	 loss=1.846769	 lr=0.007413
Epoch[167] Batch [2049]/[3760]	Speed: 117.439152 samples/sec	 accuracy=57.112805	 loss=1.846588	 lr=0.007406
Epoch[167] Batch [2099]/[3760]	Speed: 117.478036 samples/sec	 accuracy=57.058780	 loss=1.849242	 lr=0.007399
Epoch[167] Batch [2149]/[3760]	Speed: 117.165356 samples/sec	 accuracy=57.027616	 loss=1.850568	 lr=0.007393
Epoch[167] Batch [2199]/[3760]	Speed: 117.719597 samples/sec	 accuracy=57.017756	 loss=1.851010	 lr=0.007386
Epoch[167] Batch [2249]/[3760]	Speed: 117.786477 samples/sec	 accuracy=57.010417	 loss=1.850925	 lr=0.007379
Epoch[167] Batch [2299]/[3760]	Speed: 117.246111 samples/sec	 accuracy=57.016304	 loss=1.851331	 lr=0.007373
Epoch[167] Batch [2349]/[3760]	Speed: 117.503738 samples/sec	 accuracy=56.999335	 loss=1.851645	 lr=0.007366
Epoch[167] Batch [2399]/[3760]	Speed: 118.089387 samples/sec	 accuracy=56.983724	 loss=1.851858	 lr=0.007359
Epoch[167] Batch [2449]/[3760]	Speed: 116.924128 samples/sec	 accuracy=56.985969	 loss=1.851782	 lr=0.007352
Epoch[167] Batch [2499]/[3760]	Speed: 117.181056 samples/sec	 accuracy=57.007500	 loss=1.851089	 lr=0.007346
Epoch[167] Batch [2549]/[3760]	Speed: 117.540658 samples/sec	 accuracy=57.012868	 loss=1.850142	 lr=0.007339
Epoch[167] Batch [2599]/[3760]	Speed: 117.289511 samples/sec	 accuracy=56.992788	 loss=1.851119	 lr=0.007332
Epoch[167] Batch [2649]/[3760]	Speed: 117.286885 samples/sec	 accuracy=57.000000	 loss=1.850719	 lr=0.007325
Epoch[167] Batch [2699]/[3760]	Speed: 117.234902 samples/sec	 accuracy=56.995949	 loss=1.851210	 lr=0.007319
Epoch[167] Batch [2749]/[3760]	Speed: 118.232741 samples/sec	 accuracy=57.019318	 loss=1.851047	 lr=0.007312
Epoch[167] Batch [2799]/[3760]	Speed: 117.099971 samples/sec	 accuracy=57.000558	 loss=1.850936	 lr=0.007305
Epoch[167] Batch [2849]/[3760]	Speed: 117.976396 samples/sec	 accuracy=56.981908	 loss=1.852053	 lr=0.007299
Epoch[167] Batch [2899]/[3760]	Speed: 117.069538 samples/sec	 accuracy=56.986530	 loss=1.852044	 lr=0.007292
Epoch[167] Batch [2949]/[3760]	Speed: 115.217777 samples/sec	 accuracy=56.962394	 loss=1.852890	 lr=0.007285
Epoch[167] Batch [2999]/[3760]	Speed: 117.063159 samples/sec	 accuracy=56.950000	 loss=1.853829	 lr=0.007278
Epoch[167] Batch [3049]/[3760]	Speed: 117.341757 samples/sec	 accuracy=56.948258	 loss=1.853639	 lr=0.007272
Epoch[167] Batch [3099]/[3760]	Speed: 117.492184 samples/sec	 accuracy=56.958669	 loss=1.853752	 lr=0.007265
Epoch[167] Batch [3149]/[3760]	Speed: 117.272973 samples/sec	 accuracy=56.945933	 loss=1.854088	 lr=0.007258
Epoch[167] Batch [3199]/[3760]	Speed: 118.132446 samples/sec	 accuracy=56.923828	 loss=1.854863	 lr=0.007252
Epoch[167] Batch [3249]/[3760]	Speed: 117.211080 samples/sec	 accuracy=56.918269	 loss=1.855158	 lr=0.007245
Epoch[167] Batch [3299]/[3760]	Speed: 117.952499 samples/sec	 accuracy=56.902462	 loss=1.855973	 lr=0.007238
Epoch[167] Batch [3349]/[3760]	Speed: 117.863031 samples/sec	 accuracy=56.889925	 loss=1.856016	 lr=0.007232
Epoch[167] Batch [3399]/[3760]	Speed: 117.991377 samples/sec	 accuracy=56.887868	 loss=1.855861	 lr=0.007225
Epoch[167] Batch [3449]/[3760]	Speed: 117.870778 samples/sec	 accuracy=56.895380	 loss=1.855438	 lr=0.007218
Epoch[167] Batch [3499]/[3760]	Speed: 117.793280 samples/sec	 accuracy=56.903571	 loss=1.854915	 lr=0.007212
Epoch[167] Batch [3549]/[3760]	Speed: 117.017686 samples/sec	 accuracy=56.925176	 loss=1.853807	 lr=0.007205
Epoch[167] Batch [3599]/[3760]	Speed: 117.621892 samples/sec	 accuracy=56.914931	 loss=1.854123	 lr=0.007198
Epoch[167] Batch [3649]/[3760]	Speed: 117.687569 samples/sec	 accuracy=56.923373	 loss=1.853930	 lr=0.007192
Epoch[167] Batch [3699]/[3760]	Speed: 117.382989 samples/sec	 accuracy=56.934966	 loss=1.853483	 lr=0.007185
Epoch[167] Batch [3749]/[3760]	Speed: 120.469917 samples/sec	 accuracy=56.937500	 loss=1.853132	 lr=0.007178
Batch [0049]/[0303]: acc-top1=55.250000 acc-top5=77.375000
Batch [0099]/[0303]: acc-top1=52.953125 acc-top5=77.015625
Batch [0149]/[0303]: acc-top1=54.135417 acc-top5=77.239583
Batch [0199]/[0303]: acc-top1=56.921875 acc-top5=78.804688
Batch [0249]/[0303]: acc-top1=56.306250 acc-top5=78.681250
Batch [0299]/[0303]: acc-top1=55.718750 acc-top5=78.640625
[Epoch 167] training: accuracy=56.937749	 loss=1.853043
[Epoch 167] speed: 116 samples/sec	time cost: 2177.117446
[Epoch 167] validation: acc-top1=55.656972 acc-top5=78.599422 loss=2.043897
Epoch[168] Batch [0049]/[3759]	Speed: 87.566472 samples/sec	 accuracy=56.406250	 loss=1.837207	 lr=0.007170
Epoch[168] Batch [0099]/[3759]	Speed: 117.870610 samples/sec	 accuracy=56.296875	 loss=1.847077	 lr=0.007164
Epoch[168] Batch [0149]/[3759]	Speed: 117.404017 samples/sec	 accuracy=56.531250	 loss=1.838609	 lr=0.007157
Epoch[168] Batch [0199]/[3759]	Speed: 117.449091 samples/sec	 accuracy=57.023437	 loss=1.824260	 lr=0.007150
Epoch[168] Batch [0249]/[3759]	Speed: 118.174002 samples/sec	 accuracy=57.200000	 loss=1.817813	 lr=0.007144
Epoch[168] Batch [0299]/[3759]	Speed: 117.819424 samples/sec	 accuracy=57.427083	 loss=1.809831	 lr=0.007137
Epoch[168] Batch [0349]/[3759]	Speed: 117.807307 samples/sec	 accuracy=57.540179	 loss=1.815469	 lr=0.007130
Epoch[168] Batch [0399]/[3759]	Speed: 116.899384 samples/sec	 accuracy=57.492188	 loss=1.816113	 lr=0.007124
Epoch[168] Batch [0449]/[3759]	Speed: 117.685804 samples/sec	 accuracy=57.517361	 loss=1.815902	 lr=0.007117
Epoch[168] Batch [0499]/[3759]	Speed: 117.675198 samples/sec	 accuracy=57.640625	 loss=1.810418	 lr=0.007110
Epoch[168] Batch [0549]/[3759]	Speed: 115.102672 samples/sec	 accuracy=57.767045	 loss=1.806313	 lr=0.007104
Epoch[168] Batch [0599]/[3759]	Speed: 117.241859 samples/sec	 accuracy=57.723958	 loss=1.810181	 lr=0.007097
Epoch[168] Batch [0649]/[3759]	Speed: 117.234966 samples/sec	 accuracy=57.894231	 loss=1.800740	 lr=0.007091
Epoch[168] Batch [0699]/[3759]	Speed: 117.492713 samples/sec	 accuracy=57.986607	 loss=1.797446	 lr=0.007084
Epoch[168] Batch [0749]/[3759]	Speed: 117.082459 samples/sec	 accuracy=57.964583	 loss=1.798294	 lr=0.007077
Epoch[168] Batch [0799]/[3759]	Speed: 117.584481 samples/sec	 accuracy=57.988281	 loss=1.797032	 lr=0.007071
Epoch[168] Batch [0849]/[3759]	Speed: 117.527523 samples/sec	 accuracy=57.920956	 loss=1.800125	 lr=0.007064
Epoch[168] Batch [0899]/[3759]	Speed: 116.608385 samples/sec	 accuracy=57.914931	 loss=1.799522	 lr=0.007058
Epoch[168] Batch [0949]/[3759]	Speed: 117.707085 samples/sec	 accuracy=57.886513	 loss=1.802645	 lr=0.007051
Epoch[168] Batch [0999]/[3759]	Speed: 117.461469 samples/sec	 accuracy=57.826563	 loss=1.804755	 lr=0.007044
Epoch[168] Batch [1049]/[3759]	Speed: 116.790044 samples/sec	 accuracy=57.861607	 loss=1.802549	 lr=0.007038
Epoch[168] Batch [1099]/[3759]	Speed: 116.748062 samples/sec	 accuracy=57.818182	 loss=1.804689	 lr=0.007031
Epoch[168] Batch [1149]/[3759]	Speed: 118.174775 samples/sec	 accuracy=57.815217	 loss=1.807254	 lr=0.007025
Epoch[168] Batch [1199]/[3759]	Speed: 117.560814 samples/sec	 accuracy=57.803385	 loss=1.807273	 lr=0.007018
Epoch[168] Batch [1249]/[3759]	Speed: 117.133328 samples/sec	 accuracy=57.720000	 loss=1.808499	 lr=0.007011
Epoch[168] Batch [1299]/[3759]	Speed: 118.092843 samples/sec	 accuracy=57.748798	 loss=1.807328	 lr=0.007005
Epoch[168] Batch [1349]/[3759]	Speed: 117.191076 samples/sec	 accuracy=57.751157	 loss=1.807944	 lr=0.006998
Epoch[168] Batch [1399]/[3759]	Speed: 117.107989 samples/sec	 accuracy=57.751116	 loss=1.807776	 lr=0.006992
Epoch[168] Batch [1449]/[3759]	Speed: 117.588155 samples/sec	 accuracy=57.725216	 loss=1.808636	 lr=0.006985
Epoch[168] Batch [1499]/[3759]	Speed: 116.825154 samples/sec	 accuracy=57.669792	 loss=1.810147	 lr=0.006978
Epoch[168] Batch [1549]/[3759]	Speed: 117.983421 samples/sec	 accuracy=57.676411	 loss=1.809503	 lr=0.006972
Epoch[168] Batch [1599]/[3759]	Speed: 116.832699 samples/sec	 accuracy=57.678711	 loss=1.810291	 lr=0.006965
Epoch[168] Batch [1649]/[3759]	Speed: 118.428605 samples/sec	 accuracy=57.662879	 loss=1.811136	 lr=0.006959
Epoch[168] Batch [1699]/[3759]	Speed: 117.688366 samples/sec	 accuracy=57.614890	 loss=1.814034	 lr=0.006952
Epoch[168] Batch [1749]/[3759]	Speed: 117.847495 samples/sec	 accuracy=57.600000	 loss=1.814495	 lr=0.006946
Epoch[168] Batch [1799]/[3759]	Speed: 117.848575 samples/sec	 accuracy=57.640625	 loss=1.814179	 lr=0.006939
Epoch[168] Batch [1849]/[3759]	Speed: 117.518682 samples/sec	 accuracy=57.665541	 loss=1.814129	 lr=0.006933
Epoch[168] Batch [1899]/[3759]	Speed: 117.490049 samples/sec	 accuracy=57.666941	 loss=1.814823	 lr=0.006926
Epoch[168] Batch [1949]/[3759]	Speed: 117.985938 samples/sec	 accuracy=57.680288	 loss=1.815339	 lr=0.006919
Epoch[168] Batch [1999]/[3759]	Speed: 115.332360 samples/sec	 accuracy=57.661719	 loss=1.816596	 lr=0.006913
Epoch[168] Batch [2049]/[3759]	Speed: 117.380804 samples/sec	 accuracy=57.655488	 loss=1.816724	 lr=0.006906
Epoch[168] Batch [2099]/[3759]	Speed: 117.096451 samples/sec	 accuracy=57.619792	 loss=1.816951	 lr=0.006900
Epoch[168] Batch [2149]/[3759]	Speed: 117.808738 samples/sec	 accuracy=57.618459	 loss=1.817665	 lr=0.006893
Epoch[168] Batch [2199]/[3759]	Speed: 117.561005 samples/sec	 accuracy=57.611506	 loss=1.818621	 lr=0.006887
Epoch[168] Batch [2249]/[3759]	Speed: 117.485141 samples/sec	 accuracy=57.622222	 loss=1.818255	 lr=0.006880
Epoch[168] Batch [2299]/[3759]	Speed: 117.160624 samples/sec	 accuracy=57.627038	 loss=1.819125	 lr=0.006874
Epoch[168] Batch [2349]/[3759]	Speed: 117.976853 samples/sec	 accuracy=57.628324	 loss=1.818825	 lr=0.006867
Epoch[168] Batch [2399]/[3759]	Speed: 117.437829 samples/sec	 accuracy=57.605469	 loss=1.819683	 lr=0.006861
Epoch[168] Batch [2449]/[3759]	Speed: 117.892734 samples/sec	 accuracy=57.605867	 loss=1.819530	 lr=0.006854
Epoch[168] Batch [2499]/[3759]	Speed: 117.737639 samples/sec	 accuracy=57.596875	 loss=1.820156	 lr=0.006848
Epoch[168] Batch [2549]/[3759]	Speed: 117.731446 samples/sec	 accuracy=57.606618	 loss=1.819798	 lr=0.006841
Epoch[168] Batch [2599]/[3759]	Speed: 117.574794 samples/sec	 accuracy=57.603365	 loss=1.819646	 lr=0.006835
Epoch[168] Batch [2649]/[3759]	Speed: 117.888406 samples/sec	 accuracy=57.625000	 loss=1.818561	 lr=0.006828
Epoch[168] Batch [2699]/[3759]	Speed: 117.596049 samples/sec	 accuracy=57.625000	 loss=1.819365	 lr=0.006822
Epoch[168] Batch [2749]/[3759]	Speed: 117.458124 samples/sec	 accuracy=57.638636	 loss=1.818533	 lr=0.006815
Epoch[168] Batch [2799]/[3759]	Speed: 117.838607 samples/sec	 accuracy=57.662388	 loss=1.817589	 lr=0.006809
Epoch[168] Batch [2849]/[3759]	Speed: 117.341724 samples/sec	 accuracy=57.639254	 loss=1.818766	 lr=0.006802
Epoch[168] Batch [2899]/[3759]	Speed: 117.174169 samples/sec	 accuracy=57.657866	 loss=1.818267	 lr=0.006796
Epoch[168] Batch [2949]/[3759]	Speed: 117.117388 samples/sec	 accuracy=57.670551	 loss=1.818255	 lr=0.006789
Epoch[168] Batch [2999]/[3759]	Speed: 116.888372 samples/sec	 accuracy=57.671354	 loss=1.818563	 lr=0.006783
Epoch[168] Batch [3049]/[3759]	Speed: 117.740257 samples/sec	 accuracy=57.646516	 loss=1.819926	 lr=0.006776
Epoch[168] Batch [3099]/[3759]	Speed: 116.781227 samples/sec	 accuracy=57.670867	 loss=1.818890	 lr=0.006770
Epoch[168] Batch [3149]/[3759]	Speed: 117.349768 samples/sec	 accuracy=57.680556	 loss=1.819059	 lr=0.006763
Epoch[168] Batch [3199]/[3759]	Speed: 117.863220 samples/sec	 accuracy=57.667480	 loss=1.819382	 lr=0.006757
Epoch[168] Batch [3249]/[3759]	Speed: 117.138597 samples/sec	 accuracy=57.662981	 loss=1.820085	 lr=0.006750
Epoch[168] Batch [3299]/[3759]	Speed: 117.647716 samples/sec	 accuracy=57.652462	 loss=1.820851	 lr=0.006744
Epoch[168] Batch [3349]/[3759]	Speed: 117.758462 samples/sec	 accuracy=57.653451	 loss=1.820834	 lr=0.006737
Epoch[168] Batch [3399]/[3759]	Speed: 117.185218 samples/sec	 accuracy=57.631893	 loss=1.821450	 lr=0.006731
Epoch[168] Batch [3449]/[3759]	Speed: 115.653069 samples/sec	 accuracy=57.635417	 loss=1.821773	 lr=0.006724
Epoch[168] Batch [3499]/[3759]	Speed: 117.759865 samples/sec	 accuracy=57.607589	 loss=1.822571	 lr=0.006718
Epoch[168] Batch [3549]/[3759]	Speed: 117.548084 samples/sec	 accuracy=57.595070	 loss=1.822865	 lr=0.006711
Epoch[168] Batch [3599]/[3759]	Speed: 118.009490 samples/sec	 accuracy=57.602431	 loss=1.823092	 lr=0.006705
Epoch[168] Batch [3649]/[3759]	Speed: 117.497188 samples/sec	 accuracy=57.594606	 loss=1.822952	 lr=0.006698
Epoch[168] Batch [3699]/[3759]	Speed: 117.437339 samples/sec	 accuracy=57.592905	 loss=1.822875	 lr=0.006692
Epoch[168] Batch [3749]/[3759]	Speed: 121.438613 samples/sec	 accuracy=57.581250	 loss=1.823274	 lr=0.006686
Batch [0049]/[0303]: acc-top1=55.781250 acc-top5=78.000000
Batch [0099]/[0303]: acc-top1=53.343750 acc-top5=76.937500
Batch [0149]/[0303]: acc-top1=54.843750 acc-top5=77.770833
Batch [0199]/[0303]: acc-top1=57.578125 acc-top5=79.460938
Batch [0249]/[0303]: acc-top1=57.068750 acc-top5=79.281250
Batch [0299]/[0303]: acc-top1=56.484375 acc-top5=79.223958
[Epoch 168] training: accuracy=57.584713	 loss=1.823011
[Epoch 168] speed: 116 samples/sec	time cost: 2174.331985
[Epoch 168] validation: acc-top1=56.435644 acc-top5=79.218234 loss=1.969603
Epoch[169] Batch [0049]/[3760]	Speed: 87.118040 samples/sec	 accuracy=59.906250	 loss=1.723272	 lr=0.006678
Epoch[169] Batch [0099]/[3760]	Speed: 117.401533 samples/sec	 accuracy=59.187500	 loss=1.758241	 lr=0.006672
Epoch[169] Batch [0149]/[3760]	Speed: 117.459460 samples/sec	 accuracy=58.760417	 loss=1.770690	 lr=0.006665
Epoch[169] Batch [0199]/[3760]	Speed: 117.490821 samples/sec	 accuracy=58.757812	 loss=1.763429	 lr=0.006659
Epoch[169] Batch [0249]/[3760]	Speed: 116.943153 samples/sec	 accuracy=58.918750	 loss=1.755504	 lr=0.006652
Epoch[169] Batch [0299]/[3760]	Speed: 117.115760 samples/sec	 accuracy=58.682292	 loss=1.773565	 lr=0.006646
Epoch[169] Batch [0349]/[3760]	Speed: 118.302986 samples/sec	 accuracy=58.843750	 loss=1.763413	 lr=0.006639
Epoch[169] Batch [0399]/[3760]	Speed: 117.417750 samples/sec	 accuracy=58.769531	 loss=1.767260	 lr=0.006633
Epoch[169] Batch [0449]/[3760]	Speed: 117.051937 samples/sec	 accuracy=58.635417	 loss=1.773318	 lr=0.006627
Epoch[169] Batch [0499]/[3760]	Speed: 117.351456 samples/sec	 accuracy=58.587500	 loss=1.777373	 lr=0.006620
Epoch[169] Batch [0549]/[3760]	Speed: 117.334519 samples/sec	 accuracy=58.667614	 loss=1.772238	 lr=0.006614
Epoch[169] Batch [0599]/[3760]	Speed: 118.363957 samples/sec	 accuracy=58.622396	 loss=1.771445	 lr=0.006607
Epoch[169] Batch [0649]/[3760]	Speed: 117.989185 samples/sec	 accuracy=58.725962	 loss=1.768466	 lr=0.006601
Epoch[169] Batch [0699]/[3760]	Speed: 117.749918 samples/sec	 accuracy=58.839286	 loss=1.765046	 lr=0.006595
Epoch[169] Batch [0749]/[3760]	Speed: 117.169238 samples/sec	 accuracy=58.741667	 loss=1.769235	 lr=0.006588
Epoch[169] Batch [0799]/[3760]	Speed: 116.985850 samples/sec	 accuracy=58.751953	 loss=1.768979	 lr=0.006582
Epoch[169] Batch [0849]/[3760]	Speed: 117.423308 samples/sec	 accuracy=58.731618	 loss=1.769026	 lr=0.006575
Epoch[169] Batch [0899]/[3760]	Speed: 117.329057 samples/sec	 accuracy=58.612847	 loss=1.774491	 lr=0.006569
Epoch[169] Batch [0949]/[3760]	Speed: 117.640343 samples/sec	 accuracy=58.488487	 loss=1.778323	 lr=0.006563
Epoch[169] Batch [0999]/[3760]	Speed: 117.609771 samples/sec	 accuracy=58.485937	 loss=1.778600	 lr=0.006556
Epoch[169] Batch [1049]/[3760]	Speed: 117.903489 samples/sec	 accuracy=58.477679	 loss=1.779303	 lr=0.006550
Epoch[169] Batch [1099]/[3760]	Speed: 118.144766 samples/sec	 accuracy=58.458807	 loss=1.781497	 lr=0.006543
Epoch[169] Batch [1149]/[3760]	Speed: 114.877503 samples/sec	 accuracy=58.385870	 loss=1.784181	 lr=0.006537
Epoch[169] Batch [1199]/[3760]	Speed: 117.631918 samples/sec	 accuracy=58.372396	 loss=1.784637	 lr=0.006531
Epoch[169] Batch [1249]/[3760]	Speed: 117.781625 samples/sec	 accuracy=58.361250	 loss=1.786327	 lr=0.006524
Epoch[169] Batch [1299]/[3760]	Speed: 117.772227 samples/sec	 accuracy=58.332933	 loss=1.787733	 lr=0.006518
Epoch[169] Batch [1349]/[3760]	Speed: 117.171778 samples/sec	 accuracy=58.304398	 loss=1.788400	 lr=0.006512
Epoch[169] Batch [1399]/[3760]	Speed: 117.114597 samples/sec	 accuracy=58.299107	 loss=1.788978	 lr=0.006505
Epoch[169] Batch [1449]/[3760]	Speed: 117.519961 samples/sec	 accuracy=58.312500	 loss=1.789558	 lr=0.006499
Epoch[169] Batch [1499]/[3760]	Speed: 117.756964 samples/sec	 accuracy=58.279167	 loss=1.790843	 lr=0.006492
Epoch[169] Batch [1549]/[3760]	Speed: 117.605613 samples/sec	 accuracy=58.314516	 loss=1.789901	 lr=0.006486
Epoch[169] Batch [1599]/[3760]	Speed: 117.255414 samples/sec	 accuracy=58.294922	 loss=1.789109	 lr=0.006480
Epoch[169] Batch [1649]/[3760]	Speed: 117.618298 samples/sec	 accuracy=58.335227	 loss=1.788647	 lr=0.006473
Epoch[169] Batch [1699]/[3760]	Speed: 118.521604 samples/sec	 accuracy=58.315257	 loss=1.789478	 lr=0.006467
Epoch[169] Batch [1749]/[3760]	Speed: 116.935136 samples/sec	 accuracy=58.335714	 loss=1.788735	 lr=0.006461
Epoch[169] Batch [1799]/[3760]	Speed: 117.575250 samples/sec	 accuracy=58.296007	 loss=1.790443	 lr=0.006454
Epoch[169] Batch [1849]/[3760]	Speed: 117.770850 samples/sec	 accuracy=58.277027	 loss=1.792288	 lr=0.006448
Epoch[169] Batch [1899]/[3760]	Speed: 117.760890 samples/sec	 accuracy=58.282895	 loss=1.792213	 lr=0.006442
Epoch[169] Batch [1949]/[3760]	Speed: 117.364819 samples/sec	 accuracy=58.274038	 loss=1.792516	 lr=0.006435
Epoch[169] Batch [1999]/[3760]	Speed: 117.221029 samples/sec	 accuracy=58.268750	 loss=1.792677	 lr=0.006429
Epoch[169] Batch [2049]/[3760]	Speed: 117.137364 samples/sec	 accuracy=58.265244	 loss=1.793425	 lr=0.006423
Epoch[169] Batch [2099]/[3760]	Speed: 117.896773 samples/sec	 accuracy=58.272321	 loss=1.793104	 lr=0.006416
Epoch[169] Batch [2149]/[3760]	Speed: 117.615747 samples/sec	 accuracy=58.258721	 loss=1.794463	 lr=0.006410
Epoch[169] Batch [2199]/[3760]	Speed: 117.269322 samples/sec	 accuracy=58.238636	 loss=1.794535	 lr=0.006404
Epoch[169] Batch [2249]/[3760]	Speed: 116.677610 samples/sec	 accuracy=58.234722	 loss=1.794314	 lr=0.006397
Epoch[169] Batch [2299]/[3760]	Speed: 117.742514 samples/sec	 accuracy=58.202446	 loss=1.795474	 lr=0.006391
Epoch[169] Batch [2349]/[3760]	Speed: 117.710560 samples/sec	 accuracy=58.198803	 loss=1.794523	 lr=0.006385
Epoch[169] Batch [2399]/[3760]	Speed: 117.567089 samples/sec	 accuracy=58.194010	 loss=1.794075	 lr=0.006379
Epoch[169] Batch [2449]/[3760]	Speed: 117.308910 samples/sec	 accuracy=58.227041	 loss=1.792802	 lr=0.006372
Epoch[169] Batch [2499]/[3760]	Speed: 117.330403 samples/sec	 accuracy=58.228125	 loss=1.793238	 lr=0.006366
Epoch[169] Batch [2549]/[3760]	Speed: 117.866627 samples/sec	 accuracy=58.265319	 loss=1.792052	 lr=0.006360
Epoch[169] Batch [2599]/[3760]	Speed: 117.671024 samples/sec	 accuracy=58.285457	 loss=1.791694	 lr=0.006353
Epoch[169] Batch [2649]/[3760]	Speed: 115.253323 samples/sec	 accuracy=58.241745	 loss=1.793448	 lr=0.006347
Epoch[169] Batch [2699]/[3760]	Speed: 117.495143 samples/sec	 accuracy=58.241319	 loss=1.792966	 lr=0.006341
Epoch[169] Batch [2749]/[3760]	Speed: 117.735009 samples/sec	 accuracy=58.247159	 loss=1.792492	 lr=0.006334
Epoch[169] Batch [2799]/[3760]	Speed: 117.922453 samples/sec	 accuracy=58.229353	 loss=1.793229	 lr=0.006328
Epoch[169] Batch [2849]/[3760]	Speed: 117.062623 samples/sec	 accuracy=58.222039	 loss=1.793020	 lr=0.006322
Epoch[169] Batch [2899]/[3760]	Speed: 117.060399 samples/sec	 accuracy=58.231142	 loss=1.793516	 lr=0.006316
Epoch[169] Batch [2949]/[3760]	Speed: 117.472369 samples/sec	 accuracy=58.239936	 loss=1.793591	 lr=0.006309
Epoch[169] Batch [2999]/[3760]	Speed: 117.508901 samples/sec	 accuracy=58.264583	 loss=1.792332	 lr=0.006303
Epoch[169] Batch [3049]/[3760]	Speed: 117.866933 samples/sec	 accuracy=58.298668	 loss=1.791595	 lr=0.006297
Epoch[169] Batch [3099]/[3760]	Speed: 117.456161 samples/sec	 accuracy=58.273185	 loss=1.793233	 lr=0.006291
Epoch[169] Batch [3149]/[3760]	Speed: 117.382730 samples/sec	 accuracy=58.291667	 loss=1.792389	 lr=0.006284
Epoch[169] Batch [3199]/[3760]	Speed: 117.575811 samples/sec	 accuracy=58.293945	 loss=1.792558	 lr=0.006278
Epoch[169] Batch [3249]/[3760]	Speed: 117.388973 samples/sec	 accuracy=58.320192	 loss=1.792058	 lr=0.006272
Epoch[169] Batch [3299]/[3760]	Speed: 117.504901 samples/sec	 accuracy=58.302083	 loss=1.792646	 lr=0.006266
Epoch[169] Batch [3349]/[3760]	Speed: 117.480651 samples/sec	 accuracy=58.299907	 loss=1.792661	 lr=0.006259
Epoch[169] Batch [3399]/[3760]	Speed: 116.431237 samples/sec	 accuracy=58.276195	 loss=1.793180	 lr=0.006253
Epoch[169] Batch [3449]/[3760]	Speed: 117.018879 samples/sec	 accuracy=58.258605	 loss=1.793661	 lr=0.006247
Epoch[169] Batch [3499]/[3760]	Speed: 116.999923 samples/sec	 accuracy=58.254018	 loss=1.794187	 lr=0.006241
Epoch[169] Batch [3549]/[3760]	Speed: 116.865248 samples/sec	 accuracy=58.251761	 loss=1.794573	 lr=0.006234
Epoch[169] Batch [3599]/[3760]	Speed: 117.569937 samples/sec	 accuracy=58.237413	 loss=1.795111	 lr=0.006228
Epoch[169] Batch [3649]/[3760]	Speed: 117.391070 samples/sec	 accuracy=58.241010	 loss=1.795336	 lr=0.006222
Epoch[169] Batch [3699]/[3760]	Speed: 117.829634 samples/sec	 accuracy=58.233108	 loss=1.795515	 lr=0.006216
Epoch[169] Batch [3749]/[3760]	Speed: 120.460189 samples/sec	 accuracy=58.223750	 loss=1.795919	 lr=0.006209
Batch [0049]/[0303]: acc-top1=56.125000 acc-top5=76.781250
Batch [0099]/[0303]: acc-top1=53.312500 acc-top5=76.625000
Batch [0149]/[0303]: acc-top1=54.166667 acc-top5=77.364583
Batch [0199]/[0303]: acc-top1=56.914062 acc-top5=78.960938
Batch [0249]/[0303]: acc-top1=56.312500 acc-top5=78.706250
Batch [0299]/[0303]: acc-top1=55.911458 acc-top5=78.677083
[Epoch 169] training: accuracy=58.218916	 loss=1.795675
[Epoch 169] speed: 116 samples/sec	time cost: 2174.878059
[Epoch 169] validation: acc-top1=55.894183 acc-top5=78.712871 loss=2.019883
Epoch[170] Batch [0049]/[3760]	Speed: 87.138326 samples/sec	 accuracy=60.468750	 loss=1.704439	 lr=0.006202
Epoch[170] Batch [0099]/[3760]	Speed: 117.166259 samples/sec	 accuracy=59.375000	 loss=1.753280	 lr=0.006196
Epoch[170] Batch [0149]/[3760]	Speed: 117.224553 samples/sec	 accuracy=59.250000	 loss=1.767374	 lr=0.006189
Epoch[170] Batch [0199]/[3760]	Speed: 117.483485 samples/sec	 accuracy=59.070312	 loss=1.771495	 lr=0.006183
Epoch[170] Batch [0249]/[3760]	Speed: 117.325229 samples/sec	 accuracy=59.062500	 loss=1.770721	 lr=0.006177
Epoch[170] Batch [0299]/[3760]	Speed: 116.771768 samples/sec	 accuracy=59.093750	 loss=1.757641	 lr=0.006171
Epoch[170] Batch [0349]/[3760]	Speed: 115.108928 samples/sec	 accuracy=59.026786	 loss=1.755280	 lr=0.006165
Epoch[170] Batch [0399]/[3760]	Speed: 117.043361 samples/sec	 accuracy=58.906250	 loss=1.759347	 lr=0.006158
Epoch[170] Batch [0449]/[3760]	Speed: 116.777843 samples/sec	 accuracy=58.951389	 loss=1.760507	 lr=0.006152
Epoch[170] Batch [0499]/[3760]	Speed: 116.570911 samples/sec	 accuracy=58.765625	 loss=1.767312	 lr=0.006146
Epoch[170] Batch [0549]/[3760]	Speed: 117.519796 samples/sec	 accuracy=58.812500	 loss=1.759705	 lr=0.006140
Epoch[170] Batch [0599]/[3760]	Speed: 117.858853 samples/sec	 accuracy=58.919271	 loss=1.755777	 lr=0.006134
Epoch[170] Batch [0649]/[3760]	Speed: 117.020330 samples/sec	 accuracy=59.043269	 loss=1.751798	 lr=0.006127
Epoch[170] Batch [0699]/[3760]	Speed: 117.472473 samples/sec	 accuracy=59.015625	 loss=1.750683	 lr=0.006121
Epoch[170] Batch [0749]/[3760]	Speed: 117.273359 samples/sec	 accuracy=59.020833	 loss=1.748947	 lr=0.006115
Epoch[170] Batch [0799]/[3760]	Speed: 117.576385 samples/sec	 accuracy=58.884766	 loss=1.756811	 lr=0.006109
Epoch[170] Batch [0849]/[3760]	Speed: 116.903074 samples/sec	 accuracy=58.869485	 loss=1.757827	 lr=0.006103
Epoch[170] Batch [0899]/[3760]	Speed: 117.685428 samples/sec	 accuracy=58.920139	 loss=1.756769	 lr=0.006097
Epoch[170] Batch [0949]/[3760]	Speed: 117.407929 samples/sec	 accuracy=58.878289	 loss=1.759050	 lr=0.006090
Epoch[170] Batch [0999]/[3760]	Speed: 117.103817 samples/sec	 accuracy=58.843750	 loss=1.762094	 lr=0.006084
Epoch[170] Batch [1049]/[3760]	Speed: 118.053882 samples/sec	 accuracy=58.863095	 loss=1.759702	 lr=0.006078
Epoch[170] Batch [1099]/[3760]	Speed: 116.884261 samples/sec	 accuracy=58.866477	 loss=1.759664	 lr=0.006072
Epoch[170] Batch [1149]/[3760]	Speed: 117.470446 samples/sec	 accuracy=58.834239	 loss=1.761543	 lr=0.006066
Epoch[170] Batch [1199]/[3760]	Speed: 117.615885 samples/sec	 accuracy=58.830729	 loss=1.761537	 lr=0.006060
Epoch[170] Batch [1249]/[3760]	Speed: 117.241301 samples/sec	 accuracy=58.780000	 loss=1.762154	 lr=0.006053
Epoch[170] Batch [1299]/[3760]	Speed: 117.542326 samples/sec	 accuracy=58.760817	 loss=1.762526	 lr=0.006047
Epoch[170] Batch [1349]/[3760]	Speed: 117.931412 samples/sec	 accuracy=58.758102	 loss=1.762520	 lr=0.006041
Epoch[170] Batch [1399]/[3760]	Speed: 117.740571 samples/sec	 accuracy=58.763393	 loss=1.762301	 lr=0.006035
Epoch[170] Batch [1449]/[3760]	Speed: 117.559713 samples/sec	 accuracy=58.774784	 loss=1.762329	 lr=0.006029
Epoch[170] Batch [1499]/[3760]	Speed: 117.510526 samples/sec	 accuracy=58.782292	 loss=1.762450	 lr=0.006023
Epoch[170] Batch [1549]/[3760]	Speed: 117.027512 samples/sec	 accuracy=58.775202	 loss=1.763193	 lr=0.006017
Epoch[170] Batch [1599]/[3760]	Speed: 117.995112 samples/sec	 accuracy=58.776367	 loss=1.764016	 lr=0.006010
Epoch[170] Batch [1649]/[3760]	Speed: 117.622989 samples/sec	 accuracy=58.783144	 loss=1.763423	 lr=0.006004
Epoch[170] Batch [1699]/[3760]	Speed: 117.248390 samples/sec	 accuracy=58.763787	 loss=1.763804	 lr=0.005998
Epoch[170] Batch [1749]/[3760]	Speed: 117.406141 samples/sec	 accuracy=58.774107	 loss=1.763675	 lr=0.005992
Epoch[170] Batch [1799]/[3760]	Speed: 115.137525 samples/sec	 accuracy=58.799479	 loss=1.763079	 lr=0.005986
Epoch[170] Batch [1849]/[3760]	Speed: 117.511329 samples/sec	 accuracy=58.834459	 loss=1.762033	 lr=0.005980
Epoch[170] Batch [1899]/[3760]	Speed: 116.978242 samples/sec	 accuracy=58.799342	 loss=1.764226	 lr=0.005974
Epoch[170] Batch [1949]/[3760]	Speed: 117.244403 samples/sec	 accuracy=58.811699	 loss=1.764540	 lr=0.005968
Epoch[170] Batch [1999]/[3760]	Speed: 117.304670 samples/sec	 accuracy=58.800781	 loss=1.764902	 lr=0.005962
Epoch[170] Batch [2049]/[3760]	Speed: 118.218194 samples/sec	 accuracy=58.781250	 loss=1.766051	 lr=0.005955
Epoch[170] Batch [2099]/[3760]	Speed: 117.766304 samples/sec	 accuracy=58.779762	 loss=1.766627	 lr=0.005949
Epoch[170] Batch [2149]/[3760]	Speed: 117.713296 samples/sec	 accuracy=58.824855	 loss=1.765187	 lr=0.005943
Epoch[170] Batch [2199]/[3760]	Speed: 117.916101 samples/sec	 accuracy=58.879261	 loss=1.763825	 lr=0.005937
Epoch[170] Batch [2249]/[3760]	Speed: 117.847353 samples/sec	 accuracy=58.927083	 loss=1.763097	 lr=0.005931
Epoch[170] Batch [2299]/[3760]	Speed: 117.093182 samples/sec	 accuracy=58.902853	 loss=1.763461	 lr=0.005925
Epoch[170] Batch [2349]/[3760]	Speed: 117.542505 samples/sec	 accuracy=58.875665	 loss=1.764614	 lr=0.005919
Epoch[170] Batch [2399]/[3760]	Speed: 117.423828 samples/sec	 accuracy=58.888672	 loss=1.764156	 lr=0.005913
Epoch[170] Batch [2449]/[3760]	Speed: 117.456516 samples/sec	 accuracy=58.935587	 loss=1.762323	 lr=0.005907
Epoch[170] Batch [2499]/[3760]	Speed: 118.289430 samples/sec	 accuracy=58.930000	 loss=1.762625	 lr=0.005901
Epoch[170] Batch [2549]/[3760]	Speed: 117.280970 samples/sec	 accuracy=58.925245	 loss=1.762478	 lr=0.005895
Epoch[170] Batch [2599]/[3760]	Speed: 117.491436 samples/sec	 accuracy=58.899639	 loss=1.762917	 lr=0.005888
Epoch[170] Batch [2649]/[3760]	Speed: 116.323295 samples/sec	 accuracy=58.931014	 loss=1.761570	 lr=0.005882
Epoch[170] Batch [2699]/[3760]	Speed: 116.975284 samples/sec	 accuracy=58.949074	 loss=1.760602	 lr=0.005876
Epoch[170] Batch [2749]/[3760]	Speed: 117.810784 samples/sec	 accuracy=58.928977	 loss=1.761952	 lr=0.005870
Epoch[170] Batch [2799]/[3760]	Speed: 117.644802 samples/sec	 accuracy=58.925223	 loss=1.762197	 lr=0.005864
Epoch[170] Batch [2849]/[3760]	Speed: 118.033245 samples/sec	 accuracy=58.916667	 loss=1.762628	 lr=0.005858
Epoch[170] Batch [2899]/[3760]	Speed: 117.276591 samples/sec	 accuracy=58.906789	 loss=1.763104	 lr=0.005852
Epoch[170] Batch [2949]/[3760]	Speed: 118.125374 samples/sec	 accuracy=58.909958	 loss=1.762961	 lr=0.005846
Epoch[170] Batch [2999]/[3760]	Speed: 117.448232 samples/sec	 accuracy=58.886979	 loss=1.764001	 lr=0.005840
Epoch[170] Batch [3049]/[3760]	Speed: 117.443831 samples/sec	 accuracy=58.883709	 loss=1.764101	 lr=0.005834
Epoch[170] Batch [3099]/[3760]	Speed: 117.730267 samples/sec	 accuracy=58.876512	 loss=1.763780	 lr=0.005828
Epoch[170] Batch [3149]/[3760]	Speed: 117.127975 samples/sec	 accuracy=58.884921	 loss=1.763461	 lr=0.005822
Epoch[170] Batch [3199]/[3760]	Speed: 117.984896 samples/sec	 accuracy=58.886230	 loss=1.763275	 lr=0.005816
Epoch[170] Batch [3249]/[3760]	Speed: 115.177153 samples/sec	 accuracy=58.873077	 loss=1.763621	 lr=0.005810
Epoch[170] Batch [3299]/[3760]	Speed: 117.651899 samples/sec	 accuracy=58.869318	 loss=1.763345	 lr=0.005804
Epoch[170] Batch [3349]/[3760]	Speed: 116.881551 samples/sec	 accuracy=58.865205	 loss=1.762963	 lr=0.005798
Epoch[170] Batch [3399]/[3760]	Speed: 117.768733 samples/sec	 accuracy=58.870404	 loss=1.763459	 lr=0.005792
Epoch[170] Batch [3449]/[3760]	Speed: 117.102342 samples/sec	 accuracy=58.858243	 loss=1.763940	 lr=0.005786
Epoch[170] Batch [3499]/[3760]	Speed: 117.131510 samples/sec	 accuracy=58.862500	 loss=1.763876	 lr=0.005780
Epoch[170] Batch [3549]/[3760]	Speed: 117.877388 samples/sec	 accuracy=58.847711	 loss=1.764643	 lr=0.005774
Epoch[170] Batch [3599]/[3760]	Speed: 117.827579 samples/sec	 accuracy=58.852431	 loss=1.764691	 lr=0.005768
Epoch[170] Batch [3649]/[3760]	Speed: 117.637776 samples/sec	 accuracy=58.851884	 loss=1.764620	 lr=0.005762
Epoch[170] Batch [3699]/[3760]	Speed: 117.625596 samples/sec	 accuracy=58.845861	 loss=1.765007	 lr=0.005756
Epoch[170] Batch [3749]/[3760]	Speed: 120.310283 samples/sec	 accuracy=58.855000	 loss=1.764744	 lr=0.005750
Batch [0049]/[0303]: acc-top1=55.906250 acc-top5=77.812500
Batch [0099]/[0303]: acc-top1=53.171875 acc-top5=76.765625
Batch [0149]/[0303]: acc-top1=54.739583 acc-top5=77.354167
Batch [0199]/[0303]: acc-top1=57.484375 acc-top5=78.945312
Batch [0249]/[0303]: acc-top1=56.837500 acc-top5=78.787500
Batch [0299]/[0303]: acc-top1=56.104167 acc-top5=78.859375
[Epoch 170] training: accuracy=58.855967	 loss=1.764510
[Epoch 170] speed: 116 samples/sec	time cost: 2176.356266
[Epoch 170] validation: acc-top1=56.084983 acc-top5=78.867574 loss=2.002192
Epoch[171] Batch [0049]/[3759]	Speed: 86.991614 samples/sec	 accuracy=60.187500	 loss=1.714602	 lr=0.005742
Epoch[171] Batch [0099]/[3759]	Speed: 117.846595 samples/sec	 accuracy=60.000000	 loss=1.723903	 lr=0.005736
Epoch[171] Batch [0149]/[3759]	Speed: 117.499344 samples/sec	 accuracy=60.468750	 loss=1.705502	 lr=0.005730
Epoch[171] Batch [0199]/[3759]	Speed: 117.646564 samples/sec	 accuracy=60.093750	 loss=1.717030	 lr=0.005724
Epoch[171] Batch [0249]/[3759]	Speed: 117.124664 samples/sec	 accuracy=60.318750	 loss=1.710484	 lr=0.005718
Epoch[171] Batch [0299]/[3759]	Speed: 118.100386 samples/sec	 accuracy=60.218750	 loss=1.719190	 lr=0.005712
Epoch[171] Batch [0349]/[3759]	Speed: 117.093260 samples/sec	 accuracy=60.044643	 loss=1.712737	 lr=0.005706
Epoch[171] Batch [0399]/[3759]	Speed: 117.904498 samples/sec	 accuracy=59.960938	 loss=1.717145	 lr=0.005700
Epoch[171] Batch [0449]/[3759]	Speed: 117.190408 samples/sec	 accuracy=60.013889	 loss=1.716057	 lr=0.005694
Epoch[171] Batch [0499]/[3759]	Speed: 117.829163 samples/sec	 accuracy=60.065625	 loss=1.714965	 lr=0.005688
Epoch[171] Batch [0549]/[3759]	Speed: 117.353663 samples/sec	 accuracy=59.875000	 loss=1.720914	 lr=0.005682
Epoch[171] Batch [0599]/[3759]	Speed: 117.827411 samples/sec	 accuracy=59.848958	 loss=1.722749	 lr=0.005677
Epoch[171] Batch [0649]/[3759]	Speed: 117.665956 samples/sec	 accuracy=59.918269	 loss=1.722038	 lr=0.005671
Epoch[171] Batch [0699]/[3759]	Speed: 117.784663 samples/sec	 accuracy=59.796875	 loss=1.725128	 lr=0.005665
Epoch[171] Batch [0749]/[3759]	Speed: 117.329667 samples/sec	 accuracy=59.735417	 loss=1.726044	 lr=0.005659
Epoch[171] Batch [0799]/[3759]	Speed: 117.024846 samples/sec	 accuracy=59.708984	 loss=1.724982	 lr=0.005653
Epoch[171] Batch [0849]/[3759]	Speed: 115.709101 samples/sec	 accuracy=59.669118	 loss=1.728705	 lr=0.005647
Epoch[171] Batch [0899]/[3759]	Speed: 117.134066 samples/sec	 accuracy=59.605903	 loss=1.728428	 lr=0.005641
Epoch[171] Batch [0949]/[3759]	Speed: 118.038799 samples/sec	 accuracy=59.608553	 loss=1.727844	 lr=0.005635
Epoch[171] Batch [0999]/[3759]	Speed: 117.962314 samples/sec	 accuracy=59.631250	 loss=1.729033	 lr=0.005629
Epoch[171] Batch [1049]/[3759]	Speed: 117.506117 samples/sec	 accuracy=59.660714	 loss=1.727065	 lr=0.005623
Epoch[171] Batch [1099]/[3759]	Speed: 117.970237 samples/sec	 accuracy=59.626420	 loss=1.728285	 lr=0.005617
Epoch[171] Batch [1149]/[3759]	Speed: 116.280715 samples/sec	 accuracy=59.648098	 loss=1.727506	 lr=0.005611
Epoch[171] Batch [1199]/[3759]	Speed: 117.417077 samples/sec	 accuracy=59.605469	 loss=1.729432	 lr=0.005605
Epoch[171] Batch [1249]/[3759]	Speed: 117.203320 samples/sec	 accuracy=59.571250	 loss=1.730035	 lr=0.005599
Epoch[171] Batch [1299]/[3759]	Speed: 117.254405 samples/sec	 accuracy=59.585337	 loss=1.729524	 lr=0.005593
Epoch[171] Batch [1349]/[3759]	Speed: 117.190004 samples/sec	 accuracy=59.545139	 loss=1.731523	 lr=0.005587
Epoch[171] Batch [1399]/[3759]	Speed: 117.741621 samples/sec	 accuracy=59.575893	 loss=1.730680	 lr=0.005581
Epoch[171] Batch [1449]/[3759]	Speed: 117.250085 samples/sec	 accuracy=59.512931	 loss=1.733181	 lr=0.005575
Epoch[171] Batch [1499]/[3759]	Speed: 117.037917 samples/sec	 accuracy=59.501042	 loss=1.733482	 lr=0.005570
Epoch[171] Batch [1549]/[3759]	Speed: 117.840263 samples/sec	 accuracy=59.498992	 loss=1.731531	 lr=0.005564
Epoch[171] Batch [1599]/[3759]	Speed: 117.405510 samples/sec	 accuracy=59.496094	 loss=1.731221	 lr=0.005558
Epoch[171] Batch [1649]/[3759]	Speed: 116.398174 samples/sec	 accuracy=59.418561	 loss=1.734558	 lr=0.005552
Epoch[171] Batch [1699]/[3759]	Speed: 117.705628 samples/sec	 accuracy=59.424632	 loss=1.733879	 lr=0.005546
Epoch[171] Batch [1749]/[3759]	Speed: 117.835985 samples/sec	 accuracy=59.456250	 loss=1.732882	 lr=0.005540
Epoch[171] Batch [1799]/[3759]	Speed: 116.978876 samples/sec	 accuracy=59.466146	 loss=1.733281	 lr=0.005534
Epoch[171] Batch [1849]/[3759]	Speed: 117.445549 samples/sec	 accuracy=59.479730	 loss=1.732074	 lr=0.005528
Epoch[171] Batch [1899]/[3759]	Speed: 117.092252 samples/sec	 accuracy=59.486020	 loss=1.732720	 lr=0.005522
Epoch[171] Batch [1949]/[3759]	Speed: 116.976773 samples/sec	 accuracy=59.440705	 loss=1.734099	 lr=0.005516
Epoch[171] Batch [1999]/[3759]	Speed: 117.490293 samples/sec	 accuracy=59.433594	 loss=1.733822	 lr=0.005511
Epoch[171] Batch [2049]/[3759]	Speed: 116.990768 samples/sec	 accuracy=59.420732	 loss=1.735454	 lr=0.005505
Epoch[171] Batch [2099]/[3759]	Speed: 117.286729 samples/sec	 accuracy=59.443452	 loss=1.735458	 lr=0.005499
Epoch[171] Batch [2149]/[3759]	Speed: 117.258218 samples/sec	 accuracy=59.453488	 loss=1.735126	 lr=0.005493
Epoch[171] Batch [2199]/[3759]	Speed: 117.566808 samples/sec	 accuracy=59.473722	 loss=1.734170	 lr=0.005487
Epoch[171] Batch [2249]/[3759]	Speed: 117.376333 samples/sec	 accuracy=59.502083	 loss=1.733594	 lr=0.005481
Epoch[171] Batch [2299]/[3759]	Speed: 115.145472 samples/sec	 accuracy=59.519701	 loss=1.733150	 lr=0.005475
Epoch[171] Batch [2349]/[3759]	Speed: 117.488200 samples/sec	 accuracy=59.519282	 loss=1.732013	 lr=0.005469
Epoch[171] Batch [2399]/[3759]	Speed: 117.795148 samples/sec	 accuracy=59.508464	 loss=1.732526	 lr=0.005464
Epoch[171] Batch [2449]/[3759]	Speed: 117.534620 samples/sec	 accuracy=59.512755	 loss=1.732061	 lr=0.005458
Epoch[171] Batch [2499]/[3759]	Speed: 117.019517 samples/sec	 accuracy=59.503125	 loss=1.733021	 lr=0.005452
Epoch[171] Batch [2549]/[3759]	Speed: 118.220507 samples/sec	 accuracy=59.512868	 loss=1.732300	 lr=0.005446
Epoch[171] Batch [2599]/[3759]	Speed: 117.492870 samples/sec	 accuracy=59.516827	 loss=1.731871	 lr=0.005440
Epoch[171] Batch [2649]/[3759]	Speed: 117.651197 samples/sec	 accuracy=59.516509	 loss=1.732454	 lr=0.005434
Epoch[171] Batch [2699]/[3759]	Speed: 117.034686 samples/sec	 accuracy=59.524306	 loss=1.732036	 lr=0.005428
Epoch[171] Batch [2749]/[3759]	Speed: 117.222205 samples/sec	 accuracy=59.522727	 loss=1.732793	 lr=0.005423
Epoch[171] Batch [2799]/[3759]	Speed: 116.951837 samples/sec	 accuracy=59.522879	 loss=1.732710	 lr=0.005417
Epoch[171] Batch [2849]/[3759]	Speed: 117.113122 samples/sec	 accuracy=59.530154	 loss=1.733142	 lr=0.005411
Epoch[171] Batch [2899]/[3759]	Speed: 117.770068 samples/sec	 accuracy=59.523168	 loss=1.733884	 lr=0.005405
Epoch[171] Batch [2949]/[3759]	Speed: 116.770600 samples/sec	 accuracy=59.525424	 loss=1.733971	 lr=0.005399
Epoch[171] Batch [2999]/[3759]	Speed: 116.925677 samples/sec	 accuracy=59.526042	 loss=1.733631	 lr=0.005393
Epoch[171] Batch [3049]/[3759]	Speed: 117.429195 samples/sec	 accuracy=59.505635	 loss=1.734678	 lr=0.005388
Epoch[171] Batch [3099]/[3759]	Speed: 117.367779 samples/sec	 accuracy=59.489415	 loss=1.735327	 lr=0.005382
Epoch[171] Batch [3149]/[3759]	Speed: 117.353917 samples/sec	 accuracy=59.480655	 loss=1.735839	 lr=0.005376
Epoch[171] Batch [3199]/[3759]	Speed: 117.387541 samples/sec	 accuracy=59.465332	 loss=1.736682	 lr=0.005370
Epoch[171] Batch [3249]/[3759]	Speed: 117.737469 samples/sec	 accuracy=59.469231	 loss=1.736413	 lr=0.005364
Epoch[171] Batch [3299]/[3759]	Speed: 117.512557 samples/sec	 accuracy=59.484375	 loss=1.736625	 lr=0.005359
Epoch[171] Batch [3349]/[3759]	Speed: 117.513194 samples/sec	 accuracy=59.457090	 loss=1.737796	 lr=0.005353
Epoch[171] Batch [3399]/[3759]	Speed: 118.018038 samples/sec	 accuracy=59.460018	 loss=1.737771	 lr=0.005347
Epoch[171] Batch [3449]/[3759]	Speed: 117.797457 samples/sec	 accuracy=59.449728	 loss=1.738375	 lr=0.005341
Epoch[171] Batch [3499]/[3759]	Speed: 118.253093 samples/sec	 accuracy=59.448661	 loss=1.738437	 lr=0.005335
Epoch[171] Batch [3549]/[3759]	Speed: 117.059689 samples/sec	 accuracy=59.444542	 loss=1.738936	 lr=0.005330
Epoch[171] Batch [3599]/[3759]	Speed: 117.709742 samples/sec	 accuracy=59.421007	 loss=1.740059	 lr=0.005324
Epoch[171] Batch [3649]/[3759]	Speed: 117.427968 samples/sec	 accuracy=59.412243	 loss=1.740357	 lr=0.005318
Epoch[171] Batch [3699]/[3759]	Speed: 117.109960 samples/sec	 accuracy=59.409206	 loss=1.740777	 lr=0.005312
Epoch[171] Batch [3749]/[3759]	Speed: 121.261427 samples/sec	 accuracy=59.404167	 loss=1.740985	 lr=0.005306
Batch [0049]/[0303]: acc-top1=55.562500 acc-top5=77.625000
Batch [0099]/[0303]: acc-top1=53.421875 acc-top5=77.281250
Batch [0149]/[0303]: acc-top1=55.500000 acc-top5=78.031250
Batch [0199]/[0303]: acc-top1=58.312500 acc-top5=79.804688
Batch [0249]/[0303]: acc-top1=57.812500 acc-top5=79.668750
Batch [0299]/[0303]: acc-top1=57.250000 acc-top5=79.687500
[Epoch 171] training: accuracy=59.404513	 loss=1.741174
[Epoch 171] speed: 116 samples/sec	time cost: 2175.928563
[Epoch 171] validation: acc-top1=57.193688 acc-top5=79.672030 loss=1.956444
Epoch[172] Batch [0049]/[3760]	Speed: 88.345412 samples/sec	 accuracy=60.781250	 loss=1.693393	 lr=0.005300
Epoch[172] Batch [0099]/[3760]	Speed: 117.646101 samples/sec	 accuracy=60.296875	 loss=1.700644	 lr=0.005294
Epoch[172] Batch [0149]/[3760]	Speed: 117.859135 samples/sec	 accuracy=60.489583	 loss=1.701721	 lr=0.005288
Epoch[172] Batch [0199]/[3760]	Speed: 117.970647 samples/sec	 accuracy=60.750000	 loss=1.687833	 lr=0.005282
Epoch[172] Batch [0249]/[3760]	Speed: 117.166689 samples/sec	 accuracy=60.562500	 loss=1.689672	 lr=0.005276
Epoch[172] Batch [0299]/[3760]	Speed: 117.481095 samples/sec	 accuracy=60.692708	 loss=1.679687	 lr=0.005271
Epoch[172] Batch [0349]/[3760]	Speed: 116.915861 samples/sec	 accuracy=60.571429	 loss=1.680924	 lr=0.005265
Epoch[172] Batch [0399]/[3760]	Speed: 116.921440 samples/sec	 accuracy=60.566406	 loss=1.680959	 lr=0.005259
Epoch[172] Batch [0449]/[3760]	Speed: 117.270638 samples/sec	 accuracy=60.715278	 loss=1.675260	 lr=0.005253
Epoch[172] Batch [0499]/[3760]	Speed: 117.185718 samples/sec	 accuracy=60.609375	 loss=1.679617	 lr=0.005248
Epoch[172] Batch [0549]/[3760]	Speed: 117.021551 samples/sec	 accuracy=60.647727	 loss=1.678211	 lr=0.005242
Epoch[172] Batch [0599]/[3760]	Speed: 117.117152 samples/sec	 accuracy=60.575521	 loss=1.683712	 lr=0.005236
Epoch[172] Batch [0649]/[3760]	Speed: 117.584178 samples/sec	 accuracy=60.521635	 loss=1.687090	 lr=0.005230
Epoch[172] Batch [0699]/[3760]	Speed: 116.885707 samples/sec	 accuracy=60.502232	 loss=1.688520	 lr=0.005225
Epoch[172] Batch [0749]/[3760]	Speed: 117.637998 samples/sec	 accuracy=60.452083	 loss=1.691997	 lr=0.005219
Epoch[172] Batch [0799]/[3760]	Speed: 117.738718 samples/sec	 accuracy=60.427734	 loss=1.694954	 lr=0.005213
Epoch[172] Batch [0849]/[3760]	Speed: 117.390035 samples/sec	 accuracy=60.448529	 loss=1.694785	 lr=0.005207
Epoch[172] Batch [0899]/[3760]	Speed: 116.812532 samples/sec	 accuracy=60.442708	 loss=1.692477	 lr=0.005202
Epoch[172] Batch [0949]/[3760]	Speed: 117.541295 samples/sec	 accuracy=60.381579	 loss=1.695030	 lr=0.005196
Epoch[172] Batch [0999]/[3760]	Speed: 117.644671 samples/sec	 accuracy=60.354688	 loss=1.697762	 lr=0.005190
Epoch[172] Batch [1049]/[3760]	Speed: 117.533053 samples/sec	 accuracy=60.334821	 loss=1.700071	 lr=0.005185
Epoch[172] Batch [1099]/[3760]	Speed: 117.901849 samples/sec	 accuracy=60.342330	 loss=1.699003	 lr=0.005179
Epoch[172] Batch [1149]/[3760]	Speed: 117.448727 samples/sec	 accuracy=60.355978	 loss=1.696169	 lr=0.005173
Epoch[172] Batch [1199]/[3760]	Speed: 117.040650 samples/sec	 accuracy=60.377604	 loss=1.694490	 lr=0.005167
Epoch[172] Batch [1249]/[3760]	Speed: 117.895358 samples/sec	 accuracy=60.370000	 loss=1.695717	 lr=0.005162
Epoch[172] Batch [1299]/[3760]	Speed: 117.380482 samples/sec	 accuracy=60.347356	 loss=1.695397	 lr=0.005156
Epoch[172] Batch [1349]/[3760]	Speed: 117.405845 samples/sec	 accuracy=60.371528	 loss=1.696830	 lr=0.005150
Epoch[172] Batch [1399]/[3760]	Speed: 117.139847 samples/sec	 accuracy=60.380580	 loss=1.695793	 lr=0.005145
Epoch[172] Batch [1449]/[3760]	Speed: 118.167164 samples/sec	 accuracy=60.366379	 loss=1.695952	 lr=0.005139
Epoch[172] Batch [1499]/[3760]	Speed: 117.380856 samples/sec	 accuracy=60.408333	 loss=1.694013	 lr=0.005133
Epoch[172] Batch [1549]/[3760]	Speed: 115.789140 samples/sec	 accuracy=60.372984	 loss=1.694850	 lr=0.005128
Epoch[172] Batch [1599]/[3760]	Speed: 117.833883 samples/sec	 accuracy=60.395508	 loss=1.693800	 lr=0.005122
Epoch[172] Batch [1649]/[3760]	Speed: 117.254719 samples/sec	 accuracy=60.345644	 loss=1.694588	 lr=0.005116
Epoch[172] Batch [1699]/[3760]	Speed: 117.295832 samples/sec	 accuracy=60.369485	 loss=1.693905	 lr=0.005110
Epoch[172] Batch [1749]/[3760]	Speed: 118.085181 samples/sec	 accuracy=60.366964	 loss=1.694167	 lr=0.005105
Epoch[172] Batch [1799]/[3760]	Speed: 116.942395 samples/sec	 accuracy=60.358507	 loss=1.694948	 lr=0.005099
Epoch[172] Batch [1849]/[3760]	Speed: 116.876220 samples/sec	 accuracy=60.348818	 loss=1.695775	 lr=0.005093
Epoch[172] Batch [1899]/[3760]	Speed: 117.652590 samples/sec	 accuracy=60.334704	 loss=1.696153	 lr=0.005088
Epoch[172] Batch [1949]/[3760]	Speed: 116.848548 samples/sec	 accuracy=60.309295	 loss=1.698118	 lr=0.005082
Epoch[172] Batch [1999]/[3760]	Speed: 116.922765 samples/sec	 accuracy=60.304688	 loss=1.698451	 lr=0.005076
Epoch[172] Batch [2049]/[3760]	Speed: 117.659523 samples/sec	 accuracy=60.310976	 loss=1.697895	 lr=0.005071
Epoch[172] Batch [2099]/[3760]	Speed: 117.035980 samples/sec	 accuracy=60.349702	 loss=1.696285	 lr=0.005065
Epoch[172] Batch [2149]/[3760]	Speed: 117.323342 samples/sec	 accuracy=60.348110	 loss=1.696336	 lr=0.005059
Epoch[172] Batch [2199]/[3760]	Speed: 117.253923 samples/sec	 accuracy=60.335938	 loss=1.696900	 lr=0.005054
Epoch[172] Batch [2249]/[3760]	Speed: 117.880124 samples/sec	 accuracy=60.329861	 loss=1.697805	 lr=0.005048
Epoch[172] Batch [2299]/[3760]	Speed: 117.077626 samples/sec	 accuracy=60.326087	 loss=1.698369	 lr=0.005043
Epoch[172] Batch [2349]/[3760]	Speed: 117.408188 samples/sec	 accuracy=60.305186	 loss=1.699754	 lr=0.005037
Epoch[172] Batch [2399]/[3760]	Speed: 118.435793 samples/sec	 accuracy=60.281250	 loss=1.700781	 lr=0.005031
Epoch[172] Batch [2449]/[3760]	Speed: 117.866319 samples/sec	 accuracy=60.257653	 loss=1.701493	 lr=0.005026
Epoch[172] Batch [2499]/[3760]	Speed: 117.546234 samples/sec	 accuracy=60.253750	 loss=1.701790	 lr=0.005020
Epoch[172] Batch [2549]/[3760]	Speed: 117.256375 samples/sec	 accuracy=60.259804	 loss=1.702072	 lr=0.005014
Epoch[172] Batch [2599]/[3760]	Speed: 116.961301 samples/sec	 accuracy=60.254207	 loss=1.701979	 lr=0.005009
Epoch[172] Batch [2649]/[3760]	Speed: 117.447104 samples/sec	 accuracy=60.247052	 loss=1.702204	 lr=0.005003
Epoch[172] Batch [2699]/[3760]	Speed: 116.700627 samples/sec	 accuracy=60.246528	 loss=1.702296	 lr=0.004997
Epoch[172] Batch [2749]/[3760]	Speed: 117.277795 samples/sec	 accuracy=60.248864	 loss=1.701493	 lr=0.004992
Epoch[172] Batch [2799]/[3760]	Speed: 117.083711 samples/sec	 accuracy=60.248884	 loss=1.701631	 lr=0.004986
Epoch[172] Batch [2849]/[3760]	Speed: 116.971777 samples/sec	 accuracy=60.221491	 loss=1.702209	 lr=0.004981
Epoch[172] Batch [2899]/[3760]	Speed: 116.886861 samples/sec	 accuracy=60.239224	 loss=1.700873	 lr=0.004975
Epoch[172] Batch [2949]/[3760]	Speed: 117.309007 samples/sec	 accuracy=60.242055	 loss=1.701376	 lr=0.004969
Epoch[172] Batch [2999]/[3760]	Speed: 116.903466 samples/sec	 accuracy=60.251562	 loss=1.700877	 lr=0.004964
Epoch[172] Batch [3049]/[3760]	Speed: 115.511349 samples/sec	 accuracy=60.275102	 loss=1.699309	 lr=0.004958
Epoch[172] Batch [3099]/[3760]	Speed: 118.015389 samples/sec	 accuracy=60.265625	 loss=1.699778	 lr=0.004953
Epoch[172] Batch [3149]/[3760]	Speed: 117.428628 samples/sec	 accuracy=60.242560	 loss=1.700401	 lr=0.004947
Epoch[172] Batch [3199]/[3760]	Speed: 117.295445 samples/sec	 accuracy=60.245605	 loss=1.699819	 lr=0.004941
Epoch[172] Batch [3249]/[3760]	Speed: 117.754697 samples/sec	 accuracy=60.230769	 loss=1.700896	 lr=0.004936
Epoch[172] Batch [3299]/[3760]	Speed: 117.538184 samples/sec	 accuracy=60.223958	 loss=1.701361	 lr=0.004930
Epoch[172] Batch [3349]/[3760]	Speed: 117.162454 samples/sec	 accuracy=60.243470	 loss=1.700894	 lr=0.004925
Epoch[172] Batch [3399]/[3760]	Speed: 116.955320 samples/sec	 accuracy=60.229779	 loss=1.701238	 lr=0.004919
Epoch[172] Batch [3449]/[3760]	Speed: 116.909807 samples/sec	 accuracy=60.242754	 loss=1.700737	 lr=0.004913
Epoch[172] Batch [3499]/[3760]	Speed: 117.900638 samples/sec	 accuracy=60.250893	 loss=1.701004	 lr=0.004908
Epoch[172] Batch [3549]/[3760]	Speed: 117.357750 samples/sec	 accuracy=60.238116	 loss=1.701358	 lr=0.004902
Epoch[172] Batch [3599]/[3760]	Speed: 117.693702 samples/sec	 accuracy=60.240885	 loss=1.701695	 lr=0.004897
Epoch[172] Batch [3649]/[3760]	Speed: 117.152987 samples/sec	 accuracy=60.223459	 loss=1.702512	 lr=0.004891
Epoch[172] Batch [3699]/[3760]	Speed: 117.290192 samples/sec	 accuracy=60.202280	 loss=1.703674	 lr=0.004886
Epoch[172] Batch [3749]/[3760]	Speed: 120.022288 samples/sec	 accuracy=60.184583	 loss=1.704277	 lr=0.004880
Batch [0049]/[0303]: acc-top1=57.250000 acc-top5=78.312500
Batch [0099]/[0303]: acc-top1=53.843750 acc-top5=77.203125
Batch [0149]/[0303]: acc-top1=55.802083 acc-top5=78.343750
Batch [0199]/[0303]: acc-top1=58.257812 acc-top5=80.093750
Batch [0249]/[0303]: acc-top1=57.831250 acc-top5=79.812500
Batch [0299]/[0303]: acc-top1=57.572917 acc-top5=79.968750
[Epoch 172] training: accuracy=60.180352	 loss=1.704418
[Epoch 172] speed: 116 samples/sec	time cost: 2176.443128
[Epoch 172] validation: acc-top1=57.487624 acc-top5=79.965965 loss=1.934698
Epoch[173] Batch [0049]/[3760]	Speed: 87.111107 samples/sec	 accuracy=60.687500	 loss=1.656857	 lr=0.004873
Epoch[173] Batch [0099]/[3760]	Speed: 117.908354 samples/sec	 accuracy=60.046875	 loss=1.674227	 lr=0.004868
Epoch[173] Batch [0149]/[3760]	Speed: 117.681390 samples/sec	 accuracy=60.739583	 loss=1.657629	 lr=0.004862
Epoch[173] Batch [0199]/[3760]	Speed: 117.000328 samples/sec	 accuracy=60.953125	 loss=1.650455	 lr=0.004857
Epoch[173] Batch [0249]/[3760]	Speed: 116.962440 samples/sec	 accuracy=60.937500	 loss=1.653864	 lr=0.004851
Epoch[173] Batch [0299]/[3760]	Speed: 117.852162 samples/sec	 accuracy=60.989583	 loss=1.664694	 lr=0.004846
Epoch[173] Batch [0349]/[3760]	Speed: 117.252185 samples/sec	 accuracy=61.156250	 loss=1.659017	 lr=0.004840
Epoch[173] Batch [0399]/[3760]	Speed: 117.666345 samples/sec	 accuracy=61.191406	 loss=1.657109	 lr=0.004835
Epoch[173] Batch [0449]/[3760]	Speed: 118.004604 samples/sec	 accuracy=61.291667	 loss=1.656038	 lr=0.004829
Epoch[173] Batch [0499]/[3760]	Speed: 117.339948 samples/sec	 accuracy=61.325000	 loss=1.652625	 lr=0.004824
Epoch[173] Batch [0549]/[3760]	Speed: 117.162451 samples/sec	 accuracy=61.267045	 loss=1.651612	 lr=0.004818
Epoch[173] Batch [0599]/[3760]	Speed: 117.569453 samples/sec	 accuracy=61.252604	 loss=1.655249	 lr=0.004812
Epoch[173] Batch [0649]/[3760]	Speed: 117.720587 samples/sec	 accuracy=61.312500	 loss=1.654644	 lr=0.004807
Epoch[173] Batch [0699]/[3760]	Speed: 117.674089 samples/sec	 accuracy=61.299107	 loss=1.655155	 lr=0.004801
Epoch[173] Batch [0749]/[3760]	Speed: 116.692132 samples/sec	 accuracy=61.160417	 loss=1.662485	 lr=0.004796
Epoch[173] Batch [0799]/[3760]	Speed: 114.214851 samples/sec	 accuracy=61.048828	 loss=1.666607	 lr=0.004790
Epoch[173] Batch [0849]/[3760]	Speed: 117.911708 samples/sec	 accuracy=61.034926	 loss=1.665327	 lr=0.004785
Epoch[173] Batch [0899]/[3760]	Speed: 117.945358 samples/sec	 accuracy=61.043403	 loss=1.665513	 lr=0.004779
Epoch[173] Batch [0949]/[3760]	Speed: 117.115834 samples/sec	 accuracy=61.067434	 loss=1.664902	 lr=0.004774
Epoch[173] Batch [0999]/[3760]	Speed: 117.533675 samples/sec	 accuracy=61.065625	 loss=1.666039	 lr=0.004768
Epoch[173] Batch [1049]/[3760]	Speed: 116.840843 samples/sec	 accuracy=61.093750	 loss=1.665416	 lr=0.004763
Epoch[173] Batch [1099]/[3760]	Speed: 117.605578 samples/sec	 accuracy=61.109375	 loss=1.663801	 lr=0.004757
Epoch[173] Batch [1149]/[3760]	Speed: 117.079511 samples/sec	 accuracy=61.039402	 loss=1.667603	 lr=0.004752
Epoch[173] Batch [1199]/[3760]	Speed: 117.684834 samples/sec	 accuracy=61.089844	 loss=1.666961	 lr=0.004746
Epoch[173] Batch [1249]/[3760]	Speed: 117.560470 samples/sec	 accuracy=61.050000	 loss=1.667117	 lr=0.004741
Epoch[173] Batch [1299]/[3760]	Speed: 116.794390 samples/sec	 accuracy=61.075721	 loss=1.664697	 lr=0.004736
Epoch[173] Batch [1349]/[3760]	Speed: 117.406280 samples/sec	 accuracy=61.081019	 loss=1.663873	 lr=0.004730
Epoch[173] Batch [1399]/[3760]	Speed: 117.520451 samples/sec	 accuracy=61.082589	 loss=1.663072	 lr=0.004725
Epoch[173] Batch [1449]/[3760]	Speed: 116.971305 samples/sec	 accuracy=61.084052	 loss=1.663816	 lr=0.004719
Epoch[173] Batch [1499]/[3760]	Speed: 117.432190 samples/sec	 accuracy=61.064583	 loss=1.664868	 lr=0.004714
Epoch[173] Batch [1549]/[3760]	Speed: 117.410086 samples/sec	 accuracy=61.019153	 loss=1.667137	 lr=0.004708
Epoch[173] Batch [1599]/[3760]	Speed: 117.465395 samples/sec	 accuracy=61.024414	 loss=1.666576	 lr=0.004703
Epoch[173] Batch [1649]/[3760]	Speed: 117.022361 samples/sec	 accuracy=61.011364	 loss=1.666188	 lr=0.004697
Epoch[173] Batch [1699]/[3760]	Speed: 117.915170 samples/sec	 accuracy=60.973346	 loss=1.667456	 lr=0.004692
Epoch[173] Batch [1749]/[3760]	Speed: 117.216540 samples/sec	 accuracy=60.983036	 loss=1.666739	 lr=0.004686
Epoch[173] Batch [1799]/[3760]	Speed: 117.870838 samples/sec	 accuracy=60.976562	 loss=1.667034	 lr=0.004681
Epoch[173] Batch [1849]/[3760]	Speed: 117.286332 samples/sec	 accuracy=60.982264	 loss=1.666799	 lr=0.004675
Epoch[173] Batch [1899]/[3760]	Speed: 117.109899 samples/sec	 accuracy=60.970395	 loss=1.666618	 lr=0.004670
Epoch[173] Batch [1949]/[3760]	Speed: 117.275198 samples/sec	 accuracy=60.951923	 loss=1.668023	 lr=0.004665
Epoch[173] Batch [1999]/[3760]	Speed: 117.168688 samples/sec	 accuracy=60.993750	 loss=1.667188	 lr=0.004659
Epoch[173] Batch [2049]/[3760]	Speed: 117.510645 samples/sec	 accuracy=60.999238	 loss=1.668058	 lr=0.004654
Epoch[173] Batch [2099]/[3760]	Speed: 116.859998 samples/sec	 accuracy=61.006696	 loss=1.668282	 lr=0.004648
Epoch[173] Batch [2149]/[3760]	Speed: 117.588141 samples/sec	 accuracy=60.992733	 loss=1.668916	 lr=0.004643
Epoch[173] Batch [2199]/[3760]	Speed: 117.826372 samples/sec	 accuracy=60.985085	 loss=1.670038	 lr=0.004637
Epoch[173] Batch [2249]/[3760]	Speed: 116.622634 samples/sec	 accuracy=60.990972	 loss=1.669501	 lr=0.004632
Epoch[173] Batch [2299]/[3760]	Speed: 117.743659 samples/sec	 accuracy=60.991848	 loss=1.669825	 lr=0.004627
Epoch[173] Batch [2349]/[3760]	Speed: 115.632116 samples/sec	 accuracy=60.957447	 loss=1.671724	 lr=0.004621
Epoch[173] Batch [2399]/[3760]	Speed: 116.621626 samples/sec	 accuracy=60.932943	 loss=1.672428	 lr=0.004616
Epoch[173] Batch [2449]/[3760]	Speed: 117.729885 samples/sec	 accuracy=60.940689	 loss=1.673440	 lr=0.004610
Epoch[173] Batch [2499]/[3760]	Speed: 118.256443 samples/sec	 accuracy=60.935000	 loss=1.673542	 lr=0.004605
Epoch[173] Batch [2549]/[3760]	Speed: 117.127842 samples/sec	 accuracy=60.920956	 loss=1.673875	 lr=0.004599
Epoch[173] Batch [2599]/[3760]	Speed: 117.370622 samples/sec	 accuracy=60.901442	 loss=1.674287	 lr=0.004594
Epoch[173] Batch [2649]/[3760]	Speed: 116.950230 samples/sec	 accuracy=60.885613	 loss=1.674681	 lr=0.004589
Epoch[173] Batch [2699]/[3760]	Speed: 117.098098 samples/sec	 accuracy=60.872685	 loss=1.674328	 lr=0.004583
Epoch[173] Batch [2749]/[3760]	Speed: 117.359548 samples/sec	 accuracy=60.862500	 loss=1.674769	 lr=0.004578
Epoch[173] Batch [2799]/[3760]	Speed: 117.256231 samples/sec	 accuracy=60.855469	 loss=1.674818	 lr=0.004572
Epoch[173] Batch [2849]/[3760]	Speed: 117.468279 samples/sec	 accuracy=60.845943	 loss=1.675228	 lr=0.004567
Epoch[173] Batch [2899]/[3760]	Speed: 117.595272 samples/sec	 accuracy=60.818427	 loss=1.676584	 lr=0.004562
Epoch[173] Batch [2949]/[3760]	Speed: 117.392202 samples/sec	 accuracy=60.839513	 loss=1.675618	 lr=0.004556
Epoch[173] Batch [2999]/[3760]	Speed: 117.585118 samples/sec	 accuracy=60.838542	 loss=1.675831	 lr=0.004551
Epoch[173] Batch [3049]/[3760]	Speed: 116.884782 samples/sec	 accuracy=60.819160	 loss=1.676373	 lr=0.004546
Epoch[173] Batch [3099]/[3760]	Speed: 116.805696 samples/sec	 accuracy=60.816532	 loss=1.676498	 lr=0.004540
Epoch[173] Batch [3149]/[3760]	Speed: 118.029013 samples/sec	 accuracy=60.801091	 loss=1.676795	 lr=0.004535
Epoch[173] Batch [3199]/[3760]	Speed: 117.090667 samples/sec	 accuracy=60.818848	 loss=1.676834	 lr=0.004529
Epoch[173] Batch [3249]/[3760]	Speed: 117.381691 samples/sec	 accuracy=60.806250	 loss=1.676799	 lr=0.004524
Epoch[173] Batch [3299]/[3760]	Speed: 116.878720 samples/sec	 accuracy=60.793561	 loss=1.677383	 lr=0.004519
Epoch[173] Batch [3349]/[3760]	Speed: 117.527682 samples/sec	 accuracy=60.782649	 loss=1.677633	 lr=0.004513
Epoch[173] Batch [3399]/[3760]	Speed: 117.418731 samples/sec	 accuracy=60.785386	 loss=1.677945	 lr=0.004508
Epoch[173] Batch [3449]/[3760]	Speed: 117.160358 samples/sec	 accuracy=60.795743	 loss=1.677612	 lr=0.004503
Epoch[173] Batch [3499]/[3760]	Speed: 117.682537 samples/sec	 accuracy=60.788839	 loss=1.677577	 lr=0.004497
Epoch[173] Batch [3549]/[3760]	Speed: 116.328419 samples/sec	 accuracy=60.773768	 loss=1.678193	 lr=0.004492
Epoch[173] Batch [3599]/[3760]	Speed: 117.285403 samples/sec	 accuracy=60.765191	 loss=1.678857	 lr=0.004487
Epoch[173] Batch [3649]/[3760]	Speed: 118.759722 samples/sec	 accuracy=60.754281	 loss=1.679326	 lr=0.004481
Epoch[173] Batch [3699]/[3760]	Speed: 117.032990 samples/sec	 accuracy=60.760135	 loss=1.679325	 lr=0.004476
Epoch[173] Batch [3749]/[3760]	Speed: 120.770491 samples/sec	 accuracy=60.737083	 loss=1.680200	 lr=0.004471
Batch [0049]/[0303]: acc-top1=58.062500 acc-top5=79.156250
Batch [0099]/[0303]: acc-top1=55.609375 acc-top5=78.203125
Batch [0149]/[0303]: acc-top1=56.562500 acc-top5=78.739583
Batch [0199]/[0303]: acc-top1=59.085938 acc-top5=80.296875
Batch [0249]/[0303]: acc-top1=58.293750 acc-top5=80.031250
Batch [0299]/[0303]: acc-top1=57.604167 acc-top5=80.052083
[Epoch 173] training: accuracy=60.740525	 loss=1.680370
[Epoch 173] speed: 116 samples/sec	time cost: 2176.454246
[Epoch 173] validation: acc-top1=57.564975 acc-top5=80.048474 loss=1.950310
Epoch[174] Batch [0049]/[3759]	Speed: 85.960135 samples/sec	 accuracy=61.312500	 loss=1.663212	 lr=0.004464
Epoch[174] Batch [0099]/[3759]	Speed: 117.766847 samples/sec	 accuracy=61.312500	 loss=1.659798	 lr=0.004459
Epoch[174] Batch [0149]/[3759]	Speed: 116.913918 samples/sec	 accuracy=60.927083	 loss=1.672939	 lr=0.004454
Epoch[174] Batch [0199]/[3759]	Speed: 117.851596 samples/sec	 accuracy=61.203125	 loss=1.661977	 lr=0.004448
Epoch[174] Batch [0249]/[3759]	Speed: 116.775708 samples/sec	 accuracy=61.418750	 loss=1.649636	 lr=0.004443
Epoch[174] Batch [0299]/[3759]	Speed: 117.431941 samples/sec	 accuracy=61.489583	 loss=1.648242	 lr=0.004438
Epoch[174] Batch [0349]/[3759]	Speed: 117.578892 samples/sec	 accuracy=61.486607	 loss=1.641690	 lr=0.004432
Epoch[174] Batch [0399]/[3759]	Speed: 117.200636 samples/sec	 accuracy=61.359375	 loss=1.643258	 lr=0.004427
Epoch[174] Batch [0449]/[3759]	Speed: 117.265274 samples/sec	 accuracy=61.420139	 loss=1.638545	 lr=0.004422
Epoch[174] Batch [0499]/[3759]	Speed: 117.582397 samples/sec	 accuracy=61.343750	 loss=1.641131	 lr=0.004416
Epoch[174] Batch [0549]/[3759]	Speed: 118.050451 samples/sec	 accuracy=61.437500	 loss=1.639492	 lr=0.004411
Epoch[174] Batch [0599]/[3759]	Speed: 116.995002 samples/sec	 accuracy=61.484375	 loss=1.641595	 lr=0.004406
Epoch[174] Batch [0649]/[3759]	Speed: 117.555149 samples/sec	 accuracy=61.478365	 loss=1.640269	 lr=0.004401
Epoch[174] Batch [0699]/[3759]	Speed: 117.799522 samples/sec	 accuracy=61.466518	 loss=1.639808	 lr=0.004395
Epoch[174] Batch [0749]/[3759]	Speed: 117.340729 samples/sec	 accuracy=61.425000	 loss=1.643059	 lr=0.004390
Epoch[174] Batch [0799]/[3759]	Speed: 117.836916 samples/sec	 accuracy=61.460938	 loss=1.643663	 lr=0.004385
Epoch[174] Batch [0849]/[3759]	Speed: 117.353046 samples/sec	 accuracy=61.555147	 loss=1.642277	 lr=0.004379
Epoch[174] Batch [0899]/[3759]	Speed: 117.220070 samples/sec	 accuracy=61.500000	 loss=1.645271	 lr=0.004374
Epoch[174] Batch [0949]/[3759]	Speed: 117.181032 samples/sec	 accuracy=61.526316	 loss=1.643902	 lr=0.004369
Epoch[174] Batch [0999]/[3759]	Speed: 117.767169 samples/sec	 accuracy=61.432812	 loss=1.646359	 lr=0.004364
Epoch[174] Batch [1049]/[3759]	Speed: 117.706014 samples/sec	 accuracy=61.447917	 loss=1.644467	 lr=0.004358
Epoch[174] Batch [1099]/[3759]	Speed: 116.818800 samples/sec	 accuracy=61.443182	 loss=1.644246	 lr=0.004353
Epoch[174] Batch [1149]/[3759]	Speed: 117.322438 samples/sec	 accuracy=61.429348	 loss=1.645430	 lr=0.004348
Epoch[174] Batch [1199]/[3759]	Speed: 117.485419 samples/sec	 accuracy=61.399740	 loss=1.645890	 lr=0.004343
Epoch[174] Batch [1249]/[3759]	Speed: 118.041925 samples/sec	 accuracy=61.381250	 loss=1.646314	 lr=0.004337
Epoch[174] Batch [1299]/[3759]	Speed: 117.403446 samples/sec	 accuracy=61.338942	 loss=1.648635	 lr=0.004332
Epoch[174] Batch [1349]/[3759]	Speed: 117.412730 samples/sec	 accuracy=61.365741	 loss=1.648259	 lr=0.004327
Epoch[174] Batch [1399]/[3759]	Speed: 118.042009 samples/sec	 accuracy=61.357143	 loss=1.647663	 lr=0.004322
Epoch[174] Batch [1449]/[3759]	Speed: 118.000486 samples/sec	 accuracy=61.340517	 loss=1.648580	 lr=0.004316
Epoch[174] Batch [1499]/[3759]	Speed: 116.952615 samples/sec	 accuracy=61.315625	 loss=1.648937	 lr=0.004311
Epoch[174] Batch [1549]/[3759]	Speed: 114.778273 samples/sec	 accuracy=61.327621	 loss=1.647785	 lr=0.004306
Epoch[174] Batch [1599]/[3759]	Speed: 117.333768 samples/sec	 accuracy=61.323242	 loss=1.646954	 lr=0.004301
Epoch[174] Batch [1649]/[3759]	Speed: 116.824640 samples/sec	 accuracy=61.326705	 loss=1.646613	 lr=0.004295
Epoch[174] Batch [1699]/[3759]	Speed: 116.830962 samples/sec	 accuracy=61.340993	 loss=1.646553	 lr=0.004290
Epoch[174] Batch [1749]/[3759]	Speed: 117.499941 samples/sec	 accuracy=61.341071	 loss=1.647678	 lr=0.004285
Epoch[174] Batch [1799]/[3759]	Speed: 117.205779 samples/sec	 accuracy=61.337674	 loss=1.647256	 lr=0.004280
Epoch[174] Batch [1849]/[3759]	Speed: 118.031086 samples/sec	 accuracy=61.361486	 loss=1.647034	 lr=0.004274
Epoch[174] Batch [1899]/[3759]	Speed: 117.972789 samples/sec	 accuracy=61.396382	 loss=1.646195	 lr=0.004269
Epoch[174] Batch [1949]/[3759]	Speed: 117.519968 samples/sec	 accuracy=61.334135	 loss=1.648951	 lr=0.004264
Epoch[174] Batch [1999]/[3759]	Speed: 117.845343 samples/sec	 accuracy=61.346875	 loss=1.649332	 lr=0.004259
Epoch[174] Batch [2049]/[3759]	Speed: 117.391578 samples/sec	 accuracy=61.358994	 loss=1.649720	 lr=0.004254
Epoch[174] Batch [2099]/[3759]	Speed: 116.870943 samples/sec	 accuracy=61.364583	 loss=1.648877	 lr=0.004248
Epoch[174] Batch [2149]/[3759]	Speed: 117.397573 samples/sec	 accuracy=61.350291	 loss=1.649745	 lr=0.004243
Epoch[174] Batch [2199]/[3759]	Speed: 117.573045 samples/sec	 accuracy=61.322443	 loss=1.650782	 lr=0.004238
Epoch[174] Batch [2249]/[3759]	Speed: 117.402233 samples/sec	 accuracy=61.355556	 loss=1.649757	 lr=0.004233
Epoch[174] Batch [2299]/[3759]	Speed: 117.955905 samples/sec	 accuracy=61.358016	 loss=1.649117	 lr=0.004228
Epoch[174] Batch [2349]/[3759]	Speed: 117.850331 samples/sec	 accuracy=61.358378	 loss=1.648792	 lr=0.004222
Epoch[174] Batch [2399]/[3759]	Speed: 117.368455 samples/sec	 accuracy=61.363281	 loss=1.648899	 lr=0.004217
Epoch[174] Batch [2449]/[3759]	Speed: 118.021162 samples/sec	 accuracy=61.378827	 loss=1.648370	 lr=0.004212
Epoch[174] Batch [2499]/[3759]	Speed: 117.247593 samples/sec	 accuracy=61.380625	 loss=1.648170	 lr=0.004207
Epoch[174] Batch [2549]/[3759]	Speed: 118.250355 samples/sec	 accuracy=61.363971	 loss=1.649049	 lr=0.004202
Epoch[174] Batch [2599]/[3759]	Speed: 116.854694 samples/sec	 accuracy=61.378606	 loss=1.648359	 lr=0.004197
Epoch[174] Batch [2649]/[3759]	Speed: 117.601169 samples/sec	 accuracy=61.392689	 loss=1.647426	 lr=0.004191
Epoch[174] Batch [2699]/[3759]	Speed: 117.324911 samples/sec	 accuracy=61.374421	 loss=1.647638	 lr=0.004186
Epoch[174] Batch [2749]/[3759]	Speed: 117.685713 samples/sec	 accuracy=61.367045	 loss=1.648575	 lr=0.004181
Epoch[174] Batch [2799]/[3759]	Speed: 117.079611 samples/sec	 accuracy=61.341518	 loss=1.649725	 lr=0.004176
Epoch[174] Batch [2849]/[3759]	Speed: 118.027336 samples/sec	 accuracy=61.337719	 loss=1.649908	 lr=0.004171
Epoch[174] Batch [2899]/[3759]	Speed: 118.040729 samples/sec	 accuracy=61.335129	 loss=1.649598	 lr=0.004166
Epoch[174] Batch [2949]/[3759]	Speed: 116.775221 samples/sec	 accuracy=61.339513	 loss=1.649712	 lr=0.004160
Epoch[174] Batch [2999]/[3759]	Speed: 117.452676 samples/sec	 accuracy=61.328125	 loss=1.650048	 lr=0.004155
Epoch[174] Batch [3049]/[3759]	Speed: 114.934930 samples/sec	 accuracy=61.334529	 loss=1.650174	 lr=0.004150
Epoch[174] Batch [3099]/[3759]	Speed: 117.175345 samples/sec	 accuracy=61.349798	 loss=1.650313	 lr=0.004145
Epoch[174] Batch [3149]/[3759]	Speed: 117.696778 samples/sec	 accuracy=61.351190	 loss=1.650270	 lr=0.004140
Epoch[174] Batch [3199]/[3759]	Speed: 117.231761 samples/sec	 accuracy=61.321289	 loss=1.651017	 lr=0.004135
Epoch[174] Batch [3249]/[3759]	Speed: 118.314404 samples/sec	 accuracy=61.330288	 loss=1.650902	 lr=0.004130
Epoch[174] Batch [3299]/[3759]	Speed: 117.206661 samples/sec	 accuracy=61.334280	 loss=1.651360	 lr=0.004124
Epoch[174] Batch [3349]/[3759]	Speed: 117.992615 samples/sec	 accuracy=61.349347	 loss=1.650595	 lr=0.004119
Epoch[174] Batch [3399]/[3759]	Speed: 117.668009 samples/sec	 accuracy=61.359375	 loss=1.650595	 lr=0.004114
Epoch[174] Batch [3449]/[3759]	Speed: 118.064271 samples/sec	 accuracy=61.358696	 loss=1.651027	 lr=0.004109
Epoch[174] Batch [3499]/[3759]	Speed: 118.229073 samples/sec	 accuracy=61.349107	 loss=1.651098	 lr=0.004104
Epoch[174] Batch [3549]/[3759]	Speed: 118.183616 samples/sec	 accuracy=61.362676	 loss=1.651013	 lr=0.004099
Epoch[174] Batch [3599]/[3759]	Speed: 117.509365 samples/sec	 accuracy=61.360243	 loss=1.650863	 lr=0.004094
Epoch[174] Batch [3649]/[3759]	Speed: 118.667245 samples/sec	 accuracy=61.351884	 loss=1.651199	 lr=0.004089
Epoch[174] Batch [3699]/[3759]	Speed: 117.974086 samples/sec	 accuracy=61.381757	 loss=1.649934	 lr=0.004083
Epoch[174] Batch [3749]/[3759]	Speed: 121.339692 samples/sec	 accuracy=61.361667	 loss=1.650309	 lr=0.004078
Batch [0049]/[0303]: acc-top1=58.437500 acc-top5=78.781250
Batch [0099]/[0303]: acc-top1=54.968750 acc-top5=78.078125
Batch [0149]/[0303]: acc-top1=56.364583 acc-top5=78.802083
Batch [0199]/[0303]: acc-top1=59.156250 acc-top5=80.375000
Batch [0249]/[0303]: acc-top1=58.481250 acc-top5=80.012500
Batch [0299]/[0303]: acc-top1=57.833333 acc-top5=79.911458
[Epoch 174] training: accuracy=61.359404	 loss=1.650370
[Epoch 174] speed: 116 samples/sec	time cost: 2174.660954
[Epoch 174] validation: acc-top1=57.766089 acc-top5=79.909241 loss=1.913496
Epoch[175] Batch [0049]/[3760]	Speed: 88.205597 samples/sec	 accuracy=61.562500	 loss=1.578664	 lr=0.004072
Epoch[175] Batch [0099]/[3760]	Speed: 117.779159 samples/sec	 accuracy=61.343750	 loss=1.596792	 lr=0.004067
Epoch[175] Batch [0149]/[3760]	Speed: 117.408010 samples/sec	 accuracy=61.375000	 loss=1.600424	 lr=0.004062
Epoch[175] Batch [0199]/[3760]	Speed: 117.354968 samples/sec	 accuracy=61.789062	 loss=1.596310	 lr=0.004057
Epoch[175] Batch [0249]/[3760]	Speed: 118.414217 samples/sec	 accuracy=61.562500	 loss=1.613582	 lr=0.004052
Epoch[175] Batch [0299]/[3760]	Speed: 117.125002 samples/sec	 accuracy=62.083333	 loss=1.601505	 lr=0.004047
Epoch[175] Batch [0349]/[3760]	Speed: 117.234642 samples/sec	 accuracy=62.147321	 loss=1.597205	 lr=0.004042
Epoch[175] Batch [0399]/[3760]	Speed: 117.141387 samples/sec	 accuracy=62.199219	 loss=1.599768	 lr=0.004037
Epoch[175] Batch [0449]/[3760]	Speed: 118.139968 samples/sec	 accuracy=61.947917	 loss=1.607405	 lr=0.004032
Epoch[175] Batch [0499]/[3760]	Speed: 117.298110 samples/sec	 accuracy=62.043750	 loss=1.605941	 lr=0.004027
Epoch[175] Batch [0549]/[3760]	Speed: 117.784546 samples/sec	 accuracy=62.005682	 loss=1.609747	 lr=0.004022
Epoch[175] Batch [0599]/[3760]	Speed: 118.035430 samples/sec	 accuracy=61.979167	 loss=1.613242	 lr=0.004016
Epoch[175] Batch [0649]/[3760]	Speed: 117.794144 samples/sec	 accuracy=62.009615	 loss=1.613042	 lr=0.004011
Epoch[175] Batch [0699]/[3760]	Speed: 117.599209 samples/sec	 accuracy=61.868304	 loss=1.617133	 lr=0.004006
Epoch[175] Batch [0749]/[3760]	Speed: 117.289545 samples/sec	 accuracy=61.897917	 loss=1.618447	 lr=0.004001
Epoch[175] Batch [0799]/[3760]	Speed: 114.559565 samples/sec	 accuracy=61.923828	 loss=1.618330	 lr=0.003996
Epoch[175] Batch [0849]/[3760]	Speed: 117.707951 samples/sec	 accuracy=61.895221	 loss=1.618023	 lr=0.003991
Epoch[175] Batch [0899]/[3760]	Speed: 117.537361 samples/sec	 accuracy=61.902778	 loss=1.617618	 lr=0.003986
Epoch[175] Batch [0949]/[3760]	Speed: 117.086322 samples/sec	 accuracy=61.901316	 loss=1.617095	 lr=0.003981
Epoch[175] Batch [0999]/[3760]	Speed: 117.673919 samples/sec	 accuracy=61.882812	 loss=1.618694	 lr=0.003976
Epoch[175] Batch [1049]/[3760]	Speed: 117.029251 samples/sec	 accuracy=61.888393	 loss=1.617575	 lr=0.003971
Epoch[175] Batch [1099]/[3760]	Speed: 117.894116 samples/sec	 accuracy=61.845170	 loss=1.621078	 lr=0.003966
Epoch[175] Batch [1149]/[3760]	Speed: 117.136921 samples/sec	 accuracy=61.850543	 loss=1.619369	 lr=0.003961
Epoch[175] Batch [1199]/[3760]	Speed: 117.198629 samples/sec	 accuracy=61.830729	 loss=1.621011	 lr=0.003956
Epoch[175] Batch [1249]/[3760]	Speed: 117.820149 samples/sec	 accuracy=61.806250	 loss=1.621620	 lr=0.003951
Epoch[175] Batch [1299]/[3760]	Speed: 117.795205 samples/sec	 accuracy=61.866587	 loss=1.620695	 lr=0.003946
Epoch[175] Batch [1349]/[3760]	Speed: 117.155275 samples/sec	 accuracy=61.844907	 loss=1.622094	 lr=0.003941
Epoch[175] Batch [1399]/[3760]	Speed: 116.761181 samples/sec	 accuracy=61.812500	 loss=1.623986	 lr=0.003936
Epoch[175] Batch [1449]/[3760]	Speed: 117.464301 samples/sec	 accuracy=61.788793	 loss=1.623925	 lr=0.003931
Epoch[175] Batch [1499]/[3760]	Speed: 116.814231 samples/sec	 accuracy=61.778125	 loss=1.625185	 lr=0.003926
Epoch[175] Batch [1549]/[3760]	Speed: 117.851604 samples/sec	 accuracy=61.764113	 loss=1.625426	 lr=0.003921
Epoch[175] Batch [1599]/[3760]	Speed: 117.923072 samples/sec	 accuracy=61.742188	 loss=1.626876	 lr=0.003916
Epoch[175] Batch [1649]/[3760]	Speed: 117.228107 samples/sec	 accuracy=61.731061	 loss=1.626970	 lr=0.003911
Epoch[175] Batch [1699]/[3760]	Speed: 117.804531 samples/sec	 accuracy=61.722426	 loss=1.627738	 lr=0.003906
Epoch[175] Batch [1749]/[3760]	Speed: 118.066183 samples/sec	 accuracy=61.725000	 loss=1.627710	 lr=0.003901
Epoch[175] Batch [1799]/[3760]	Speed: 117.755338 samples/sec	 accuracy=61.763021	 loss=1.626377	 lr=0.003896
Epoch[175] Batch [1849]/[3760]	Speed: 117.071327 samples/sec	 accuracy=61.774493	 loss=1.626345	 lr=0.003891
Epoch[175] Batch [1899]/[3760]	Speed: 117.323620 samples/sec	 accuracy=61.755757	 loss=1.626993	 lr=0.003886
Epoch[175] Batch [1949]/[3760]	Speed: 117.363984 samples/sec	 accuracy=61.761218	 loss=1.626333	 lr=0.003881
Epoch[175] Batch [1999]/[3760]	Speed: 117.020031 samples/sec	 accuracy=61.746094	 loss=1.627834	 lr=0.003876
Epoch[175] Batch [2049]/[3760]	Speed: 116.817328 samples/sec	 accuracy=61.716463	 loss=1.629301	 lr=0.003871
Epoch[175] Batch [2099]/[3760]	Speed: 117.437145 samples/sec	 accuracy=61.747768	 loss=1.628418	 lr=0.003866
Epoch[175] Batch [2149]/[3760]	Speed: 117.748362 samples/sec	 accuracy=61.716570	 loss=1.628819	 lr=0.003861
Epoch[175] Batch [2199]/[3760]	Speed: 117.585633 samples/sec	 accuracy=61.724432	 loss=1.629328	 lr=0.003856
Epoch[175] Batch [2249]/[3760]	Speed: 117.492036 samples/sec	 accuracy=61.743750	 loss=1.627576	 lr=0.003851
Epoch[175] Batch [2299]/[3760]	Speed: 117.646590 samples/sec	 accuracy=61.747283	 loss=1.626860	 lr=0.003846
Epoch[175] Batch [2349]/[3760]	Speed: 114.910038 samples/sec	 accuracy=61.761303	 loss=1.626411	 lr=0.003841
Epoch[175] Batch [2399]/[3760]	Speed: 117.714333 samples/sec	 accuracy=61.736979	 loss=1.627971	 lr=0.003836
Epoch[175] Batch [2449]/[3760]	Speed: 117.836266 samples/sec	 accuracy=61.735332	 loss=1.627770	 lr=0.003831
Epoch[175] Batch [2499]/[3760]	Speed: 117.402017 samples/sec	 accuracy=61.739375	 loss=1.627011	 lr=0.003826
Epoch[175] Batch [2549]/[3760]	Speed: 117.552935 samples/sec	 accuracy=61.756740	 loss=1.626280	 lr=0.003821
Epoch[175] Batch [2599]/[3760]	Speed: 117.179450 samples/sec	 accuracy=61.765625	 loss=1.626365	 lr=0.003816
Epoch[175] Batch [2649]/[3760]	Speed: 116.988852 samples/sec	 accuracy=61.801887	 loss=1.625092	 lr=0.003811
Epoch[175] Batch [2699]/[3760]	Speed: 117.438962 samples/sec	 accuracy=61.792245	 loss=1.626253	 lr=0.003806
Epoch[175] Batch [2749]/[3760]	Speed: 117.332003 samples/sec	 accuracy=61.788636	 loss=1.626381	 lr=0.003802
Epoch[175] Batch [2799]/[3760]	Speed: 117.099687 samples/sec	 accuracy=61.776786	 loss=1.626620	 lr=0.003797
Epoch[175] Batch [2849]/[3760]	Speed: 117.804557 samples/sec	 accuracy=61.774671	 loss=1.626904	 lr=0.003792
Epoch[175] Batch [2899]/[3760]	Speed: 117.770950 samples/sec	 accuracy=61.772629	 loss=1.627189	 lr=0.003787
Epoch[175] Batch [2949]/[3760]	Speed: 117.790684 samples/sec	 accuracy=61.774894	 loss=1.626829	 lr=0.003782
Epoch[175] Batch [2999]/[3760]	Speed: 118.094931 samples/sec	 accuracy=61.764583	 loss=1.626919	 lr=0.003777
Epoch[175] Batch [3049]/[3760]	Speed: 118.077260 samples/sec	 accuracy=61.778176	 loss=1.626610	 lr=0.003772
Epoch[175] Batch [3099]/[3760]	Speed: 117.585128 samples/sec	 accuracy=61.776714	 loss=1.626934	 lr=0.003767
Epoch[175] Batch [3149]/[3760]	Speed: 117.613458 samples/sec	 accuracy=61.796627	 loss=1.626717	 lr=0.003762
Epoch[175] Batch [3199]/[3760]	Speed: 117.328069 samples/sec	 accuracy=61.815430	 loss=1.626171	 lr=0.003757
Epoch[175] Batch [3249]/[3760]	Speed: 117.809760 samples/sec	 accuracy=61.798558	 loss=1.627093	 lr=0.003752
Epoch[175] Batch [3299]/[3760]	Speed: 117.923533 samples/sec	 accuracy=61.812973	 loss=1.626765	 lr=0.003747
Epoch[175] Batch [3349]/[3760]	Speed: 116.985719 samples/sec	 accuracy=61.812034	 loss=1.626884	 lr=0.003743
Epoch[175] Batch [3399]/[3760]	Speed: 117.970369 samples/sec	 accuracy=61.816176	 loss=1.626185	 lr=0.003738
Epoch[175] Batch [3449]/[3760]	Speed: 117.621332 samples/sec	 accuracy=61.801178	 loss=1.626598	 lr=0.003733
Epoch[175] Batch [3499]/[3760]	Speed: 117.331459 samples/sec	 accuracy=61.786607	 loss=1.627496	 lr=0.003728
Epoch[175] Batch [3549]/[3760]	Speed: 117.169117 samples/sec	 accuracy=61.794454	 loss=1.626790	 lr=0.003723
Epoch[175] Batch [3599]/[3760]	Speed: 117.792499 samples/sec	 accuracy=61.791667	 loss=1.626668	 lr=0.003718
Epoch[175] Batch [3649]/[3760]	Speed: 117.204320 samples/sec	 accuracy=61.802226	 loss=1.626561	 lr=0.003713
Epoch[175] Batch [3699]/[3760]	Speed: 117.855738 samples/sec	 accuracy=61.808699	 loss=1.626555	 lr=0.003708
Epoch[175] Batch [3749]/[3760]	Speed: 120.586436 samples/sec	 accuracy=61.829583	 loss=1.626144	 lr=0.003703
Batch [0049]/[0303]: acc-top1=58.500000 acc-top5=79.625000
Batch [0099]/[0303]: acc-top1=55.750000 acc-top5=78.125000
Batch [0149]/[0303]: acc-top1=57.062500 acc-top5=78.958333
Batch [0199]/[0303]: acc-top1=59.695312 acc-top5=80.617188
Batch [0249]/[0303]: acc-top1=59.306250 acc-top5=80.475000
Batch [0299]/[0303]: acc-top1=58.828125 acc-top5=80.406250
[Epoch 175] training: accuracy=61.836353	 loss=1.625939
[Epoch 175] speed: 116 samples/sec	time cost: 2173.599507
[Epoch 175] validation: acc-top1=58.807756 acc-top5=80.409447 loss=1.896789
Epoch[176] Batch [0049]/[3760]	Speed: 87.142427 samples/sec	 accuracy=62.718750	 loss=1.587037	 lr=0.003698
Epoch[176] Batch [0099]/[3760]	Speed: 115.525540 samples/sec	 accuracy=62.796875	 loss=1.581345	 lr=0.003693
Epoch[176] Batch [0149]/[3760]	Speed: 117.921023 samples/sec	 accuracy=62.291667	 loss=1.593612	 lr=0.003688
Epoch[176] Batch [0199]/[3760]	Speed: 117.041386 samples/sec	 accuracy=62.257812	 loss=1.587732	 lr=0.003683
Epoch[176] Batch [0249]/[3760]	Speed: 117.554282 samples/sec	 accuracy=62.218750	 loss=1.591753	 lr=0.003678
Epoch[176] Batch [0299]/[3760]	Speed: 117.377624 samples/sec	 accuracy=62.401042	 loss=1.585682	 lr=0.003673
Epoch[176] Batch [0349]/[3760]	Speed: 117.874280 samples/sec	 accuracy=62.183036	 loss=1.593665	 lr=0.003668
Epoch[176] Batch [0399]/[3760]	Speed: 117.202365 samples/sec	 accuracy=62.164062	 loss=1.599476	 lr=0.003664
Epoch[176] Batch [0449]/[3760]	Speed: 116.997678 samples/sec	 accuracy=62.315972	 loss=1.593839	 lr=0.003659
Epoch[176] Batch [0499]/[3760]	Speed: 118.639998 samples/sec	 accuracy=62.440625	 loss=1.588746	 lr=0.003654
Epoch[176] Batch [0549]/[3760]	Speed: 117.002759 samples/sec	 accuracy=62.437500	 loss=1.590253	 lr=0.003649
Epoch[176] Batch [0599]/[3760]	Speed: 117.805174 samples/sec	 accuracy=62.440104	 loss=1.590612	 lr=0.003644
Epoch[176] Batch [0649]/[3760]	Speed: 117.436626 samples/sec	 accuracy=62.555288	 loss=1.588009	 lr=0.003639
Epoch[176] Batch [0699]/[3760]	Speed: 118.214521 samples/sec	 accuracy=62.616071	 loss=1.586455	 lr=0.003635
Epoch[176] Batch [0749]/[3760]	Speed: 117.721215 samples/sec	 accuracy=62.625000	 loss=1.585755	 lr=0.003630
Epoch[176] Batch [0799]/[3760]	Speed: 117.758412 samples/sec	 accuracy=62.664062	 loss=1.585097	 lr=0.003625
Epoch[176] Batch [0849]/[3760]	Speed: 117.687545 samples/sec	 accuracy=62.746324	 loss=1.582788	 lr=0.003620
Epoch[176] Batch [0899]/[3760]	Speed: 117.884833 samples/sec	 accuracy=62.737847	 loss=1.582435	 lr=0.003615
Epoch[176] Batch [0949]/[3760]	Speed: 117.297726 samples/sec	 accuracy=62.731908	 loss=1.581202	 lr=0.003611
Epoch[176] Batch [0999]/[3760]	Speed: 117.077157 samples/sec	 accuracy=62.693750	 loss=1.583986	 lr=0.003606
Epoch[176] Batch [1049]/[3760]	Speed: 117.323014 samples/sec	 accuracy=62.696429	 loss=1.584228	 lr=0.003601
Epoch[176] Batch [1099]/[3760]	Speed: 117.846062 samples/sec	 accuracy=62.717330	 loss=1.581976	 lr=0.003596
Epoch[176] Batch [1149]/[3760]	Speed: 117.503719 samples/sec	 accuracy=62.725543	 loss=1.580981	 lr=0.003591
Epoch[176] Batch [1199]/[3760]	Speed: 117.294338 samples/sec	 accuracy=62.774740	 loss=1.580471	 lr=0.003586
Epoch[176] Batch [1249]/[3760]	Speed: 118.008049 samples/sec	 accuracy=62.758750	 loss=1.580660	 lr=0.003582
Epoch[176] Batch [1299]/[3760]	Speed: 116.653819 samples/sec	 accuracy=62.731971	 loss=1.581427	 lr=0.003577
Epoch[176] Batch [1349]/[3760]	Speed: 117.341992 samples/sec	 accuracy=62.706019	 loss=1.581437	 lr=0.003572
Epoch[176] Batch [1399]/[3760]	Speed: 117.306479 samples/sec	 accuracy=62.783482	 loss=1.577980	 lr=0.003567
Epoch[176] Batch [1449]/[3760]	Speed: 116.757839 samples/sec	 accuracy=62.758621	 loss=1.578838	 lr=0.003563
Epoch[176] Batch [1499]/[3760]	Speed: 117.676865 samples/sec	 accuracy=62.731250	 loss=1.579675	 lr=0.003558
Epoch[176] Batch [1549]/[3760]	Speed: 117.634915 samples/sec	 accuracy=62.666331	 loss=1.583326	 lr=0.003553
Epoch[176] Batch [1599]/[3760]	Speed: 117.761081 samples/sec	 accuracy=62.684570	 loss=1.583472	 lr=0.003548
Epoch[176] Batch [1649]/[3760]	Speed: 115.078347 samples/sec	 accuracy=62.682765	 loss=1.583756	 lr=0.003543
Epoch[176] Batch [1699]/[3760]	Speed: 117.707395 samples/sec	 accuracy=62.645221	 loss=1.584339	 lr=0.003539
Epoch[176] Batch [1749]/[3760]	Speed: 117.408345 samples/sec	 accuracy=62.657143	 loss=1.583080	 lr=0.003534
Epoch[176] Batch [1799]/[3760]	Speed: 117.604070 samples/sec	 accuracy=62.723958	 loss=1.580423	 lr=0.003529
Epoch[176] Batch [1849]/[3760]	Speed: 117.471284 samples/sec	 accuracy=62.686655	 loss=1.581822	 lr=0.003524
Epoch[176] Batch [1899]/[3760]	Speed: 117.462905 samples/sec	 accuracy=62.675164	 loss=1.582099	 lr=0.003520
Epoch[176] Batch [1949]/[3760]	Speed: 117.264479 samples/sec	 accuracy=62.705929	 loss=1.581287	 lr=0.003515
Epoch[176] Batch [1999]/[3760]	Speed: 117.854509 samples/sec	 accuracy=62.675000	 loss=1.582583	 lr=0.003510
Epoch[176] Batch [2049]/[3760]	Speed: 116.398844 samples/sec	 accuracy=62.624238	 loss=1.583673	 lr=0.003505
Epoch[176] Batch [2099]/[3760]	Speed: 116.886710 samples/sec	 accuracy=62.605655	 loss=1.585105	 lr=0.003501
Epoch[176] Batch [2149]/[3760]	Speed: 117.485446 samples/sec	 accuracy=62.579942	 loss=1.585260	 lr=0.003496
Epoch[176] Batch [2199]/[3760]	Speed: 117.853810 samples/sec	 accuracy=62.578125	 loss=1.586050	 lr=0.003491
Epoch[176] Batch [2249]/[3760]	Speed: 117.087951 samples/sec	 accuracy=62.565972	 loss=1.586257	 lr=0.003486
Epoch[176] Batch [2299]/[3760]	Speed: 117.438045 samples/sec	 accuracy=62.567255	 loss=1.586522	 lr=0.003482
Epoch[176] Batch [2349]/[3760]	Speed: 116.852528 samples/sec	 accuracy=62.579122	 loss=1.587019	 lr=0.003477
Epoch[176] Batch [2399]/[3760]	Speed: 116.261693 samples/sec	 accuracy=62.570964	 loss=1.587206	 lr=0.003472
Epoch[176] Batch [2449]/[3760]	Speed: 118.071376 samples/sec	 accuracy=62.589286	 loss=1.586187	 lr=0.003468
Epoch[176] Batch [2499]/[3760]	Speed: 117.197420 samples/sec	 accuracy=62.561875	 loss=1.587473	 lr=0.003463
Epoch[176] Batch [2549]/[3760]	Speed: 117.148246 samples/sec	 accuracy=62.553309	 loss=1.587754	 lr=0.003458
Epoch[176] Batch [2599]/[3760]	Speed: 117.300901 samples/sec	 accuracy=62.555889	 loss=1.587844	 lr=0.003453
Epoch[176] Batch [2649]/[3760]	Speed: 117.030725 samples/sec	 accuracy=62.577241	 loss=1.587050	 lr=0.003449
Epoch[176] Batch [2699]/[3760]	Speed: 117.585450 samples/sec	 accuracy=62.570602	 loss=1.587427	 lr=0.003444
Epoch[176] Batch [2749]/[3760]	Speed: 116.964939 samples/sec	 accuracy=62.551705	 loss=1.587767	 lr=0.003439
Epoch[176] Batch [2799]/[3760]	Speed: 117.999708 samples/sec	 accuracy=62.561384	 loss=1.587640	 lr=0.003435
Epoch[176] Batch [2849]/[3760]	Speed: 117.594182 samples/sec	 accuracy=62.570724	 loss=1.587016	 lr=0.003430
Epoch[176] Batch [2899]/[3760]	Speed: 117.297495 samples/sec	 accuracy=62.558190	 loss=1.587396	 lr=0.003425
Epoch[176] Batch [2949]/[3760]	Speed: 117.565216 samples/sec	 accuracy=62.543962	 loss=1.587959	 lr=0.003421
Epoch[176] Batch [2999]/[3760]	Speed: 117.839892 samples/sec	 accuracy=62.554167	 loss=1.588045	 lr=0.003416
Epoch[176] Batch [3049]/[3760]	Speed: 118.091405 samples/sec	 accuracy=62.561988	 loss=1.586642	 lr=0.003411
Epoch[176] Batch [3099]/[3760]	Speed: 117.287343 samples/sec	 accuracy=62.570060	 loss=1.585917	 lr=0.003406
Epoch[176] Batch [3149]/[3760]	Speed: 114.576402 samples/sec	 accuracy=62.592758	 loss=1.585047	 lr=0.003402
Epoch[176] Batch [3199]/[3760]	Speed: 118.200826 samples/sec	 accuracy=62.592285	 loss=1.585310	 lr=0.003397
Epoch[176] Batch [3249]/[3760]	Speed: 117.325156 samples/sec	 accuracy=62.595673	 loss=1.585106	 lr=0.003392
Epoch[176] Batch [3299]/[3760]	Speed: 117.218823 samples/sec	 accuracy=62.598011	 loss=1.585134	 lr=0.003388
Epoch[176] Batch [3349]/[3760]	Speed: 117.406108 samples/sec	 accuracy=62.586754	 loss=1.585103	 lr=0.003383
Epoch[176] Batch [3399]/[3760]	Speed: 117.211411 samples/sec	 accuracy=62.584099	 loss=1.585167	 lr=0.003378
Epoch[176] Batch [3449]/[3760]	Speed: 117.891615 samples/sec	 accuracy=62.575181	 loss=1.586179	 lr=0.003374
Epoch[176] Batch [3499]/[3760]	Speed: 117.465412 samples/sec	 accuracy=62.558482	 loss=1.587079	 lr=0.003369
Epoch[176] Batch [3549]/[3760]	Speed: 117.332199 samples/sec	 accuracy=62.563820	 loss=1.587261	 lr=0.003364
Epoch[176] Batch [3599]/[3760]	Speed: 117.195878 samples/sec	 accuracy=62.565972	 loss=1.587133	 lr=0.003360
Epoch[176] Batch [3649]/[3760]	Speed: 117.280212 samples/sec	 accuracy=62.548373	 loss=1.587263	 lr=0.003355
Epoch[176] Batch [3699]/[3760]	Speed: 117.289517 samples/sec	 accuracy=62.529139	 loss=1.587766	 lr=0.003351
Epoch[176] Batch [3749]/[3760]	Speed: 120.325464 samples/sec	 accuracy=62.533333	 loss=1.587710	 lr=0.003346
Batch [0049]/[0303]: acc-top1=58.281250 acc-top5=79.000000
Batch [0099]/[0303]: acc-top1=55.546875 acc-top5=78.000000
Batch [0149]/[0303]: acc-top1=56.625000 acc-top5=78.677083
Batch [0199]/[0303]: acc-top1=59.250000 acc-top5=80.273438
Batch [0249]/[0303]: acc-top1=58.762500 acc-top5=80.100000
Batch [0299]/[0303]: acc-top1=58.421875 acc-top5=80.130208
[Epoch 176] training: accuracy=62.536154	 loss=1.587529
[Epoch 176] speed: 116 samples/sec	time cost: 2176.080476
[Epoch 176] validation: acc-top1=58.400371 acc-top5=80.161922 loss=1.911087
Epoch[177] Batch [0049]/[3759]	Speed: 87.458843 samples/sec	 accuracy=63.031250	 loss=1.574159	 lr=0.003340
Epoch[177] Batch [0099]/[3759]	Speed: 117.238103 samples/sec	 accuracy=63.437500	 loss=1.553590	 lr=0.003336
Epoch[177] Batch [0149]/[3759]	Speed: 117.266604 samples/sec	 accuracy=63.510417	 loss=1.551535	 lr=0.003331
Epoch[177] Batch [0199]/[3759]	Speed: 117.341825 samples/sec	 accuracy=63.382812	 loss=1.553463	 lr=0.003326
Epoch[177] Batch [0249]/[3759]	Speed: 117.821842 samples/sec	 accuracy=63.425000	 loss=1.548999	 lr=0.003322
Epoch[177] Batch [0299]/[3759]	Speed: 117.267571 samples/sec	 accuracy=63.645833	 loss=1.542568	 lr=0.003317
Epoch[177] Batch [0349]/[3759]	Speed: 117.605755 samples/sec	 accuracy=63.450893	 loss=1.544714	 lr=0.003313
Epoch[177] Batch [0399]/[3759]	Speed: 116.870889 samples/sec	 accuracy=63.351562	 loss=1.548109	 lr=0.003308
Epoch[177] Batch [0449]/[3759]	Speed: 117.099152 samples/sec	 accuracy=63.388889	 loss=1.549086	 lr=0.003303
Epoch[177] Batch [0499]/[3759]	Speed: 117.004629 samples/sec	 accuracy=63.506250	 loss=1.549141	 lr=0.003299
Epoch[177] Batch [0549]/[3759]	Speed: 116.796984 samples/sec	 accuracy=63.417614	 loss=1.553366	 lr=0.003294
Epoch[177] Batch [0599]/[3759]	Speed: 117.753432 samples/sec	 accuracy=63.531250	 loss=1.551161	 lr=0.003290
Epoch[177] Batch [0649]/[3759]	Speed: 117.162022 samples/sec	 accuracy=63.528846	 loss=1.550706	 lr=0.003285
Epoch[177] Batch [0699]/[3759]	Speed: 117.693800 samples/sec	 accuracy=63.562500	 loss=1.548084	 lr=0.003280
Epoch[177] Batch [0749]/[3759]	Speed: 117.081914 samples/sec	 accuracy=63.654167	 loss=1.546725	 lr=0.003276
Epoch[177] Batch [0799]/[3759]	Speed: 116.549103 samples/sec	 accuracy=63.625000	 loss=1.545954	 lr=0.003271
Epoch[177] Batch [0849]/[3759]	Speed: 118.192493 samples/sec	 accuracy=63.571691	 loss=1.550067	 lr=0.003267
Epoch[177] Batch [0899]/[3759]	Speed: 118.185581 samples/sec	 accuracy=63.565972	 loss=1.548099	 lr=0.003262
Epoch[177] Batch [0949]/[3759]	Speed: 114.848554 samples/sec	 accuracy=63.557566	 loss=1.550418	 lr=0.003257
Epoch[177] Batch [0999]/[3759]	Speed: 117.341893 samples/sec	 accuracy=63.540625	 loss=1.549851	 lr=0.003253
Epoch[177] Batch [1049]/[3759]	Speed: 117.495517 samples/sec	 accuracy=63.544643	 loss=1.548116	 lr=0.003248
Epoch[177] Batch [1099]/[3759]	Speed: 117.238873 samples/sec	 accuracy=63.526989	 loss=1.548343	 lr=0.003244
Epoch[177] Batch [1149]/[3759]	Speed: 117.250088 samples/sec	 accuracy=63.452446	 loss=1.549817	 lr=0.003239
Epoch[177] Batch [1199]/[3759]	Speed: 117.472104 samples/sec	 accuracy=63.416667	 loss=1.551338	 lr=0.003235
Epoch[177] Batch [1249]/[3759]	Speed: 117.717150 samples/sec	 accuracy=63.458750	 loss=1.551291	 lr=0.003230
Epoch[177] Batch [1299]/[3759]	Speed: 117.278537 samples/sec	 accuracy=63.489183	 loss=1.551014	 lr=0.003225
Epoch[177] Batch [1349]/[3759]	Speed: 117.803473 samples/sec	 accuracy=63.496528	 loss=1.549504	 lr=0.003221
Epoch[177] Batch [1399]/[3759]	Speed: 117.516103 samples/sec	 accuracy=63.474330	 loss=1.551435	 lr=0.003216
Epoch[177] Batch [1449]/[3759]	Speed: 116.981363 samples/sec	 accuracy=63.439655	 loss=1.552356	 lr=0.003212
Epoch[177] Batch [1499]/[3759]	Speed: 117.439492 samples/sec	 accuracy=63.501042	 loss=1.549549	 lr=0.003207
Epoch[177] Batch [1549]/[3759]	Speed: 116.655988 samples/sec	 accuracy=63.496976	 loss=1.549916	 lr=0.003203
Epoch[177] Batch [1599]/[3759]	Speed: 117.276058 samples/sec	 accuracy=63.457031	 loss=1.551734	 lr=0.003198
Epoch[177] Batch [1649]/[3759]	Speed: 117.267876 samples/sec	 accuracy=63.456439	 loss=1.551040	 lr=0.003194
Epoch[177] Batch [1699]/[3759]	Speed: 117.464889 samples/sec	 accuracy=63.453125	 loss=1.551344	 lr=0.003189
Epoch[177] Batch [1749]/[3759]	Speed: 116.851999 samples/sec	 accuracy=63.437500	 loss=1.551645	 lr=0.003185
Epoch[177] Batch [1799]/[3759]	Speed: 117.393011 samples/sec	 accuracy=63.420139	 loss=1.553241	 lr=0.003180
Epoch[177] Batch [1849]/[3759]	Speed: 117.251657 samples/sec	 accuracy=63.423986	 loss=1.553450	 lr=0.003175
Epoch[177] Batch [1899]/[3759]	Speed: 117.276454 samples/sec	 accuracy=63.380757	 loss=1.555068	 lr=0.003171
Epoch[177] Batch [1949]/[3759]	Speed: 117.673474 samples/sec	 accuracy=63.387821	 loss=1.554498	 lr=0.003166
Epoch[177] Batch [1999]/[3759]	Speed: 117.056423 samples/sec	 accuracy=63.362500	 loss=1.556797	 lr=0.003162
Epoch[177] Batch [2049]/[3759]	Speed: 117.270759 samples/sec	 accuracy=63.333079	 loss=1.557500	 lr=0.003157
Epoch[177] Batch [2099]/[3759]	Speed: 116.673165 samples/sec	 accuracy=63.347470	 loss=1.556373	 lr=0.003153
Epoch[177] Batch [2149]/[3759]	Speed: 118.006717 samples/sec	 accuracy=63.348110	 loss=1.556636	 lr=0.003148
Epoch[177] Batch [2199]/[3759]	Speed: 117.802602 samples/sec	 accuracy=63.348011	 loss=1.556853	 lr=0.003144
Epoch[177] Batch [2249]/[3759]	Speed: 117.300684 samples/sec	 accuracy=63.330556	 loss=1.556912	 lr=0.003139
Epoch[177] Batch [2299]/[3759]	Speed: 117.140117 samples/sec	 accuracy=63.312500	 loss=1.557592	 lr=0.003135
Epoch[177] Batch [2349]/[3759]	Speed: 117.035371 samples/sec	 accuracy=63.342420	 loss=1.556861	 lr=0.003130
Epoch[177] Batch [2399]/[3759]	Speed: 117.347579 samples/sec	 accuracy=63.345052	 loss=1.556212	 lr=0.003126
Epoch[177] Batch [2449]/[3759]	Speed: 117.161352 samples/sec	 accuracy=63.386480	 loss=1.555373	 lr=0.003121
Epoch[177] Batch [2499]/[3759]	Speed: 115.583767 samples/sec	 accuracy=63.368750	 loss=1.555659	 lr=0.003117
Epoch[177] Batch [2549]/[3759]	Speed: 117.036853 samples/sec	 accuracy=63.374387	 loss=1.555424	 lr=0.003112
Epoch[177] Batch [2599]/[3759]	Speed: 117.570938 samples/sec	 accuracy=63.350361	 loss=1.556415	 lr=0.003108
Epoch[177] Batch [2649]/[3759]	Speed: 117.465142 samples/sec	 accuracy=63.333137	 loss=1.557147	 lr=0.003104
Epoch[177] Batch [2699]/[3759]	Speed: 116.926126 samples/sec	 accuracy=63.366319	 loss=1.555890	 lr=0.003099
Epoch[177] Batch [2749]/[3759]	Speed: 117.652361 samples/sec	 accuracy=63.341477	 loss=1.556470	 lr=0.003095
Epoch[177] Batch [2799]/[3759]	Speed: 116.682630 samples/sec	 accuracy=63.330915	 loss=1.557080	 lr=0.003090
Epoch[177] Batch [2849]/[3759]	Speed: 117.311567 samples/sec	 accuracy=63.326754	 loss=1.557460	 lr=0.003086
Epoch[177] Batch [2899]/[3759]	Speed: 117.315276 samples/sec	 accuracy=63.329741	 loss=1.557435	 lr=0.003081
Epoch[177] Batch [2949]/[3759]	Speed: 117.460078 samples/sec	 accuracy=63.311970	 loss=1.557409	 lr=0.003077
Epoch[177] Batch [2999]/[3759]	Speed: 117.189719 samples/sec	 accuracy=63.320833	 loss=1.557204	 lr=0.003072
Epoch[177] Batch [3049]/[3759]	Speed: 117.507800 samples/sec	 accuracy=63.292520	 loss=1.557924	 lr=0.003068
Epoch[177] Batch [3099]/[3759]	Speed: 117.635046 samples/sec	 accuracy=63.296875	 loss=1.557664	 lr=0.003063
Epoch[177] Batch [3149]/[3759]	Speed: 117.000045 samples/sec	 accuracy=63.292659	 loss=1.558278	 lr=0.003059
Epoch[177] Batch [3199]/[3759]	Speed: 117.719533 samples/sec	 accuracy=63.295410	 loss=1.558099	 lr=0.003054
Epoch[177] Batch [3249]/[3759]	Speed: 117.189790 samples/sec	 accuracy=63.311058	 loss=1.558418	 lr=0.003050
Epoch[177] Batch [3299]/[3759]	Speed: 117.535344 samples/sec	 accuracy=63.288352	 loss=1.559821	 lr=0.003046
Epoch[177] Batch [3349]/[3759]	Speed: 117.124620 samples/sec	 accuracy=63.277519	 loss=1.559984	 lr=0.003041
Epoch[177] Batch [3399]/[3759]	Speed: 117.837256 samples/sec	 accuracy=63.272978	 loss=1.559853	 lr=0.003037
Epoch[177] Batch [3449]/[3759]	Speed: 118.076203 samples/sec	 accuracy=63.277174	 loss=1.559968	 lr=0.003032
Epoch[177] Batch [3499]/[3759]	Speed: 116.553358 samples/sec	 accuracy=63.263839	 loss=1.560612	 lr=0.003028
Epoch[177] Batch [3549]/[3759]	Speed: 117.264095 samples/sec	 accuracy=63.267165	 loss=1.560852	 lr=0.003024
Epoch[177] Batch [3599]/[3759]	Speed: 117.868011 samples/sec	 accuracy=63.273872	 loss=1.560455	 lr=0.003019
Epoch[177] Batch [3649]/[3759]	Speed: 117.086022 samples/sec	 accuracy=63.266695	 loss=1.560915	 lr=0.003015
Epoch[177] Batch [3699]/[3759]	Speed: 117.345032 samples/sec	 accuracy=63.253801	 loss=1.561574	 lr=0.003010
Epoch[177] Batch [3749]/[3759]	Speed: 120.739581 samples/sec	 accuracy=63.267917	 loss=1.561275	 lr=0.003006
Batch [0049]/[0303]: acc-top1=59.312500 acc-top5=79.000000
Batch [0099]/[0303]: acc-top1=56.296875 acc-top5=78.796875
Batch [0149]/[0303]: acc-top1=57.312500 acc-top5=79.250000
Batch [0199]/[0303]: acc-top1=60.023438 acc-top5=80.875000
Batch [0249]/[0303]: acc-top1=59.550000 acc-top5=80.718750
Batch [0299]/[0303]: acc-top1=59.317708 acc-top5=80.755208
[Epoch 177] training: accuracy=63.273144	 loss=1.561091
[Epoch 177] speed: 116 samples/sec	time cost: 2177.727140
[Epoch 177] validation: acc-top1=59.235767 acc-top5=80.775578 loss=1.871210
Epoch[178] Batch [0049]/[3760]	Speed: 87.452087 samples/sec	 accuracy=64.562500	 loss=1.482893	 lr=0.003001
Epoch[178] Batch [0099]/[3760]	Speed: 117.328287 samples/sec	 accuracy=63.234375	 loss=1.535110	 lr=0.002996
Epoch[178] Batch [0149]/[3760]	Speed: 117.553898 samples/sec	 accuracy=63.468750	 loss=1.536759	 lr=0.002992
Epoch[178] Batch [0199]/[3760]	Speed: 117.182479 samples/sec	 accuracy=63.414062	 loss=1.541041	 lr=0.002987
Epoch[178] Batch [0249]/[3760]	Speed: 115.423554 samples/sec	 accuracy=63.631250	 loss=1.537328	 lr=0.002983
Epoch[178] Batch [0299]/[3760]	Speed: 118.033508 samples/sec	 accuracy=63.770833	 loss=1.531733	 lr=0.002979
Epoch[178] Batch [0349]/[3760]	Speed: 117.489408 samples/sec	 accuracy=64.116071	 loss=1.512038	 lr=0.002974
Epoch[178] Batch [0399]/[3760]	Speed: 117.686409 samples/sec	 accuracy=64.035156	 loss=1.517932	 lr=0.002970
Epoch[178] Batch [0449]/[3760]	Speed: 117.501234 samples/sec	 accuracy=63.958333	 loss=1.519990	 lr=0.002966
Epoch[178] Batch [0499]/[3760]	Speed: 117.174798 samples/sec	 accuracy=63.918750	 loss=1.522612	 lr=0.002961
Epoch[178] Batch [0549]/[3760]	Speed: 117.381452 samples/sec	 accuracy=63.985795	 loss=1.522420	 lr=0.002957
Epoch[178] Batch [0599]/[3760]	Speed: 117.795980 samples/sec	 accuracy=63.958333	 loss=1.526632	 lr=0.002952
Epoch[178] Batch [0649]/[3760]	Speed: 117.557133 samples/sec	 accuracy=63.903846	 loss=1.528582	 lr=0.002948
Epoch[178] Batch [0699]/[3760]	Speed: 117.734841 samples/sec	 accuracy=63.850446	 loss=1.532656	 lr=0.002944
Epoch[178] Batch [0749]/[3760]	Speed: 117.543446 samples/sec	 accuracy=63.991667	 loss=1.525555	 lr=0.002939
Epoch[178] Batch [0799]/[3760]	Speed: 117.753000 samples/sec	 accuracy=64.019531	 loss=1.524810	 lr=0.002935
Epoch[178] Batch [0849]/[3760]	Speed: 117.310625 samples/sec	 accuracy=64.014706	 loss=1.524278	 lr=0.002931
Epoch[178] Batch [0899]/[3760]	Speed: 117.602257 samples/sec	 accuracy=63.968750	 loss=1.524031	 lr=0.002926
Epoch[178] Batch [0949]/[3760]	Speed: 117.812942 samples/sec	 accuracy=64.072368	 loss=1.520419	 lr=0.002922
Epoch[178] Batch [0999]/[3760]	Speed: 117.508978 samples/sec	 accuracy=64.067188	 loss=1.520837	 lr=0.002918
Epoch[178] Batch [1049]/[3760]	Speed: 117.608673 samples/sec	 accuracy=64.056548	 loss=1.522377	 lr=0.002913
Epoch[178] Batch [1099]/[3760]	Speed: 117.776609 samples/sec	 accuracy=64.066761	 loss=1.522728	 lr=0.002909
Epoch[178] Batch [1149]/[3760]	Speed: 117.608434 samples/sec	 accuracy=64.069293	 loss=1.522688	 lr=0.002905
Epoch[178] Batch [1199]/[3760]	Speed: 117.641428 samples/sec	 accuracy=64.074219	 loss=1.522222	 lr=0.002900
Epoch[178] Batch [1249]/[3760]	Speed: 116.848090 samples/sec	 accuracy=64.116250	 loss=1.520459	 lr=0.002896
Epoch[178] Batch [1299]/[3760]	Speed: 117.666775 samples/sec	 accuracy=64.112981	 loss=1.520122	 lr=0.002892
Epoch[178] Batch [1349]/[3760]	Speed: 117.472125 samples/sec	 accuracy=64.109954	 loss=1.519773	 lr=0.002887
Epoch[178] Batch [1399]/[3760]	Speed: 117.314156 samples/sec	 accuracy=64.097098	 loss=1.521069	 lr=0.002883
Epoch[178] Batch [1449]/[3760]	Speed: 117.666637 samples/sec	 accuracy=64.132543	 loss=1.520999	 lr=0.002879
Epoch[178] Batch [1499]/[3760]	Speed: 117.178074 samples/sec	 accuracy=64.145833	 loss=1.520375	 lr=0.002874
Epoch[178] Batch [1549]/[3760]	Speed: 117.381721 samples/sec	 accuracy=64.131048	 loss=1.521467	 lr=0.002870
Epoch[178] Batch [1599]/[3760]	Speed: 117.030621 samples/sec	 accuracy=64.097656	 loss=1.522598	 lr=0.002866
Epoch[178] Batch [1649]/[3760]	Speed: 117.079269 samples/sec	 accuracy=64.077652	 loss=1.522301	 lr=0.002861
Epoch[178] Batch [1699]/[3760]	Speed: 118.299266 samples/sec	 accuracy=64.043199	 loss=1.522546	 lr=0.002857
Epoch[178] Batch [1749]/[3760]	Speed: 118.080151 samples/sec	 accuracy=64.016071	 loss=1.523047	 lr=0.002853
Epoch[178] Batch [1799]/[3760]	Speed: 115.199497 samples/sec	 accuracy=64.006076	 loss=1.523812	 lr=0.002849
Epoch[178] Batch [1849]/[3760]	Speed: 117.138329 samples/sec	 accuracy=63.972973	 loss=1.524689	 lr=0.002844
Epoch[178] Batch [1899]/[3760]	Speed: 117.771410 samples/sec	 accuracy=63.962993	 loss=1.525822	 lr=0.002840
Epoch[178] Batch [1949]/[3760]	Speed: 117.373685 samples/sec	 accuracy=63.959135	 loss=1.526791	 lr=0.002836
Epoch[178] Batch [1999]/[3760]	Speed: 117.541944 samples/sec	 accuracy=63.964844	 loss=1.526819	 lr=0.002831
Epoch[178] Batch [2049]/[3760]	Speed: 117.220355 samples/sec	 accuracy=63.969512	 loss=1.527348	 lr=0.002827
Epoch[178] Batch [2099]/[3760]	Speed: 117.653958 samples/sec	 accuracy=63.943452	 loss=1.527466	 lr=0.002823
Epoch[178] Batch [2149]/[3760]	Speed: 117.939701 samples/sec	 accuracy=63.953488	 loss=1.526263	 lr=0.002819
Epoch[178] Batch [2199]/[3760]	Speed: 117.564199 samples/sec	 accuracy=63.975852	 loss=1.525323	 lr=0.002814
Epoch[178] Batch [2249]/[3760]	Speed: 116.830022 samples/sec	 accuracy=63.975000	 loss=1.525156	 lr=0.002810
Epoch[178] Batch [2299]/[3760]	Speed: 117.135068 samples/sec	 accuracy=63.990489	 loss=1.524243	 lr=0.002806
Epoch[178] Batch [2349]/[3760]	Speed: 117.246813 samples/sec	 accuracy=64.007314	 loss=1.524264	 lr=0.002802
Epoch[178] Batch [2399]/[3760]	Speed: 117.832470 samples/sec	 accuracy=64.030599	 loss=1.523779	 lr=0.002797
Epoch[178] Batch [2449]/[3760]	Speed: 117.012131 samples/sec	 accuracy=64.019770	 loss=1.524025	 lr=0.002793
Epoch[178] Batch [2499]/[3760]	Speed: 116.999373 samples/sec	 accuracy=64.024375	 loss=1.523355	 lr=0.002789
Epoch[178] Batch [2549]/[3760]	Speed: 117.612543 samples/sec	 accuracy=63.992647	 loss=1.524373	 lr=0.002785
Epoch[178] Batch [2599]/[3760]	Speed: 117.792127 samples/sec	 accuracy=63.981370	 loss=1.524674	 lr=0.002780
Epoch[178] Batch [2649]/[3760]	Speed: 118.539299 samples/sec	 accuracy=63.993514	 loss=1.524130	 lr=0.002776
Epoch[178] Batch [2699]/[3760]	Speed: 117.816593 samples/sec	 accuracy=64.002894	 loss=1.523785	 lr=0.002772
Epoch[178] Batch [2749]/[3760]	Speed: 117.975817 samples/sec	 accuracy=64.038636	 loss=1.522649	 lr=0.002768
Epoch[178] Batch [2799]/[3760]	Speed: 117.263287 samples/sec	 accuracy=64.044643	 loss=1.521250	 lr=0.002763
Epoch[178] Batch [2849]/[3760]	Speed: 117.957920 samples/sec	 accuracy=64.033443	 loss=1.521575	 lr=0.002759
Epoch[178] Batch [2899]/[3760]	Speed: 117.693044 samples/sec	 accuracy=64.038254	 loss=1.521264	 lr=0.002755
Epoch[178] Batch [2949]/[3760]	Speed: 117.672290 samples/sec	 accuracy=64.032839	 loss=1.521014	 lr=0.002751
Epoch[178] Batch [2999]/[3760]	Speed: 117.750211 samples/sec	 accuracy=64.045312	 loss=1.520736	 lr=0.002746
Epoch[178] Batch [3049]/[3760]	Speed: 117.518116 samples/sec	 accuracy=64.031250	 loss=1.521377	 lr=0.002742
Epoch[178] Batch [3099]/[3760]	Speed: 117.488896 samples/sec	 accuracy=64.048891	 loss=1.521018	 lr=0.002738
Epoch[178] Batch [3149]/[3760]	Speed: 117.633828 samples/sec	 accuracy=64.055060	 loss=1.520904	 lr=0.002734
Epoch[178] Batch [3199]/[3760]	Speed: 117.493418 samples/sec	 accuracy=64.062500	 loss=1.520823	 lr=0.002730
Epoch[178] Batch [3249]/[3760]	Speed: 117.532927 samples/sec	 accuracy=64.062019	 loss=1.520826	 lr=0.002725
Epoch[178] Batch [3299]/[3760]	Speed: 115.328733 samples/sec	 accuracy=64.064394	 loss=1.520636	 lr=0.002721
Epoch[178] Batch [3349]/[3760]	Speed: 117.574005 samples/sec	 accuracy=64.062034	 loss=1.521134	 lr=0.002717
Epoch[178] Batch [3399]/[3760]	Speed: 117.211998 samples/sec	 accuracy=64.065717	 loss=1.521011	 lr=0.002713
Epoch[178] Batch [3449]/[3760]	Speed: 117.443868 samples/sec	 accuracy=64.053442	 loss=1.521604	 lr=0.002709
Epoch[178] Batch [3499]/[3760]	Speed: 118.487628 samples/sec	 accuracy=64.058482	 loss=1.521472	 lr=0.002704
Epoch[178] Batch [3549]/[3760]	Speed: 117.880279 samples/sec	 accuracy=64.041373	 loss=1.522373	 lr=0.002700
Epoch[178] Batch [3599]/[3760]	Speed: 117.542000 samples/sec	 accuracy=64.039497	 loss=1.522594	 lr=0.002696
Epoch[178] Batch [3649]/[3760]	Speed: 117.697666 samples/sec	 accuracy=64.042808	 loss=1.523048	 lr=0.002692
Epoch[178] Batch [3699]/[3760]	Speed: 117.603732 samples/sec	 accuracy=64.042652	 loss=1.522723	 lr=0.002688
Epoch[178] Batch [3749]/[3760]	Speed: 119.847544 samples/sec	 accuracy=64.040000	 loss=1.522749	 lr=0.002684
Batch [0049]/[0303]: acc-top1=59.125000 acc-top5=79.093750
Batch [0099]/[0303]: acc-top1=56.859375 acc-top5=78.937500
Batch [0149]/[0303]: acc-top1=57.968750 acc-top5=79.531250
Batch [0199]/[0303]: acc-top1=60.421875 acc-top5=81.015625
Batch [0249]/[0303]: acc-top1=60.012500 acc-top5=80.881250
Batch [0299]/[0303]: acc-top1=59.520833 acc-top5=80.812500
[Epoch 178] training: accuracy=64.044631	 loss=1.522729
[Epoch 178] speed: 116 samples/sec	time cost: 2174.832463
[Epoch 178] validation: acc-top1=59.442038 acc-top5=80.785891 loss=1.886012
Epoch[179] Batch [0049]/[3760]	Speed: 86.863586 samples/sec	 accuracy=64.406250	 loss=1.510059	 lr=0.002679
Epoch[179] Batch [0099]/[3760]	Speed: 118.067316 samples/sec	 accuracy=65.640625	 loss=1.467030	 lr=0.002674
Epoch[179] Batch [0149]/[3760]	Speed: 117.942093 samples/sec	 accuracy=65.687500	 loss=1.454882	 lr=0.002670
Epoch[179] Batch [0199]/[3760]	Speed: 117.213408 samples/sec	 accuracy=65.609375	 loss=1.463975	 lr=0.002666
Epoch[179] Batch [0249]/[3760]	Speed: 117.498626 samples/sec	 accuracy=65.312500	 loss=1.462831	 lr=0.002662
Epoch[179] Batch [0299]/[3760]	Speed: 116.343194 samples/sec	 accuracy=65.312500	 loss=1.473628	 lr=0.002658
Epoch[179] Batch [0349]/[3760]	Speed: 117.563254 samples/sec	 accuracy=65.285714	 loss=1.474757	 lr=0.002654
Epoch[179] Batch [0399]/[3760]	Speed: 116.859077 samples/sec	 accuracy=65.406250	 loss=1.469555	 lr=0.002650
Epoch[179] Batch [0449]/[3760]	Speed: 117.448799 samples/sec	 accuracy=65.326389	 loss=1.472520	 lr=0.002645
Epoch[179] Batch [0499]/[3760]	Speed: 117.339220 samples/sec	 accuracy=65.265625	 loss=1.474066	 lr=0.002641
Epoch[179] Batch [0549]/[3760]	Speed: 117.376068 samples/sec	 accuracy=65.159091	 loss=1.476331	 lr=0.002637
Epoch[179] Batch [0599]/[3760]	Speed: 117.185265 samples/sec	 accuracy=65.119792	 loss=1.481072	 lr=0.002633
Epoch[179] Batch [0649]/[3760]	Speed: 117.598890 samples/sec	 accuracy=64.956731	 loss=1.488083	 lr=0.002629
Epoch[179] Batch [0699]/[3760]	Speed: 117.474336 samples/sec	 accuracy=65.029018	 loss=1.485630	 lr=0.002625
Epoch[179] Batch [0749]/[3760]	Speed: 117.640582 samples/sec	 accuracy=65.054167	 loss=1.484778	 lr=0.002621
Epoch[179] Batch [0799]/[3760]	Speed: 117.330021 samples/sec	 accuracy=65.041016	 loss=1.485835	 lr=0.002616
Epoch[179] Batch [0849]/[3760]	Speed: 117.091284 samples/sec	 accuracy=64.988971	 loss=1.486424	 lr=0.002612
Epoch[179] Batch [0899]/[3760]	Speed: 117.219844 samples/sec	 accuracy=64.892361	 loss=1.488137	 lr=0.002608
Epoch[179] Batch [0949]/[3760]	Speed: 115.743368 samples/sec	 accuracy=64.875000	 loss=1.488916	 lr=0.002604
Epoch[179] Batch [0999]/[3760]	Speed: 117.210996 samples/sec	 accuracy=64.842188	 loss=1.490416	 lr=0.002600
Epoch[179] Batch [1049]/[3760]	Speed: 117.383686 samples/sec	 accuracy=64.870536	 loss=1.489475	 lr=0.002596
Epoch[179] Batch [1099]/[3760]	Speed: 117.080356 samples/sec	 accuracy=64.833807	 loss=1.490084	 lr=0.002592
Epoch[179] Batch [1149]/[3760]	Speed: 117.970466 samples/sec	 accuracy=64.841033	 loss=1.490379	 lr=0.002588
Epoch[179] Batch [1199]/[3760]	Speed: 116.566056 samples/sec	 accuracy=64.815104	 loss=1.491274	 lr=0.002584
Epoch[179] Batch [1249]/[3760]	Speed: 117.411306 samples/sec	 accuracy=64.848750	 loss=1.490383	 lr=0.002580
Epoch[179] Batch [1299]/[3760]	Speed: 117.419937 samples/sec	 accuracy=64.864183	 loss=1.488917	 lr=0.002575
Epoch[179] Batch [1349]/[3760]	Speed: 117.493285 samples/sec	 accuracy=64.853009	 loss=1.488743	 lr=0.002571
Epoch[179] Batch [1399]/[3760]	Speed: 117.406112 samples/sec	 accuracy=64.856027	 loss=1.489758	 lr=0.002567
Epoch[179] Batch [1449]/[3760]	Speed: 117.388659 samples/sec	 accuracy=64.873922	 loss=1.489022	 lr=0.002563
Epoch[179] Batch [1499]/[3760]	Speed: 116.964888 samples/sec	 accuracy=64.876042	 loss=1.489020	 lr=0.002559
Epoch[179] Batch [1549]/[3760]	Speed: 116.991037 samples/sec	 accuracy=64.890121	 loss=1.487450	 lr=0.002555
Epoch[179] Batch [1599]/[3760]	Speed: 117.341641 samples/sec	 accuracy=64.873047	 loss=1.486974	 lr=0.002551
Epoch[179] Batch [1649]/[3760]	Speed: 117.410775 samples/sec	 accuracy=64.794508	 loss=1.490748	 lr=0.002547
Epoch[179] Batch [1699]/[3760]	Speed: 116.855358 samples/sec	 accuracy=64.766544	 loss=1.492209	 lr=0.002543
Epoch[179] Batch [1749]/[3760]	Speed: 117.464490 samples/sec	 accuracy=64.808036	 loss=1.490481	 lr=0.002539
Epoch[179] Batch [1799]/[3760]	Speed: 117.629317 samples/sec	 accuracy=64.776042	 loss=1.490530	 lr=0.002535
Epoch[179] Batch [1849]/[3760]	Speed: 117.739640 samples/sec	 accuracy=64.837838	 loss=1.488811	 lr=0.002531
Epoch[179] Batch [1899]/[3760]	Speed: 117.587047 samples/sec	 accuracy=64.819901	 loss=1.489221	 lr=0.002527
Epoch[179] Batch [1949]/[3760]	Speed: 117.532574 samples/sec	 accuracy=64.794872	 loss=1.489899	 lr=0.002523
Epoch[179] Batch [1999]/[3760]	Speed: 117.432432 samples/sec	 accuracy=64.775000	 loss=1.489796	 lr=0.002519
Epoch[179] Batch [2049]/[3760]	Speed: 117.715361 samples/sec	 accuracy=64.767530	 loss=1.489515	 lr=0.002515
Epoch[179] Batch [2099]/[3760]	Speed: 118.006056 samples/sec	 accuracy=64.753720	 loss=1.489564	 lr=0.002510
Epoch[179] Batch [2149]/[3760]	Speed: 117.430989 samples/sec	 accuracy=64.721657	 loss=1.490544	 lr=0.002506
Epoch[179] Batch [2199]/[3760]	Speed: 116.935110 samples/sec	 accuracy=64.691761	 loss=1.491837	 lr=0.002502
Epoch[179] Batch [2249]/[3760]	Speed: 117.784745 samples/sec	 accuracy=64.689583	 loss=1.492022	 lr=0.002498
Epoch[179] Batch [2299]/[3760]	Speed: 117.777215 samples/sec	 accuracy=64.710598	 loss=1.491799	 lr=0.002494
Epoch[179] Batch [2349]/[3760]	Speed: 117.978305 samples/sec	 accuracy=64.710106	 loss=1.491688	 lr=0.002490
Epoch[179] Batch [2399]/[3760]	Speed: 116.825686 samples/sec	 accuracy=64.723958	 loss=1.491519	 lr=0.002486
Epoch[179] Batch [2449]/[3760]	Speed: 117.260450 samples/sec	 accuracy=64.720026	 loss=1.491513	 lr=0.002482
Epoch[179] Batch [2499]/[3760]	Speed: 114.747120 samples/sec	 accuracy=64.718750	 loss=1.492093	 lr=0.002478
Epoch[179] Batch [2549]/[3760]	Speed: 116.510761 samples/sec	 accuracy=64.719975	 loss=1.491644	 lr=0.002474
Epoch[179] Batch [2599]/[3760]	Speed: 117.173984 samples/sec	 accuracy=64.724159	 loss=1.490881	 lr=0.002470
Epoch[179] Batch [2649]/[3760]	Speed: 117.143059 samples/sec	 accuracy=64.754717	 loss=1.490067	 lr=0.002466
Epoch[179] Batch [2699]/[3760]	Speed: 117.594804 samples/sec	 accuracy=64.756366	 loss=1.489840	 lr=0.002462
Epoch[179] Batch [2749]/[3760]	Speed: 117.845806 samples/sec	 accuracy=64.772159	 loss=1.489322	 lr=0.002458
Epoch[179] Batch [2799]/[3760]	Speed: 117.334241 samples/sec	 accuracy=64.765625	 loss=1.489499	 lr=0.002454
Epoch[179] Batch [2849]/[3760]	Speed: 117.825060 samples/sec	 accuracy=64.754386	 loss=1.489558	 lr=0.002450
Epoch[179] Batch [2899]/[3760]	Speed: 118.073472 samples/sec	 accuracy=64.771552	 loss=1.489221	 lr=0.002446
Epoch[179] Batch [2949]/[3760]	Speed: 117.670969 samples/sec	 accuracy=64.758475	 loss=1.490205	 lr=0.002442
Epoch[179] Batch [2999]/[3760]	Speed: 118.162638 samples/sec	 accuracy=64.743750	 loss=1.490665	 lr=0.002438
Epoch[179] Batch [3049]/[3760]	Speed: 117.497868 samples/sec	 accuracy=64.754611	 loss=1.490938	 lr=0.002434
Epoch[179] Batch [3099]/[3760]	Speed: 118.257551 samples/sec	 accuracy=64.746472	 loss=1.490658	 lr=0.002430
Epoch[179] Batch [3149]/[3760]	Speed: 117.216219 samples/sec	 accuracy=64.734127	 loss=1.490712	 lr=0.002426
Epoch[179] Batch [3199]/[3760]	Speed: 117.776489 samples/sec	 accuracy=64.740723	 loss=1.490767	 lr=0.002422
Epoch[179] Batch [3249]/[3760]	Speed: 117.657995 samples/sec	 accuracy=64.758173	 loss=1.490353	 lr=0.002419
Epoch[179] Batch [3299]/[3760]	Speed: 117.511302 samples/sec	 accuracy=64.735322	 loss=1.490699	 lr=0.002415
Epoch[179] Batch [3349]/[3760]	Speed: 117.608478 samples/sec	 accuracy=64.729011	 loss=1.491029	 lr=0.002411
Epoch[179] Batch [3399]/[3760]	Speed: 118.480061 samples/sec	 accuracy=64.716912	 loss=1.491683	 lr=0.002407
Epoch[179] Batch [3449]/[3760]	Speed: 117.422931 samples/sec	 accuracy=64.736413	 loss=1.491198	 lr=0.002403
Epoch[179] Batch [3499]/[3760]	Speed: 117.492869 samples/sec	 accuracy=64.759821	 loss=1.490016	 lr=0.002399
Epoch[179] Batch [3549]/[3760]	Speed: 117.435921 samples/sec	 accuracy=64.746039	 loss=1.490685	 lr=0.002395
Epoch[179] Batch [3599]/[3760]	Speed: 117.724506 samples/sec	 accuracy=64.746094	 loss=1.490091	 lr=0.002391
Epoch[179] Batch [3649]/[3760]	Speed: 116.617916 samples/sec	 accuracy=64.765839	 loss=1.489419	 lr=0.002387
Epoch[179] Batch [3699]/[3760]	Speed: 117.231574 samples/sec	 accuracy=64.754223	 loss=1.489736	 lr=0.002383
Epoch[179] Batch [3749]/[3760]	Speed: 120.310784 samples/sec	 accuracy=64.765833	 loss=1.489118	 lr=0.002379
Batch [0049]/[0303]: acc-top1=59.750000 acc-top5=80.031250
Batch [0099]/[0303]: acc-top1=56.656250 acc-top5=79.343750
Batch [0149]/[0303]: acc-top1=57.895833 acc-top5=79.833333
Batch [0199]/[0303]: acc-top1=60.570312 acc-top5=81.390625
Batch [0249]/[0303]: acc-top1=60.037500 acc-top5=81.150000
Batch [0299]/[0303]: acc-top1=59.562500 acc-top5=81.083333
[Epoch 179] training: accuracy=64.773105	 loss=1.488884
[Epoch 179] speed: 116 samples/sec	time cost: 2174.995111
[Epoch 179] validation: acc-top1=59.524546 acc-top5=81.105611 loss=1.868040
Epoch[180] Batch [0049]/[3759]	Speed: 88.148112 samples/sec	 accuracy=67.093750	 loss=1.402440	 lr=0.002374
Epoch[180] Batch [0099]/[3759]	Speed: 117.820304 samples/sec	 accuracy=66.390625	 loss=1.431654	 lr=0.002370
Epoch[180] Batch [0149]/[3759]	Speed: 117.666368 samples/sec	 accuracy=66.281250	 loss=1.425746	 lr=0.002366
Epoch[180] Batch [0199]/[3759]	Speed: 117.266290 samples/sec	 accuracy=66.234375	 loss=1.428104	 lr=0.002363
Epoch[180] Batch [0249]/[3759]	Speed: 115.545120 samples/sec	 accuracy=66.056250	 loss=1.430367	 lr=0.002359
Epoch[180] Batch [0299]/[3759]	Speed: 117.072065 samples/sec	 accuracy=65.994792	 loss=1.434745	 lr=0.002355
Epoch[180] Batch [0349]/[3759]	Speed: 117.450760 samples/sec	 accuracy=65.937500	 loss=1.435639	 lr=0.002351
Epoch[180] Batch [0399]/[3759]	Speed: 117.379919 samples/sec	 accuracy=65.859375	 loss=1.439363	 lr=0.002347
Epoch[180] Batch [0449]/[3759]	Speed: 117.340983 samples/sec	 accuracy=65.760417	 loss=1.444641	 lr=0.002343
Epoch[180] Batch [0499]/[3759]	Speed: 117.795315 samples/sec	 accuracy=65.693750	 loss=1.445306	 lr=0.002339
Epoch[180] Batch [0549]/[3759]	Speed: 117.521720 samples/sec	 accuracy=65.625000	 loss=1.445764	 lr=0.002335
Epoch[180] Batch [0599]/[3759]	Speed: 117.753078 samples/sec	 accuracy=65.627604	 loss=1.447302	 lr=0.002331
Epoch[180] Batch [0649]/[3759]	Speed: 117.286518 samples/sec	 accuracy=65.694712	 loss=1.443694	 lr=0.002327
Epoch[180] Batch [0699]/[3759]	Speed: 117.127192 samples/sec	 accuracy=65.700893	 loss=1.445082	 lr=0.002324
Epoch[180] Batch [0749]/[3759]	Speed: 116.891691 samples/sec	 accuracy=65.760417	 loss=1.444554	 lr=0.002320
Epoch[180] Batch [0799]/[3759]	Speed: 117.149391 samples/sec	 accuracy=65.750000	 loss=1.444985	 lr=0.002316
Epoch[180] Batch [0849]/[3759]	Speed: 117.486038 samples/sec	 accuracy=65.751838	 loss=1.443969	 lr=0.002312
Epoch[180] Batch [0899]/[3759]	Speed: 118.171655 samples/sec	 accuracy=65.711806	 loss=1.445895	 lr=0.002308
Epoch[180] Batch [0949]/[3759]	Speed: 116.995939 samples/sec	 accuracy=65.730263	 loss=1.446543	 lr=0.002304
Epoch[180] Batch [0999]/[3759]	Speed: 116.663459 samples/sec	 accuracy=65.726562	 loss=1.446600	 lr=0.002300
Epoch[180] Batch [1049]/[3759]	Speed: 117.324984 samples/sec	 accuracy=65.732143	 loss=1.448934	 lr=0.002296
Epoch[180] Batch [1099]/[3759]	Speed: 117.755740 samples/sec	 accuracy=65.703125	 loss=1.451016	 lr=0.002293
Epoch[180] Batch [1149]/[3759]	Speed: 117.284722 samples/sec	 accuracy=65.683424	 loss=1.451972	 lr=0.002289
Epoch[180] Batch [1199]/[3759]	Speed: 117.406123 samples/sec	 accuracy=65.675781	 loss=1.452539	 lr=0.002285
Epoch[180] Batch [1249]/[3759]	Speed: 117.940138 samples/sec	 accuracy=65.683750	 loss=1.451733	 lr=0.002281
Epoch[180] Batch [1299]/[3759]	Speed: 117.623845 samples/sec	 accuracy=65.694712	 loss=1.452470	 lr=0.002277
Epoch[180] Batch [1349]/[3759]	Speed: 117.164293 samples/sec	 accuracy=65.646991	 loss=1.454038	 lr=0.002273
Epoch[180] Batch [1399]/[3759]	Speed: 117.359495 samples/sec	 accuracy=65.578125	 loss=1.455791	 lr=0.002269
Epoch[180] Batch [1449]/[3759]	Speed: 118.147607 samples/sec	 accuracy=65.571121	 loss=1.455694	 lr=0.002266
Epoch[180] Batch [1499]/[3759]	Speed: 117.557382 samples/sec	 accuracy=65.522917	 loss=1.457496	 lr=0.002262
Epoch[180] Batch [1549]/[3759]	Speed: 117.304169 samples/sec	 accuracy=65.497984	 loss=1.458871	 lr=0.002258
Epoch[180] Batch [1599]/[3759]	Speed: 119.000927 samples/sec	 accuracy=65.486328	 loss=1.458664	 lr=0.002254
Epoch[180] Batch [1649]/[3759]	Speed: 117.030560 samples/sec	 accuracy=65.475379	 loss=1.460105	 lr=0.002250
Epoch[180] Batch [1699]/[3759]	Speed: 118.218485 samples/sec	 accuracy=65.448529	 loss=1.460424	 lr=0.002246
Epoch[180] Batch [1749]/[3759]	Speed: 117.054369 samples/sec	 accuracy=65.465179	 loss=1.459867	 lr=0.002243
Epoch[180] Batch [1799]/[3759]	Speed: 115.514304 samples/sec	 accuracy=65.480035	 loss=1.460059	 lr=0.002239
Epoch[180] Batch [1849]/[3759]	Speed: 118.375013 samples/sec	 accuracy=65.435811	 loss=1.460430	 lr=0.002235
Epoch[180] Batch [1899]/[3759]	Speed: 117.105014 samples/sec	 accuracy=65.422697	 loss=1.460858	 lr=0.002231
Epoch[180] Batch [1949]/[3759]	Speed: 118.183922 samples/sec	 accuracy=65.455929	 loss=1.459196	 lr=0.002227
Epoch[180] Batch [1999]/[3759]	Speed: 117.456341 samples/sec	 accuracy=65.510156	 loss=1.457238	 lr=0.002224
Epoch[180] Batch [2049]/[3759]	Speed: 117.728687 samples/sec	 accuracy=65.494665	 loss=1.457933	 lr=0.002220
Epoch[180] Batch [2099]/[3759]	Speed: 117.372059 samples/sec	 accuracy=65.510417	 loss=1.456576	 lr=0.002216
Epoch[180] Batch [2149]/[3759]	Speed: 117.344287 samples/sec	 accuracy=65.521802	 loss=1.455907	 lr=0.002212
Epoch[180] Batch [2199]/[3759]	Speed: 117.518572 samples/sec	 accuracy=65.536932	 loss=1.456055	 lr=0.002208
Epoch[180] Batch [2249]/[3759]	Speed: 118.273382 samples/sec	 accuracy=65.550694	 loss=1.456026	 lr=0.002205
Epoch[180] Batch [2299]/[3759]	Speed: 117.542070 samples/sec	 accuracy=65.546196	 loss=1.455420	 lr=0.002201
Epoch[180] Batch [2349]/[3759]	Speed: 117.818208 samples/sec	 accuracy=65.571144	 loss=1.454973	 lr=0.002197
Epoch[180] Batch [2399]/[3759]	Speed: 117.845304 samples/sec	 accuracy=65.575521	 loss=1.455072	 lr=0.002193
Epoch[180] Batch [2449]/[3759]	Speed: 117.223495 samples/sec	 accuracy=65.577806	 loss=1.455590	 lr=0.002189
Epoch[180] Batch [2499]/[3759]	Speed: 117.400061 samples/sec	 accuracy=65.576875	 loss=1.455965	 lr=0.002186
Epoch[180] Batch [2549]/[3759]	Speed: 117.603479 samples/sec	 accuracy=65.585784	 loss=1.456123	 lr=0.002182
Epoch[180] Batch [2599]/[3759]	Speed: 117.536893 samples/sec	 accuracy=65.588942	 loss=1.456257	 lr=0.002178
Epoch[180] Batch [2649]/[3759]	Speed: 117.652680 samples/sec	 accuracy=65.586085	 loss=1.456427	 lr=0.002174
Epoch[180] Batch [2699]/[3759]	Speed: 117.754000 samples/sec	 accuracy=65.593171	 loss=1.455861	 lr=0.002171
Epoch[180] Batch [2749]/[3759]	Speed: 117.493009 samples/sec	 accuracy=65.581818	 loss=1.455925	 lr=0.002167
Epoch[180] Batch [2799]/[3759]	Speed: 117.285169 samples/sec	 accuracy=65.571987	 loss=1.456645	 lr=0.002163
Epoch[180] Batch [2849]/[3759]	Speed: 117.714105 samples/sec	 accuracy=65.559211	 loss=1.456451	 lr=0.002159
Epoch[180] Batch [2899]/[3759]	Speed: 116.889711 samples/sec	 accuracy=65.525862	 loss=1.458002	 lr=0.002156
Epoch[180] Batch [2949]/[3759]	Speed: 116.743722 samples/sec	 accuracy=65.524364	 loss=1.457537	 lr=0.002152
Epoch[180] Batch [2999]/[3759]	Speed: 117.701475 samples/sec	 accuracy=65.519271	 loss=1.457943	 lr=0.002148
Epoch[180] Batch [3049]/[3759]	Speed: 117.780770 samples/sec	 accuracy=65.486168	 loss=1.458901	 lr=0.002144
Epoch[180] Batch [3099]/[3759]	Speed: 117.776657 samples/sec	 accuracy=65.475302	 loss=1.459292	 lr=0.002141
Epoch[180] Batch [3149]/[3759]	Speed: 117.079697 samples/sec	 accuracy=65.483135	 loss=1.459066	 lr=0.002137
Epoch[180] Batch [3199]/[3759]	Speed: 117.320396 samples/sec	 accuracy=65.456543	 loss=1.460048	 lr=0.002133
Epoch[180] Batch [3249]/[3759]	Speed: 118.052283 samples/sec	 accuracy=65.457212	 loss=1.460567	 lr=0.002129
Epoch[180] Batch [3299]/[3759]	Speed: 117.423363 samples/sec	 accuracy=65.450758	 loss=1.460878	 lr=0.002126
Epoch[180] Batch [3349]/[3759]	Speed: 116.602005 samples/sec	 accuracy=65.446828	 loss=1.460662	 lr=0.002122
Epoch[180] Batch [3399]/[3759]	Speed: 115.369445 samples/sec	 accuracy=65.438419	 loss=1.461164	 lr=0.002118
Epoch[180] Batch [3449]/[3759]	Speed: 117.286816 samples/sec	 accuracy=65.456522	 loss=1.461109	 lr=0.002115
Epoch[180] Batch [3499]/[3759]	Speed: 117.690617 samples/sec	 accuracy=65.445536	 loss=1.461789	 lr=0.002111
Epoch[180] Batch [3549]/[3759]	Speed: 117.410202 samples/sec	 accuracy=65.451144	 loss=1.460956	 lr=0.002107
Epoch[180] Batch [3599]/[3759]	Speed: 116.718254 samples/sec	 accuracy=65.449653	 loss=1.461144	 lr=0.002103
Epoch[180] Batch [3649]/[3759]	Speed: 117.203632 samples/sec	 accuracy=65.452911	 loss=1.460977	 lr=0.002100
Epoch[180] Batch [3699]/[3759]	Speed: 117.754666 samples/sec	 accuracy=65.443412	 loss=1.461216	 lr=0.002096
Epoch[180] Batch [3749]/[3759]	Speed: 120.496068 samples/sec	 accuracy=65.432500	 loss=1.461385	 lr=0.002092
Batch [0049]/[0303]: acc-top1=59.312500 acc-top5=79.343750
Batch [0099]/[0303]: acc-top1=57.156250 acc-top5=79.343750
Batch [0149]/[0303]: acc-top1=58.281250 acc-top5=80.052083
Batch [0199]/[0303]: acc-top1=61.054688 acc-top5=81.640625
Batch [0249]/[0303]: acc-top1=60.387500 acc-top5=81.400000
Batch [0299]/[0303]: acc-top1=59.973958 acc-top5=81.359375
[Epoch 180] training: accuracy=65.432545	 loss=1.461284
[Epoch 180] speed: 116 samples/sec	time cost: 2173.938993
[Epoch 180] validation: acc-top1=59.900990 acc-top5=81.347979 loss=1.840802
Epoch[181] Batch [0049]/[3760]	Speed: 87.653219 samples/sec	 accuracy=65.531250	 loss=1.454491	 lr=0.002088
Epoch[181] Batch [0099]/[3760]	Speed: 117.798935 samples/sec	 accuracy=65.531250	 loss=1.469911	 lr=0.002084
Epoch[181] Batch [0149]/[3760]	Speed: 117.461800 samples/sec	 accuracy=66.468750	 loss=1.441731	 lr=0.002081
Epoch[181] Batch [0199]/[3760]	Speed: 117.666358 samples/sec	 accuracy=66.234375	 loss=1.448162	 lr=0.002077
Epoch[181] Batch [0249]/[3760]	Speed: 117.796852 samples/sec	 accuracy=66.106250	 loss=1.449195	 lr=0.002073
Epoch[181] Batch [0299]/[3760]	Speed: 117.728830 samples/sec	 accuracy=66.135417	 loss=1.445432	 lr=0.002070
Epoch[181] Batch [0349]/[3760]	Speed: 118.126193 samples/sec	 accuracy=66.174107	 loss=1.443817	 lr=0.002066
Epoch[181] Batch [0399]/[3760]	Speed: 117.094333 samples/sec	 accuracy=65.898438	 loss=1.448806	 lr=0.002062
Epoch[181] Batch [0449]/[3760]	Speed: 117.810458 samples/sec	 accuracy=66.034722	 loss=1.442992	 lr=0.002059
Epoch[181] Batch [0499]/[3760]	Speed: 117.529293 samples/sec	 accuracy=65.981250	 loss=1.441569	 lr=0.002055
Epoch[181] Batch [0549]/[3760]	Speed: 117.768608 samples/sec	 accuracy=66.042614	 loss=1.437177	 lr=0.002051
Epoch[181] Batch [0599]/[3760]	Speed: 117.597195 samples/sec	 accuracy=66.033854	 loss=1.436471	 lr=0.002048
Epoch[181] Batch [0649]/[3760]	Speed: 117.625113 samples/sec	 accuracy=66.045673	 loss=1.437512	 lr=0.002044
Epoch[181] Batch [0699]/[3760]	Speed: 117.490964 samples/sec	 accuracy=66.004464	 loss=1.439560	 lr=0.002040
Epoch[181] Batch [0749]/[3760]	Speed: 117.663424 samples/sec	 accuracy=66.004167	 loss=1.439531	 lr=0.002037
Epoch[181] Batch [0799]/[3760]	Speed: 117.499161 samples/sec	 accuracy=66.037109	 loss=1.438567	 lr=0.002033
Epoch[181] Batch [0849]/[3760]	Speed: 117.820668 samples/sec	 accuracy=66.077206	 loss=1.438180	 lr=0.002029
Epoch[181] Batch [0899]/[3760]	Speed: 117.850472 samples/sec	 accuracy=66.152778	 loss=1.434334	 lr=0.002026
Epoch[181] Batch [0949]/[3760]	Speed: 117.839439 samples/sec	 accuracy=66.111842	 loss=1.436780	 lr=0.002022
Epoch[181] Batch [0999]/[3760]	Speed: 117.725231 samples/sec	 accuracy=66.064062	 loss=1.437765	 lr=0.002019
Epoch[181] Batch [1049]/[3760]	Speed: 116.995689 samples/sec	 accuracy=66.013393	 loss=1.438735	 lr=0.002015
Epoch[181] Batch [1099]/[3760]	Speed: 117.946653 samples/sec	 accuracy=66.015625	 loss=1.439620	 lr=0.002011
Epoch[181] Batch [1149]/[3760]	Speed: 114.419033 samples/sec	 accuracy=66.012228	 loss=1.438062	 lr=0.002008
Epoch[181] Batch [1199]/[3760]	Speed: 116.561441 samples/sec	 accuracy=65.983073	 loss=1.436975	 lr=0.002004
Epoch[181] Batch [1249]/[3760]	Speed: 116.724455 samples/sec	 accuracy=65.931250	 loss=1.438945	 lr=0.002000
Epoch[181] Batch [1299]/[3760]	Speed: 117.662876 samples/sec	 accuracy=65.911058	 loss=1.440180	 lr=0.001997
Epoch[181] Batch [1349]/[3760]	Speed: 117.170946 samples/sec	 accuracy=65.959491	 loss=1.439847	 lr=0.001993
Epoch[181] Batch [1399]/[3760]	Speed: 117.205269 samples/sec	 accuracy=65.940848	 loss=1.439747	 lr=0.001990
Epoch[181] Batch [1449]/[3760]	Speed: 117.504429 samples/sec	 accuracy=65.964440	 loss=1.438430	 lr=0.001986
Epoch[181] Batch [1499]/[3760]	Speed: 117.208464 samples/sec	 accuracy=65.937500	 loss=1.439945	 lr=0.001982
Epoch[181] Batch [1549]/[3760]	Speed: 117.309709 samples/sec	 accuracy=65.965726	 loss=1.439678	 lr=0.001979
Epoch[181] Batch [1599]/[3760]	Speed: 117.722441 samples/sec	 accuracy=65.971680	 loss=1.438739	 lr=0.001975
Epoch[181] Batch [1649]/[3760]	Speed: 117.617692 samples/sec	 accuracy=65.968750	 loss=1.439696	 lr=0.001972
Epoch[181] Batch [1699]/[3760]	Speed: 117.446223 samples/sec	 accuracy=65.975184	 loss=1.438204	 lr=0.001968
Epoch[181] Batch [1749]/[3760]	Speed: 116.732513 samples/sec	 accuracy=65.980357	 loss=1.437281	 lr=0.001964
Epoch[181] Batch [1799]/[3760]	Speed: 117.006972 samples/sec	 accuracy=65.971354	 loss=1.438125	 lr=0.001961
Epoch[181] Batch [1849]/[3760]	Speed: 117.198543 samples/sec	 accuracy=65.994932	 loss=1.437508	 lr=0.001957
Epoch[181] Batch [1899]/[3760]	Speed: 117.689109 samples/sec	 accuracy=66.018914	 loss=1.436838	 lr=0.001954
Epoch[181] Batch [1949]/[3760]	Speed: 117.780111 samples/sec	 accuracy=66.036859	 loss=1.435945	 lr=0.001950
Epoch[181] Batch [1999]/[3760]	Speed: 117.895687 samples/sec	 accuracy=66.053125	 loss=1.435468	 lr=0.001947
Epoch[181] Batch [2049]/[3760]	Speed: 118.088607 samples/sec	 accuracy=66.041159	 loss=1.435796	 lr=0.001943
Epoch[181] Batch [2099]/[3760]	Speed: 117.853891 samples/sec	 accuracy=66.052827	 loss=1.435619	 lr=0.001940
Epoch[181] Batch [2149]/[3760]	Speed: 117.611127 samples/sec	 accuracy=66.046512	 loss=1.435618	 lr=0.001936
Epoch[181] Batch [2199]/[3760]	Speed: 117.384491 samples/sec	 accuracy=66.058949	 loss=1.435592	 lr=0.001932
Epoch[181] Batch [2249]/[3760]	Speed: 117.033119 samples/sec	 accuracy=66.034722	 loss=1.436064	 lr=0.001929
Epoch[181] Batch [2299]/[3760]	Speed: 117.604004 samples/sec	 accuracy=66.059103	 loss=1.435971	 lr=0.001925
Epoch[181] Batch [2349]/[3760]	Speed: 117.546933 samples/sec	 accuracy=66.049867	 loss=1.436691	 lr=0.001922
Epoch[181] Batch [2399]/[3760]	Speed: 117.253016 samples/sec	 accuracy=66.013672	 loss=1.437782	 lr=0.001918
Epoch[181] Batch [2449]/[3760]	Speed: 117.870210 samples/sec	 accuracy=66.014031	 loss=1.437935	 lr=0.001915
Epoch[181] Batch [2499]/[3760]	Speed: 117.381048 samples/sec	 accuracy=66.042500	 loss=1.437564	 lr=0.001911
Epoch[181] Batch [2549]/[3760]	Speed: 117.630402 samples/sec	 accuracy=66.041054	 loss=1.437997	 lr=0.001908
Epoch[181] Batch [2599]/[3760]	Speed: 117.407337 samples/sec	 accuracy=66.058894	 loss=1.437625	 lr=0.001904
Epoch[181] Batch [2649]/[3760]	Speed: 116.961995 samples/sec	 accuracy=66.052476	 loss=1.437381	 lr=0.001901
Epoch[181] Batch [2699]/[3760]	Speed: 115.502549 samples/sec	 accuracy=66.098380	 loss=1.435816	 lr=0.001897
Epoch[181] Batch [2749]/[3760]	Speed: 117.603651 samples/sec	 accuracy=66.086364	 loss=1.435562	 lr=0.001894
Epoch[181] Batch [2799]/[3760]	Speed: 117.362697 samples/sec	 accuracy=66.099888	 loss=1.435446	 lr=0.001890
Epoch[181] Batch [2849]/[3760]	Speed: 118.043188 samples/sec	 accuracy=66.125548	 loss=1.435002	 lr=0.001887
Epoch[181] Batch [2899]/[3760]	Speed: 117.116905 samples/sec	 accuracy=66.118534	 loss=1.435347	 lr=0.001883
Epoch[181] Batch [2949]/[3760]	Speed: 117.193322 samples/sec	 accuracy=66.115996	 loss=1.435914	 lr=0.001879
Epoch[181] Batch [2999]/[3760]	Speed: 117.790528 samples/sec	 accuracy=66.108333	 loss=1.436109	 lr=0.001876
Epoch[181] Batch [3049]/[3760]	Speed: 118.088535 samples/sec	 accuracy=66.117316	 loss=1.436040	 lr=0.001872
Epoch[181] Batch [3099]/[3760]	Speed: 118.365082 samples/sec	 accuracy=66.119456	 loss=1.435448	 lr=0.001869
Epoch[181] Batch [3149]/[3760]	Speed: 117.582314 samples/sec	 accuracy=66.100198	 loss=1.435761	 lr=0.001866
Epoch[181] Batch [3199]/[3760]	Speed: 117.812019 samples/sec	 accuracy=66.115234	 loss=1.435361	 lr=0.001862
Epoch[181] Batch [3249]/[3760]	Speed: 117.757662 samples/sec	 accuracy=66.117788	 loss=1.435272	 lr=0.001859
Epoch[181] Batch [3299]/[3760]	Speed: 117.444909 samples/sec	 accuracy=66.121212	 loss=1.435249	 lr=0.001855
Epoch[181] Batch [3349]/[3760]	Speed: 117.911828 samples/sec	 accuracy=66.117537	 loss=1.435150	 lr=0.001852
Epoch[181] Batch [3399]/[3760]	Speed: 118.067613 samples/sec	 accuracy=66.121324	 loss=1.435053	 lr=0.001848
Epoch[181] Batch [3449]/[3760]	Speed: 116.954630 samples/sec	 accuracy=66.115489	 loss=1.434715	 lr=0.001845
Epoch[181] Batch [3499]/[3760]	Speed: 117.269896 samples/sec	 accuracy=66.116964	 loss=1.434998	 lr=0.001841
Epoch[181] Batch [3549]/[3760]	Speed: 118.006170 samples/sec	 accuracy=66.135563	 loss=1.433749	 lr=0.001838
Epoch[181] Batch [3599]/[3760]	Speed: 118.122229 samples/sec	 accuracy=66.144965	 loss=1.432922	 lr=0.001834
Epoch[181] Batch [3649]/[3760]	Speed: 117.004153 samples/sec	 accuracy=66.149401	 loss=1.432445	 lr=0.001831
Epoch[181] Batch [3699]/[3760]	Speed: 117.898783 samples/sec	 accuracy=66.157939	 loss=1.432362	 lr=0.001827
Epoch[181] Batch [3749]/[3760]	Speed: 120.924573 samples/sec	 accuracy=66.146667	 loss=1.432169	 lr=0.001824
Batch [0049]/[0303]: acc-top1=59.062500 acc-top5=79.687500
Batch [0099]/[0303]: acc-top1=57.375000 acc-top5=79.765625
Batch [0149]/[0303]: acc-top1=58.687500 acc-top5=80.468750
Batch [0199]/[0303]: acc-top1=61.304688 acc-top5=81.882812
Batch [0249]/[0303]: acc-top1=60.531250 acc-top5=81.612500
Batch [0299]/[0303]: acc-top1=60.067708 acc-top5=81.661458
[Epoch 181] training: accuracy=66.148604	 loss=1.432057
[Epoch 181] speed: 116 samples/sec	time cost: 2173.177061
[Epoch 181] validation: acc-top1=60.004125 acc-top5=81.693482 loss=1.846985
Epoch[182] Batch [0049]/[3760]	Speed: 87.946698 samples/sec	 accuracy=66.093750	 loss=1.395789	 lr=0.001820
Epoch[182] Batch [0099]/[3760]	Speed: 118.159111 samples/sec	 accuracy=66.750000	 loss=1.372465	 lr=0.001816
Epoch[182] Batch [0149]/[3760]	Speed: 118.254475 samples/sec	 accuracy=67.125000	 loss=1.372665	 lr=0.001813
Epoch[182] Batch [0199]/[3760]	Speed: 116.997584 samples/sec	 accuracy=67.226562	 loss=1.370906	 lr=0.001809
Epoch[182] Batch [0249]/[3760]	Speed: 116.929973 samples/sec	 accuracy=67.387500	 loss=1.364568	 lr=0.001806
Epoch[182] Batch [0299]/[3760]	Speed: 117.953886 samples/sec	 accuracy=67.161458	 loss=1.379579	 lr=0.001803
Epoch[182] Batch [0349]/[3760]	Speed: 117.229463 samples/sec	 accuracy=67.263393	 loss=1.371964	 lr=0.001799
Epoch[182] Batch [0399]/[3760]	Speed: 115.270359 samples/sec	 accuracy=67.355469	 loss=1.364989	 lr=0.001796
Epoch[182] Batch [0449]/[3760]	Speed: 117.752372 samples/sec	 accuracy=67.166667	 loss=1.377232	 lr=0.001792
Epoch[182] Batch [0499]/[3760]	Speed: 117.496774 samples/sec	 accuracy=67.206250	 loss=1.373055	 lr=0.001789
Epoch[182] Batch [0549]/[3760]	Speed: 117.733088 samples/sec	 accuracy=67.130682	 loss=1.381278	 lr=0.001785
Epoch[182] Batch [0599]/[3760]	Speed: 117.225285 samples/sec	 accuracy=67.015625	 loss=1.383105	 lr=0.001782
Epoch[182] Batch [0649]/[3760]	Speed: 117.773258 samples/sec	 accuracy=66.968750	 loss=1.386064	 lr=0.001779
Epoch[182] Batch [0699]/[3760]	Speed: 117.167356 samples/sec	 accuracy=66.975446	 loss=1.386774	 lr=0.001775
Epoch[182] Batch [0749]/[3760]	Speed: 117.646764 samples/sec	 accuracy=66.983333	 loss=1.389091	 lr=0.001772
Epoch[182] Batch [0799]/[3760]	Speed: 118.219131 samples/sec	 accuracy=66.927734	 loss=1.391342	 lr=0.001768
Epoch[182] Batch [0849]/[3760]	Speed: 118.237362 samples/sec	 accuracy=66.906250	 loss=1.391690	 lr=0.001765
Epoch[182] Batch [0899]/[3760]	Speed: 117.536227 samples/sec	 accuracy=66.897569	 loss=1.392666	 lr=0.001762
Epoch[182] Batch [0949]/[3760]	Speed: 117.710617 samples/sec	 accuracy=66.901316	 loss=1.393680	 lr=0.001758
Epoch[182] Batch [0999]/[3760]	Speed: 117.576342 samples/sec	 accuracy=66.900000	 loss=1.395703	 lr=0.001755
Epoch[182] Batch [1049]/[3760]	Speed: 116.769269 samples/sec	 accuracy=66.909226	 loss=1.393346	 lr=0.001751
Epoch[182] Batch [1099]/[3760]	Speed: 117.581159 samples/sec	 accuracy=66.896307	 loss=1.392858	 lr=0.001748
Epoch[182] Batch [1149]/[3760]	Speed: 116.802361 samples/sec	 accuracy=66.930707	 loss=1.392899	 lr=0.001745
Epoch[182] Batch [1199]/[3760]	Speed: 117.323722 samples/sec	 accuracy=66.917969	 loss=1.393013	 lr=0.001741
Epoch[182] Batch [1249]/[3760]	Speed: 117.237295 samples/sec	 accuracy=66.917500	 loss=1.393146	 lr=0.001738
Epoch[182] Batch [1299]/[3760]	Speed: 117.889236 samples/sec	 accuracy=66.852163	 loss=1.394472	 lr=0.001735
Epoch[182] Batch [1349]/[3760]	Speed: 117.415696 samples/sec	 accuracy=66.833333	 loss=1.395771	 lr=0.001731
Epoch[182] Batch [1399]/[3760]	Speed: 117.727125 samples/sec	 accuracy=66.813616	 loss=1.396261	 lr=0.001728
Epoch[182] Batch [1449]/[3760]	Speed: 118.038305 samples/sec	 accuracy=66.830819	 loss=1.395219	 lr=0.001724
Epoch[182] Batch [1499]/[3760]	Speed: 117.752117 samples/sec	 accuracy=66.835417	 loss=1.395348	 lr=0.001721
Epoch[182] Batch [1549]/[3760]	Speed: 117.246960 samples/sec	 accuracy=66.792339	 loss=1.396732	 lr=0.001718
Epoch[182] Batch [1599]/[3760]	Speed: 118.106871 samples/sec	 accuracy=66.818359	 loss=1.396046	 lr=0.001714
Epoch[182] Batch [1649]/[3760]	Speed: 118.094509 samples/sec	 accuracy=66.840909	 loss=1.395891	 lr=0.001711
Epoch[182] Batch [1699]/[3760]	Speed: 117.910602 samples/sec	 accuracy=66.852941	 loss=1.394625	 lr=0.001708
Epoch[182] Batch [1749]/[3760]	Speed: 117.297273 samples/sec	 accuracy=66.873214	 loss=1.394532	 lr=0.001704
Epoch[182] Batch [1799]/[3760]	Speed: 117.699610 samples/sec	 accuracy=66.869792	 loss=1.396287	 lr=0.001701
Epoch[182] Batch [1849]/[3760]	Speed: 118.181290 samples/sec	 accuracy=66.807432	 loss=1.398816	 lr=0.001698
Epoch[182] Batch [1899]/[3760]	Speed: 117.689420 samples/sec	 accuracy=66.785362	 loss=1.399841	 lr=0.001694
Epoch[182] Batch [1949]/[3760]	Speed: 118.147633 samples/sec	 accuracy=66.780449	 loss=1.400090	 lr=0.001691
Epoch[182] Batch [1999]/[3760]	Speed: 115.437622 samples/sec	 accuracy=66.763281	 loss=1.400864	 lr=0.001688
Epoch[182] Batch [2049]/[3760]	Speed: 117.689073 samples/sec	 accuracy=66.739329	 loss=1.402199	 lr=0.001684
Epoch[182] Batch [2099]/[3760]	Speed: 116.929864 samples/sec	 accuracy=66.762649	 loss=1.401677	 lr=0.001681
Epoch[182] Batch [2149]/[3760]	Speed: 117.101226 samples/sec	 accuracy=66.776890	 loss=1.401420	 lr=0.001678
Epoch[182] Batch [2199]/[3760]	Speed: 117.471332 samples/sec	 accuracy=66.753551	 loss=1.402599	 lr=0.001674
Epoch[182] Batch [2249]/[3760]	Speed: 117.407296 samples/sec	 accuracy=66.725000	 loss=1.403504	 lr=0.001671
Epoch[182] Batch [2299]/[3760]	Speed: 117.815422 samples/sec	 accuracy=66.730299	 loss=1.403511	 lr=0.001668
Epoch[182] Batch [2349]/[3760]	Speed: 117.152558 samples/sec	 accuracy=66.730053	 loss=1.403317	 lr=0.001665
Epoch[182] Batch [2399]/[3760]	Speed: 117.987739 samples/sec	 accuracy=66.726562	 loss=1.403453	 lr=0.001661
Epoch[182] Batch [2449]/[3760]	Speed: 117.342556 samples/sec	 accuracy=66.721301	 loss=1.404423	 lr=0.001658
Epoch[182] Batch [2499]/[3760]	Speed: 117.492691 samples/sec	 accuracy=66.728750	 loss=1.404265	 lr=0.001655
Epoch[182] Batch [2549]/[3760]	Speed: 117.150698 samples/sec	 accuracy=66.758578	 loss=1.403430	 lr=0.001651
Epoch[182] Batch [2599]/[3760]	Speed: 117.757281 samples/sec	 accuracy=66.778245	 loss=1.402439	 lr=0.001648
Epoch[182] Batch [2649]/[3760]	Speed: 117.705452 samples/sec	 accuracy=66.775354	 loss=1.402546	 lr=0.001645
Epoch[182] Batch [2699]/[3760]	Speed: 117.382953 samples/sec	 accuracy=66.760417	 loss=1.402738	 lr=0.001641
Epoch[182] Batch [2749]/[3760]	Speed: 117.656139 samples/sec	 accuracy=66.753977	 loss=1.402519	 lr=0.001638
Epoch[182] Batch [2799]/[3760]	Speed: 117.461905 samples/sec	 accuracy=66.749442	 loss=1.402615	 lr=0.001635
Epoch[182] Batch [2849]/[3760]	Speed: 117.402624 samples/sec	 accuracy=66.722588	 loss=1.403887	 lr=0.001632
Epoch[182] Batch [2899]/[3760]	Speed: 117.578114 samples/sec	 accuracy=66.716056	 loss=1.404064	 lr=0.001628
Epoch[182] Batch [2949]/[3760]	Speed: 116.787602 samples/sec	 accuracy=66.706568	 loss=1.404604	 lr=0.001625
Epoch[182] Batch [2999]/[3760]	Speed: 117.484040 samples/sec	 accuracy=66.720833	 loss=1.404237	 lr=0.001622
Epoch[182] Batch [3049]/[3760]	Speed: 117.587677 samples/sec	 accuracy=66.721824	 loss=1.403410	 lr=0.001619
Epoch[182] Batch [3099]/[3760]	Speed: 117.333929 samples/sec	 accuracy=66.732863	 loss=1.403169	 lr=0.001615
Epoch[182] Batch [3149]/[3760]	Speed: 116.564082 samples/sec	 accuracy=66.741071	 loss=1.403201	 lr=0.001612
Epoch[182] Batch [3199]/[3760]	Speed: 116.899392 samples/sec	 accuracy=66.730957	 loss=1.403360	 lr=0.001609
Epoch[182] Batch [3249]/[3760]	Speed: 117.307091 samples/sec	 accuracy=66.729327	 loss=1.403160	 lr=0.001606
Epoch[182] Batch [3299]/[3760]	Speed: 117.839869 samples/sec	 accuracy=66.720644	 loss=1.403675	 lr=0.001602
Epoch[182] Batch [3349]/[3760]	Speed: 117.351599 samples/sec	 accuracy=66.702892	 loss=1.404133	 lr=0.001599
Epoch[182] Batch [3399]/[3760]	Speed: 117.687238 samples/sec	 accuracy=66.679688	 loss=1.404886	 lr=0.001596
Epoch[182] Batch [3449]/[3760]	Speed: 117.263075 samples/sec	 accuracy=66.675725	 loss=1.404927	 lr=0.001593
Epoch[182] Batch [3499]/[3760]	Speed: 117.094574 samples/sec	 accuracy=66.689286	 loss=1.404765	 lr=0.001589
Epoch[182] Batch [3549]/[3760]	Speed: 114.649674 samples/sec	 accuracy=66.675176	 loss=1.405399	 lr=0.001586
Epoch[182] Batch [3599]/[3760]	Speed: 117.497511 samples/sec	 accuracy=66.692274	 loss=1.404789	 lr=0.001583
Epoch[182] Batch [3649]/[3760]	Speed: 117.222130 samples/sec	 accuracy=66.705051	 loss=1.404372	 lr=0.001580
Epoch[182] Batch [3699]/[3760]	Speed: 117.759672 samples/sec	 accuracy=66.714527	 loss=1.403973	 lr=0.001577
Epoch[182] Batch [3749]/[3760]	Speed: 119.895799 samples/sec	 accuracy=66.723750	 loss=1.403840	 lr=0.001573
Batch [0049]/[0303]: acc-top1=60.500000 acc-top5=80.125000
Batch [0099]/[0303]: acc-top1=57.656250 acc-top5=79.625000
Batch [0149]/[0303]: acc-top1=58.822917 acc-top5=80.312500
Batch [0199]/[0303]: acc-top1=61.726563 acc-top5=81.890625
Batch [0249]/[0303]: acc-top1=61.093750 acc-top5=81.706250
Batch [0299]/[0303]: acc-top1=60.484375 acc-top5=81.687500
[Epoch 182] training: accuracy=66.728723	 loss=1.403556
[Epoch 182] speed: 116 samples/sec	time cost: 2174.435491
[Epoch 182] validation: acc-top1=60.421823 acc-top5=81.703795 loss=1.831159
Epoch[183] Batch [0049]/[3759]	Speed: 87.550341 samples/sec	 accuracy=67.968750	 loss=1.352516	 lr=0.001570
Epoch[183] Batch [0099]/[3759]	Speed: 118.447999 samples/sec	 accuracy=68.031250	 loss=1.360421	 lr=0.001566
Epoch[183] Batch [0149]/[3759]	Speed: 116.854555 samples/sec	 accuracy=67.677083	 loss=1.372821	 lr=0.001563
Epoch[183] Batch [0199]/[3759]	Speed: 117.470291 samples/sec	 accuracy=67.875000	 loss=1.362396	 lr=0.001560
Epoch[183] Batch [0249]/[3759]	Speed: 117.141032 samples/sec	 accuracy=67.931250	 loss=1.358839	 lr=0.001557
Epoch[183] Batch [0299]/[3759]	Speed: 117.251792 samples/sec	 accuracy=67.765625	 loss=1.359015	 lr=0.001554
Epoch[183] Batch [0349]/[3759]	Speed: 117.546352 samples/sec	 accuracy=67.834821	 loss=1.357671	 lr=0.001550
Epoch[183] Batch [0399]/[3759]	Speed: 117.827680 samples/sec	 accuracy=67.761719	 loss=1.357122	 lr=0.001547
Epoch[183] Batch [0449]/[3759]	Speed: 117.302073 samples/sec	 accuracy=68.000000	 loss=1.351573	 lr=0.001544
Epoch[183] Batch [0499]/[3759]	Speed: 118.514847 samples/sec	 accuracy=68.025000	 loss=1.348184	 lr=0.001541
Epoch[183] Batch [0549]/[3759]	Speed: 117.084738 samples/sec	 accuracy=67.985795	 loss=1.351474	 lr=0.001538
Epoch[183] Batch [0599]/[3759]	Speed: 117.285494 samples/sec	 accuracy=67.984375	 loss=1.352277	 lr=0.001534
Epoch[183] Batch [0649]/[3759]	Speed: 117.579923 samples/sec	 accuracy=67.975962	 loss=1.351161	 lr=0.001531
Epoch[183] Batch [0699]/[3759]	Speed: 117.612138 samples/sec	 accuracy=67.919643	 loss=1.353681	 lr=0.001528
Epoch[183] Batch [0749]/[3759]	Speed: 117.377050 samples/sec	 accuracy=67.925000	 loss=1.351770	 lr=0.001525
Epoch[183] Batch [0799]/[3759]	Speed: 117.813818 samples/sec	 accuracy=67.914062	 loss=1.352170	 lr=0.001522
Epoch[183] Batch [0849]/[3759]	Speed: 117.307743 samples/sec	 accuracy=67.919118	 loss=1.350913	 lr=0.001519
Epoch[183] Batch [0899]/[3759]	Speed: 117.621302 samples/sec	 accuracy=67.996528	 loss=1.349090	 lr=0.001515
Epoch[183] Batch [0949]/[3759]	Speed: 117.877253 samples/sec	 accuracy=67.963816	 loss=1.351182	 lr=0.001512
Epoch[183] Batch [0999]/[3759]	Speed: 117.583291 samples/sec	 accuracy=67.887500	 loss=1.354527	 lr=0.001509
Epoch[183] Batch [1049]/[3759]	Speed: 117.896644 samples/sec	 accuracy=67.931548	 loss=1.354186	 lr=0.001506
Epoch[183] Batch [1099]/[3759]	Speed: 117.616270 samples/sec	 accuracy=67.907670	 loss=1.354251	 lr=0.001503
Epoch[183] Batch [1149]/[3759]	Speed: 117.744877 samples/sec	 accuracy=67.923913	 loss=1.354064	 lr=0.001500
Epoch[183] Batch [1199]/[3759]	Speed: 117.191384 samples/sec	 accuracy=67.904948	 loss=1.353724	 lr=0.001497
Epoch[183] Batch [1249]/[3759]	Speed: 117.426515 samples/sec	 accuracy=67.921250	 loss=1.351747	 lr=0.001494
Epoch[183] Batch [1299]/[3759]	Speed: 117.422501 samples/sec	 accuracy=67.877404	 loss=1.351863	 lr=0.001490
Epoch[183] Batch [1349]/[3759]	Speed: 118.242153 samples/sec	 accuracy=67.835648	 loss=1.352219	 lr=0.001487
Epoch[183] Batch [1399]/[3759]	Speed: 115.480069 samples/sec	 accuracy=67.813616	 loss=1.354192	 lr=0.001484
Epoch[183] Batch [1449]/[3759]	Speed: 117.674552 samples/sec	 accuracy=67.808190	 loss=1.354762	 lr=0.001481
Epoch[183] Batch [1499]/[3759]	Speed: 117.710412 samples/sec	 accuracy=67.764583	 loss=1.356983	 lr=0.001478
Epoch[183] Batch [1549]/[3759]	Speed: 117.654434 samples/sec	 accuracy=67.781250	 loss=1.357045	 lr=0.001475
Epoch[183] Batch [1599]/[3759]	Speed: 117.149848 samples/sec	 accuracy=67.734375	 loss=1.358992	 lr=0.001472
Epoch[183] Batch [1649]/[3759]	Speed: 118.038625 samples/sec	 accuracy=67.666667	 loss=1.362055	 lr=0.001469
Epoch[183] Batch [1699]/[3759]	Speed: 117.915668 samples/sec	 accuracy=67.635110	 loss=1.362931	 lr=0.001465
Epoch[183] Batch [1749]/[3759]	Speed: 118.647873 samples/sec	 accuracy=67.608929	 loss=1.365210	 lr=0.001462
Epoch[183] Batch [1799]/[3759]	Speed: 117.501827 samples/sec	 accuracy=67.636285	 loss=1.363689	 lr=0.001459
Epoch[183] Batch [1849]/[3759]	Speed: 116.182587 samples/sec	 accuracy=67.647804	 loss=1.362979	 lr=0.001456
Epoch[183] Batch [1899]/[3759]	Speed: 118.013020 samples/sec	 accuracy=67.616776	 loss=1.364661	 lr=0.001453
Epoch[183] Batch [1949]/[3759]	Speed: 117.867686 samples/sec	 accuracy=67.572115	 loss=1.365341	 lr=0.001450
Epoch[183] Batch [1999]/[3759]	Speed: 117.684250 samples/sec	 accuracy=67.584375	 loss=1.365361	 lr=0.001447
Epoch[183] Batch [2049]/[3759]	Speed: 117.134904 samples/sec	 accuracy=67.570122	 loss=1.365613	 lr=0.001444
Epoch[183] Batch [2099]/[3759]	Speed: 117.990235 samples/sec	 accuracy=67.597470	 loss=1.364139	 lr=0.001441
Epoch[183] Batch [2149]/[3759]	Speed: 117.904106 samples/sec	 accuracy=67.617006	 loss=1.363347	 lr=0.001438
Epoch[183] Batch [2199]/[3759]	Speed: 116.882062 samples/sec	 accuracy=67.635653	 loss=1.362524	 lr=0.001435
Epoch[183] Batch [2249]/[3759]	Speed: 117.434088 samples/sec	 accuracy=67.652778	 loss=1.361469	 lr=0.001432
Epoch[183] Batch [2299]/[3759]	Speed: 117.732749 samples/sec	 accuracy=67.661685	 loss=1.361971	 lr=0.001429
Epoch[183] Batch [2349]/[3759]	Speed: 117.376481 samples/sec	 accuracy=67.678191	 loss=1.361360	 lr=0.001425
Epoch[183] Batch [2399]/[3759]	Speed: 116.495341 samples/sec	 accuracy=67.657552	 loss=1.362350	 lr=0.001422
Epoch[183] Batch [2449]/[3759]	Speed: 117.129931 samples/sec	 accuracy=67.627551	 loss=1.363821	 lr=0.001419
Epoch[183] Batch [2499]/[3759]	Speed: 116.919082 samples/sec	 accuracy=67.613750	 loss=1.363787	 lr=0.001416
Epoch[183] Batch [2549]/[3759]	Speed: 117.674597 samples/sec	 accuracy=67.635417	 loss=1.363918	 lr=0.001413
Epoch[183] Batch [2599]/[3759]	Speed: 117.820793 samples/sec	 accuracy=67.623197	 loss=1.364797	 lr=0.001410
Epoch[183] Batch [2649]/[3759]	Speed: 118.349260 samples/sec	 accuracy=67.596108	 loss=1.365096	 lr=0.001407
Epoch[183] Batch [2699]/[3759]	Speed: 117.634464 samples/sec	 accuracy=67.612847	 loss=1.364747	 lr=0.001404
Epoch[183] Batch [2749]/[3759]	Speed: 116.957077 samples/sec	 accuracy=67.619886	 loss=1.364463	 lr=0.001401
Epoch[183] Batch [2799]/[3759]	Speed: 117.202625 samples/sec	 accuracy=67.622210	 loss=1.364104	 lr=0.001398
Epoch[183] Batch [2849]/[3759]	Speed: 117.963898 samples/sec	 accuracy=67.610197	 loss=1.365310	 lr=0.001395
Epoch[183] Batch [2899]/[3759]	Speed: 117.004266 samples/sec	 accuracy=67.591595	 loss=1.365841	 lr=0.001392
Epoch[183] Batch [2949]/[3759]	Speed: 117.165516 samples/sec	 accuracy=67.574153	 loss=1.367024	 lr=0.001389
Epoch[183] Batch [2999]/[3759]	Speed: 114.892769 samples/sec	 accuracy=67.576563	 loss=1.367073	 lr=0.001386
Epoch[183] Batch [3049]/[3759]	Speed: 118.105560 samples/sec	 accuracy=67.557889	 loss=1.367189	 lr=0.001383
Epoch[183] Batch [3099]/[3759]	Speed: 116.629311 samples/sec	 accuracy=67.574597	 loss=1.366660	 lr=0.001380
Epoch[183] Batch [3149]/[3759]	Speed: 117.704610 samples/sec	 accuracy=67.578869	 loss=1.366599	 lr=0.001377
Epoch[183] Batch [3199]/[3759]	Speed: 117.269612 samples/sec	 accuracy=67.567383	 loss=1.367336	 lr=0.001374
Epoch[183] Batch [3249]/[3759]	Speed: 117.938020 samples/sec	 accuracy=67.581250	 loss=1.366874	 lr=0.001371
Epoch[183] Batch [3299]/[3759]	Speed: 117.155063 samples/sec	 accuracy=67.575284	 loss=1.367015	 lr=0.001368
Epoch[183] Batch [3349]/[3759]	Speed: 117.419546 samples/sec	 accuracy=67.583489	 loss=1.366969	 lr=0.001365
Epoch[183] Batch [3399]/[3759]	Speed: 117.711750 samples/sec	 accuracy=67.564798	 loss=1.367818	 lr=0.001362
Epoch[183] Batch [3449]/[3759]	Speed: 117.557814 samples/sec	 accuracy=67.566576	 loss=1.367497	 lr=0.001359
Epoch[183] Batch [3499]/[3759]	Speed: 117.505190 samples/sec	 accuracy=67.546875	 loss=1.368411	 lr=0.001356
Epoch[183] Batch [3549]/[3759]	Speed: 118.002008 samples/sec	 accuracy=67.539173	 loss=1.368505	 lr=0.001353
Epoch[183] Batch [3599]/[3759]	Speed: 118.141742 samples/sec	 accuracy=67.536024	 loss=1.368688	 lr=0.001350
Epoch[183] Batch [3649]/[3759]	Speed: 117.730070 samples/sec	 accuracy=67.537243	 loss=1.368844	 lr=0.001347
Epoch[183] Batch [3699]/[3759]	Speed: 117.368630 samples/sec	 accuracy=67.522804	 loss=1.369121	 lr=0.001344
Epoch[183] Batch [3749]/[3759]	Speed: 120.574821 samples/sec	 accuracy=67.536667	 loss=1.368616	 lr=0.001341
Batch [0049]/[0303]: acc-top1=60.281250 acc-top5=79.687500
Batch [0099]/[0303]: acc-top1=57.640625 acc-top5=79.625000
Batch [0149]/[0303]: acc-top1=58.937500 acc-top5=80.375000
Batch [0199]/[0303]: acc-top1=61.453125 acc-top5=81.914062
Batch [0249]/[0303]: acc-top1=60.937500 acc-top5=81.687500
Batch [0299]/[0303]: acc-top1=60.536458 acc-top5=81.744792
[Epoch 183] training: accuracy=67.539572	 loss=1.368658
[Epoch 183] speed: 116 samples/sec	time cost: 2173.346551
[Epoch 183] validation: acc-top1=60.483705 acc-top5=81.770833 loss=1.834068
Epoch[184] Batch [0049]/[3760]	Speed: 87.484382 samples/sec	 accuracy=67.281250	 loss=1.347841	 lr=0.001338
Epoch[184] Batch [0099]/[3760]	Speed: 118.110972 samples/sec	 accuracy=68.250000	 loss=1.337432	 lr=0.001335
Epoch[184] Batch [0149]/[3760]	Speed: 116.660884 samples/sec	 accuracy=68.104167	 loss=1.355214	 lr=0.001332
Epoch[184] Batch [0199]/[3760]	Speed: 117.907586 samples/sec	 accuracy=68.390625	 loss=1.337156	 lr=0.001329
Epoch[184] Batch [0249]/[3760]	Speed: 117.492417 samples/sec	 accuracy=68.187500	 loss=1.338602	 lr=0.001326
Epoch[184] Batch [0299]/[3760]	Speed: 117.160760 samples/sec	 accuracy=68.036458	 loss=1.338536	 lr=0.001323
Epoch[184] Batch [0349]/[3760]	Speed: 118.401293 samples/sec	 accuracy=67.897321	 loss=1.347407	 lr=0.001320
Epoch[184] Batch [0399]/[3760]	Speed: 117.169320 samples/sec	 accuracy=67.828125	 loss=1.353680	 lr=0.001317
Epoch[184] Batch [0449]/[3760]	Speed: 117.787848 samples/sec	 accuracy=68.034722	 loss=1.345555	 lr=0.001314
Epoch[184] Batch [0499]/[3760]	Speed: 117.653998 samples/sec	 accuracy=68.003125	 loss=1.349169	 lr=0.001311
Epoch[184] Batch [0549]/[3760]	Speed: 117.664491 samples/sec	 accuracy=67.991477	 loss=1.349906	 lr=0.001308
Epoch[184] Batch [0599]/[3760]	Speed: 117.705475 samples/sec	 accuracy=68.020833	 loss=1.349010	 lr=0.001305
Epoch[184] Batch [0649]/[3760]	Speed: 118.213832 samples/sec	 accuracy=67.947115	 loss=1.352314	 lr=0.001302
Epoch[184] Batch [0699]/[3760]	Speed: 117.699088 samples/sec	 accuracy=67.975446	 loss=1.352543	 lr=0.001299
Epoch[184] Batch [0749]/[3760]	Speed: 117.332423 samples/sec	 accuracy=67.995833	 loss=1.353625	 lr=0.001296
Epoch[184] Batch [0799]/[3760]	Speed: 114.772854 samples/sec	 accuracy=68.017578	 loss=1.353822	 lr=0.001294
Epoch[184] Batch [0849]/[3760]	Speed: 117.573674 samples/sec	 accuracy=68.027574	 loss=1.352433	 lr=0.001291
Epoch[184] Batch [0899]/[3760]	Speed: 117.185656 samples/sec	 accuracy=67.968750	 loss=1.354416	 lr=0.001288
Epoch[184] Batch [0949]/[3760]	Speed: 118.195417 samples/sec	 accuracy=67.888158	 loss=1.355457	 lr=0.001285
Epoch[184] Batch [0999]/[3760]	Speed: 117.531075 samples/sec	 accuracy=67.876563	 loss=1.352888	 lr=0.001282
Epoch[184] Batch [1049]/[3760]	Speed: 118.012259 samples/sec	 accuracy=67.846726	 loss=1.354665	 lr=0.001279
Epoch[184] Batch [1099]/[3760]	Speed: 117.454514 samples/sec	 accuracy=67.873580	 loss=1.351361	 lr=0.001276
Epoch[184] Batch [1149]/[3760]	Speed: 117.401831 samples/sec	 accuracy=67.873641	 loss=1.352201	 lr=0.001273
Epoch[184] Batch [1199]/[3760]	Speed: 118.145482 samples/sec	 accuracy=67.910156	 loss=1.350118	 lr=0.001270
Epoch[184] Batch [1249]/[3760]	Speed: 117.545964 samples/sec	 accuracy=67.885000	 loss=1.351285	 lr=0.001267
Epoch[184] Batch [1299]/[3760]	Speed: 117.158502 samples/sec	 accuracy=67.903846	 loss=1.351095	 lr=0.001265
Epoch[184] Batch [1349]/[3760]	Speed: 117.819382 samples/sec	 accuracy=67.920139	 loss=1.351161	 lr=0.001262
Epoch[184] Batch [1399]/[3760]	Speed: 117.481728 samples/sec	 accuracy=67.957589	 loss=1.350336	 lr=0.001259
Epoch[184] Batch [1449]/[3760]	Speed: 116.907241 samples/sec	 accuracy=68.016164	 loss=1.347634	 lr=0.001256
Epoch[184] Batch [1499]/[3760]	Speed: 117.765519 samples/sec	 accuracy=68.038542	 loss=1.346550	 lr=0.001253
Epoch[184] Batch [1549]/[3760]	Speed: 117.504435 samples/sec	 accuracy=68.071573	 loss=1.344921	 lr=0.001250
Epoch[184] Batch [1599]/[3760]	Speed: 118.039502 samples/sec	 accuracy=68.071289	 loss=1.343718	 lr=0.001247
Epoch[184] Batch [1649]/[3760]	Speed: 118.308141 samples/sec	 accuracy=68.055871	 loss=1.344148	 lr=0.001244
Epoch[184] Batch [1699]/[3760]	Speed: 118.369393 samples/sec	 accuracy=68.036765	 loss=1.344385	 lr=0.001242
Epoch[184] Batch [1749]/[3760]	Speed: 117.686918 samples/sec	 accuracy=68.062500	 loss=1.343202	 lr=0.001239
Epoch[184] Batch [1799]/[3760]	Speed: 118.329434 samples/sec	 accuracy=68.051215	 loss=1.344408	 lr=0.001236
Epoch[184] Batch [1849]/[3760]	Speed: 117.721950 samples/sec	 accuracy=68.021959	 loss=1.344734	 lr=0.001233
Epoch[184] Batch [1899]/[3760]	Speed: 118.173485 samples/sec	 accuracy=68.037007	 loss=1.343120	 lr=0.001230
Epoch[184] Batch [1949]/[3760]	Speed: 117.571365 samples/sec	 accuracy=68.010417	 loss=1.344249	 lr=0.001227
Epoch[184] Batch [1999]/[3760]	Speed: 117.814177 samples/sec	 accuracy=68.029688	 loss=1.343887	 lr=0.001225
Epoch[184] Batch [2049]/[3760]	Speed: 116.787227 samples/sec	 accuracy=68.029726	 loss=1.344139	 lr=0.001222
Epoch[184] Batch [2099]/[3760]	Speed: 117.121470 samples/sec	 accuracy=68.020089	 loss=1.344033	 lr=0.001219
Epoch[184] Batch [2149]/[3760]	Speed: 117.836117 samples/sec	 accuracy=68.024709	 loss=1.344279	 lr=0.001216
Epoch[184] Batch [2199]/[3760]	Speed: 117.527087 samples/sec	 accuracy=68.032670	 loss=1.343297	 lr=0.001213
Epoch[184] Batch [2249]/[3760]	Speed: 117.922063 samples/sec	 accuracy=68.052083	 loss=1.342489	 lr=0.001210
Epoch[184] Batch [2299]/[3760]	Speed: 117.641640 samples/sec	 accuracy=68.052989	 loss=1.342994	 lr=0.001208
Epoch[184] Batch [2349]/[3760]	Speed: 114.469012 samples/sec	 accuracy=68.067819	 loss=1.341718	 lr=0.001205
Epoch[184] Batch [2399]/[3760]	Speed: 117.348412 samples/sec	 accuracy=68.071615	 loss=1.341480	 lr=0.001202
Epoch[184] Batch [2449]/[3760]	Speed: 117.912366 samples/sec	 accuracy=68.070153	 loss=1.341338	 lr=0.001199
Epoch[184] Batch [2499]/[3760]	Speed: 117.126527 samples/sec	 accuracy=68.061875	 loss=1.340730	 lr=0.001196
Epoch[184] Batch [2549]/[3760]	Speed: 117.595215 samples/sec	 accuracy=68.046569	 loss=1.341385	 lr=0.001193
Epoch[184] Batch [2599]/[3760]	Speed: 117.575580 samples/sec	 accuracy=68.055288	 loss=1.341298	 lr=0.001191
Epoch[184] Batch [2649]/[3760]	Speed: 117.107845 samples/sec	 accuracy=68.068396	 loss=1.340653	 lr=0.001188
Epoch[184] Batch [2699]/[3760]	Speed: 117.077012 samples/sec	 accuracy=68.049190	 loss=1.341229	 lr=0.001185
Epoch[184] Batch [2749]/[3760]	Speed: 117.838176 samples/sec	 accuracy=68.043182	 loss=1.341394	 lr=0.001182
Epoch[184] Batch [2799]/[3760]	Speed: 118.265704 samples/sec	 accuracy=68.059152	 loss=1.341048	 lr=0.001180
Epoch[184] Batch [2849]/[3760]	Speed: 117.084804 samples/sec	 accuracy=68.025768	 loss=1.342026	 lr=0.001177
Epoch[184] Batch [2899]/[3760]	Speed: 117.338065 samples/sec	 accuracy=68.030711	 loss=1.341684	 lr=0.001174
Epoch[184] Batch [2949]/[3760]	Speed: 117.608277 samples/sec	 accuracy=68.057733	 loss=1.340851	 lr=0.001171
Epoch[184] Batch [2999]/[3760]	Speed: 117.701263 samples/sec	 accuracy=68.084896	 loss=1.339921	 lr=0.001168
Epoch[184] Batch [3049]/[3760]	Speed: 118.192328 samples/sec	 accuracy=68.079918	 loss=1.339300	 lr=0.001166
Epoch[184] Batch [3099]/[3760]	Speed: 117.727509 samples/sec	 accuracy=68.087198	 loss=1.339237	 lr=0.001163
Epoch[184] Batch [3149]/[3760]	Speed: 116.914310 samples/sec	 accuracy=68.074901	 loss=1.339679	 lr=0.001160
Epoch[184] Batch [3199]/[3760]	Speed: 117.749923 samples/sec	 accuracy=68.071777	 loss=1.340073	 lr=0.001157
Epoch[184] Batch [3249]/[3760]	Speed: 117.243228 samples/sec	 accuracy=68.073077	 loss=1.340262	 lr=0.001155
Epoch[184] Batch [3299]/[3760]	Speed: 117.869196 samples/sec	 accuracy=68.086174	 loss=1.339917	 lr=0.001152
Epoch[184] Batch [3349]/[3760]	Speed: 117.446179 samples/sec	 accuracy=68.090019	 loss=1.339855	 lr=0.001149
Epoch[184] Batch [3399]/[3760]	Speed: 117.540328 samples/sec	 accuracy=68.104320	 loss=1.339235	 lr=0.001146
Epoch[184] Batch [3449]/[3760]	Speed: 117.913942 samples/sec	 accuracy=68.081069	 loss=1.340135	 lr=0.001144
Epoch[184] Batch [3499]/[3760]	Speed: 116.881949 samples/sec	 accuracy=68.071429	 loss=1.340780	 lr=0.001141
Epoch[184] Batch [3549]/[3760]	Speed: 118.045706 samples/sec	 accuracy=68.069982	 loss=1.340608	 lr=0.001138
Epoch[184] Batch [3599]/[3760]	Speed: 117.738118 samples/sec	 accuracy=68.072917	 loss=1.340802	 lr=0.001135
Epoch[184] Batch [3649]/[3760]	Speed: 117.929842 samples/sec	 accuracy=68.085616	 loss=1.340304	 lr=0.001133
Epoch[184] Batch [3699]/[3760]	Speed: 117.128491 samples/sec	 accuracy=68.087838	 loss=1.340248	 lr=0.001130
Epoch[184] Batch [3749]/[3760]	Speed: 119.600544 samples/sec	 accuracy=68.097500	 loss=1.340297	 lr=0.001127
Batch [0049]/[0303]: acc-top1=60.562500 acc-top5=80.406250
Batch [0099]/[0303]: acc-top1=57.562500 acc-top5=79.671875
Batch [0149]/[0303]: acc-top1=59.197917 acc-top5=80.427083
Batch [0199]/[0303]: acc-top1=61.718750 acc-top5=82.117188
Batch [0249]/[0303]: acc-top1=61.362500 acc-top5=81.912500
Batch [0299]/[0303]: acc-top1=60.927083 acc-top5=81.947917
[Epoch 184] training: accuracy=68.100898	 loss=1.340196
[Epoch 184] speed: 117 samples/sec	time cost: 2172.770996
[Epoch 184] validation: acc-top1=60.854992 acc-top5=81.956477 loss=1.833053
Epoch[185] Batch [0049]/[3760]	Speed: 86.547243 samples/sec	 accuracy=69.218750	 loss=1.298284	 lr=0.001124
Epoch[185] Batch [0099]/[3760]	Speed: 117.789464 samples/sec	 accuracy=69.093750	 loss=1.301981	 lr=0.001121
Epoch[185] Batch [0149]/[3760]	Speed: 115.723882 samples/sec	 accuracy=69.177083	 loss=1.309717	 lr=0.001119
Epoch[185] Batch [0199]/[3760]	Speed: 116.790799 samples/sec	 accuracy=69.375000	 loss=1.297657	 lr=0.001116
Epoch[185] Batch [0249]/[3760]	Speed: 118.268235 samples/sec	 accuracy=69.218750	 loss=1.303910	 lr=0.001113
Epoch[185] Batch [0299]/[3760]	Speed: 117.374003 samples/sec	 accuracy=69.291667	 loss=1.303078	 lr=0.001110
Epoch[185] Batch [0349]/[3760]	Speed: 117.660241 samples/sec	 accuracy=69.049107	 loss=1.309597	 lr=0.001108
Epoch[185] Batch [0399]/[3760]	Speed: 117.390794 samples/sec	 accuracy=69.089844	 loss=1.311343	 lr=0.001105
Epoch[185] Batch [0449]/[3760]	Speed: 117.692102 samples/sec	 accuracy=69.010417	 loss=1.311840	 lr=0.001102
Epoch[185] Batch [0499]/[3760]	Speed: 117.679732 samples/sec	 accuracy=69.015625	 loss=1.311507	 lr=0.001100
Epoch[185] Batch [0549]/[3760]	Speed: 117.011701 samples/sec	 accuracy=68.897727	 loss=1.316417	 lr=0.001097
Epoch[185] Batch [0599]/[3760]	Speed: 117.447559 samples/sec	 accuracy=68.955729	 loss=1.314854	 lr=0.001094
Epoch[185] Batch [0649]/[3760]	Speed: 116.510257 samples/sec	 accuracy=68.884615	 loss=1.316617	 lr=0.001092
Epoch[185] Batch [0699]/[3760]	Speed: 117.860979 samples/sec	 accuracy=68.930804	 loss=1.318194	 lr=0.001089
Epoch[185] Batch [0749]/[3760]	Speed: 117.401103 samples/sec	 accuracy=68.964583	 loss=1.317490	 lr=0.001086
Epoch[185] Batch [0799]/[3760]	Speed: 116.552541 samples/sec	 accuracy=69.080078	 loss=1.314668	 lr=0.001084
Epoch[185] Batch [0849]/[3760]	Speed: 117.667567 samples/sec	 accuracy=69.014706	 loss=1.315261	 lr=0.001081
Epoch[185] Batch [0899]/[3760]	Speed: 117.374643 samples/sec	 accuracy=69.005208	 loss=1.315161	 lr=0.001078
Epoch[185] Batch [0949]/[3760]	Speed: 117.707191 samples/sec	 accuracy=68.991776	 loss=1.315614	 lr=0.001076
Epoch[185] Batch [0999]/[3760]	Speed: 116.689407 samples/sec	 accuracy=68.967188	 loss=1.315630	 lr=0.001073
Epoch[185] Batch [1049]/[3760]	Speed: 117.295687 samples/sec	 accuracy=68.943452	 loss=1.316859	 lr=0.001070
Epoch[185] Batch [1099]/[3760]	Speed: 117.362087 samples/sec	 accuracy=68.921875	 loss=1.317173	 lr=0.001068
Epoch[185] Batch [1149]/[3760]	Speed: 117.042544 samples/sec	 accuracy=68.922554	 loss=1.317752	 lr=0.001065
Epoch[185] Batch [1199]/[3760]	Speed: 117.760553 samples/sec	 accuracy=68.916667	 loss=1.317971	 lr=0.001062
Epoch[185] Batch [1249]/[3760]	Speed: 117.820696 samples/sec	 accuracy=68.887500	 loss=1.317648	 lr=0.001060
Epoch[185] Batch [1299]/[3760]	Speed: 117.586993 samples/sec	 accuracy=68.856971	 loss=1.319065	 lr=0.001057
Epoch[185] Batch [1349]/[3760]	Speed: 117.540270 samples/sec	 accuracy=68.847222	 loss=1.318638	 lr=0.001054
Epoch[185] Batch [1399]/[3760]	Speed: 117.130114 samples/sec	 accuracy=68.833705	 loss=1.318520	 lr=0.001052
Epoch[185] Batch [1449]/[3760]	Speed: 117.811911 samples/sec	 accuracy=68.857759	 loss=1.316547	 lr=0.001049
Epoch[185] Batch [1499]/[3760]	Speed: 117.672018 samples/sec	 accuracy=68.861458	 loss=1.316246	 lr=0.001046
Epoch[185] Batch [1549]/[3760]	Speed: 117.442564 samples/sec	 accuracy=68.843750	 loss=1.316873	 lr=0.001044
Epoch[185] Batch [1599]/[3760]	Speed: 117.505478 samples/sec	 accuracy=68.874023	 loss=1.315842	 lr=0.001041
Epoch[185] Batch [1649]/[3760]	Speed: 117.159990 samples/sec	 accuracy=68.862689	 loss=1.316780	 lr=0.001039
Epoch[185] Batch [1699]/[3760]	Speed: 118.009897 samples/sec	 accuracy=68.816176	 loss=1.318663	 lr=0.001036
Epoch[185] Batch [1749]/[3760]	Speed: 115.146911 samples/sec	 accuracy=68.852679	 loss=1.317301	 lr=0.001033
Epoch[185] Batch [1799]/[3760]	Speed: 116.692316 samples/sec	 accuracy=68.894965	 loss=1.315956	 lr=0.001031
Epoch[185] Batch [1849]/[3760]	Speed: 117.633405 samples/sec	 accuracy=68.872466	 loss=1.316984	 lr=0.001028
Epoch[185] Batch [1899]/[3760]	Speed: 116.231102 samples/sec	 accuracy=68.865132	 loss=1.316533	 lr=0.001026
Epoch[185] Batch [1949]/[3760]	Speed: 117.816695 samples/sec	 accuracy=68.855769	 loss=1.316797	 lr=0.001023
Epoch[185] Batch [1999]/[3760]	Speed: 117.054121 samples/sec	 accuracy=68.874219	 loss=1.315934	 lr=0.001020
Epoch[185] Batch [2049]/[3760]	Speed: 117.648055 samples/sec	 accuracy=68.881098	 loss=1.315785	 lr=0.001018
Epoch[185] Batch [2099]/[3760]	Speed: 117.498150 samples/sec	 accuracy=68.874256	 loss=1.314903	 lr=0.001015
Epoch[185] Batch [2149]/[3760]	Speed: 116.985547 samples/sec	 accuracy=68.900436	 loss=1.314538	 lr=0.001013
Epoch[185] Batch [2199]/[3760]	Speed: 117.858190 samples/sec	 accuracy=68.904830	 loss=1.313875	 lr=0.001010
Epoch[185] Batch [2249]/[3760]	Speed: 117.595246 samples/sec	 accuracy=68.923611	 loss=1.312995	 lr=0.001007
Epoch[185] Batch [2299]/[3760]	Speed: 117.351534 samples/sec	 accuracy=68.925951	 loss=1.312059	 lr=0.001005
Epoch[185] Batch [2349]/[3760]	Speed: 117.774433 samples/sec	 accuracy=68.919548	 loss=1.312880	 lr=0.001002
Epoch[185] Batch [2399]/[3760]	Speed: 117.347978 samples/sec	 accuracy=68.916016	 loss=1.312973	 lr=0.001000
Epoch[185] Batch [2449]/[3760]	Speed: 117.198547 samples/sec	 accuracy=68.894770	 loss=1.312956	 lr=0.000997
Epoch[185] Batch [2499]/[3760]	Speed: 118.008652 samples/sec	 accuracy=68.903750	 loss=1.312186	 lr=0.000995
Epoch[185] Batch [2549]/[3760]	Speed: 116.840468 samples/sec	 accuracy=68.886642	 loss=1.312207	 lr=0.000992
Epoch[185] Batch [2599]/[3760]	Speed: 118.068815 samples/sec	 accuracy=68.897837	 loss=1.311439	 lr=0.000989
Epoch[185] Batch [2649]/[3760]	Speed: 117.170765 samples/sec	 accuracy=68.896816	 loss=1.311954	 lr=0.000987
Epoch[185] Batch [2699]/[3760]	Speed: 117.948644 samples/sec	 accuracy=68.893519	 loss=1.312416	 lr=0.000984
Epoch[185] Batch [2749]/[3760]	Speed: 117.122525 samples/sec	 accuracy=68.873864	 loss=1.312986	 lr=0.000982
Epoch[185] Batch [2799]/[3760]	Speed: 117.454625 samples/sec	 accuracy=68.880580	 loss=1.312449	 lr=0.000979
Epoch[185] Batch [2849]/[3760]	Speed: 117.331958 samples/sec	 accuracy=68.854167	 loss=1.313319	 lr=0.000977
Epoch[185] Batch [2899]/[3760]	Speed: 118.103865 samples/sec	 accuracy=68.848599	 loss=1.313271	 lr=0.000974
Epoch[185] Batch [2949]/[3760]	Speed: 117.755563 samples/sec	 accuracy=68.844809	 loss=1.313687	 lr=0.000972
Epoch[185] Batch [2999]/[3760]	Speed: 117.059064 samples/sec	 accuracy=68.840625	 loss=1.313522	 lr=0.000969
Epoch[185] Batch [3049]/[3760]	Speed: 117.280790 samples/sec	 accuracy=68.840676	 loss=1.313987	 lr=0.000967
Epoch[185] Batch [3099]/[3760]	Speed: 117.776563 samples/sec	 accuracy=68.824597	 loss=1.313883	 lr=0.000964
Epoch[185] Batch [3149]/[3760]	Speed: 117.283353 samples/sec	 accuracy=68.835813	 loss=1.313212	 lr=0.000962
Epoch[185] Batch [3199]/[3760]	Speed: 117.478213 samples/sec	 accuracy=68.825684	 loss=1.313336	 lr=0.000959
Epoch[185] Batch [3249]/[3760]	Speed: 117.261508 samples/sec	 accuracy=68.815385	 loss=1.313462	 lr=0.000957
Epoch[185] Batch [3299]/[3760]	Speed: 117.198339 samples/sec	 accuracy=68.821023	 loss=1.312893	 lr=0.000954
Epoch[185] Batch [3349]/[3760]	Speed: 114.740561 samples/sec	 accuracy=68.835821	 loss=1.312547	 lr=0.000952
Epoch[185] Batch [3399]/[3760]	Speed: 117.120368 samples/sec	 accuracy=68.824449	 loss=1.312702	 lr=0.000949
Epoch[185] Batch [3449]/[3760]	Speed: 117.041931 samples/sec	 accuracy=68.829257	 loss=1.311994	 lr=0.000947
Epoch[185] Batch [3499]/[3760]	Speed: 117.926133 samples/sec	 accuracy=68.823214	 loss=1.312275	 lr=0.000944
Epoch[185] Batch [3549]/[3760]	Speed: 117.760529 samples/sec	 accuracy=68.815581	 loss=1.312571	 lr=0.000942
Epoch[185] Batch [3599]/[3760]	Speed: 117.205530 samples/sec	 accuracy=68.812934	 loss=1.312843	 lr=0.000939
Epoch[185] Batch [3649]/[3760]	Speed: 117.076269 samples/sec	 accuracy=68.813784	 loss=1.312543	 lr=0.000937
Epoch[185] Batch [3699]/[3760]	Speed: 117.254838 samples/sec	 accuracy=68.810389	 loss=1.312330	 lr=0.000934
Epoch[185] Batch [3749]/[3760]	Speed: 120.815964 samples/sec	 accuracy=68.821250	 loss=1.312120	 lr=0.000932
Batch [0049]/[0303]: acc-top1=60.593750 acc-top5=80.125000
Batch [0099]/[0303]: acc-top1=57.937500 acc-top5=79.687500
Batch [0149]/[0303]: acc-top1=59.312500 acc-top5=80.541667
Batch [0199]/[0303]: acc-top1=61.875000 acc-top5=82.265625
Batch [0249]/[0303]: acc-top1=61.443750 acc-top5=82.050000
Batch [0299]/[0303]: acc-top1=60.963542 acc-top5=82.036458
[Epoch 185] training: accuracy=68.812749	 loss=1.312498
[Epoch 185] speed: 116 samples/sec	time cost: 2177.214112
[Epoch 185] validation: acc-top1=60.927186 acc-top5=82.054455 loss=1.814389
Epoch[186] Batch [0049]/[3759]	Speed: 85.510932 samples/sec	 accuracy=69.031250	 loss=1.271642	 lr=0.000929
Epoch[186] Batch [0099]/[3759]	Speed: 117.337719 samples/sec	 accuracy=68.687500	 loss=1.275012	 lr=0.000926
Epoch[186] Batch [0149]/[3759]	Speed: 117.716947 samples/sec	 accuracy=68.802083	 loss=1.288904	 lr=0.000924
Epoch[186] Batch [0199]/[3759]	Speed: 116.666461 samples/sec	 accuracy=69.054688	 loss=1.288190	 lr=0.000921
Epoch[186] Batch [0249]/[3759]	Speed: 117.553845 samples/sec	 accuracy=69.100000	 loss=1.282605	 lr=0.000919
Epoch[186] Batch [0299]/[3759]	Speed: 116.674770 samples/sec	 accuracy=69.135417	 loss=1.282655	 lr=0.000916
Epoch[186] Batch [0349]/[3759]	Speed: 117.607351 samples/sec	 accuracy=69.361607	 loss=1.274236	 lr=0.000914
Epoch[186] Batch [0399]/[3759]	Speed: 117.185027 samples/sec	 accuracy=69.394531	 loss=1.275403	 lr=0.000911
Epoch[186] Batch [0449]/[3759]	Speed: 117.260680 samples/sec	 accuracy=69.461806	 loss=1.273575	 lr=0.000909
Epoch[186] Batch [0499]/[3759]	Speed: 117.751657 samples/sec	 accuracy=69.406250	 loss=1.277236	 lr=0.000907
Epoch[186] Batch [0549]/[3759]	Speed: 117.724135 samples/sec	 accuracy=69.468750	 loss=1.277717	 lr=0.000904
Epoch[186] Batch [0599]/[3759]	Speed: 117.607276 samples/sec	 accuracy=69.427083	 loss=1.279837	 lr=0.000902
Epoch[186] Batch [0649]/[3759]	Speed: 117.657369 samples/sec	 accuracy=69.437500	 loss=1.278959	 lr=0.000899
Epoch[186] Batch [0699]/[3759]	Speed: 117.983863 samples/sec	 accuracy=69.363839	 loss=1.280416	 lr=0.000897
Epoch[186] Batch [0749]/[3759]	Speed: 117.215696 samples/sec	 accuracy=69.343750	 loss=1.279330	 lr=0.000894
Epoch[186] Batch [0799]/[3759]	Speed: 117.673237 samples/sec	 accuracy=69.273438	 loss=1.283134	 lr=0.000892
Epoch[186] Batch [0849]/[3759]	Speed: 117.059585 samples/sec	 accuracy=69.281250	 loss=1.284904	 lr=0.000889
Epoch[186] Batch [0899]/[3759]	Speed: 117.625850 samples/sec	 accuracy=69.244792	 loss=1.287724	 lr=0.000887
Epoch[186] Batch [0949]/[3759]	Speed: 117.642749 samples/sec	 accuracy=69.266447	 loss=1.287474	 lr=0.000885
Epoch[186] Batch [0999]/[3759]	Speed: 117.822802 samples/sec	 accuracy=69.204687	 loss=1.288002	 lr=0.000882
Epoch[186] Batch [1049]/[3759]	Speed: 117.463300 samples/sec	 accuracy=69.282738	 loss=1.285909	 lr=0.000880
Epoch[186] Batch [1099]/[3759]	Speed: 117.449489 samples/sec	 accuracy=69.325284	 loss=1.285709	 lr=0.000877
Epoch[186] Batch [1149]/[3759]	Speed: 117.512373 samples/sec	 accuracy=69.350543	 loss=1.285385	 lr=0.000875
Epoch[186] Batch [1199]/[3759]	Speed: 115.004022 samples/sec	 accuracy=69.385417	 loss=1.284706	 lr=0.000873
Epoch[186] Batch [1249]/[3759]	Speed: 117.651050 samples/sec	 accuracy=69.411250	 loss=1.283252	 lr=0.000870
Epoch[186] Batch [1299]/[3759]	Speed: 117.126919 samples/sec	 accuracy=69.330529	 loss=1.284891	 lr=0.000868
Epoch[186] Batch [1349]/[3759]	Speed: 117.193861 samples/sec	 accuracy=69.326389	 loss=1.284482	 lr=0.000865
Epoch[186] Batch [1399]/[3759]	Speed: 117.160088 samples/sec	 accuracy=69.314732	 loss=1.284236	 lr=0.000863
Epoch[186] Batch [1449]/[3759]	Speed: 116.607779 samples/sec	 accuracy=69.233836	 loss=1.287567	 lr=0.000861
Epoch[186] Batch [1499]/[3759]	Speed: 117.450572 samples/sec	 accuracy=69.291667	 loss=1.284319	 lr=0.000858
Epoch[186] Batch [1549]/[3759]	Speed: 117.847812 samples/sec	 accuracy=69.287298	 loss=1.284817	 lr=0.000856
Epoch[186] Batch [1599]/[3759]	Speed: 117.742577 samples/sec	 accuracy=69.281250	 loss=1.284969	 lr=0.000854
Epoch[186] Batch [1649]/[3759]	Speed: 117.066580 samples/sec	 accuracy=69.286932	 loss=1.284925	 lr=0.000851
Epoch[186] Batch [1699]/[3759]	Speed: 117.583565 samples/sec	 accuracy=69.305147	 loss=1.284885	 lr=0.000849
Epoch[186] Batch [1749]/[3759]	Speed: 117.431629 samples/sec	 accuracy=69.320536	 loss=1.284973	 lr=0.000846
Epoch[186] Batch [1799]/[3759]	Speed: 117.691192 samples/sec	 accuracy=69.313368	 loss=1.286694	 lr=0.000844
Epoch[186] Batch [1849]/[3759]	Speed: 118.057108 samples/sec	 accuracy=69.326858	 loss=1.286423	 lr=0.000842
Epoch[186] Batch [1899]/[3759]	Speed: 118.220333 samples/sec	 accuracy=69.350329	 loss=1.285472	 lr=0.000839
Epoch[186] Batch [1949]/[3759]	Speed: 117.582611 samples/sec	 accuracy=69.322115	 loss=1.286512	 lr=0.000837
Epoch[186] Batch [1999]/[3759]	Speed: 117.281378 samples/sec	 accuracy=69.340625	 loss=1.286299	 lr=0.000835
Epoch[186] Batch [2049]/[3759]	Speed: 118.186390 samples/sec	 accuracy=69.349085	 loss=1.285920	 lr=0.000832
Epoch[186] Batch [2099]/[3759]	Speed: 117.228695 samples/sec	 accuracy=69.383185	 loss=1.284492	 lr=0.000830
Epoch[186] Batch [2149]/[3759]	Speed: 117.153565 samples/sec	 accuracy=69.372093	 loss=1.285203	 lr=0.000828
Epoch[186] Batch [2199]/[3759]	Speed: 116.927051 samples/sec	 accuracy=69.374290	 loss=1.285289	 lr=0.000825
Epoch[186] Batch [2249]/[3759]	Speed: 117.481273 samples/sec	 accuracy=69.398611	 loss=1.284243	 lr=0.000823
Epoch[186] Batch [2299]/[3759]	Speed: 116.646844 samples/sec	 accuracy=69.379076	 loss=1.284791	 lr=0.000821
Epoch[186] Batch [2349]/[3759]	Speed: 117.379930 samples/sec	 accuracy=69.388963	 loss=1.284496	 lr=0.000818
Epoch[186] Batch [2399]/[3759]	Speed: 117.563007 samples/sec	 accuracy=69.358073	 loss=1.285137	 lr=0.000816
Epoch[186] Batch [2449]/[3759]	Speed: 117.394455 samples/sec	 accuracy=69.355867	 loss=1.285952	 lr=0.000814
Epoch[186] Batch [2499]/[3759]	Speed: 118.340023 samples/sec	 accuracy=69.323750	 loss=1.287133	 lr=0.000811
Epoch[186] Batch [2549]/[3759]	Speed: 117.526107 samples/sec	 accuracy=69.332108	 loss=1.286208	 lr=0.000809
Epoch[186] Batch [2599]/[3759]	Speed: 117.694109 samples/sec	 accuracy=69.329327	 loss=1.285994	 lr=0.000807
Epoch[186] Batch [2649]/[3759]	Speed: 116.825383 samples/sec	 accuracy=69.340212	 loss=1.285600	 lr=0.000804
Epoch[186] Batch [2699]/[3759]	Speed: 116.786401 samples/sec	 accuracy=69.343171	 loss=1.285577	 lr=0.000802
Epoch[186] Batch [2749]/[3759]	Speed: 117.452156 samples/sec	 accuracy=69.363068	 loss=1.285156	 lr=0.000800
Epoch[186] Batch [2799]/[3759]	Speed: 114.099424 samples/sec	 accuracy=69.356027	 loss=1.285625	 lr=0.000798
Epoch[186] Batch [2849]/[3759]	Speed: 117.464627 samples/sec	 accuracy=69.363487	 loss=1.284988	 lr=0.000795
Epoch[186] Batch [2899]/[3759]	Speed: 117.794104 samples/sec	 accuracy=69.373922	 loss=1.284486	 lr=0.000793
Epoch[186] Batch [2949]/[3759]	Speed: 116.851149 samples/sec	 accuracy=69.368644	 loss=1.284535	 lr=0.000791
Epoch[186] Batch [2999]/[3759]	Speed: 117.038288 samples/sec	 accuracy=69.358854	 loss=1.284911	 lr=0.000788
Epoch[186] Batch [3049]/[3759]	Speed: 118.064725 samples/sec	 accuracy=69.359631	 loss=1.285304	 lr=0.000786
Epoch[186] Batch [3099]/[3759]	Speed: 117.184158 samples/sec	 accuracy=69.365927	 loss=1.285554	 lr=0.000784
Epoch[186] Batch [3149]/[3759]	Speed: 117.072640 samples/sec	 accuracy=69.386409	 loss=1.285267	 lr=0.000782
Epoch[186] Batch [3199]/[3759]	Speed: 117.853410 samples/sec	 accuracy=69.399414	 loss=1.284892	 lr=0.000779
Epoch[186] Batch [3249]/[3759]	Speed: 117.091394 samples/sec	 accuracy=69.380288	 loss=1.285453	 lr=0.000777
Epoch[186] Batch [3299]/[3759]	Speed: 116.965025 samples/sec	 accuracy=69.393466	 loss=1.284693	 lr=0.000775
Epoch[186] Batch [3349]/[3759]	Speed: 116.868018 samples/sec	 accuracy=69.396455	 loss=1.284045	 lr=0.000772
Epoch[186] Batch [3399]/[3759]	Speed: 116.536057 samples/sec	 accuracy=69.393842	 loss=1.283711	 lr=0.000770
Epoch[186] Batch [3449]/[3759]	Speed: 117.678993 samples/sec	 accuracy=69.398551	 loss=1.283656	 lr=0.000768
Epoch[186] Batch [3499]/[3759]	Speed: 117.456212 samples/sec	 accuracy=69.392857	 loss=1.284443	 lr=0.000766
Epoch[186] Batch [3549]/[3759]	Speed: 117.000641 samples/sec	 accuracy=69.403169	 loss=1.284014	 lr=0.000763
Epoch[186] Batch [3599]/[3759]	Speed: 117.332400 samples/sec	 accuracy=69.420139	 loss=1.283119	 lr=0.000761
Epoch[186] Batch [3649]/[3759]	Speed: 117.789175 samples/sec	 accuracy=69.420805	 loss=1.283018	 lr=0.000759
Epoch[186] Batch [3699]/[3759]	Speed: 116.959101 samples/sec	 accuracy=69.415963	 loss=1.283792	 lr=0.000757
Epoch[186] Batch [3749]/[3759]	Speed: 120.802598 samples/sec	 accuracy=69.415000	 loss=1.283796	 lr=0.000755
Batch [0049]/[0303]: acc-top1=60.937500 acc-top5=80.531250
Batch [0099]/[0303]: acc-top1=58.750000 acc-top5=80.281250
Batch [0149]/[0303]: acc-top1=60.010417 acc-top5=80.833333
Batch [0199]/[0303]: acc-top1=62.406250 acc-top5=82.382812
Batch [0249]/[0303]: acc-top1=61.906250 acc-top5=82.062500
Batch [0299]/[0303]: acc-top1=61.359375 acc-top5=82.109375
[Epoch 186] training: accuracy=69.412992	 loss=1.283830
[Epoch 186] speed: 116 samples/sec	time cost: 2177.014346
[Epoch 186] validation: acc-top1=61.267533 acc-top5=82.131807 loss=1.791538
Epoch[187] Batch [0049]/[3760]	Speed: 87.263752 samples/sec	 accuracy=69.187500	 loss=1.303880	 lr=0.000752
Epoch[187] Batch [0099]/[3760]	Speed: 117.791793 samples/sec	 accuracy=69.343750	 loss=1.285741	 lr=0.000750
Epoch[187] Batch [0149]/[3760]	Speed: 117.630146 samples/sec	 accuracy=69.260417	 loss=1.286782	 lr=0.000747
Epoch[187] Batch [0199]/[3760]	Speed: 117.308209 samples/sec	 accuracy=69.476562	 loss=1.283167	 lr=0.000745
Epoch[187] Batch [0249]/[3760]	Speed: 117.136811 samples/sec	 accuracy=69.631250	 loss=1.275008	 lr=0.000743
Epoch[187] Batch [0299]/[3760]	Speed: 116.844419 samples/sec	 accuracy=69.838542	 loss=1.274271	 lr=0.000741
Epoch[187] Batch [0349]/[3760]	Speed: 117.122467 samples/sec	 accuracy=70.035714	 loss=1.267712	 lr=0.000739
Epoch[187] Batch [0399]/[3760]	Speed: 117.009465 samples/sec	 accuracy=70.035156	 loss=1.265684	 lr=0.000736
Epoch[187] Batch [0449]/[3760]	Speed: 117.657816 samples/sec	 accuracy=69.958333	 loss=1.266047	 lr=0.000734
Epoch[187] Batch [0499]/[3760]	Speed: 117.305708 samples/sec	 accuracy=69.818750	 loss=1.272000	 lr=0.000732
Epoch[187] Batch [0549]/[3760]	Speed: 117.024046 samples/sec	 accuracy=69.920455	 loss=1.267999	 lr=0.000730
Epoch[187] Batch [0599]/[3760]	Speed: 117.153698 samples/sec	 accuracy=69.877604	 loss=1.269789	 lr=0.000728
Epoch[187] Batch [0649]/[3760]	Speed: 114.990290 samples/sec	 accuracy=69.737981	 loss=1.272757	 lr=0.000725
Epoch[187] Batch [0699]/[3760]	Speed: 117.685148 samples/sec	 accuracy=69.696429	 loss=1.274679	 lr=0.000723
Epoch[187] Batch [0749]/[3760]	Speed: 117.825080 samples/sec	 accuracy=69.625000	 loss=1.274107	 lr=0.000721
Epoch[187] Batch [0799]/[3760]	Speed: 118.132774 samples/sec	 accuracy=69.501953	 loss=1.275979	 lr=0.000719
Epoch[187] Batch [0849]/[3760]	Speed: 117.383676 samples/sec	 accuracy=69.529412	 loss=1.275688	 lr=0.000717
Epoch[187] Batch [0899]/[3760]	Speed: 117.325884 samples/sec	 accuracy=69.545139	 loss=1.273972	 lr=0.000714
Epoch[187] Batch [0949]/[3760]	Speed: 117.212394 samples/sec	 accuracy=69.603618	 loss=1.271792	 lr=0.000712
Epoch[187] Batch [0999]/[3760]	Speed: 118.224476 samples/sec	 accuracy=69.582813	 loss=1.271948	 lr=0.000710
Epoch[187] Batch [1049]/[3760]	Speed: 117.837055 samples/sec	 accuracy=69.480655	 loss=1.275454	 lr=0.000708
Epoch[187] Batch [1099]/[3760]	Speed: 117.452576 samples/sec	 accuracy=69.536932	 loss=1.272719	 lr=0.000706
Epoch[187] Batch [1149]/[3760]	Speed: 118.280103 samples/sec	 accuracy=69.528533	 loss=1.273615	 lr=0.000704
Epoch[187] Batch [1199]/[3760]	Speed: 116.928375 samples/sec	 accuracy=69.507812	 loss=1.274866	 lr=0.000702
Epoch[187] Batch [1249]/[3760]	Speed: 117.577164 samples/sec	 accuracy=69.592500	 loss=1.271696	 lr=0.000699
Epoch[187] Batch [1299]/[3760]	Speed: 117.317452 samples/sec	 accuracy=69.602163	 loss=1.272548	 lr=0.000697
Epoch[187] Batch [1349]/[3760]	Speed: 117.111581 samples/sec	 accuracy=69.594907	 loss=1.273254	 lr=0.000695
Epoch[187] Batch [1399]/[3760]	Speed: 117.300813 samples/sec	 accuracy=69.588170	 loss=1.273130	 lr=0.000693
Epoch[187] Batch [1449]/[3760]	Speed: 117.063780 samples/sec	 accuracy=69.599138	 loss=1.272674	 lr=0.000691
Epoch[187] Batch [1499]/[3760]	Speed: 116.833457 samples/sec	 accuracy=69.627083	 loss=1.271343	 lr=0.000689
Epoch[187] Batch [1549]/[3760]	Speed: 117.505600 samples/sec	 accuracy=69.661290	 loss=1.269927	 lr=0.000687
Epoch[187] Batch [1599]/[3760]	Speed: 116.989077 samples/sec	 accuracy=69.666016	 loss=1.269821	 lr=0.000684
Epoch[187] Batch [1649]/[3760]	Speed: 117.667299 samples/sec	 accuracy=69.682765	 loss=1.269146	 lr=0.000682
Epoch[187] Batch [1699]/[3760]	Speed: 118.587290 samples/sec	 accuracy=69.672794	 loss=1.269704	 lr=0.000680
Epoch[187] Batch [1749]/[3760]	Speed: 117.613278 samples/sec	 accuracy=69.685714	 loss=1.268602	 lr=0.000678
Epoch[187] Batch [1799]/[3760]	Speed: 117.300856 samples/sec	 accuracy=69.700521	 loss=1.268283	 lr=0.000676
Epoch[187] Batch [1849]/[3760]	Speed: 117.180479 samples/sec	 accuracy=69.690034	 loss=1.267488	 lr=0.000674
Epoch[187] Batch [1899]/[3760]	Speed: 117.828120 samples/sec	 accuracy=69.744243	 loss=1.265426	 lr=0.000672
Epoch[187] Batch [1949]/[3760]	Speed: 117.585829 samples/sec	 accuracy=69.776442	 loss=1.263819	 lr=0.000670
Epoch[187] Batch [1999]/[3760]	Speed: 118.047797 samples/sec	 accuracy=69.787500	 loss=1.263458	 lr=0.000667
Epoch[187] Batch [2049]/[3760]	Speed: 117.873488 samples/sec	 accuracy=69.784299	 loss=1.263386	 lr=0.000665
Epoch[187] Batch [2099]/[3760]	Speed: 117.166180 samples/sec	 accuracy=69.761161	 loss=1.263563	 lr=0.000663
Epoch[187] Batch [2149]/[3760]	Speed: 117.746750 samples/sec	 accuracy=69.787791	 loss=1.263005	 lr=0.000661
Epoch[187] Batch [2199]/[3760]	Speed: 117.254878 samples/sec	 accuracy=69.796875	 loss=1.262903	 lr=0.000659
Epoch[187] Batch [2249]/[3760]	Speed: 114.880930 samples/sec	 accuracy=69.790972	 loss=1.263267	 lr=0.000657
Epoch[187] Batch [2299]/[3760]	Speed: 118.446656 samples/sec	 accuracy=69.787364	 loss=1.263273	 lr=0.000655
Epoch[187] Batch [2349]/[3760]	Speed: 118.050484 samples/sec	 accuracy=69.796543	 loss=1.263610	 lr=0.000653
Epoch[187] Batch [2399]/[3760]	Speed: 117.447157 samples/sec	 accuracy=69.805990	 loss=1.263296	 lr=0.000651
Epoch[187] Batch [2449]/[3760]	Speed: 117.489252 samples/sec	 accuracy=69.820791	 loss=1.263013	 lr=0.000649
Epoch[187] Batch [2499]/[3760]	Speed: 118.350517 samples/sec	 accuracy=69.847500	 loss=1.262027	 lr=0.000647
Epoch[187] Batch [2549]/[3760]	Speed: 117.673276 samples/sec	 accuracy=69.829044	 loss=1.262183	 lr=0.000645
Epoch[187] Batch [2599]/[3760]	Speed: 117.344446 samples/sec	 accuracy=69.823317	 loss=1.262576	 lr=0.000643
Epoch[187] Batch [2649]/[3760]	Speed: 117.106675 samples/sec	 accuracy=69.824292	 loss=1.262761	 lr=0.000640
Epoch[187] Batch [2699]/[3760]	Speed: 117.194153 samples/sec	 accuracy=69.814815	 loss=1.263220	 lr=0.000638
Epoch[187] Batch [2749]/[3760]	Speed: 116.946655 samples/sec	 accuracy=69.791477	 loss=1.263986	 lr=0.000636
Epoch[187] Batch [2799]/[3760]	Speed: 117.267354 samples/sec	 accuracy=69.812500	 loss=1.263535	 lr=0.000634
Epoch[187] Batch [2849]/[3760]	Speed: 117.596709 samples/sec	 accuracy=69.819627	 loss=1.263537	 lr=0.000632
Epoch[187] Batch [2899]/[3760]	Speed: 116.605485 samples/sec	 accuracy=69.818966	 loss=1.263434	 lr=0.000630
Epoch[187] Batch [2949]/[3760]	Speed: 117.830770 samples/sec	 accuracy=69.801377	 loss=1.264395	 lr=0.000628
Epoch[187] Batch [2999]/[3760]	Speed: 117.088811 samples/sec	 accuracy=69.815104	 loss=1.263979	 lr=0.000626
Epoch[187] Batch [3049]/[3760]	Speed: 117.568071 samples/sec	 accuracy=69.797131	 loss=1.264523	 lr=0.000624
Epoch[187] Batch [3099]/[3760]	Speed: 116.787575 samples/sec	 accuracy=69.813004	 loss=1.264093	 lr=0.000622
Epoch[187] Batch [3149]/[3760]	Speed: 117.575006 samples/sec	 accuracy=69.827381	 loss=1.263170	 lr=0.000620
Epoch[187] Batch [3199]/[3760]	Speed: 117.517209 samples/sec	 accuracy=69.840332	 loss=1.263034	 lr=0.000618
Epoch[187] Batch [3249]/[3760]	Speed: 117.783253 samples/sec	 accuracy=69.841827	 loss=1.262883	 lr=0.000616
Epoch[187] Batch [3299]/[3760]	Speed: 118.098923 samples/sec	 accuracy=69.821970	 loss=1.263086	 lr=0.000614
Epoch[187] Batch [3349]/[3760]	Speed: 117.692669 samples/sec	 accuracy=69.847948	 loss=1.261919	 lr=0.000612
Epoch[187] Batch [3399]/[3760]	Speed: 117.761116 samples/sec	 accuracy=69.842831	 loss=1.262815	 lr=0.000610
Epoch[187] Batch [3449]/[3760]	Speed: 117.823117 samples/sec	 accuracy=69.848732	 loss=1.262788	 lr=0.000608
Epoch[187] Batch [3499]/[3760]	Speed: 117.665312 samples/sec	 accuracy=69.839286	 loss=1.263300	 lr=0.000606
Epoch[187] Batch [3549]/[3760]	Speed: 117.778963 samples/sec	 accuracy=69.841109	 loss=1.263150	 lr=0.000604
Epoch[187] Batch [3599]/[3760]	Speed: 117.746707 samples/sec	 accuracy=69.819878	 loss=1.264041	 lr=0.000602
Epoch[187] Batch [3649]/[3760]	Speed: 117.226254 samples/sec	 accuracy=69.842894	 loss=1.263734	 lr=0.000600
Epoch[187] Batch [3699]/[3760]	Speed: 117.711820 samples/sec	 accuracy=69.842061	 loss=1.263400	 lr=0.000598
Epoch[187] Batch [3749]/[3760]	Speed: 120.880624 samples/sec	 accuracy=69.848333	 loss=1.262913	 lr=0.000596
Batch [0049]/[0303]: acc-top1=61.531250 acc-top5=80.687500
Batch [0099]/[0303]: acc-top1=58.546875 acc-top5=80.406250
Batch [0149]/[0303]: acc-top1=59.802083 acc-top5=81.104167
Batch [0199]/[0303]: acc-top1=62.468750 acc-top5=82.710938
Batch [0249]/[0303]: acc-top1=61.937500 acc-top5=82.418750
Batch [0299]/[0303]: acc-top1=61.510417 acc-top5=82.421875
[Epoch 187] training: accuracy=69.850399	 loss=1.262938
[Epoch 187] speed: 116 samples/sec	time cost: 2174.815980
[Epoch 187] validation: acc-top1=61.437706 acc-top5=82.451526 loss=1.783212
Epoch[188] Batch [0049]/[3760]	Speed: 84.374080 samples/sec	 accuracy=70.562500	 loss=1.236904	 lr=0.000594
Epoch[188] Batch [0099]/[3760]	Speed: 117.773433 samples/sec	 accuracy=70.609375	 loss=1.230807	 lr=0.000592
Epoch[188] Batch [0149]/[3760]	Speed: 117.750387 samples/sec	 accuracy=70.604167	 loss=1.231449	 lr=0.000590
Epoch[188] Batch [0199]/[3760]	Speed: 117.880911 samples/sec	 accuracy=70.578125	 loss=1.231045	 lr=0.000588
Epoch[188] Batch [0249]/[3760]	Speed: 117.439794 samples/sec	 accuracy=70.537500	 loss=1.228125	 lr=0.000586
Epoch[188] Batch [0299]/[3760]	Speed: 117.080979 samples/sec	 accuracy=70.322917	 loss=1.235725	 lr=0.000584
Epoch[188] Batch [0349]/[3760]	Speed: 117.386638 samples/sec	 accuracy=70.540179	 loss=1.230716	 lr=0.000582
Epoch[188] Batch [0399]/[3760]	Speed: 116.962079 samples/sec	 accuracy=70.484375	 loss=1.235691	 lr=0.000580
Epoch[188] Batch [0449]/[3760]	Speed: 117.831531 samples/sec	 accuracy=70.496528	 loss=1.236246	 lr=0.000578
Epoch[188] Batch [0499]/[3760]	Speed: 117.402476 samples/sec	 accuracy=70.600000	 loss=1.233770	 lr=0.000576
Epoch[188] Batch [0549]/[3760]	Speed: 117.309421 samples/sec	 accuracy=70.497159	 loss=1.237700	 lr=0.000574
Epoch[188] Batch [0599]/[3760]	Speed: 117.376593 samples/sec	 accuracy=70.502604	 loss=1.234841	 lr=0.000572
Epoch[188] Batch [0649]/[3760]	Speed: 117.197932 samples/sec	 accuracy=70.435096	 loss=1.237197	 lr=0.000570
Epoch[188] Batch [0699]/[3760]	Speed: 117.720363 samples/sec	 accuracy=70.444196	 loss=1.238486	 lr=0.000568
Epoch[188] Batch [0749]/[3760]	Speed: 116.756495 samples/sec	 accuracy=70.508333	 loss=1.234985	 lr=0.000566
Epoch[188] Batch [0799]/[3760]	Speed: 117.851917 samples/sec	 accuracy=70.496094	 loss=1.235311	 lr=0.000564
Epoch[188] Batch [0849]/[3760]	Speed: 116.949867 samples/sec	 accuracy=70.538603	 loss=1.234087	 lr=0.000562
Epoch[188] Batch [0899]/[3760]	Speed: 117.903266 samples/sec	 accuracy=70.567708	 loss=1.232003	 lr=0.000560
Epoch[188] Batch [0949]/[3760]	Speed: 117.767084 samples/sec	 accuracy=70.498355	 loss=1.232700	 lr=0.000558
Epoch[188] Batch [0999]/[3760]	Speed: 116.728717 samples/sec	 accuracy=70.431250	 loss=1.234591	 lr=0.000557
Epoch[188] Batch [1049]/[3760]	Speed: 117.024864 samples/sec	 accuracy=70.361607	 loss=1.238383	 lr=0.000555
Epoch[188] Batch [1099]/[3760]	Speed: 117.208090 samples/sec	 accuracy=70.360795	 loss=1.239091	 lr=0.000553
Epoch[188] Batch [1149]/[3760]	Speed: 117.730761 samples/sec	 accuracy=70.370924	 loss=1.238758	 lr=0.000551
Epoch[188] Batch [1199]/[3760]	Speed: 117.044219 samples/sec	 accuracy=70.403646	 loss=1.237784	 lr=0.000549
Epoch[188] Batch [1249]/[3760]	Speed: 118.219031 samples/sec	 accuracy=70.393750	 loss=1.238053	 lr=0.000547
Epoch[188] Batch [1299]/[3760]	Speed: 117.318148 samples/sec	 accuracy=70.365385	 loss=1.239868	 lr=0.000545
Epoch[188] Batch [1349]/[3760]	Speed: 117.565066 samples/sec	 accuracy=70.365741	 loss=1.240412	 lr=0.000543
Epoch[188] Batch [1399]/[3760]	Speed: 117.470802 samples/sec	 accuracy=70.359375	 loss=1.241929	 lr=0.000541
Epoch[188] Batch [1449]/[3760]	Speed: 117.365480 samples/sec	 accuracy=70.379310	 loss=1.241512	 lr=0.000539
Epoch[188] Batch [1499]/[3760]	Speed: 117.183933 samples/sec	 accuracy=70.371875	 loss=1.241826	 lr=0.000538
Epoch[188] Batch [1549]/[3760]	Speed: 117.502730 samples/sec	 accuracy=70.383065	 loss=1.241075	 lr=0.000536
Epoch[188] Batch [1599]/[3760]	Speed: 117.625558 samples/sec	 accuracy=70.401367	 loss=1.240611	 lr=0.000534
Epoch[188] Batch [1649]/[3760]	Speed: 114.344033 samples/sec	 accuracy=70.451705	 loss=1.238801	 lr=0.000532
Epoch[188] Batch [1699]/[3760]	Speed: 118.359290 samples/sec	 accuracy=70.437500	 loss=1.238983	 lr=0.000530
Epoch[188] Batch [1749]/[3760]	Speed: 116.820292 samples/sec	 accuracy=70.438393	 loss=1.239211	 lr=0.000528
Epoch[188] Batch [1799]/[3760]	Speed: 116.905631 samples/sec	 accuracy=70.434028	 loss=1.240345	 lr=0.000526
Epoch[188] Batch [1849]/[3760]	Speed: 117.550791 samples/sec	 accuracy=70.386824	 loss=1.241765	 lr=0.000524
Epoch[188] Batch [1899]/[3760]	Speed: 117.107391 samples/sec	 accuracy=70.370066	 loss=1.242571	 lr=0.000523
Epoch[188] Batch [1949]/[3760]	Speed: 117.271045 samples/sec	 accuracy=70.380609	 loss=1.242193	 lr=0.000521
Epoch[188] Batch [1999]/[3760]	Speed: 117.094003 samples/sec	 accuracy=70.356250	 loss=1.243487	 lr=0.000519
Epoch[188] Batch [2049]/[3760]	Speed: 116.543019 samples/sec	 accuracy=70.395579	 loss=1.241743	 lr=0.000517
Epoch[188] Batch [2099]/[3760]	Speed: 116.574374 samples/sec	 accuracy=70.384673	 loss=1.242134	 lr=0.000515
Epoch[188] Batch [2149]/[3760]	Speed: 117.452341 samples/sec	 accuracy=70.384448	 loss=1.242074	 lr=0.000513
Epoch[188] Batch [2199]/[3760]	Speed: 117.312027 samples/sec	 accuracy=70.372869	 loss=1.242791	 lr=0.000511
Epoch[188] Batch [2249]/[3760]	Speed: 116.851158 samples/sec	 accuracy=70.397917	 loss=1.242570	 lr=0.000510
Epoch[188] Batch [2299]/[3760]	Speed: 116.856239 samples/sec	 accuracy=70.419837	 loss=1.241706	 lr=0.000508
Epoch[188] Batch [2349]/[3760]	Speed: 117.583075 samples/sec	 accuracy=70.451463	 loss=1.240326	 lr=0.000506
Epoch[188] Batch [2399]/[3760]	Speed: 117.116791 samples/sec	 accuracy=70.430339	 loss=1.241442	 lr=0.000504
Epoch[188] Batch [2449]/[3760]	Speed: 116.735594 samples/sec	 accuracy=70.427934	 loss=1.241850	 lr=0.000502
Epoch[188] Batch [2499]/[3760]	Speed: 117.338339 samples/sec	 accuracy=70.433125	 loss=1.241745	 lr=0.000500
Epoch[188] Batch [2549]/[3760]	Speed: 117.973536 samples/sec	 accuracy=70.426471	 loss=1.241873	 lr=0.000499
Epoch[188] Batch [2599]/[3760]	Speed: 118.573537 samples/sec	 accuracy=70.443510	 loss=1.241604	 lr=0.000497
Epoch[188] Batch [2649]/[3760]	Speed: 117.510881 samples/sec	 accuracy=70.449882	 loss=1.241036	 lr=0.000495
Epoch[188] Batch [2699]/[3760]	Speed: 117.393347 samples/sec	 accuracy=70.462963	 loss=1.240561	 lr=0.000493
Epoch[188] Batch [2749]/[3760]	Speed: 118.437289 samples/sec	 accuracy=70.448295	 loss=1.241211	 lr=0.000491
Epoch[188] Batch [2799]/[3760]	Speed: 116.941053 samples/sec	 accuracy=70.452009	 loss=1.241181	 lr=0.000490
Epoch[188] Batch [2849]/[3760]	Speed: 117.458062 samples/sec	 accuracy=70.450658	 loss=1.240860	 lr=0.000488
Epoch[188] Batch [2899]/[3760]	Speed: 117.325831 samples/sec	 accuracy=70.443427	 loss=1.241134	 lr=0.000486
Epoch[188] Batch [2949]/[3760]	Speed: 117.551072 samples/sec	 accuracy=70.463453	 loss=1.240508	 lr=0.000484
Epoch[188] Batch [2999]/[3760]	Speed: 117.119104 samples/sec	 accuracy=70.465104	 loss=1.240239	 lr=0.000482
Epoch[188] Batch [3049]/[3760]	Speed: 117.894047 samples/sec	 accuracy=70.481045	 loss=1.240580	 lr=0.000481
Epoch[188] Batch [3099]/[3760]	Speed: 117.195752 samples/sec	 accuracy=70.482359	 loss=1.240440	 lr=0.000479
Epoch[188] Batch [3149]/[3760]	Speed: 116.882093 samples/sec	 accuracy=70.485119	 loss=1.240733	 lr=0.000477
Epoch[188] Batch [3199]/[3760]	Speed: 117.150022 samples/sec	 accuracy=70.470215	 loss=1.241762	 lr=0.000475
Epoch[188] Batch [3249]/[3760]	Speed: 117.612143 samples/sec	 accuracy=70.471635	 loss=1.241735	 lr=0.000474
Epoch[188] Batch [3299]/[3760]	Speed: 114.891806 samples/sec	 accuracy=70.470170	 loss=1.242039	 lr=0.000472
Epoch[188] Batch [3349]/[3760]	Speed: 117.756189 samples/sec	 accuracy=70.479478	 loss=1.241521	 lr=0.000470
Epoch[188] Batch [3399]/[3760]	Speed: 117.157537 samples/sec	 accuracy=70.465533	 loss=1.241864	 lr=0.000468
Epoch[188] Batch [3449]/[3760]	Speed: 118.140989 samples/sec	 accuracy=70.468750	 loss=1.241531	 lr=0.000466
Epoch[188] Batch [3499]/[3760]	Speed: 117.996472 samples/sec	 accuracy=70.475000	 loss=1.241744	 lr=0.000465
Epoch[188] Batch [3549]/[3760]	Speed: 117.327011 samples/sec	 accuracy=70.491637	 loss=1.241125	 lr=0.000463
Epoch[188] Batch [3599]/[3760]	Speed: 117.607050 samples/sec	 accuracy=70.493490	 loss=1.241253	 lr=0.000461
Epoch[188] Batch [3649]/[3760]	Speed: 116.978218 samples/sec	 accuracy=70.506421	 loss=1.240900	 lr=0.000459
Epoch[188] Batch [3699]/[3760]	Speed: 117.227575 samples/sec	 accuracy=70.510135	 loss=1.241218	 lr=0.000458
Epoch[188] Batch [3749]/[3760]	Speed: 120.230964 samples/sec	 accuracy=70.505417	 loss=1.241792	 lr=0.000456
Batch [0049]/[0303]: acc-top1=62.093750 acc-top5=81.187500
Batch [0099]/[0303]: acc-top1=58.937500 acc-top5=80.718750
Batch [0149]/[0303]: acc-top1=60.083333 acc-top5=81.187500
Batch [0199]/[0303]: acc-top1=62.765625 acc-top5=82.828125
Batch [0249]/[0303]: acc-top1=62.162500 acc-top5=82.600000
Batch [0299]/[0303]: acc-top1=61.770833 acc-top5=82.526042
[Epoch 188] training: accuracy=70.504488	 loss=1.241889
[Epoch 188] speed: 116 samples/sec	time cost: 2177.899157
[Epoch 188] validation: acc-top1=61.705858 acc-top5=82.544348 loss=1.788911
Epoch[189] Batch [0049]/[3759]	Speed: 87.019374 samples/sec	 accuracy=70.812500	 loss=1.248467	 lr=0.000454
Epoch[189] Batch [0099]/[3759]	Speed: 117.104634 samples/sec	 accuracy=71.046875	 loss=1.240999	 lr=0.000452
Epoch[189] Batch [0149]/[3759]	Speed: 117.602089 samples/sec	 accuracy=70.458333	 loss=1.263322	 lr=0.000450
Epoch[189] Batch [0199]/[3759]	Speed: 117.037419 samples/sec	 accuracy=70.531250	 loss=1.256315	 lr=0.000449
Epoch[189] Batch [0249]/[3759]	Speed: 117.546448 samples/sec	 accuracy=70.956250	 loss=1.236655	 lr=0.000447
Epoch[189] Batch [0299]/[3759]	Speed: 117.768015 samples/sec	 accuracy=71.187500	 loss=1.228161	 lr=0.000445
Epoch[189] Batch [0349]/[3759]	Speed: 117.741738 samples/sec	 accuracy=71.133929	 loss=1.231576	 lr=0.000444
Epoch[189] Batch [0399]/[3759]	Speed: 117.379817 samples/sec	 accuracy=70.992188	 loss=1.234328	 lr=0.000442
Epoch[189] Batch [0449]/[3759]	Speed: 117.697228 samples/sec	 accuracy=71.013889	 loss=1.238542	 lr=0.000440
Epoch[189] Batch [0499]/[3759]	Speed: 117.100872 samples/sec	 accuracy=71.106250	 loss=1.233077	 lr=0.000438
Epoch[189] Batch [0549]/[3759]	Speed: 117.249623 samples/sec	 accuracy=71.184659	 loss=1.232064	 lr=0.000437
Epoch[189] Batch [0599]/[3759]	Speed: 117.327209 samples/sec	 accuracy=71.057292	 loss=1.233854	 lr=0.000435
Epoch[189] Batch [0649]/[3759]	Speed: 117.765080 samples/sec	 accuracy=71.141827	 loss=1.230109	 lr=0.000433
Epoch[189] Batch [0699]/[3759]	Speed: 116.900750 samples/sec	 accuracy=71.156250	 loss=1.224831	 lr=0.000432
Epoch[189] Batch [0749]/[3759]	Speed: 117.399023 samples/sec	 accuracy=71.185417	 loss=1.224884	 lr=0.000430
Epoch[189] Batch [0799]/[3759]	Speed: 117.742055 samples/sec	 accuracy=71.138672	 loss=1.225650	 lr=0.000428
Epoch[189] Batch [0849]/[3759]	Speed: 117.188382 samples/sec	 accuracy=71.044118	 loss=1.229134	 lr=0.000427
Epoch[189] Batch [0899]/[3759]	Speed: 118.095348 samples/sec	 accuracy=70.951389	 loss=1.232550	 lr=0.000425
Epoch[189] Batch [0949]/[3759]	Speed: 118.129263 samples/sec	 accuracy=70.978618	 loss=1.230792	 lr=0.000423
Epoch[189] Batch [0999]/[3759]	Speed: 116.554989 samples/sec	 accuracy=70.946875	 loss=1.231806	 lr=0.000422
Epoch[189] Batch [1049]/[3759]	Speed: 117.708966 samples/sec	 accuracy=70.986607	 loss=1.230722	 lr=0.000420
Epoch[189] Batch [1099]/[3759]	Speed: 118.149577 samples/sec	 accuracy=71.018466	 loss=1.227896	 lr=0.000418
Epoch[189] Batch [1149]/[3759]	Speed: 114.078035 samples/sec	 accuracy=71.043478	 loss=1.226124	 lr=0.000417
Epoch[189] Batch [1199]/[3759]	Speed: 118.040841 samples/sec	 accuracy=71.040365	 loss=1.226527	 lr=0.000415
Epoch[189] Batch [1249]/[3759]	Speed: 118.022246 samples/sec	 accuracy=71.025000	 loss=1.226237	 lr=0.000413
Epoch[189] Batch [1299]/[3759]	Speed: 117.317384 samples/sec	 accuracy=71.079327	 loss=1.223681	 lr=0.000412
Epoch[189] Batch [1349]/[3759]	Speed: 117.364709 samples/sec	 accuracy=71.093750	 loss=1.223784	 lr=0.000410
Epoch[189] Batch [1399]/[3759]	Speed: 117.838078 samples/sec	 accuracy=71.094866	 loss=1.222989	 lr=0.000408
Epoch[189] Batch [1449]/[3759]	Speed: 118.296304 samples/sec	 accuracy=71.099138	 loss=1.221554	 lr=0.000407
Epoch[189] Batch [1499]/[3759]	Speed: 117.968298 samples/sec	 accuracy=71.119792	 loss=1.220882	 lr=0.000405
Epoch[189] Batch [1549]/[3759]	Speed: 117.448380 samples/sec	 accuracy=71.110887	 loss=1.220613	 lr=0.000403
Epoch[189] Batch [1599]/[3759]	Speed: 117.481568 samples/sec	 accuracy=71.081055	 loss=1.220597	 lr=0.000402
Epoch[189] Batch [1649]/[3759]	Speed: 118.126472 samples/sec	 accuracy=71.095644	 loss=1.219605	 lr=0.000400
Epoch[189] Batch [1699]/[3759]	Speed: 117.546124 samples/sec	 accuracy=71.093750	 loss=1.219116	 lr=0.000398
Epoch[189] Batch [1749]/[3759]	Speed: 116.645722 samples/sec	 accuracy=71.115179	 loss=1.217317	 lr=0.000397
Epoch[189] Batch [1799]/[3759]	Speed: 117.863465 samples/sec	 accuracy=71.119792	 loss=1.217309	 lr=0.000395
Epoch[189] Batch [1849]/[3759]	Speed: 117.697070 samples/sec	 accuracy=71.140203	 loss=1.215670	 lr=0.000394
Epoch[189] Batch [1899]/[3759]	Speed: 117.624546 samples/sec	 accuracy=71.167763	 loss=1.214766	 lr=0.000392
Epoch[189] Batch [1949]/[3759]	Speed: 117.833458 samples/sec	 accuracy=71.164263	 loss=1.214803	 lr=0.000390
Epoch[189] Batch [1999]/[3759]	Speed: 117.372761 samples/sec	 accuracy=71.164844	 loss=1.214079	 lr=0.000389
Epoch[189] Batch [2049]/[3759]	Speed: 117.846986 samples/sec	 accuracy=71.169207	 loss=1.214299	 lr=0.000387
Epoch[189] Batch [2099]/[3759]	Speed: 117.896916 samples/sec	 accuracy=71.179315	 loss=1.213599	 lr=0.000386
Epoch[189] Batch [2149]/[3759]	Speed: 116.749013 samples/sec	 accuracy=71.138081	 loss=1.214636	 lr=0.000384
Epoch[189] Batch [2199]/[3759]	Speed: 116.786282 samples/sec	 accuracy=71.129261	 loss=1.213910	 lr=0.000382
Epoch[189] Batch [2249]/[3759]	Speed: 117.918413 samples/sec	 accuracy=71.081944	 loss=1.215680	 lr=0.000381
Epoch[189] Batch [2299]/[3759]	Speed: 117.909193 samples/sec	 accuracy=71.084918	 loss=1.215406	 lr=0.000379
Epoch[189] Batch [2349]/[3759]	Speed: 116.557072 samples/sec	 accuracy=71.115027	 loss=1.214208	 lr=0.000378
Epoch[189] Batch [2399]/[3759]	Speed: 118.057284 samples/sec	 accuracy=71.070312	 loss=1.215319	 lr=0.000376
Epoch[189] Batch [2449]/[3759]	Speed: 117.523897 samples/sec	 accuracy=71.086735	 loss=1.215102	 lr=0.000374
Epoch[189] Batch [2499]/[3759]	Speed: 118.063441 samples/sec	 accuracy=71.085625	 loss=1.214975	 lr=0.000373
Epoch[189] Batch [2549]/[3759]	Speed: 118.202627 samples/sec	 accuracy=71.087010	 loss=1.215151	 lr=0.000371
Epoch[189] Batch [2599]/[3759]	Speed: 117.685776 samples/sec	 accuracy=71.075120	 loss=1.215968	 lr=0.000370
Epoch[189] Batch [2649]/[3759]	Speed: 117.221389 samples/sec	 accuracy=71.058962	 loss=1.216299	 lr=0.000368
Epoch[189] Batch [2699]/[3759]	Speed: 118.169686 samples/sec	 accuracy=71.068866	 loss=1.216348	 lr=0.000367
Epoch[189] Batch [2749]/[3759]	Speed: 118.083685 samples/sec	 accuracy=71.057386	 loss=1.216812	 lr=0.000365
Epoch[189] Batch [2799]/[3759]	Speed: 114.755578 samples/sec	 accuracy=71.080915	 loss=1.216245	 lr=0.000364
Epoch[189] Batch [2849]/[3759]	Speed: 117.355151 samples/sec	 accuracy=71.084430	 loss=1.216371	 lr=0.000362
Epoch[189] Batch [2899]/[3759]	Speed: 116.945452 samples/sec	 accuracy=71.055496	 loss=1.217229	 lr=0.000360
Epoch[189] Batch [2949]/[3759]	Speed: 117.565909 samples/sec	 accuracy=71.052436	 loss=1.217399	 lr=0.000359
Epoch[189] Batch [2999]/[3759]	Speed: 117.212206 samples/sec	 accuracy=71.050521	 loss=1.217711	 lr=0.000357
Epoch[189] Batch [3049]/[3759]	Speed: 116.409878 samples/sec	 accuracy=71.045594	 loss=1.217772	 lr=0.000356
Epoch[189] Batch [3099]/[3759]	Speed: 117.121358 samples/sec	 accuracy=71.027722	 loss=1.218382	 lr=0.000354
Epoch[189] Batch [3149]/[3759]	Speed: 118.528199 samples/sec	 accuracy=70.999504	 loss=1.218775	 lr=0.000353
Epoch[189] Batch [3199]/[3759]	Speed: 117.380174 samples/sec	 accuracy=70.950195	 loss=1.220875	 lr=0.000351
Epoch[189] Batch [3249]/[3759]	Speed: 117.819104 samples/sec	 accuracy=70.964904	 loss=1.220349	 lr=0.000350
Epoch[189] Batch [3299]/[3759]	Speed: 117.547281 samples/sec	 accuracy=70.940341	 loss=1.221505	 lr=0.000348
Epoch[189] Batch [3349]/[3759]	Speed: 117.617682 samples/sec	 accuracy=70.940299	 loss=1.221617	 lr=0.000347
Epoch[189] Batch [3399]/[3759]	Speed: 117.392403 samples/sec	 accuracy=70.917739	 loss=1.222230	 lr=0.000345
Epoch[189] Batch [3449]/[3759]	Speed: 117.473380 samples/sec	 accuracy=70.903533	 loss=1.222458	 lr=0.000344
Epoch[189] Batch [3499]/[3759]	Speed: 117.616087 samples/sec	 accuracy=70.884821	 loss=1.222623	 lr=0.000342
Epoch[189] Batch [3549]/[3759]	Speed: 118.148297 samples/sec	 accuracy=70.862676	 loss=1.223332	 lr=0.000341
Epoch[189] Batch [3599]/[3759]	Speed: 117.410873 samples/sec	 accuracy=70.828993	 loss=1.224451	 lr=0.000339
Epoch[189] Batch [3649]/[3759]	Speed: 118.223463 samples/sec	 accuracy=70.821918	 loss=1.224939	 lr=0.000338
Epoch[189] Batch [3699]/[3759]	Speed: 117.480218 samples/sec	 accuracy=70.826858	 loss=1.224746	 lr=0.000336
Epoch[189] Batch [3749]/[3759]	Speed: 120.579516 samples/sec	 accuracy=70.809583	 loss=1.225355	 lr=0.000335
Batch [0049]/[0303]: acc-top1=61.625000 acc-top5=81.343750
Batch [0099]/[0303]: acc-top1=58.828125 acc-top5=80.468750
Batch [0149]/[0303]: acc-top1=60.197917 acc-top5=80.937500
Batch [0199]/[0303]: acc-top1=62.820312 acc-top5=82.656250
Batch [0249]/[0303]: acc-top1=62.262500 acc-top5=82.450000
Batch [0299]/[0303]: acc-top1=61.869792 acc-top5=82.458333
[Epoch 189] training: accuracy=70.807978	 loss=1.225358
[Epoch 189] speed: 116 samples/sec	time cost: 2173.204129
[Epoch 189] validation: acc-top1=61.798680 acc-top5=82.487624 loss=1.780793
Epoch[190] Batch [0049]/[3760]	Speed: 87.424764 samples/sec	 accuracy=71.125000	 loss=1.224312	 lr=0.000333
Epoch[190] Batch [0099]/[3760]	Speed: 117.864280 samples/sec	 accuracy=71.750000	 loss=1.202070	 lr=0.000331
Epoch[190] Batch [0149]/[3760]	Speed: 116.879339 samples/sec	 accuracy=71.666667	 loss=1.200833	 lr=0.000330
Epoch[190] Batch [0199]/[3760]	Speed: 117.552252 samples/sec	 accuracy=71.867188	 loss=1.184860	 lr=0.000328
Epoch[190] Batch [0249]/[3760]	Speed: 117.613171 samples/sec	 accuracy=72.025000	 loss=1.183047	 lr=0.000327
Epoch[190] Batch [0299]/[3760]	Speed: 117.513936 samples/sec	 accuracy=71.776042	 loss=1.186443	 lr=0.000325
Epoch[190] Batch [0349]/[3760]	Speed: 117.829513 samples/sec	 accuracy=71.736607	 loss=1.184105	 lr=0.000324
Epoch[190] Batch [0399]/[3760]	Speed: 118.122873 samples/sec	 accuracy=71.535156	 loss=1.189780	 lr=0.000323
Epoch[190] Batch [0449]/[3760]	Speed: 116.648941 samples/sec	 accuracy=71.427083	 loss=1.192576	 lr=0.000321
Epoch[190] Batch [0499]/[3760]	Speed: 116.981655 samples/sec	 accuracy=71.475000	 loss=1.191681	 lr=0.000320
Epoch[190] Batch [0549]/[3760]	Speed: 117.992320 samples/sec	 accuracy=71.457386	 loss=1.197135	 lr=0.000318
Epoch[190] Batch [0599]/[3760]	Speed: 117.297195 samples/sec	 accuracy=71.403646	 loss=1.198302	 lr=0.000317
Epoch[190] Batch [0649]/[3760]	Speed: 115.718535 samples/sec	 accuracy=71.365385	 loss=1.200492	 lr=0.000315
Epoch[190] Batch [0699]/[3760]	Speed: 117.401983 samples/sec	 accuracy=71.383929	 loss=1.199614	 lr=0.000314
Epoch[190] Batch [0749]/[3760]	Speed: 117.341751 samples/sec	 accuracy=71.402083	 loss=1.199514	 lr=0.000312
Epoch[190] Batch [0799]/[3760]	Speed: 117.948202 samples/sec	 accuracy=71.433594	 loss=1.200407	 lr=0.000311
Epoch[190] Batch [0849]/[3760]	Speed: 117.619244 samples/sec	 accuracy=71.437500	 loss=1.199301	 lr=0.000310
Epoch[190] Batch [0899]/[3760]	Speed: 117.424356 samples/sec	 accuracy=71.453125	 loss=1.197985	 lr=0.000308
Epoch[190] Batch [0949]/[3760]	Speed: 116.649068 samples/sec	 accuracy=71.429276	 loss=1.198521	 lr=0.000307
Epoch[190] Batch [0999]/[3760]	Speed: 117.785684 samples/sec	 accuracy=71.407813	 loss=1.198825	 lr=0.000305
Epoch[190] Batch [1049]/[3760]	Speed: 116.622090 samples/sec	 accuracy=71.367560	 loss=1.200329	 lr=0.000304
Epoch[190] Batch [1099]/[3760]	Speed: 116.987589 samples/sec	 accuracy=71.318182	 loss=1.202888	 lr=0.000302
Epoch[190] Batch [1149]/[3760]	Speed: 117.335810 samples/sec	 accuracy=71.316576	 loss=1.202845	 lr=0.000301
Epoch[190] Batch [1199]/[3760]	Speed: 117.789128 samples/sec	 accuracy=71.299479	 loss=1.204122	 lr=0.000300
Epoch[190] Batch [1249]/[3760]	Speed: 117.313869 samples/sec	 accuracy=71.313750	 loss=1.203990	 lr=0.000298
Epoch[190] Batch [1299]/[3760]	Speed: 117.830487 samples/sec	 accuracy=71.324519	 loss=1.203169	 lr=0.000297
Epoch[190] Batch [1349]/[3760]	Speed: 117.261290 samples/sec	 accuracy=71.285880	 loss=1.204066	 lr=0.000295
Epoch[190] Batch [1399]/[3760]	Speed: 117.310349 samples/sec	 accuracy=71.301339	 loss=1.204224	 lr=0.000294
Epoch[190] Batch [1449]/[3760]	Speed: 117.514264 samples/sec	 accuracy=71.306034	 loss=1.204084	 lr=0.000293
Epoch[190] Batch [1499]/[3760]	Speed: 117.706782 samples/sec	 accuracy=71.295833	 loss=1.204729	 lr=0.000291
Epoch[190] Batch [1549]/[3760]	Speed: 117.498411 samples/sec	 accuracy=71.265121	 loss=1.205359	 lr=0.000290
Epoch[190] Batch [1599]/[3760]	Speed: 118.041851 samples/sec	 accuracy=71.244141	 loss=1.205748	 lr=0.000288
Epoch[190] Batch [1649]/[3760]	Speed: 117.800927 samples/sec	 accuracy=71.224432	 loss=1.206814	 lr=0.000287
Epoch[190] Batch [1699]/[3760]	Speed: 118.146755 samples/sec	 accuracy=71.203125	 loss=1.208674	 lr=0.000286
Epoch[190] Batch [1749]/[3760]	Speed: 117.577180 samples/sec	 accuracy=71.196429	 loss=1.208320	 lr=0.000284
Epoch[190] Batch [1799]/[3760]	Speed: 117.689447 samples/sec	 accuracy=71.207465	 loss=1.208335	 lr=0.000283
Epoch[190] Batch [1849]/[3760]	Speed: 118.053286 samples/sec	 accuracy=71.214527	 loss=1.208234	 lr=0.000282
Epoch[190] Batch [1899]/[3760]	Speed: 117.210133 samples/sec	 accuracy=71.190789	 loss=1.209053	 lr=0.000280
Epoch[190] Batch [1949]/[3760]	Speed: 117.157344 samples/sec	 accuracy=71.189103	 loss=1.208665	 lr=0.000279
Epoch[190] Batch [1999]/[3760]	Speed: 117.524292 samples/sec	 accuracy=71.215625	 loss=1.207774	 lr=0.000277
Epoch[190] Batch [2049]/[3760]	Speed: 117.428530 samples/sec	 accuracy=71.223323	 loss=1.207415	 lr=0.000276
Epoch[190] Batch [2099]/[3760]	Speed: 117.313438 samples/sec	 accuracy=71.212798	 loss=1.207295	 lr=0.000275
Epoch[190] Batch [2149]/[3760]	Speed: 117.411952 samples/sec	 accuracy=71.213663	 loss=1.207716	 lr=0.000273
Epoch[190] Batch [2199]/[3760]	Speed: 117.189198 samples/sec	 accuracy=71.205256	 loss=1.207591	 lr=0.000272
Epoch[190] Batch [2249]/[3760]	Speed: 117.478247 samples/sec	 accuracy=71.205556	 loss=1.207415	 lr=0.000271
Epoch[190] Batch [2299]/[3760]	Speed: 115.285883 samples/sec	 accuracy=71.203125	 loss=1.207368	 lr=0.000269
Epoch[190] Batch [2349]/[3760]	Speed: 117.752639 samples/sec	 accuracy=71.196809	 loss=1.208337	 lr=0.000268
Epoch[190] Batch [2399]/[3760]	Speed: 117.598195 samples/sec	 accuracy=71.179036	 loss=1.208503	 lr=0.000267
Epoch[190] Batch [2449]/[3760]	Speed: 117.753248 samples/sec	 accuracy=71.148597	 loss=1.209654	 lr=0.000265
Epoch[190] Batch [2499]/[3760]	Speed: 117.508108 samples/sec	 accuracy=71.150000	 loss=1.208839	 lr=0.000264
Epoch[190] Batch [2549]/[3760]	Speed: 117.701647 samples/sec	 accuracy=71.140319	 loss=1.209139	 lr=0.000263
Epoch[190] Batch [2599]/[3760]	Speed: 117.667951 samples/sec	 accuracy=71.171875	 loss=1.208362	 lr=0.000261
Epoch[190] Batch [2649]/[3760]	Speed: 118.060537 samples/sec	 accuracy=71.152712	 loss=1.209282	 lr=0.000260
Epoch[190] Batch [2699]/[3760]	Speed: 117.249440 samples/sec	 accuracy=71.136574	 loss=1.209962	 lr=0.000259
Epoch[190] Batch [2749]/[3760]	Speed: 117.473506 samples/sec	 accuracy=71.146591	 loss=1.209427	 lr=0.000257
Epoch[190] Batch [2799]/[3760]	Speed: 117.833271 samples/sec	 accuracy=71.158482	 loss=1.209426	 lr=0.000256
Epoch[190] Batch [2849]/[3760]	Speed: 117.737706 samples/sec	 accuracy=71.155702	 loss=1.209281	 lr=0.000255
Epoch[190] Batch [2899]/[3760]	Speed: 117.685393 samples/sec	 accuracy=71.138470	 loss=1.209566	 lr=0.000254
Epoch[190] Batch [2949]/[3760]	Speed: 117.555766 samples/sec	 accuracy=71.138771	 loss=1.209724	 lr=0.000252
Epoch[190] Batch [2999]/[3760]	Speed: 117.943936 samples/sec	 accuracy=71.133854	 loss=1.209857	 lr=0.000251
Epoch[190] Batch [3049]/[3760]	Speed: 117.521646 samples/sec	 accuracy=71.135246	 loss=1.210381	 lr=0.000250
Epoch[190] Batch [3099]/[3760]	Speed: 117.101048 samples/sec	 accuracy=71.145665	 loss=1.210106	 lr=0.000248
Epoch[190] Batch [3149]/[3760]	Speed: 117.550526 samples/sec	 accuracy=71.149306	 loss=1.209986	 lr=0.000247
Epoch[190] Batch [3199]/[3760]	Speed: 117.063751 samples/sec	 accuracy=71.145508	 loss=1.210365	 lr=0.000246
Epoch[190] Batch [3249]/[3760]	Speed: 117.493655 samples/sec	 accuracy=71.143269	 loss=1.210160	 lr=0.000245
Epoch[190] Batch [3299]/[3760]	Speed: 118.060385 samples/sec	 accuracy=71.134943	 loss=1.210820	 lr=0.000243
Epoch[190] Batch [3349]/[3760]	Speed: 117.272246 samples/sec	 accuracy=71.148321	 loss=1.210236	 lr=0.000242
Epoch[190] Batch [3399]/[3760]	Speed: 117.941257 samples/sec	 accuracy=71.147059	 loss=1.210201	 lr=0.000241
Epoch[190] Batch [3449]/[3760]	Speed: 118.044925 samples/sec	 accuracy=71.142210	 loss=1.210163	 lr=0.000239
Epoch[190] Batch [3499]/[3760]	Speed: 117.546913 samples/sec	 accuracy=71.131696	 loss=1.210772	 lr=0.000238
Epoch[190] Batch [3549]/[3760]	Speed: 117.491223 samples/sec	 accuracy=71.140405	 loss=1.210873	 lr=0.000237
Epoch[190] Batch [3599]/[3760]	Speed: 117.371829 samples/sec	 accuracy=71.160156	 loss=1.209661	 lr=0.000236
Epoch[190] Batch [3649]/[3760]	Speed: 118.166289 samples/sec	 accuracy=71.149829	 loss=1.209861	 lr=0.000234
Epoch[190] Batch [3699]/[3760]	Speed: 117.250888 samples/sec	 accuracy=71.176098	 loss=1.208633	 lr=0.000233
Epoch[190] Batch [3749]/[3760]	Speed: 120.154306 samples/sec	 accuracy=71.172917	 loss=1.208608	 lr=0.000232
Batch [0049]/[0303]: acc-top1=62.125000 acc-top5=81.531250
Batch [0099]/[0303]: acc-top1=59.281250 acc-top5=80.796875
Batch [0149]/[0303]: acc-top1=60.354167 acc-top5=81.250000
Batch [0199]/[0303]: acc-top1=62.906250 acc-top5=82.851562
Batch [0249]/[0303]: acc-top1=62.368750 acc-top5=82.637500
Batch [0299]/[0303]: acc-top1=61.864583 acc-top5=82.593750
[Epoch 190] training: accuracy=71.170628	 loss=1.208517
[Epoch 190] speed: 116 samples/sec	time cost: 2174.467564
[Epoch 190] validation: acc-top1=61.798680 acc-top5=82.611386 loss=1.777737
Epoch[191] Batch [0049]/[3760]	Speed: 87.216997 samples/sec	 accuracy=71.500000	 loss=1.177858	 lr=0.000230
Epoch[191] Batch [0099]/[3760]	Speed: 116.705904 samples/sec	 accuracy=71.125000	 loss=1.191624	 lr=0.000229
Epoch[191] Batch [0149]/[3760]	Speed: 114.938026 samples/sec	 accuracy=71.093750	 loss=1.203242	 lr=0.000228
Epoch[191] Batch [0199]/[3760]	Speed: 116.980132 samples/sec	 accuracy=71.125000	 loss=1.205212	 lr=0.000227
Epoch[191] Batch [0249]/[3760]	Speed: 117.649630 samples/sec	 accuracy=71.281250	 loss=1.198347	 lr=0.000226
Epoch[191] Batch [0299]/[3760]	Speed: 117.237001 samples/sec	 accuracy=71.213542	 loss=1.199296	 lr=0.000224
Epoch[191] Batch [0349]/[3760]	Speed: 116.756371 samples/sec	 accuracy=71.151786	 loss=1.204060	 lr=0.000223
Epoch[191] Batch [0399]/[3760]	Speed: 117.332803 samples/sec	 accuracy=71.316406	 loss=1.191600	 lr=0.000222
Epoch[191] Batch [0449]/[3760]	Speed: 117.650068 samples/sec	 accuracy=71.288194	 loss=1.195883	 lr=0.000221
Epoch[191] Batch [0499]/[3760]	Speed: 117.442985 samples/sec	 accuracy=71.387500	 loss=1.191811	 lr=0.000219
Epoch[191] Batch [0549]/[3760]	Speed: 117.987604 samples/sec	 accuracy=71.375000	 loss=1.193571	 lr=0.000218
Epoch[191] Batch [0599]/[3760]	Speed: 117.526730 samples/sec	 accuracy=71.247396	 loss=1.199669	 lr=0.000217
Epoch[191] Batch [0649]/[3760]	Speed: 117.379603 samples/sec	 accuracy=71.218750	 loss=1.200321	 lr=0.000216
Epoch[191] Batch [0699]/[3760]	Speed: 117.611629 samples/sec	 accuracy=71.203125	 loss=1.204577	 lr=0.000215
Epoch[191] Batch [0749]/[3760]	Speed: 117.919045 samples/sec	 accuracy=71.277083	 loss=1.201283	 lr=0.000213
Epoch[191] Batch [0799]/[3760]	Speed: 117.799493 samples/sec	 accuracy=71.300781	 loss=1.200238	 lr=0.000212
Epoch[191] Batch [0849]/[3760]	Speed: 117.093753 samples/sec	 accuracy=71.332721	 loss=1.198946	 lr=0.000211
Epoch[191] Batch [0899]/[3760]	Speed: 117.451873 samples/sec	 accuracy=71.284722	 loss=1.201754	 lr=0.000210
Epoch[191] Batch [0949]/[3760]	Speed: 116.934228 samples/sec	 accuracy=71.266447	 loss=1.201989	 lr=0.000209
Epoch[191] Batch [0999]/[3760]	Speed: 117.048821 samples/sec	 accuracy=71.292187	 loss=1.200933	 lr=0.000208
Epoch[191] Batch [1049]/[3760]	Speed: 117.393622 samples/sec	 accuracy=71.303571	 loss=1.201061	 lr=0.000206
Epoch[191] Batch [1099]/[3760]	Speed: 117.420297 samples/sec	 accuracy=71.285511	 loss=1.201397	 lr=0.000205
Epoch[191] Batch [1149]/[3760]	Speed: 117.053491 samples/sec	 accuracy=71.270380	 loss=1.201542	 lr=0.000204
Epoch[191] Batch [1199]/[3760]	Speed: 117.192602 samples/sec	 accuracy=71.277344	 loss=1.201111	 lr=0.000203
Epoch[191] Batch [1249]/[3760]	Speed: 117.377842 samples/sec	 accuracy=71.282500	 loss=1.201223	 lr=0.000202
Epoch[191] Batch [1299]/[3760]	Speed: 117.653613 samples/sec	 accuracy=71.288462	 loss=1.200041	 lr=0.000201
Epoch[191] Batch [1349]/[3760]	Speed: 117.940393 samples/sec	 accuracy=71.245370	 loss=1.200594	 lr=0.000199
Epoch[191] Batch [1399]/[3760]	Speed: 117.504237 samples/sec	 accuracy=71.254464	 loss=1.199905	 lr=0.000198
Epoch[191] Batch [1449]/[3760]	Speed: 116.792303 samples/sec	 accuracy=71.283405	 loss=1.199134	 lr=0.000197
Epoch[191] Batch [1499]/[3760]	Speed: 117.823305 samples/sec	 accuracy=71.295833	 loss=1.198670	 lr=0.000196
Epoch[191] Batch [1549]/[3760]	Speed: 117.013628 samples/sec	 accuracy=71.277218	 loss=1.199167	 lr=0.000195
Epoch[191] Batch [1599]/[3760]	Speed: 117.813229 samples/sec	 accuracy=71.274414	 loss=1.198952	 lr=0.000194
Epoch[191] Batch [1649]/[3760]	Speed: 117.388190 samples/sec	 accuracy=71.277462	 loss=1.198618	 lr=0.000193
Epoch[191] Batch [1699]/[3760]	Speed: 117.985985 samples/sec	 accuracy=71.262868	 loss=1.198952	 lr=0.000191
Epoch[191] Batch [1749]/[3760]	Speed: 114.512977 samples/sec	 accuracy=71.241071	 loss=1.199915	 lr=0.000190
Epoch[191] Batch [1799]/[3760]	Speed: 117.191406 samples/sec	 accuracy=71.238715	 loss=1.199435	 lr=0.000189
Epoch[191] Batch [1849]/[3760]	Speed: 118.081919 samples/sec	 accuracy=71.246622	 loss=1.198856	 lr=0.000188
Epoch[191] Batch [1899]/[3760]	Speed: 117.344469 samples/sec	 accuracy=71.236842	 loss=1.199087	 lr=0.000187
Epoch[191] Batch [1949]/[3760]	Speed: 117.061849 samples/sec	 accuracy=71.220353	 loss=1.199827	 lr=0.000186
Epoch[191] Batch [1999]/[3760]	Speed: 117.127365 samples/sec	 accuracy=71.233594	 loss=1.199548	 lr=0.000185
Epoch[191] Batch [2049]/[3760]	Speed: 116.930277 samples/sec	 accuracy=71.254573	 loss=1.198591	 lr=0.000184
Epoch[191] Batch [2099]/[3760]	Speed: 116.671929 samples/sec	 accuracy=71.279762	 loss=1.198002	 lr=0.000183
Epoch[191] Batch [2149]/[3760]	Speed: 117.538666 samples/sec	 accuracy=71.273983	 loss=1.197415	 lr=0.000181
Epoch[191] Batch [2199]/[3760]	Speed: 117.359893 samples/sec	 accuracy=71.262784	 loss=1.198169	 lr=0.000180
Epoch[191] Batch [2249]/[3760]	Speed: 117.059316 samples/sec	 accuracy=71.239583	 loss=1.199712	 lr=0.000179
Epoch[191] Batch [2299]/[3760]	Speed: 116.921914 samples/sec	 accuracy=71.237092	 loss=1.199937	 lr=0.000178
Epoch[191] Batch [2349]/[3760]	Speed: 117.163931 samples/sec	 accuracy=71.269282	 loss=1.199013	 lr=0.000177
Epoch[191] Batch [2399]/[3760]	Speed: 117.211836 samples/sec	 accuracy=71.263672	 loss=1.198964	 lr=0.000176
Epoch[191] Batch [2449]/[3760]	Speed: 116.766410 samples/sec	 accuracy=71.253189	 loss=1.200265	 lr=0.000175
Epoch[191] Batch [2499]/[3760]	Speed: 117.302516 samples/sec	 accuracy=71.253750	 loss=1.200656	 lr=0.000174
Epoch[191] Batch [2549]/[3760]	Speed: 116.587929 samples/sec	 accuracy=71.248775	 loss=1.200796	 lr=0.000173
Epoch[191] Batch [2599]/[3760]	Speed: 117.726073 samples/sec	 accuracy=71.251202	 loss=1.200375	 lr=0.000172
Epoch[191] Batch [2649]/[3760]	Speed: 117.270450 samples/sec	 accuracy=71.274764	 loss=1.200043	 lr=0.000171
Epoch[191] Batch [2699]/[3760]	Speed: 117.128978 samples/sec	 accuracy=71.275463	 loss=1.200027	 lr=0.000170
Epoch[191] Batch [2749]/[3760]	Speed: 117.522990 samples/sec	 accuracy=71.303409	 loss=1.199721	 lr=0.000169
Epoch[191] Batch [2799]/[3760]	Speed: 117.569071 samples/sec	 accuracy=71.323661	 loss=1.198823	 lr=0.000167
Epoch[191] Batch [2849]/[3760]	Speed: 116.685393 samples/sec	 accuracy=71.346491	 loss=1.198037	 lr=0.000166
Epoch[191] Batch [2899]/[3760]	Speed: 117.060512 samples/sec	 accuracy=71.359914	 loss=1.197733	 lr=0.000165
Epoch[191] Batch [2949]/[3760]	Speed: 117.454141 samples/sec	 accuracy=71.369174	 loss=1.197207	 lr=0.000164
Epoch[191] Batch [2999]/[3760]	Speed: 117.670342 samples/sec	 accuracy=71.364583	 loss=1.197849	 lr=0.000163
Epoch[191] Batch [3049]/[3760]	Speed: 117.006566 samples/sec	 accuracy=71.371414	 loss=1.197108	 lr=0.000162
Epoch[191] Batch [3099]/[3760]	Speed: 117.280481 samples/sec	 accuracy=71.382560	 loss=1.197473	 lr=0.000161
Epoch[191] Batch [3149]/[3760]	Speed: 117.241874 samples/sec	 accuracy=71.365575	 loss=1.198049	 lr=0.000160
Epoch[191] Batch [3199]/[3760]	Speed: 116.957677 samples/sec	 accuracy=71.346191	 loss=1.198583	 lr=0.000159
Epoch[191] Batch [3249]/[3760]	Speed: 117.689194 samples/sec	 accuracy=71.356731	 loss=1.198405	 lr=0.000158
Epoch[191] Batch [3299]/[3760]	Speed: 118.043037 samples/sec	 accuracy=71.369792	 loss=1.197553	 lr=0.000157
Epoch[191] Batch [3349]/[3760]	Speed: 116.746845 samples/sec	 accuracy=71.379664	 loss=1.197151	 lr=0.000156
Epoch[191] Batch [3399]/[3760]	Speed: 115.562948 samples/sec	 accuracy=71.395680	 loss=1.196967	 lr=0.000155
Epoch[191] Batch [3449]/[3760]	Speed: 117.576162 samples/sec	 accuracy=71.412138	 loss=1.196490	 lr=0.000154
Epoch[191] Batch [3499]/[3760]	Speed: 117.227047 samples/sec	 accuracy=71.418750	 loss=1.196236	 lr=0.000153
Epoch[191] Batch [3549]/[3760]	Speed: 117.874252 samples/sec	 accuracy=71.415053	 loss=1.196308	 lr=0.000152
Epoch[191] Batch [3599]/[3760]	Speed: 118.127060 samples/sec	 accuracy=71.402344	 loss=1.196485	 lr=0.000151
Epoch[191] Batch [3649]/[3760]	Speed: 117.594234 samples/sec	 accuracy=71.421661	 loss=1.196313	 lr=0.000150
Epoch[191] Batch [3699]/[3760]	Speed: 116.938349 samples/sec	 accuracy=71.440034	 loss=1.196135	 lr=0.000149
Epoch[191] Batch [3749]/[3760]	Speed: 120.601667 samples/sec	 accuracy=71.410417	 loss=1.196664	 lr=0.000148
Batch [0049]/[0303]: acc-top1=61.718750 acc-top5=80.875000
Batch [0099]/[0303]: acc-top1=59.015625 acc-top5=80.578125
Batch [0149]/[0303]: acc-top1=60.510417 acc-top5=81.229167
Batch [0199]/[0303]: acc-top1=63.195312 acc-top5=82.937500
Batch [0249]/[0303]: acc-top1=62.593750 acc-top5=82.637500
Batch [0299]/[0303]: acc-top1=62.067708 acc-top5=82.598958
[Epoch 191] training: accuracy=71.410406	 loss=1.196732
[Epoch 191] speed: 116 samples/sec	time cost: 2177.984783
[Epoch 191] validation: acc-top1=61.984323 acc-top5=82.616543 loss=1.775270
Epoch[192] Batch [0049]/[3759]	Speed: 88.590522 samples/sec	 accuracy=70.812500	 loss=1.204265	 lr=0.000147
Epoch[192] Batch [0099]/[3759]	Speed: 117.206915 samples/sec	 accuracy=70.609375	 loss=1.221849	 lr=0.000146
Epoch[192] Batch [0149]/[3759]	Speed: 117.162776 samples/sec	 accuracy=70.843750	 loss=1.220836	 lr=0.000145
Epoch[192] Batch [0199]/[3759]	Speed: 117.987518 samples/sec	 accuracy=71.000000	 loss=1.217065	 lr=0.000144
Epoch[192] Batch [0249]/[3759]	Speed: 118.098504 samples/sec	 accuracy=71.056250	 loss=1.219576	 lr=0.000143
Epoch[192] Batch [0299]/[3759]	Speed: 117.017374 samples/sec	 accuracy=71.135417	 loss=1.212959	 lr=0.000142
Epoch[192] Batch [0349]/[3759]	Speed: 117.768577 samples/sec	 accuracy=71.191964	 loss=1.211472	 lr=0.000141
Epoch[192] Batch [0399]/[3759]	Speed: 117.739363 samples/sec	 accuracy=71.066406	 loss=1.212627	 lr=0.000140
Epoch[192] Batch [0449]/[3759]	Speed: 117.555060 samples/sec	 accuracy=71.312500	 loss=1.200706	 lr=0.000139
Epoch[192] Batch [0499]/[3759]	Speed: 117.531746 samples/sec	 accuracy=71.287500	 loss=1.204676	 lr=0.000138
Epoch[192] Batch [0549]/[3759]	Speed: 117.889678 samples/sec	 accuracy=71.281250	 loss=1.207781	 lr=0.000137
Epoch[192] Batch [0599]/[3759]	Speed: 117.385791 samples/sec	 accuracy=71.473958	 loss=1.199046	 lr=0.000136
Epoch[192] Batch [0649]/[3759]	Speed: 117.691997 samples/sec	 accuracy=71.502404	 loss=1.200440	 lr=0.000135
Epoch[192] Batch [0699]/[3759]	Speed: 116.665583 samples/sec	 accuracy=71.607143	 loss=1.196643	 lr=0.000134
Epoch[192] Batch [0749]/[3759]	Speed: 117.490002 samples/sec	 accuracy=71.710417	 loss=1.195490	 lr=0.000133
Epoch[192] Batch [0799]/[3759]	Speed: 118.032645 samples/sec	 accuracy=71.730469	 loss=1.193263	 lr=0.000132
Epoch[192] Batch [0849]/[3759]	Speed: 118.014290 samples/sec	 accuracy=71.670956	 loss=1.193905	 lr=0.000131
Epoch[192] Batch [0899]/[3759]	Speed: 117.673047 samples/sec	 accuracy=71.703125	 loss=1.193685	 lr=0.000131
Epoch[192] Batch [0949]/[3759]	Speed: 116.581414 samples/sec	 accuracy=71.738487	 loss=1.192099	 lr=0.000130
Epoch[192] Batch [0999]/[3759]	Speed: 117.425906 samples/sec	 accuracy=71.682812	 loss=1.192797	 lr=0.000129
Epoch[192] Batch [1049]/[3759]	Speed: 117.729764 samples/sec	 accuracy=71.741071	 loss=1.191660	 lr=0.000128
Epoch[192] Batch [1099]/[3759]	Speed: 114.828655 samples/sec	 accuracy=71.717330	 loss=1.192842	 lr=0.000127
Epoch[192] Batch [1149]/[3759]	Speed: 117.403658 samples/sec	 accuracy=71.607337	 loss=1.195953	 lr=0.000126
Epoch[192] Batch [1199]/[3759]	Speed: 118.178823 samples/sec	 accuracy=71.630208	 loss=1.194059	 lr=0.000125
Epoch[192] Batch [1249]/[3759]	Speed: 117.566493 samples/sec	 accuracy=71.630000	 loss=1.194203	 lr=0.000124
Epoch[192] Batch [1299]/[3759]	Speed: 117.664913 samples/sec	 accuracy=71.683894	 loss=1.192055	 lr=0.000123
Epoch[192] Batch [1349]/[3759]	Speed: 117.608202 samples/sec	 accuracy=71.699074	 loss=1.191856	 lr=0.000122
Epoch[192] Batch [1399]/[3759]	Speed: 117.826061 samples/sec	 accuracy=71.674107	 loss=1.193517	 lr=0.000121
Epoch[192] Batch [1449]/[3759]	Speed: 117.415231 samples/sec	 accuracy=71.646552	 loss=1.193296	 lr=0.000120
Epoch[192] Batch [1499]/[3759]	Speed: 117.880039 samples/sec	 accuracy=71.627083	 loss=1.194238	 lr=0.000120
Epoch[192] Batch [1549]/[3759]	Speed: 117.726305 samples/sec	 accuracy=71.617944	 loss=1.193973	 lr=0.000119
Epoch[192] Batch [1599]/[3759]	Speed: 117.579170 samples/sec	 accuracy=71.594727	 loss=1.194974	 lr=0.000118
Epoch[192] Batch [1649]/[3759]	Speed: 118.156379 samples/sec	 accuracy=71.601326	 loss=1.193529	 lr=0.000117
Epoch[192] Batch [1699]/[3759]	Speed: 117.132696 samples/sec	 accuracy=71.599265	 loss=1.193762	 lr=0.000116
Epoch[192] Batch [1749]/[3759]	Speed: 118.311317 samples/sec	 accuracy=71.627679	 loss=1.192278	 lr=0.000115
Epoch[192] Batch [1799]/[3759]	Speed: 117.086941 samples/sec	 accuracy=71.654514	 loss=1.190403	 lr=0.000114
Epoch[192] Batch [1849]/[3759]	Speed: 117.186094 samples/sec	 accuracy=71.645270	 loss=1.191091	 lr=0.000113
Epoch[192] Batch [1899]/[3759]	Speed: 117.640388 samples/sec	 accuracy=71.654605	 loss=1.190639	 lr=0.000113
Epoch[192] Batch [1949]/[3759]	Speed: 117.605544 samples/sec	 accuracy=71.642628	 loss=1.190756	 lr=0.000112
Epoch[192] Batch [1999]/[3759]	Speed: 117.666408 samples/sec	 accuracy=71.614063	 loss=1.191737	 lr=0.000111
Epoch[192] Batch [2049]/[3759]	Speed: 117.790419 samples/sec	 accuracy=71.625762	 loss=1.191662	 lr=0.000110
Epoch[192] Batch [2099]/[3759]	Speed: 117.362295 samples/sec	 accuracy=71.612351	 loss=1.192382	 lr=0.000109
Epoch[192] Batch [2149]/[3759]	Speed: 117.495817 samples/sec	 accuracy=71.630087	 loss=1.192056	 lr=0.000108
Epoch[192] Batch [2199]/[3759]	Speed: 117.555253 samples/sec	 accuracy=71.628551	 loss=1.191498	 lr=0.000107
Epoch[192] Batch [2249]/[3759]	Speed: 117.776155 samples/sec	 accuracy=71.604861	 loss=1.192568	 lr=0.000107
Epoch[192] Batch [2299]/[3759]	Speed: 116.699893 samples/sec	 accuracy=71.588315	 loss=1.192731	 lr=0.000106
Epoch[192] Batch [2349]/[3759]	Speed: 118.339797 samples/sec	 accuracy=71.581782	 loss=1.193583	 lr=0.000105
Epoch[192] Batch [2399]/[3759]	Speed: 117.775412 samples/sec	 accuracy=71.604167	 loss=1.192342	 lr=0.000104
Epoch[192] Batch [2449]/[3759]	Speed: 117.427057 samples/sec	 accuracy=71.607143	 loss=1.192393	 lr=0.000103
Epoch[192] Batch [2499]/[3759]	Speed: 118.084452 samples/sec	 accuracy=71.612500	 loss=1.191453	 lr=0.000102
Epoch[192] Batch [2549]/[3759]	Speed: 118.123236 samples/sec	 accuracy=71.598652	 loss=1.191794	 lr=0.000102
Epoch[192] Batch [2599]/[3759]	Speed: 117.947768 samples/sec	 accuracy=71.601562	 loss=1.191452	 lr=0.000101
Epoch[192] Batch [2649]/[3759]	Speed: 117.464269 samples/sec	 accuracy=71.594929	 loss=1.191393	 lr=0.000100
Epoch[192] Batch [2699]/[3759]	Speed: 116.903619 samples/sec	 accuracy=71.573495	 loss=1.191250	 lr=0.000099
Epoch[192] Batch [2749]/[3759]	Speed: 115.041007 samples/sec	 accuracy=71.556250	 loss=1.191619	 lr=0.000098
Epoch[192] Batch [2799]/[3759]	Speed: 118.103552 samples/sec	 accuracy=71.545201	 loss=1.191583	 lr=0.000098
Epoch[192] Batch [2849]/[3759]	Speed: 117.840462 samples/sec	 accuracy=71.542215	 loss=1.191307	 lr=0.000097
Epoch[192] Batch [2899]/[3759]	Speed: 116.734328 samples/sec	 accuracy=71.551724	 loss=1.191242	 lr=0.000096
Epoch[192] Batch [2949]/[3759]	Speed: 117.855960 samples/sec	 accuracy=71.572034	 loss=1.190944	 lr=0.000095
Epoch[192] Batch [2999]/[3759]	Speed: 117.184826 samples/sec	 accuracy=71.597396	 loss=1.189716	 lr=0.000094
Epoch[192] Batch [3049]/[3759]	Speed: 116.797590 samples/sec	 accuracy=71.582992	 loss=1.189538	 lr=0.000094
Epoch[192] Batch [3099]/[3759]	Speed: 117.440072 samples/sec	 accuracy=71.563508	 loss=1.190252	 lr=0.000093
Epoch[192] Batch [3149]/[3759]	Speed: 118.049194 samples/sec	 accuracy=71.553571	 loss=1.190722	 lr=0.000092
Epoch[192] Batch [3199]/[3759]	Speed: 117.988661 samples/sec	 accuracy=71.557617	 loss=1.191129	 lr=0.000091
Epoch[192] Batch [3249]/[3759]	Speed: 118.116566 samples/sec	 accuracy=71.550481	 loss=1.191138	 lr=0.000090
Epoch[192] Batch [3299]/[3759]	Speed: 117.552026 samples/sec	 accuracy=71.548769	 loss=1.191359	 lr=0.000090
Epoch[192] Batch [3349]/[3759]	Speed: 117.862954 samples/sec	 accuracy=71.547575	 loss=1.191001	 lr=0.000089
Epoch[192] Batch [3399]/[3759]	Speed: 118.065211 samples/sec	 accuracy=71.541360	 loss=1.190995	 lr=0.000088
Epoch[192] Batch [3449]/[3759]	Speed: 117.655022 samples/sec	 accuracy=71.548007	 loss=1.191212	 lr=0.000087
Epoch[192] Batch [3499]/[3759]	Speed: 117.656112 samples/sec	 accuracy=71.546429	 loss=1.191049	 lr=0.000087
Epoch[192] Batch [3549]/[3759]	Speed: 117.566252 samples/sec	 accuracy=71.527729	 loss=1.191682	 lr=0.000086
Epoch[192] Batch [3599]/[3759]	Speed: 118.207585 samples/sec	 accuracy=71.516493	 loss=1.192336	 lr=0.000085
Epoch[192] Batch [3649]/[3759]	Speed: 117.282243 samples/sec	 accuracy=71.503853	 loss=1.193022	 lr=0.000084
Epoch[192] Batch [3699]/[3759]	Speed: 117.499907 samples/sec	 accuracy=71.492821	 loss=1.193285	 lr=0.000084
Epoch[192] Batch [3749]/[3759]	Speed: 120.748697 samples/sec	 accuracy=71.505000	 loss=1.192914	 lr=0.000083
Batch [0049]/[0303]: acc-top1=61.937500 acc-top5=81.250000
Batch [0099]/[0303]: acc-top1=59.046875 acc-top5=80.703125
Batch [0149]/[0303]: acc-top1=60.468750 acc-top5=81.270833
Batch [0199]/[0303]: acc-top1=63.085938 acc-top5=83.007812
Batch [0249]/[0303]: acc-top1=62.531250 acc-top5=82.837500
Batch [0299]/[0303]: acc-top1=62.093750 acc-top5=82.843750
[Epoch 192] training: accuracy=71.506302	 loss=1.193012
[Epoch 192] speed: 117 samples/sec	time cost: 2171.590899
[Epoch 192] validation: acc-top1=62.041048 acc-top5=82.869224 loss=1.771651
Epoch[193] Batch [0049]/[3760]	Speed: 87.229481 samples/sec	 accuracy=71.718750	 loss=1.197348	 lr=0.000082
Epoch[193] Batch [0099]/[3760]	Speed: 117.429783 samples/sec	 accuracy=71.468750	 loss=1.193966	 lr=0.000081
Epoch[193] Batch [0149]/[3760]	Speed: 117.805648 samples/sec	 accuracy=71.760417	 loss=1.175130	 lr=0.000080
Epoch[193] Batch [0199]/[3760]	Speed: 116.928644 samples/sec	 accuracy=71.578125	 loss=1.184360	 lr=0.000080
Epoch[193] Batch [0249]/[3760]	Speed: 118.220296 samples/sec	 accuracy=71.462500	 loss=1.185426	 lr=0.000079
Epoch[193] Batch [0299]/[3760]	Speed: 117.456320 samples/sec	 accuracy=71.421875	 loss=1.187356	 lr=0.000078
Epoch[193] Batch [0349]/[3760]	Speed: 118.058745 samples/sec	 accuracy=71.464286	 loss=1.190527	 lr=0.000078
Epoch[193] Batch [0399]/[3760]	Speed: 117.264596 samples/sec	 accuracy=71.390625	 loss=1.193699	 lr=0.000077
Epoch[193] Batch [0449]/[3760]	Speed: 117.153431 samples/sec	 accuracy=71.454861	 loss=1.194992	 lr=0.000076
Epoch[193] Batch [0499]/[3760]	Speed: 118.246781 samples/sec	 accuracy=71.568750	 loss=1.190757	 lr=0.000075
Epoch[193] Batch [0549]/[3760]	Speed: 117.760071 samples/sec	 accuracy=71.548295	 loss=1.188871	 lr=0.000075
Epoch[193] Batch [0599]/[3760]	Speed: 117.518774 samples/sec	 accuracy=71.546875	 loss=1.187590	 lr=0.000074
Epoch[193] Batch [0649]/[3760]	Speed: 114.776424 samples/sec	 accuracy=71.555288	 loss=1.187712	 lr=0.000073
Epoch[193] Batch [0699]/[3760]	Speed: 117.924337 samples/sec	 accuracy=71.569196	 loss=1.190596	 lr=0.000073
Epoch[193] Batch [0749]/[3760]	Speed: 117.657076 samples/sec	 accuracy=71.620833	 loss=1.188755	 lr=0.000072
Epoch[193] Batch [0799]/[3760]	Speed: 117.071512 samples/sec	 accuracy=71.582031	 loss=1.187727	 lr=0.000071
Epoch[193] Batch [0849]/[3760]	Speed: 117.428820 samples/sec	 accuracy=71.580882	 loss=1.190635	 lr=0.000071
Epoch[193] Batch [0899]/[3760]	Speed: 117.070885 samples/sec	 accuracy=71.600694	 loss=1.189836	 lr=0.000070
Epoch[193] Batch [0949]/[3760]	Speed: 117.361665 samples/sec	 accuracy=71.649671	 loss=1.189331	 lr=0.000069
Epoch[193] Batch [0999]/[3760]	Speed: 117.351428 samples/sec	 accuracy=71.665625	 loss=1.188003	 lr=0.000069
Epoch[193] Batch [1049]/[3760]	Speed: 117.787513 samples/sec	 accuracy=71.651786	 loss=1.188287	 lr=0.000068
Epoch[193] Batch [1099]/[3760]	Speed: 118.024858 samples/sec	 accuracy=71.678977	 loss=1.187396	 lr=0.000067
Epoch[193] Batch [1149]/[3760]	Speed: 117.456433 samples/sec	 accuracy=71.721467	 loss=1.186242	 lr=0.000067
Epoch[193] Batch [1199]/[3760]	Speed: 117.894426 samples/sec	 accuracy=71.759115	 loss=1.184983	 lr=0.000066
Epoch[193] Batch [1249]/[3760]	Speed: 117.643553 samples/sec	 accuracy=71.733750	 loss=1.184443	 lr=0.000065
Epoch[193] Batch [1299]/[3760]	Speed: 117.076387 samples/sec	 accuracy=71.751202	 loss=1.184548	 lr=0.000065
Epoch[193] Batch [1349]/[3760]	Speed: 117.017797 samples/sec	 accuracy=71.755787	 loss=1.184278	 lr=0.000064
Epoch[193] Batch [1399]/[3760]	Speed: 117.306743 samples/sec	 accuracy=71.758929	 loss=1.183574	 lr=0.000063
Epoch[193] Batch [1449]/[3760]	Speed: 117.276944 samples/sec	 accuracy=71.746767	 loss=1.183289	 lr=0.000063
Epoch[193] Batch [1499]/[3760]	Speed: 117.470428 samples/sec	 accuracy=71.743750	 loss=1.183611	 lr=0.000062
Epoch[193] Batch [1549]/[3760]	Speed: 117.921066 samples/sec	 accuracy=71.757056	 loss=1.183895	 lr=0.000061
Epoch[193] Batch [1599]/[3760]	Speed: 118.030314 samples/sec	 accuracy=71.764648	 loss=1.183468	 lr=0.000061
Epoch[193] Batch [1649]/[3760]	Speed: 117.161700 samples/sec	 accuracy=71.776515	 loss=1.182158	 lr=0.000060
Epoch[193] Batch [1699]/[3760]	Speed: 117.239474 samples/sec	 accuracy=71.777574	 loss=1.181905	 lr=0.000059
Epoch[193] Batch [1749]/[3760]	Speed: 116.310227 samples/sec	 accuracy=71.791071	 loss=1.180759	 lr=0.000059
Epoch[193] Batch [1799]/[3760]	Speed: 117.262959 samples/sec	 accuracy=71.779514	 loss=1.180606	 lr=0.000058
Epoch[193] Batch [1849]/[3760]	Speed: 117.919342 samples/sec	 accuracy=71.789696	 loss=1.180056	 lr=0.000058
Epoch[193] Batch [1899]/[3760]	Speed: 117.260082 samples/sec	 accuracy=71.773849	 loss=1.180170	 lr=0.000057
Epoch[193] Batch [1949]/[3760]	Speed: 118.474835 samples/sec	 accuracy=71.807692	 loss=1.178488	 lr=0.000056
Epoch[193] Batch [1999]/[3760]	Speed: 117.589677 samples/sec	 accuracy=71.803906	 loss=1.178851	 lr=0.000056
Epoch[193] Batch [2049]/[3760]	Speed: 117.499608 samples/sec	 accuracy=71.830793	 loss=1.178087	 lr=0.000055
Epoch[193] Batch [2099]/[3760]	Speed: 118.176126 samples/sec	 accuracy=71.821429	 loss=1.179104	 lr=0.000054
Epoch[193] Batch [2149]/[3760]	Speed: 118.054233 samples/sec	 accuracy=71.844477	 loss=1.178681	 lr=0.000054
Epoch[193] Batch [2199]/[3760]	Speed: 117.110612 samples/sec	 accuracy=71.865767	 loss=1.178291	 lr=0.000053
Epoch[193] Batch [2249]/[3760]	Speed: 117.872947 samples/sec	 accuracy=71.864583	 loss=1.178761	 lr=0.000053
Epoch[193] Batch [2299]/[3760]	Speed: 117.976124 samples/sec	 accuracy=71.857337	 loss=1.179208	 lr=0.000052
Epoch[193] Batch [2349]/[3760]	Speed: 114.664864 samples/sec	 accuracy=71.861037	 loss=1.179248	 lr=0.000052
Epoch[193] Batch [2399]/[3760]	Speed: 117.418370 samples/sec	 accuracy=71.865885	 loss=1.178967	 lr=0.000051
Epoch[193] Batch [2449]/[3760]	Speed: 116.902120 samples/sec	 accuracy=71.896046	 loss=1.177661	 lr=0.000050
Epoch[193] Batch [2499]/[3760]	Speed: 117.690660 samples/sec	 accuracy=71.892500	 loss=1.178135	 lr=0.000050
Epoch[193] Batch [2549]/[3760]	Speed: 117.971316 samples/sec	 accuracy=71.893382	 loss=1.178604	 lr=0.000049
Epoch[193] Batch [2599]/[3760]	Speed: 117.657379 samples/sec	 accuracy=71.871394	 loss=1.179270	 lr=0.000049
Epoch[193] Batch [2649]/[3760]	Speed: 117.123013 samples/sec	 accuracy=71.860849	 loss=1.179928	 lr=0.000048
Epoch[193] Batch [2699]/[3760]	Speed: 117.805797 samples/sec	 accuracy=71.865162	 loss=1.180035	 lr=0.000048
Epoch[193] Batch [2749]/[3760]	Speed: 118.787634 samples/sec	 accuracy=71.892614	 loss=1.179578	 lr=0.000047
Epoch[193] Batch [2799]/[3760]	Speed: 118.083951 samples/sec	 accuracy=71.906808	 loss=1.179224	 lr=0.000046
Epoch[193] Batch [2849]/[3760]	Speed: 117.692266 samples/sec	 accuracy=71.903509	 loss=1.179128	 lr=0.000046
Epoch[193] Batch [2899]/[3760]	Speed: 117.748900 samples/sec	 accuracy=71.911099	 loss=1.177999	 lr=0.000045
Epoch[193] Batch [2949]/[3760]	Speed: 118.144182 samples/sec	 accuracy=71.907839	 loss=1.178290	 lr=0.000045
Epoch[193] Batch [2999]/[3760]	Speed: 117.647868 samples/sec	 accuracy=71.901563	 loss=1.178790	 lr=0.000044
Epoch[193] Batch [3049]/[3760]	Speed: 117.525878 samples/sec	 accuracy=71.900102	 loss=1.179269	 lr=0.000044
Epoch[193] Batch [3099]/[3760]	Speed: 117.865719 samples/sec	 accuracy=71.876512	 loss=1.180092	 lr=0.000043
Epoch[193] Batch [3149]/[3760]	Speed: 117.388171 samples/sec	 accuracy=71.866567	 loss=1.180373	 lr=0.000043
Epoch[193] Batch [3199]/[3760]	Speed: 118.001956 samples/sec	 accuracy=71.860352	 loss=1.180837	 lr=0.000042
Epoch[193] Batch [3249]/[3760]	Speed: 117.878432 samples/sec	 accuracy=71.856731	 loss=1.181135	 lr=0.000042
Epoch[193] Batch [3299]/[3760]	Speed: 117.809526 samples/sec	 accuracy=71.867898	 loss=1.180923	 lr=0.000041
Epoch[193] Batch [3349]/[3760]	Speed: 117.581828 samples/sec	 accuracy=71.877799	 loss=1.180488	 lr=0.000040
Epoch[193] Batch [3399]/[3760]	Speed: 118.152181 samples/sec	 accuracy=71.901195	 loss=1.179211	 lr=0.000040
Epoch[193] Batch [3449]/[3760]	Speed: 117.521575 samples/sec	 accuracy=71.892210	 loss=1.179564	 lr=0.000039
Epoch[193] Batch [3499]/[3760]	Speed: 117.394422 samples/sec	 accuracy=71.891964	 loss=1.179767	 lr=0.000039
Epoch[193] Batch [3549]/[3760]	Speed: 117.685291 samples/sec	 accuracy=71.887324	 loss=1.179993	 lr=0.000038
Epoch[193] Batch [3599]/[3760]	Speed: 117.778863 samples/sec	 accuracy=71.881510	 loss=1.180155	 lr=0.000038
Epoch[193] Batch [3649]/[3760]	Speed: 117.650352 samples/sec	 accuracy=71.855736	 loss=1.180922	 lr=0.000037
Epoch[193] Batch [3699]/[3760]	Speed: 117.688107 samples/sec	 accuracy=71.859375	 loss=1.180100	 lr=0.000037
Epoch[193] Batch [3749]/[3760]	Speed: 121.274941 samples/sec	 accuracy=71.858750	 loss=1.179867	 lr=0.000036
Batch [0049]/[0303]: acc-top1=62.500000 acc-top5=81.062500
Batch [0099]/[0303]: acc-top1=59.671875 acc-top5=80.656250
Batch [0149]/[0303]: acc-top1=60.770833 acc-top5=81.270833
Batch [0199]/[0303]: acc-top1=63.367188 acc-top5=82.945312
Batch [0249]/[0303]: acc-top1=62.731250 acc-top5=82.718750
Batch [0299]/[0303]: acc-top1=62.234375 acc-top5=82.671875
[Epoch 193] training: accuracy=71.857131	 loss=1.180037
[Epoch 193] speed: 117 samples/sec	time cost: 2173.027952
[Epoch 193] validation: acc-top1=62.175124 acc-top5=82.688738 loss=1.775081
Epoch[194] Batch [0049]/[3759]	Speed: 86.598461 samples/sec	 accuracy=72.687500	 loss=1.157436	 lr=0.000036
Epoch[194] Batch [0099]/[3759]	Speed: 118.001132 samples/sec	 accuracy=71.890625	 loss=1.175416	 lr=0.000035
Epoch[194] Batch [0149]/[3759]	Speed: 117.445996 samples/sec	 accuracy=71.708333	 loss=1.188906	 lr=0.000035
Epoch[194] Batch [0199]/[3759]	Speed: 117.599318 samples/sec	 accuracy=71.593750	 loss=1.185123	 lr=0.000034
Epoch[194] Batch [0249]/[3759]	Speed: 115.018863 samples/sec	 accuracy=71.937500	 loss=1.173088	 lr=0.000034
Epoch[194] Batch [0299]/[3759]	Speed: 117.486961 samples/sec	 accuracy=71.958333	 loss=1.167198	 lr=0.000033
Epoch[194] Batch [0349]/[3759]	Speed: 117.583448 samples/sec	 accuracy=72.107143	 loss=1.166004	 lr=0.000033
Epoch[194] Batch [0399]/[3759]	Speed: 116.771528 samples/sec	 accuracy=72.210938	 loss=1.164921	 lr=0.000033
Epoch[194] Batch [0449]/[3759]	Speed: 117.548032 samples/sec	 accuracy=72.190972	 loss=1.166561	 lr=0.000032
Epoch[194] Batch [0499]/[3759]	Speed: 117.734891 samples/sec	 accuracy=72.071875	 loss=1.170673	 lr=0.000032
Epoch[194] Batch [0549]/[3759]	Speed: 117.446275 samples/sec	 accuracy=72.082386	 loss=1.169471	 lr=0.000031
Epoch[194] Batch [0599]/[3759]	Speed: 117.510190 samples/sec	 accuracy=72.166667	 loss=1.163150	 lr=0.000031
Epoch[194] Batch [0649]/[3759]	Speed: 117.555817 samples/sec	 accuracy=72.091346	 loss=1.164197	 lr=0.000030
Epoch[194] Batch [0699]/[3759]	Speed: 117.862906 samples/sec	 accuracy=72.064732	 loss=1.164917	 lr=0.000030
Epoch[194] Batch [0749]/[3759]	Speed: 117.657732 samples/sec	 accuracy=72.118750	 loss=1.165248	 lr=0.000029
Epoch[194] Batch [0799]/[3759]	Speed: 117.549639 samples/sec	 accuracy=72.076172	 loss=1.166878	 lr=0.000029
Epoch[194] Batch [0849]/[3759]	Speed: 118.000282 samples/sec	 accuracy=72.049632	 loss=1.167472	 lr=0.000028
Epoch[194] Batch [0899]/[3759]	Speed: 117.376447 samples/sec	 accuracy=72.095486	 loss=1.167124	 lr=0.000028
Epoch[194] Batch [0949]/[3759]	Speed: 117.658264 samples/sec	 accuracy=72.062500	 loss=1.170742	 lr=0.000028
Epoch[194] Batch [0999]/[3759]	Speed: 117.598372 samples/sec	 accuracy=72.073437	 loss=1.171245	 lr=0.000027
Epoch[194] Batch [1049]/[3759]	Speed: 118.072182 samples/sec	 accuracy=72.037202	 loss=1.172166	 lr=0.000027
Epoch[194] Batch [1099]/[3759]	Speed: 117.091297 samples/sec	 accuracy=72.107955	 loss=1.170088	 lr=0.000026
Epoch[194] Batch [1149]/[3759]	Speed: 118.090123 samples/sec	 accuracy=72.078804	 loss=1.171864	 lr=0.000026
Epoch[194] Batch [1199]/[3759]	Speed: 117.216652 samples/sec	 accuracy=72.044271	 loss=1.172856	 lr=0.000025
Epoch[194] Batch [1249]/[3759]	Speed: 117.053891 samples/sec	 accuracy=72.055000	 loss=1.172259	 lr=0.000025
Epoch[194] Batch [1299]/[3759]	Speed: 117.710257 samples/sec	 accuracy=71.978365	 loss=1.176542	 lr=0.000025
Epoch[194] Batch [1349]/[3759]	Speed: 117.815735 samples/sec	 accuracy=71.913194	 loss=1.177637	 lr=0.000024
Epoch[194] Batch [1399]/[3759]	Speed: 117.603696 samples/sec	 accuracy=71.924107	 loss=1.177779	 lr=0.000024
Epoch[194] Batch [1449]/[3759]	Speed: 118.532150 samples/sec	 accuracy=71.879310	 loss=1.178319	 lr=0.000023
Epoch[194] Batch [1499]/[3759]	Speed: 117.760780 samples/sec	 accuracy=71.859375	 loss=1.179255	 lr=0.000023
Epoch[194] Batch [1549]/[3759]	Speed: 117.756082 samples/sec	 accuracy=71.831653	 loss=1.178877	 lr=0.000023
Epoch[194] Batch [1599]/[3759]	Speed: 117.282831 samples/sec	 accuracy=71.888672	 loss=1.175770	 lr=0.000022
Epoch[194] Batch [1649]/[3759]	Speed: 117.949550 samples/sec	 accuracy=71.889205	 loss=1.175342	 lr=0.000022
Epoch[194] Batch [1699]/[3759]	Speed: 117.061791 samples/sec	 accuracy=71.936581	 loss=1.173672	 lr=0.000022
Epoch[194] Batch [1749]/[3759]	Speed: 117.399541 samples/sec	 accuracy=71.931250	 loss=1.173592	 lr=0.000021
Epoch[194] Batch [1799]/[3759]	Speed: 117.660480 samples/sec	 accuracy=71.947049	 loss=1.173021	 lr=0.000021
Epoch[194] Batch [1849]/[3759]	Speed: 118.477035 samples/sec	 accuracy=71.983953	 loss=1.172092	 lr=0.000020
Epoch[194] Batch [1899]/[3759]	Speed: 114.868151 samples/sec	 accuracy=72.008224	 loss=1.171961	 lr=0.000020
Epoch[194] Batch [1949]/[3759]	Speed: 117.046645 samples/sec	 accuracy=72.008814	 loss=1.172184	 lr=0.000020
Epoch[194] Batch [1999]/[3759]	Speed: 118.031047 samples/sec	 accuracy=71.986719	 loss=1.173296	 lr=0.000019
Epoch[194] Batch [2049]/[3759]	Speed: 117.597972 samples/sec	 accuracy=71.996951	 loss=1.173050	 lr=0.000019
Epoch[194] Batch [2099]/[3759]	Speed: 117.458468 samples/sec	 accuracy=71.982143	 loss=1.173528	 lr=0.000019
Epoch[194] Batch [2149]/[3759]	Speed: 118.021895 samples/sec	 accuracy=71.973837	 loss=1.173889	 lr=0.000018
Epoch[194] Batch [2199]/[3759]	Speed: 117.612720 samples/sec	 accuracy=71.963778	 loss=1.174375	 lr=0.000018
Epoch[194] Batch [2249]/[3759]	Speed: 117.394079 samples/sec	 accuracy=71.964583	 loss=1.174144	 lr=0.000018
Epoch[194] Batch [2299]/[3759]	Speed: 117.486986 samples/sec	 accuracy=71.936821	 loss=1.175133	 lr=0.000017
Epoch[194] Batch [2349]/[3759]	Speed: 116.990084 samples/sec	 accuracy=71.924202	 loss=1.175417	 lr=0.000017
Epoch[194] Batch [2399]/[3759]	Speed: 116.776115 samples/sec	 accuracy=71.940104	 loss=1.174708	 lr=0.000017
Epoch[194] Batch [2449]/[3759]	Speed: 118.729717 samples/sec	 accuracy=71.950893	 loss=1.174175	 lr=0.000016
Epoch[194] Batch [2499]/[3759]	Speed: 117.126144 samples/sec	 accuracy=71.960000	 loss=1.173228	 lr=0.000016
Epoch[194] Batch [2549]/[3759]	Speed: 117.383641 samples/sec	 accuracy=71.945466	 loss=1.173594	 lr=0.000016
Epoch[194] Batch [2599]/[3759]	Speed: 117.805877 samples/sec	 accuracy=71.923077	 loss=1.174590	 lr=0.000015
Epoch[194] Batch [2649]/[3759]	Speed: 117.101723 samples/sec	 accuracy=71.938090	 loss=1.173520	 lr=0.000015
Epoch[194] Batch [2699]/[3759]	Speed: 117.833669 samples/sec	 accuracy=71.923032	 loss=1.173551	 lr=0.000015
Epoch[194] Batch [2749]/[3759]	Speed: 116.827019 samples/sec	 accuracy=71.932955	 loss=1.172984	 lr=0.000014
Epoch[194] Batch [2799]/[3759]	Speed: 118.230646 samples/sec	 accuracy=71.926339	 loss=1.172586	 lr=0.000014
Epoch[194] Batch [2849]/[3759]	Speed: 117.253187 samples/sec	 accuracy=71.944079	 loss=1.171890	 lr=0.000014
Epoch[194] Batch [2899]/[3759]	Speed: 118.100388 samples/sec	 accuracy=71.945043	 loss=1.171228	 lr=0.000013
Epoch[194] Batch [2949]/[3759]	Speed: 117.676347 samples/sec	 accuracy=71.944915	 loss=1.171377	 lr=0.000013
Epoch[194] Batch [2999]/[3759]	Speed: 118.134728 samples/sec	 accuracy=71.959375	 loss=1.170788	 lr=0.000013
Epoch[194] Batch [3049]/[3759]	Speed: 117.222042 samples/sec	 accuracy=71.955430	 loss=1.171056	 lr=0.000013
Epoch[194] Batch [3099]/[3759]	Speed: 117.014220 samples/sec	 accuracy=71.965222	 loss=1.170454	 lr=0.000012
Epoch[194] Batch [3149]/[3759]	Speed: 117.731928 samples/sec	 accuracy=71.963294	 loss=1.170495	 lr=0.000012
Epoch[194] Batch [3199]/[3759]	Speed: 115.796894 samples/sec	 accuracy=71.951172	 loss=1.170797	 lr=0.000012
Epoch[194] Batch [3249]/[3759]	Speed: 118.022351 samples/sec	 accuracy=71.968750	 loss=1.170318	 lr=0.000011
Epoch[194] Batch [3299]/[3759]	Speed: 117.262480 samples/sec	 accuracy=71.983902	 loss=1.170467	 lr=0.000011
Epoch[194] Batch [3349]/[3759]	Speed: 117.243222 samples/sec	 accuracy=71.979011	 loss=1.171004	 lr=0.000011
Epoch[194] Batch [3399]/[3759]	Speed: 117.452405 samples/sec	 accuracy=71.996324	 loss=1.170551	 lr=0.000011
Epoch[194] Batch [3449]/[3759]	Speed: 117.120007 samples/sec	 accuracy=71.991395	 loss=1.170686	 lr=0.000010
Epoch[194] Batch [3499]/[3759]	Speed: 117.393973 samples/sec	 accuracy=71.977232	 loss=1.171331	 lr=0.000010
Epoch[194] Batch [3549]/[3759]	Speed: 114.585676 samples/sec	 accuracy=71.975792	 loss=1.171478	 lr=0.000010
Epoch[194] Batch [3599]/[3759]	Speed: 117.302059 samples/sec	 accuracy=71.976128	 loss=1.171409	 lr=0.000010
Epoch[194] Batch [3649]/[3759]	Speed: 118.049368 samples/sec	 accuracy=71.985445	 loss=1.170842	 lr=0.000009
Epoch[194] Batch [3699]/[3759]	Speed: 117.354715 samples/sec	 accuracy=71.992821	 loss=1.170490	 lr=0.000009
Epoch[194] Batch [3749]/[3759]	Speed: 120.339195 samples/sec	 accuracy=71.974583	 loss=1.170986	 lr=0.000009
Batch [0049]/[0303]: acc-top1=61.656250 acc-top5=81.125000
Batch [0099]/[0303]: acc-top1=58.921875 acc-top5=80.687500
Batch [0149]/[0303]: acc-top1=60.281250 acc-top5=81.427083
Batch [0199]/[0303]: acc-top1=63.015625 acc-top5=82.968750
Batch [0249]/[0303]: acc-top1=62.412500 acc-top5=82.668750
Batch [0299]/[0303]: acc-top1=61.942708 acc-top5=82.598958
[Epoch 194] training: accuracy=71.971851	 loss=1.171014
[Epoch 194] speed: 116 samples/sec	time cost: 2174.293273
[Epoch 194] validation: acc-top1=61.876031 acc-top5=82.632013 loss=1.773839
Epoch[195] Batch [0049]/[3760]	Speed: 86.529962 samples/sec	 accuracy=72.218750	 loss=1.167758	 lr=0.000009
Epoch[195] Batch [0099]/[3760]	Speed: 118.137047 samples/sec	 accuracy=72.093750	 loss=1.172270	 lr=0.000008
Epoch[195] Batch [0149]/[3760]	Speed: 117.755141 samples/sec	 accuracy=72.354167	 loss=1.162029	 lr=0.000008
Epoch[195] Batch [0199]/[3760]	Speed: 117.531713 samples/sec	 accuracy=72.273438	 loss=1.165299	 lr=0.000008
Epoch[195] Batch [0249]/[3760]	Speed: 117.008411 samples/sec	 accuracy=72.381250	 loss=1.162976	 lr=0.000008
Epoch[195] Batch [0299]/[3760]	Speed: 117.447947 samples/sec	 accuracy=72.208333	 loss=1.168528	 lr=0.000007
Epoch[195] Batch [0349]/[3760]	Speed: 117.406728 samples/sec	 accuracy=72.156250	 loss=1.171128	 lr=0.000007
Epoch[195] Batch [0399]/[3760]	Speed: 117.792764 samples/sec	 accuracy=72.234375	 loss=1.168748	 lr=0.000007
Epoch[195] Batch [0449]/[3760]	Speed: 117.132396 samples/sec	 accuracy=72.270833	 loss=1.169595	 lr=0.000007
Epoch[195] Batch [0499]/[3760]	Speed: 117.201648 samples/sec	 accuracy=72.293750	 loss=1.165297	 lr=0.000007
Epoch[195] Batch [0549]/[3760]	Speed: 117.864655 samples/sec	 accuracy=72.215909	 loss=1.169514	 lr=0.000006
Epoch[195] Batch [0599]/[3760]	Speed: 117.338099 samples/sec	 accuracy=72.223958	 loss=1.166097	 lr=0.000006
Epoch[195] Batch [0649]/[3760]	Speed: 116.711553 samples/sec	 accuracy=72.276442	 loss=1.164602	 lr=0.000006
Epoch[195] Batch [0699]/[3760]	Speed: 117.950558 samples/sec	 accuracy=72.229911	 loss=1.166589	 lr=0.000006
Epoch[195] Batch [0749]/[3760]	Speed: 117.167449 samples/sec	 accuracy=72.202083	 loss=1.166267	 lr=0.000006
Epoch[195] Batch [0799]/[3760]	Speed: 117.319180 samples/sec	 accuracy=72.218750	 loss=1.168314	 lr=0.000005
Epoch[195] Batch [0849]/[3760]	Speed: 117.809873 samples/sec	 accuracy=72.218750	 loss=1.169406	 lr=0.000005
Epoch[195] Batch [0899]/[3760]	Speed: 117.428933 samples/sec	 accuracy=72.244792	 loss=1.170084	 lr=0.000005
Epoch[195] Batch [0949]/[3760]	Speed: 117.768434 samples/sec	 accuracy=72.276316	 loss=1.169186	 lr=0.000005
Epoch[195] Batch [0999]/[3760]	Speed: 117.071279 samples/sec	 accuracy=72.275000	 loss=1.168662	 lr=0.000005
Epoch[195] Batch [1049]/[3760]	Speed: 117.387863 samples/sec	 accuracy=72.254464	 loss=1.168636	 lr=0.000004
Epoch[195] Batch [1099]/[3760]	Speed: 117.610276 samples/sec	 accuracy=72.265625	 loss=1.168408	 lr=0.000004
Epoch[195] Batch [1149]/[3760]	Speed: 117.781411 samples/sec	 accuracy=72.260870	 loss=1.168051	 lr=0.000004
Epoch[195] Batch [1199]/[3760]	Speed: 117.831806 samples/sec	 accuracy=72.255208	 loss=1.168382	 lr=0.000004
Epoch[195] Batch [1249]/[3760]	Speed: 117.572081 samples/sec	 accuracy=72.236250	 loss=1.170087	 lr=0.000004
Epoch[195] Batch [1299]/[3760]	Speed: 117.519356 samples/sec	 accuracy=72.204327	 loss=1.170704	 lr=0.000004
Epoch[195] Batch [1349]/[3760]	Speed: 117.601525 samples/sec	 accuracy=72.185185	 loss=1.170492	 lr=0.000003
Epoch[195] Batch [1399]/[3760]	Speed: 116.449154 samples/sec	 accuracy=72.187500	 loss=1.170190	 lr=0.000003
Epoch[195] Batch [1449]/[3760]	Speed: 114.572448 samples/sec	 accuracy=72.179957	 loss=1.169575	 lr=0.000003
Epoch[195] Batch [1499]/[3760]	Speed: 117.210733 samples/sec	 accuracy=72.172917	 loss=1.169143	 lr=0.000003
Epoch[195] Batch [1549]/[3760]	Speed: 117.714092 samples/sec	 accuracy=72.159274	 loss=1.168575	 lr=0.000003
Epoch[195] Batch [1599]/[3760]	Speed: 117.748743 samples/sec	 accuracy=72.157227	 loss=1.167325	 lr=0.000003
Epoch[195] Batch [1649]/[3760]	Speed: 117.677247 samples/sec	 accuracy=72.109848	 loss=1.168487	 lr=0.000003
Epoch[195] Batch [1699]/[3760]	Speed: 116.777724 samples/sec	 accuracy=72.092831	 loss=1.169940	 lr=0.000002
Epoch[195] Batch [1749]/[3760]	Speed: 117.283294 samples/sec	 accuracy=72.098214	 loss=1.169101	 lr=0.000002
Epoch[195] Batch [1799]/[3760]	Speed: 117.695938 samples/sec	 accuracy=72.096354	 loss=1.169380	 lr=0.000002
Epoch[195] Batch [1849]/[3760]	Speed: 117.545767 samples/sec	 accuracy=72.116554	 loss=1.169473	 lr=0.000002
Epoch[195] Batch [1899]/[3760]	Speed: 117.366145 samples/sec	 accuracy=72.098684	 loss=1.170031	 lr=0.000002
Epoch[195] Batch [1949]/[3760]	Speed: 117.926088 samples/sec	 accuracy=72.092147	 loss=1.170942	 lr=0.000002
Epoch[195] Batch [1999]/[3760]	Speed: 117.516905 samples/sec	 accuracy=72.072656	 loss=1.171210	 lr=0.000002
Epoch[195] Batch [2049]/[3760]	Speed: 117.214476 samples/sec	 accuracy=72.073171	 loss=1.171453	 lr=0.000002
Epoch[195] Batch [2099]/[3760]	Speed: 117.905478 samples/sec	 accuracy=72.070685	 loss=1.171797	 lr=0.000002
Epoch[195] Batch [2149]/[3760]	Speed: 116.804921 samples/sec	 accuracy=72.091570	 loss=1.171075	 lr=0.000001
Epoch[195] Batch [2199]/[3760]	Speed: 117.624635 samples/sec	 accuracy=72.110085	 loss=1.171332	 lr=0.000001
Epoch[195] Batch [2249]/[3760]	Speed: 117.406248 samples/sec	 accuracy=72.101389	 loss=1.171816	 lr=0.000001
Epoch[195] Batch [2299]/[3760]	Speed: 117.721944 samples/sec	 accuracy=72.101223	 loss=1.171303	 lr=0.000001
Epoch[195] Batch [2349]/[3760]	Speed: 116.975173 samples/sec	 accuracy=72.071809	 loss=1.172092	 lr=0.000001
Epoch[195] Batch [2399]/[3760]	Speed: 117.656004 samples/sec	 accuracy=72.079427	 loss=1.171699	 lr=0.000001
Epoch[195] Batch [2449]/[3760]	Speed: 117.016069 samples/sec	 accuracy=72.068240	 loss=1.171747	 lr=0.000001
Epoch[195] Batch [2499]/[3760]	Speed: 116.876487 samples/sec	 accuracy=72.080625	 loss=1.171235	 lr=0.000001
Epoch[195] Batch [2549]/[3760]	Speed: 116.325130 samples/sec	 accuracy=72.075980	 loss=1.171170	 lr=0.000001
Epoch[195] Batch [2599]/[3760]	Speed: 117.020107 samples/sec	 accuracy=72.074519	 loss=1.171267	 lr=0.000001
Epoch[195] Batch [2649]/[3760]	Speed: 116.586616 samples/sec	 accuracy=72.066627	 loss=1.171952	 lr=0.000001
Epoch[195] Batch [2699]/[3760]	Speed: 116.855746 samples/sec	 accuracy=72.093171	 loss=1.171187	 lr=0.000001
Epoch[195] Batch [2749]/[3760]	Speed: 117.261750 samples/sec	 accuracy=72.086932	 loss=1.171119	 lr=0.000001
Epoch[195] Batch [2799]/[3760]	Speed: 117.086707 samples/sec	 accuracy=72.087612	 loss=1.170899	 lr=0.000000
Epoch[195] Batch [2849]/[3760]	Speed: 117.567770 samples/sec	 accuracy=72.069627	 loss=1.171303	 lr=0.000000
Epoch[195] Batch [2899]/[3760]	Speed: 117.337819 samples/sec	 accuracy=72.039871	 loss=1.172012	 lr=0.000000
Epoch[195] Batch [2949]/[3760]	Speed: 117.307939 samples/sec	 accuracy=72.043432	 loss=1.171940	 lr=0.000000
Epoch[195] Batch [2999]/[3760]	Speed: 116.957445 samples/sec	 accuracy=72.051042	 loss=1.171507	 lr=0.000000
Epoch[195] Batch [3049]/[3760]	Speed: 117.106341 samples/sec	 accuracy=72.068648	 loss=1.170693	 lr=0.000000
Epoch[195] Batch [3099]/[3760]	Speed: 117.397166 samples/sec	 accuracy=72.050403	 loss=1.171355	 lr=0.000000
Epoch[195] Batch [3149]/[3760]	Speed: 114.620479 samples/sec	 accuracy=72.039187	 loss=1.172157	 lr=0.000000
Epoch[195] Batch [3199]/[3760]	Speed: 116.894661 samples/sec	 accuracy=72.051758	 loss=1.172086	 lr=0.000000
Epoch[195] Batch [3249]/[3760]	Speed: 117.826622 samples/sec	 accuracy=72.048558	 loss=1.172383	 lr=0.000000
Epoch[195] Batch [3299]/[3760]	Speed: 117.157209 samples/sec	 accuracy=72.045928	 loss=1.172613	 lr=0.000000
Epoch[195] Batch [3349]/[3760]	Speed: 116.814469 samples/sec	 accuracy=72.059701	 loss=1.171789	 lr=0.000000
Epoch[195] Batch [3399]/[3760]	Speed: 117.729522 samples/sec	 accuracy=72.086857	 loss=1.170244	 lr=0.000000
Epoch[195] Batch [3449]/[3760]	Speed: 116.573897 samples/sec	 accuracy=72.096920	 loss=1.169342	 lr=0.000000
Epoch[195] Batch [3499]/[3760]	Speed: 117.904211 samples/sec	 accuracy=72.085714	 loss=1.169757	 lr=0.000000
Epoch[195] Batch [3549]/[3760]	Speed: 117.370865 samples/sec	 accuracy=72.065141	 loss=1.170367	 lr=0.000000
Epoch[195] Batch [3599]/[3760]	Speed: 117.719684 samples/sec	 accuracy=72.053385	 loss=1.170922	 lr=0.000000
Epoch[195] Batch [3649]/[3760]	Speed: 116.642633 samples/sec	 accuracy=72.062928	 loss=1.170800	 lr=0.000000
Epoch[195] Batch [3699]/[3760]	Speed: 117.633284 samples/sec	 accuracy=72.070524	 loss=1.170093	 lr=0.000000
Epoch[195] Batch [3749]/[3760]	Speed: 120.754018 samples/sec	 accuracy=72.073750	 loss=1.169812	 lr=0.000000
Batch [0049]/[0303]: acc-top1=62.156250 acc-top5=81.156250
Batch [0099]/[0303]: acc-top1=59.437500 acc-top5=80.484375
Batch [0149]/[0303]: acc-top1=60.562500 acc-top5=81.156250
Batch [0199]/[0303]: acc-top1=63.218750 acc-top5=82.898438
Batch [0249]/[0303]: acc-top1=62.743750 acc-top5=82.637500
Batch [0299]/[0303]: acc-top1=62.166667 acc-top5=82.619792
[Epoch 195] training: accuracy=72.071559	 loss=1.170077
[Epoch 195] speed: 116 samples/sec	time cost: 2177.198365
[Epoch 195] validation: acc-top1=62.113243 acc-top5=82.652640 loss=1.779143