AMP: true
AUG:
  AUTO_AUGMENT: rand-m9-mstd0.5-inc1
  COLOR_JITTER: 0.4
  CUTMIX: 1.0
  CUTMIX_MINMAX: null
  MESA: 1.0
  MIXUP: 0.8
  MIXUP_MODE: batch
  MIXUP_PROB: 1.0
  MIXUP_SWITCH_PROB: 0.5
  RECOUNT: 1
  REMODE: pixel
  REPROB: 0.25
BASE:
- ''
DATA:
  BATCH_SIZE: 256
  CACHE_MODE: part
  DATASET: imagenet
  DATA_PATH: /data/imagenet
  IMG_SIZE: 256
  INTERPOLATION: bicubic
  NUM_WORKERS: 16
  PIN_MEMORY: true
  ZIP_MODE: false
EVAL_MODE: false
LOCAL_RANK: 0
MODEL:
  DROP_PATH_RATE: 0.2
  DROP_RATE: 0.0
  LABEL_SMOOTHING: 0.1
  MLLA:
    APE: false
    DEPTHS:
    - 2
    - 4
    - 8
    - 4
    EMBED_DIM: 48
    EXPANSION_RATIO: 2.5
    IN_CHANS: 3
    MLP_RATIO: 4.0
    NUM_HEADS:
    - 2
    - 4
    - 8
    - 16
    PATCH_SIZE: 4
    QKV_BIAS: true
  NAME: mlla_mini
  NUM_CLASSES: 1000
  RESUME: ''
  TYPE: mlla
OUTPUT: output/300e/mlla_mini/recconv_5x5_nearest_interp
PRINT_FREQ: 100
SAVE_FREQ: 1
SEED: 0
TAG: recconv_5x5_nearest_interp
TEST:
  CROP: true
THROUGHPUT_MODE: false
TRAIN:
  AUTO_RESUME: true
  BASE_LR: 0.001
  CLIP_GRAD: 5.0
  COOLDOWN_EPOCHS: 0
  EPOCHS: 300
  LR_SCHEDULER:
    DECAY_EPOCHS: 30
    DECAY_RATE: 0.1
    NAME: cosine
  MIN_LR: 1.0e-05
  OPTIMIZER:
    BETAS:
    - 0.9
    - 0.999
    EPS: 1.0e-08
    MOMENTUM: 0.9
    NAME: adamw
  START_EPOCH: 0
  USE_CHECKPOINT: false
  WARMUP_EPOCHS: 20
  WARMUP_LR: 1.0e-06
  WEIGHT_DECAY: 0.05

Creating model:mlla/mlla_mini
MLLA(
  (patch_embed): Stem(
    (conv1): ConvLayer(
      (conv): Conv2d(3, 24, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (norm): BatchNorm2d(24, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (act): ReLU()
    )
    (conv2): Sequential(
      (0): ConvLayer(
        (conv): Conv2d(24, 24, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (norm): BatchNorm2d(24, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (act): ReLU()
      )
      (1): ConvLayer(
        (conv): Conv2d(24, 24, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
        (norm): BatchNorm2d(24, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (conv3): Sequential(
      (0): ConvLayer(
        (conv): Conv2d(24, 192, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
        (norm): BatchNorm2d(192, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (act): ReLU()
      )
      (1): ConvLayer(
        (conv): Conv2d(192, 48, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (norm): BatchNorm2d(48, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
  )
  (pos_drop): Dropout(p=0.0, inplace=False)
  (layers): ModuleList(
    (0): BasicLayer(
      dim=48, level=4, depth=2
      (blocks): ModuleList(
        (0): MLLABlock(
          dim=48, mlp_ratio=4.0, level=4, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(48, 48, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=48)
          (norm1): LayerNorm((48,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=48, out_features=120, bias=True)
          (o_proj): Conv2d(60, 48, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(60, 60, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=60, bias=False)
            (convs): ModuleList(
              (0-4): 5 x Conv2d(60, 60, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=60, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=48)
          (norm2): LayerNorm((48,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=48, out_features=192, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=192, out_features=48, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): Identity()
        )
        (1): MLLABlock(
          dim=48, mlp_ratio=4.0, level=4, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(48, 48, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=48)
          (norm1): LayerNorm((48,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=48, out_features=120, bias=True)
          (o_proj): Conv2d(60, 48, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(60, 60, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=60, bias=False)
            (convs): ModuleList(
              (0-4): 5 x Conv2d(60, 60, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=60, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=48)
          (norm2): LayerNorm((48,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=48, out_features=192, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=192, out_features=48, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): DropPath(drop_prob=0.012)
        )
      )
      (downsample): MLLABlock(
        dim=48, mlp_ratio=4.0, level=3, downsample=True, expansion_ratio=2.5
        (cpe1): Conv2d(48, 96, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=48)
        (norm1): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
        (i_proj): Linear(in_features=96, out_features=240, bias=True)
        (o_proj): Conv2d(120, 96, kernel_size=(1, 1), stride=(1, 1))
        (act): GELU(approximate='none')
        (agg): RecConv2d(
          (down): Conv2d(120, 120, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=120, bias=False)
          (convs): ModuleList(
            (0-3): 4 x Conv2d(120, 120, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=120, bias=False)
          )
          (up): Upsample(scale_factor=2.0, mode='nearest')
        )
        (cpe2): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=96)
        (norm2): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=96, out_features=384, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=384, out_features=96, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
        (drop_path): DropPath(drop_prob=0.012)
      )
    )
    (1): BasicLayer(
      dim=96, level=3, depth=4
      (blocks): ModuleList(
        (0): MLLABlock(
          dim=96, mlp_ratio=4.0, level=3, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(96, 96, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=96)
          (norm1): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=96, out_features=240, bias=True)
          (o_proj): Conv2d(120, 96, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(120, 120, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=120, bias=False)
            (convs): ModuleList(
              (0-3): 4 x Conv2d(120, 120, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=120, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=96)
          (norm2): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=96, out_features=384, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=384, out_features=96, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): DropPath(drop_prob=0.024)
        )
        (1): MLLABlock(
          dim=96, mlp_ratio=4.0, level=3, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(96, 96, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=96)
          (norm1): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=96, out_features=240, bias=True)
          (o_proj): Conv2d(120, 96, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(120, 120, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=120, bias=False)
            (convs): ModuleList(
              (0-3): 4 x Conv2d(120, 120, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=120, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=96)
          (norm2): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=96, out_features=384, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=384, out_features=96, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): DropPath(drop_prob=0.035)
        )
        (2): MLLABlock(
          dim=96, mlp_ratio=4.0, level=3, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(96, 96, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=96)
          (norm1): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=96, out_features=240, bias=True)
          (o_proj): Conv2d(120, 96, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(120, 120, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=120, bias=False)
            (convs): ModuleList(
              (0-3): 4 x Conv2d(120, 120, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=120, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=96)
          (norm2): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=96, out_features=384, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=384, out_features=96, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): DropPath(drop_prob=0.047)
        )
        (3): MLLABlock(
          dim=96, mlp_ratio=4.0, level=3, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(96, 96, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=96)
          (norm1): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=96, out_features=240, bias=True)
          (o_proj): Conv2d(120, 96, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(120, 120, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=120, bias=False)
            (convs): ModuleList(
              (0-3): 4 x Conv2d(120, 120, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=120, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=96)
          (norm2): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=96, out_features=384, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=384, out_features=96, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): DropPath(drop_prob=0.059)
        )
      )
      (downsample): MLLABlock(
        dim=96, mlp_ratio=4.0, level=2, downsample=True, expansion_ratio=2.5
        (cpe1): Conv2d(96, 192, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=96)
        (norm1): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
        (i_proj): Linear(in_features=192, out_features=480, bias=True)
        (o_proj): Conv2d(240, 192, kernel_size=(1, 1), stride=(1, 1))
        (act): GELU(approximate='none')
        (agg): RecConv2d(
          (down): Conv2d(240, 240, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=240, bias=False)
          (convs): ModuleList(
            (0-2): 3 x Conv2d(240, 240, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=240, bias=False)
          )
          (up): Upsample(scale_factor=2.0, mode='nearest')
        )
        (cpe2): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
        (norm2): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=192, out_features=768, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=768, out_features=192, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
        (drop_path): DropPath(drop_prob=0.059)
      )
    )
    (2): BasicLayer(
      dim=192, level=2, depth=8
      (blocks): ModuleList(
        (0): MLLABlock(
          dim=192, mlp_ratio=4.0, level=2, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(192, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=192)
          (norm1): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=192, out_features=480, bias=True)
          (o_proj): Conv2d(240, 192, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(240, 240, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=240, bias=False)
            (convs): ModuleList(
              (0-2): 3 x Conv2d(240, 240, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=240, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
          (norm2): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=192, out_features=768, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=768, out_features=192, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): DropPath(drop_prob=0.071)
        )
        (1): MLLABlock(
          dim=192, mlp_ratio=4.0, level=2, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(192, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=192)
          (norm1): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=192, out_features=480, bias=True)
          (o_proj): Conv2d(240, 192, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(240, 240, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=240, bias=False)
            (convs): ModuleList(
              (0-2): 3 x Conv2d(240, 240, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=240, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
          (norm2): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=192, out_features=768, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=768, out_features=192, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): DropPath(drop_prob=0.082)
        )
        (2): MLLABlock(
          dim=192, mlp_ratio=4.0, level=2, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(192, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=192)
          (norm1): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=192, out_features=480, bias=True)
          (o_proj): Conv2d(240, 192, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(240, 240, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=240, bias=False)
            (convs): ModuleList(
              (0-2): 3 x Conv2d(240, 240, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=240, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
          (norm2): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=192, out_features=768, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=768, out_features=192, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): DropPath(drop_prob=0.094)
        )
        (3): MLLABlock(
          dim=192, mlp_ratio=4.0, level=2, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(192, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=192)
          (norm1): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=192, out_features=480, bias=True)
          (o_proj): Conv2d(240, 192, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(240, 240, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=240, bias=False)
            (convs): ModuleList(
              (0-2): 3 x Conv2d(240, 240, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=240, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
          (norm2): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=192, out_features=768, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=768, out_features=192, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): DropPath(drop_prob=0.106)
        )
        (4): MLLABlock(
          dim=192, mlp_ratio=4.0, level=2, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(192, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=192)
          (norm1): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=192, out_features=480, bias=True)
          (o_proj): Conv2d(240, 192, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(240, 240, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=240, bias=False)
            (convs): ModuleList(
              (0-2): 3 x Conv2d(240, 240, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=240, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
          (norm2): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=192, out_features=768, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=768, out_features=192, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): DropPath(drop_prob=0.118)
        )
        (5): MLLABlock(
          dim=192, mlp_ratio=4.0, level=2, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(192, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=192)
          (norm1): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=192, out_features=480, bias=True)
          (o_proj): Conv2d(240, 192, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(240, 240, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=240, bias=False)
            (convs): ModuleList(
              (0-2): 3 x Conv2d(240, 240, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=240, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
          (norm2): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=192, out_features=768, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=768, out_features=192, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): DropPath(drop_prob=0.129)
        )
        (6): MLLABlock(
          dim=192, mlp_ratio=4.0, level=2, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(192, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=192)
          (norm1): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=192, out_features=480, bias=True)
          (o_proj): Conv2d(240, 192, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(240, 240, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=240, bias=False)
            (convs): ModuleList(
              (0-2): 3 x Conv2d(240, 240, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=240, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
          (norm2): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=192, out_features=768, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=768, out_features=192, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): DropPath(drop_prob=0.141)
        )
        (7): MLLABlock(
          dim=192, mlp_ratio=4.0, level=2, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(192, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=192)
          (norm1): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=192, out_features=480, bias=True)
          (o_proj): Conv2d(240, 192, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(240, 240, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=240, bias=False)
            (convs): ModuleList(
              (0-2): 3 x Conv2d(240, 240, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=240, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
          (norm2): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=192, out_features=768, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=768, out_features=192, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): DropPath(drop_prob=0.153)
        )
      )
      (downsample): MLLABlock(
        dim=192, mlp_ratio=4.0, level=1, downsample=True, expansion_ratio=2.5
        (cpe1): Conv2d(192, 384, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=192)
        (norm1): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (i_proj): Linear(in_features=384, out_features=960, bias=True)
        (o_proj): Conv2d(480, 384, kernel_size=(1, 1), stride=(1, 1))
        (act): GELU(approximate='none')
        (agg): RecConv2d(
          (down): Conv2d(480, 480, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=480, bias=False)
          (convs): ModuleList(
            (0-1): 2 x Conv2d(480, 480, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=480, bias=False)
          )
          (up): Upsample(scale_factor=2.0, mode='nearest')
        )
        (cpe2): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=384)
        (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=384, out_features=1536, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=1536, out_features=384, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
        (drop_path): DropPath(drop_prob=0.153)
      )
    )
    (3): BasicLayer(
      dim=384, level=1, depth=4
      (blocks): ModuleList(
        (0): MLLABlock(
          dim=384, mlp_ratio=4.0, level=1, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(384, 384, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=384)
          (norm1): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=384, out_features=960, bias=True)
          (o_proj): Conv2d(480, 384, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(480, 480, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=480, bias=False)
            (convs): ModuleList(
              (0-1): 2 x Conv2d(480, 480, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=480, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=384)
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=1536, out_features=384, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): DropPath(drop_prob=0.165)
        )
        (1): MLLABlock(
          dim=384, mlp_ratio=4.0, level=1, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(384, 384, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=384)
          (norm1): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=384, out_features=960, bias=True)
          (o_proj): Conv2d(480, 384, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(480, 480, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=480, bias=False)
            (convs): ModuleList(
              (0-1): 2 x Conv2d(480, 480, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=480, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=384)
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=1536, out_features=384, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): DropPath(drop_prob=0.176)
        )
        (2): MLLABlock(
          dim=384, mlp_ratio=4.0, level=1, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(384, 384, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=384)
          (norm1): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=384, out_features=960, bias=True)
          (o_proj): Conv2d(480, 384, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(480, 480, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=480, bias=False)
            (convs): ModuleList(
              (0-1): 2 x Conv2d(480, 480, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=480, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=384)
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=1536, out_features=384, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): DropPath(drop_prob=0.188)
        )
        (3): MLLABlock(
          dim=384, mlp_ratio=4.0, level=1, downsample=False, expansion_ratio=2.5
          (cpe1): Conv2d(384, 384, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=384)
          (norm1): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (i_proj): Linear(in_features=384, out_features=960, bias=True)
          (o_proj): Conv2d(480, 384, kernel_size=(1, 1), stride=(1, 1))
          (act): GELU(approximate='none')
          (agg): RecConv2d(
            (down): Conv2d(480, 480, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=480, bias=False)
            (convs): ModuleList(
              (0-1): 2 x Conv2d(480, 480, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=480, bias=False)
            )
            (up): Upsample(scale_factor=2.0, mode='nearest')
          )
          (cpe2): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=384)
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=1536, out_features=384, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
          (drop_path): DropPath(drop_prob=0.200)
        )
      )
      (downsample): Identity()
    )
  )
  (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
  (head): Linear(in_features=384, out_features=1000, bias=True)
)

EPOCH 1
 * Acc@1 0.538 Acc@5 2.194
Accuracy of the network on the 50000 test images: 0.5%
 * Acc@1 0.102 Acc@5 0.506
Accuracy of the ema network on the 50000 test images: 0.1%
--
EPOCH 2
 * Acc@1 3.434 Acc@5 11.224
Accuracy of the network on the 50000 test images: 3.4%
 * Acc@1 0.092 Acc@5 0.496
Accuracy of the ema network on the 50000 test images: 0.1%
--
EPOCH 3
 * Acc@1 9.746 Acc@5 24.846
Accuracy of the network on the 50000 test images: 9.7%
 * Acc@1 0.100 Acc@5 0.528
Accuracy of the ema network on the 50000 test images: 0.1%
--
EPOCH 4
 * Acc@1 17.308 Acc@5 38.096
Accuracy of the network on the 50000 test images: 17.3%
 * Acc@1 0.138 Acc@5 0.552
Accuracy of the ema network on the 50000 test images: 0.1%
--
EPOCH 5
 * Acc@1 24.316 Acc@5 47.986
Accuracy of the network on the 50000 test images: 24.3%
 * Acc@1 0.134 Acc@5 0.638
Accuracy of the ema network on the 50000 test images: 0.1%
--
EPOCH 6
 * Acc@1 31.850 Acc@5 57.468
Accuracy of the network on the 50000 test images: 31.9%
 * Acc@1 0.128 Acc@5 0.688
Accuracy of the ema network on the 50000 test images: 0.1%
--
EPOCH 7
 * Acc@1 37.986 Acc@5 63.916
Accuracy of the network on the 50000 test images: 38.0%
 * Acc@1 0.128 Acc@5 0.676
Accuracy of the ema network on the 50000 test images: 0.1%
--
EPOCH 8
 * Acc@1 42.442 Acc@5 68.390
Accuracy of the network on the 50000 test images: 42.4%
 * Acc@1 0.176 Acc@5 0.824
Accuracy of the ema network on the 50000 test images: 0.2%
--
EPOCH 9
 * Acc@1 47.194 Acc@5 73.098
Accuracy of the network on the 50000 test images: 47.2%
 * Acc@1 0.216 Acc@5 1.028
Accuracy of the ema network on the 50000 test images: 0.2%
--
EPOCH 10
 * Acc@1 50.500 Acc@5 75.884
Accuracy of the network on the 50000 test images: 50.5%
 * Acc@1 0.226 Acc@5 1.216
Accuracy of the ema network on the 50000 test images: 0.2%
--
EPOCH 11
 * Acc@1 53.242 Acc@5 78.324
Accuracy of the network on the 50000 test images: 53.2%
 * Acc@1 0.256 Acc@5 1.444
Accuracy of the ema network on the 50000 test images: 0.3%
--
EPOCH 12
 * Acc@1 54.240 Acc@5 79.098
Accuracy of the network on the 50000 test images: 54.2%
 * Acc@1 0.404 Acc@5 1.802
Accuracy of the ema network on the 50000 test images: 0.4%
--
EPOCH 13
 * Acc@1 56.962 Acc@5 81.124
Accuracy of the network on the 50000 test images: 57.0%
 * Acc@1 0.464 Acc@5 2.340
Accuracy of the ema network on the 50000 test images: 0.5%
--
EPOCH 14
 * Acc@1 58.122 Acc@5 81.982
Accuracy of the network on the 50000 test images: 58.1%
 * Acc@1 0.764 Acc@5 3.728
Accuracy of the ema network on the 50000 test images: 0.8%
--
EPOCH 15
 * Acc@1 59.928 Acc@5 83.450
Accuracy of the network on the 50000 test images: 59.9%
 * Acc@1 1.794 Acc@5 6.880
Accuracy of the ema network on the 50000 test images: 1.8%
--
EPOCH 16
 * Acc@1 61.212 Acc@5 84.276
Accuracy of the network on the 50000 test images: 61.2%
 * Acc@1 4.260 Acc@5 13.110
Accuracy of the ema network on the 50000 test images: 4.3%
--
EPOCH 17
 * Acc@1 62.294 Acc@5 85.094
Accuracy of the network on the 50000 test images: 62.3%
 * Acc@1 8.424 Acc@5 21.732
Accuracy of the ema network on the 50000 test images: 8.4%
--
EPOCH 18
 * Acc@1 62.344 Acc@5 85.310
Accuracy of the network on the 50000 test images: 62.3%
 * Acc@1 13.646 Acc@5 31.144
Accuracy of the ema network on the 50000 test images: 13.6%
--
EPOCH 19
 * Acc@1 63.544 Acc@5 85.986
Accuracy of the network on the 50000 test images: 63.5%
 * Acc@1 19.540 Acc@5 40.278
Accuracy of the ema network on the 50000 test images: 19.5%
--
EPOCH 20
 * Acc@1 63.794 Acc@5 86.446
Accuracy of the network on the 50000 test images: 63.8%
 * Acc@1 25.406 Acc@5 48.494
Accuracy of the ema network on the 50000 test images: 25.4%
--
EPOCH 21
 * Acc@1 64.566 Acc@5 86.764
Accuracy of the network on the 50000 test images: 64.6%
 * Acc@1 31.062 Acc@5 55.658
Accuracy of the ema network on the 50000 test images: 31.1%
--
EPOCH 22
 * Acc@1 66.216 Acc@5 87.770
Accuracy of the network on the 50000 test images: 66.2%
 * Acc@1 36.450 Acc@5 61.960
Accuracy of the ema network on the 50000 test images: 36.5%
--
EPOCH 23
 * Acc@1 66.570 Acc@5 88.034
Accuracy of the network on the 50000 test images: 66.6%
 * Acc@1 41.196 Acc@5 67.136
Accuracy of the ema network on the 50000 test images: 41.2%
--
EPOCH 24
 * Acc@1 67.092 Acc@5 88.148
Accuracy of the network on the 50000 test images: 67.1%
 * Acc@1 45.560 Acc@5 71.294
Accuracy of the ema network on the 50000 test images: 45.6%
--
EPOCH 25
 * Acc@1 67.714 Acc@5 88.706
Accuracy of the network on the 50000 test images: 67.7%
 * Acc@1 49.302 Acc@5 74.790
Accuracy of the ema network on the 50000 test images: 49.3%
--
EPOCH 26
 * Acc@1 68.096 Acc@5 88.692
Accuracy of the network on the 50000 test images: 68.1%
 * Acc@1 52.658 Acc@5 77.706
Accuracy of the ema network on the 50000 test images: 52.7%
--
EPOCH 27
 * Acc@1 68.478 Acc@5 89.120
Accuracy of the network on the 50000 test images: 68.5%
 * Acc@1 55.626 Acc@5 80.178
Accuracy of the ema network on the 50000 test images: 55.6%
--
EPOCH 28
 * Acc@1 68.904 Acc@5 89.318
Accuracy of the network on the 50000 test images: 68.9%
 * Acc@1 58.326 Acc@5 82.174
Accuracy of the ema network on the 50000 test images: 58.3%
--
EPOCH 29
 * Acc@1 69.148 Acc@5 89.426
Accuracy of the network on the 50000 test images: 69.1%
 * Acc@1 60.580 Acc@5 83.758
Accuracy of the ema network on the 50000 test images: 60.6%
--
EPOCH 30
 * Acc@1 69.230 Acc@5 89.714
Accuracy of the network on the 50000 test images: 69.2%
 * Acc@1 62.418 Acc@5 85.128
Accuracy of the ema network on the 50000 test images: 62.4%
--
EPOCH 31
 * Acc@1 69.996 Acc@5 90.156
Accuracy of the network on the 50000 test images: 70.0%
 * Acc@1 64.038 Acc@5 86.220
Accuracy of the ema network on the 50000 test images: 64.0%
--
EPOCH 32
 * Acc@1 70.184 Acc@5 90.148
Accuracy of the network on the 50000 test images: 70.2%
 * Acc@1 65.502 Acc@5 87.246
Accuracy of the ema network on the 50000 test images: 65.5%
--
EPOCH 33
 * Acc@1 70.600 Acc@5 90.470
Accuracy of the network on the 50000 test images: 70.6%
 * Acc@1 66.750 Acc@5 88.054
Accuracy of the ema network on the 50000 test images: 66.8%
--
EPOCH 34
 * Acc@1 70.902 Acc@5 90.582
Accuracy of the network on the 50000 test images: 70.9%
 * Acc@1 67.870 Acc@5 88.734
Accuracy of the ema network on the 50000 test images: 67.9%
--
EPOCH 35
 * Acc@1 70.700 Acc@5 90.576
Accuracy of the network on the 50000 test images: 70.7%
 * Acc@1 68.806 Acc@5 89.252
Accuracy of the ema network on the 50000 test images: 68.8%
--
EPOCH 36
 * Acc@1 70.934 Acc@5 90.594
Accuracy of the network on the 50000 test images: 70.9%
 * Acc@1 69.618 Acc@5 89.754
Accuracy of the ema network on the 50000 test images: 69.6%
--
EPOCH 37
 * Acc@1 71.766 Acc@5 90.954
Accuracy of the network on the 50000 test images: 71.8%
 * Acc@1 70.320 Acc@5 90.158
Accuracy of the ema network on the 50000 test images: 70.3%
--
EPOCH 38
 * Acc@1 70.982 Acc@5 90.682
Accuracy of the network on the 50000 test images: 71.0%
 * Acc@1 71.028 Acc@5 90.516
Accuracy of the ema network on the 50000 test images: 71.0%
--
EPOCH 39
 * Acc@1 71.532 Acc@5 91.000
Accuracy of the network on the 50000 test images: 71.5%
 * Acc@1 71.528 Acc@5 90.816
Accuracy of the ema network on the 50000 test images: 71.5%
--
EPOCH 40
 * Acc@1 72.132 Acc@5 91.208
Accuracy of the network on the 50000 test images: 72.1%
 * Acc@1 71.996 Acc@5 91.072
Accuracy of the ema network on the 50000 test images: 72.0%
--
EPOCH 41
 * Acc@1 71.936 Acc@5 91.078
Accuracy of the network on the 50000 test images: 71.9%
 * Acc@1 72.496 Acc@5 91.280
Accuracy of the ema network on the 50000 test images: 72.5%
--
EPOCH 42
 * Acc@1 72.394 Acc@5 91.386
Accuracy of the network on the 50000 test images: 72.4%
 * Acc@1 72.942 Acc@5 91.544
Accuracy of the ema network on the 50000 test images: 72.9%
--
EPOCH 43
 * Acc@1 72.408 Acc@5 91.290
Accuracy of the network on the 50000 test images: 72.4%
 * Acc@1 73.232 Acc@5 91.750
Accuracy of the ema network on the 50000 test images: 73.2%
--
EPOCH 44
 * Acc@1 71.854 Acc@5 91.148
Accuracy of the network on the 50000 test images: 71.9%
 * Acc@1 73.578 Acc@5 91.908
Accuracy of the ema network on the 50000 test images: 73.6%
--
EPOCH 45
 * Acc@1 72.610 Acc@5 91.518
Accuracy of the network on the 50000 test images: 72.6%
 * Acc@1 73.852 Acc@5 92.072
Accuracy of the ema network on the 50000 test images: 73.9%
--
EPOCH 46
 * Acc@1 72.786 Acc@5 91.526
Accuracy of the network on the 50000 test images: 72.8%
 * Acc@1 74.126 Acc@5 92.206
Accuracy of the ema network on the 50000 test images: 74.1%
--
EPOCH 47
 * Acc@1 72.798 Acc@5 91.716
Accuracy of the network on the 50000 test images: 72.8%
 * Acc@1 74.394 Acc@5 92.312
Accuracy of the ema network on the 50000 test images: 74.4%
--
EPOCH 48
 * Acc@1 72.814 Acc@5 91.494
Accuracy of the network on the 50000 test images: 72.8%
 * Acc@1 74.582 Acc@5 92.430
Accuracy of the ema network on the 50000 test images: 74.6%
--
EPOCH 49
 * Acc@1 73.094 Acc@5 91.820
Accuracy of the network on the 50000 test images: 73.1%
 * Acc@1 74.812 Acc@5 92.512
Accuracy of the ema network on the 50000 test images: 74.8%
--
EPOCH 50
 * Acc@1 72.938 Acc@5 91.786
Accuracy of the network on the 50000 test images: 72.9%
 * Acc@1 74.982 Acc@5 92.612
Accuracy of the ema network on the 50000 test images: 75.0%
--
EPOCH 51
 * Acc@1 73.090 Acc@5 91.762
Accuracy of the network on the 50000 test images: 73.1%
 * Acc@1 75.196 Acc@5 92.712
Accuracy of the ema network on the 50000 test images: 75.2%
--
EPOCH 52
 * Acc@1 73.062 Acc@5 91.750
Accuracy of the network on the 50000 test images: 73.1%
 * Acc@1 75.404 Acc@5 92.820
Accuracy of the ema network on the 50000 test images: 75.4%
--
EPOCH 53
 * Acc@1 73.618 Acc@5 92.028
Accuracy of the network on the 50000 test images: 73.6%
 * Acc@1 75.514 Acc@5 92.866
Accuracy of the ema network on the 50000 test images: 75.5%
--
EPOCH 54
 * Acc@1 73.492 Acc@5 91.868
Accuracy of the network on the 50000 test images: 73.5%
 * Acc@1 75.604 Acc@5 92.962
Accuracy of the ema network on the 50000 test images: 75.6%
--
EPOCH 55
 * Acc@1 73.344 Acc@5 91.878
Accuracy of the network on the 50000 test images: 73.3%
 * Acc@1 75.756 Acc@5 93.016
Accuracy of the ema network on the 50000 test images: 75.8%
--
EPOCH 56
 * Acc@1 73.576 Acc@5 92.000
Accuracy of the network on the 50000 test images: 73.6%
 * Acc@1 75.880 Acc@5 93.088
Accuracy of the ema network on the 50000 test images: 75.9%
--
EPOCH 57
 * Acc@1 73.478 Acc@5 91.998
Accuracy of the network on the 50000 test images: 73.5%
 * Acc@1 75.960 Acc@5 93.164
Accuracy of the ema network on the 50000 test images: 76.0%
--
EPOCH 58
 * Acc@1 73.890 Acc@5 92.188
Accuracy of the network on the 50000 test images: 73.9%
 * Acc@1 76.050 Acc@5 93.212
Accuracy of the ema network on the 50000 test images: 76.1%
--
EPOCH 59
 * Acc@1 73.858 Acc@5 92.180
Accuracy of the network on the 50000 test images: 73.9%
 * Acc@1 76.184 Acc@5 93.260
Accuracy of the ema network on the 50000 test images: 76.2%
--
EPOCH 60
 * Acc@1 74.000 Acc@5 92.230
Accuracy of the network on the 50000 test images: 74.0%
 * Acc@1 76.264 Acc@5 93.338
Accuracy of the ema network on the 50000 test images: 76.3%
--
EPOCH 61
 * Acc@1 73.940 Acc@5 92.186
Accuracy of the network on the 50000 test images: 73.9%
 * Acc@1 76.378 Acc@5 93.374
Accuracy of the ema network on the 50000 test images: 76.4%
--
EPOCH 62
 * Acc@1 73.878 Acc@5 92.008
Accuracy of the network on the 50000 test images: 73.9%
 * Acc@1 76.474 Acc@5 93.488
Accuracy of the ema network on the 50000 test images: 76.5%
--
EPOCH 63
 * Acc@1 73.716 Acc@5 92.128
Accuracy of the network on the 50000 test images: 73.7%
 * Acc@1 76.542 Acc@5 93.500
Accuracy of the ema network on the 50000 test images: 76.5%
--
EPOCH 64
 * Acc@1 74.164 Acc@5 92.240
Accuracy of the network on the 50000 test images: 74.2%
 * Acc@1 76.622 Acc@5 93.546
Accuracy of the ema network on the 50000 test images: 76.6%
--
EPOCH 65
 * Acc@1 73.644 Acc@5 92.170
Accuracy of the network on the 50000 test images: 73.6%
 * Acc@1 76.706 Acc@5 93.580
Accuracy of the ema network on the 50000 test images: 76.7%
--
EPOCH 66
 * Acc@1 74.148 Acc@5 92.374
Accuracy of the network on the 50000 test images: 74.1%
 * Acc@1 76.796 Acc@5 93.636
Accuracy of the ema network on the 50000 test images: 76.8%
--
EPOCH 67
 * Acc@1 74.576 Acc@5 92.566
Accuracy of the network on the 50000 test images: 74.6%
 * Acc@1 76.892 Acc@5 93.680
Accuracy of the ema network on the 50000 test images: 76.9%
--
EPOCH 68
 * Acc@1 74.564 Acc@5 92.674
Accuracy of the network on the 50000 test images: 74.6%
 * Acc@1 76.930 Acc@5 93.716
Accuracy of the ema network on the 50000 test images: 76.9%
--
EPOCH 69
 * Acc@1 74.296 Acc@5 92.472
Accuracy of the network on the 50000 test images: 74.3%
 * Acc@1 76.988 Acc@5 93.758
Accuracy of the ema network on the 50000 test images: 77.0%
--
EPOCH 70
 * Acc@1 74.292 Acc@5 92.480
Accuracy of the network on the 50000 test images: 74.3%
 * Acc@1 77.058 Acc@5 93.784
Accuracy of the ema network on the 50000 test images: 77.1%
--
EPOCH 71
 * Acc@1 74.402 Acc@5 92.550
Accuracy of the network on the 50000 test images: 74.4%
 * Acc@1 77.118 Acc@5 93.810
Accuracy of the ema network on the 50000 test images: 77.1%
--
EPOCH 72
 * Acc@1 74.598 Acc@5 92.660
Accuracy of the network on the 50000 test images: 74.6%
 * Acc@1 77.154 Acc@5 93.862
Accuracy of the ema network on the 50000 test images: 77.2%
--
EPOCH 73
 * Acc@1 74.378 Acc@5 92.534
Accuracy of the network on the 50000 test images: 74.4%
 * Acc@1 77.204 Acc@5 93.898
Accuracy of the ema network on the 50000 test images: 77.2%
--
EPOCH 74
 * Acc@1 74.594 Acc@5 92.824
Accuracy of the network on the 50000 test images: 74.6%
 * Acc@1 77.290 Acc@5 93.928
Accuracy of the ema network on the 50000 test images: 77.3%
--
EPOCH 75
 * Acc@1 74.914 Acc@5 92.746
Accuracy of the network on the 50000 test images: 74.9%
 * Acc@1 77.348 Acc@5 93.952
Accuracy of the ema network on the 50000 test images: 77.3%
--
EPOCH 76
 * Acc@1 75.736 Acc@5 93.064
Accuracy of the network on the 50000 test images: 75.7%
 * Acc@1 77.414 Acc@5 93.976
Accuracy of the ema network on the 50000 test images: 77.4%
--
EPOCH 77
 * Acc@1 75.536 Acc@5 93.014
Accuracy of the network on the 50000 test images: 75.5%
 * Acc@1 77.432 Acc@5 93.998
Accuracy of the ema network on the 50000 test images: 77.4%
--
EPOCH 78
 * Acc@1 75.554 Acc@5 93.004
Accuracy of the network on the 50000 test images: 75.6%
 * Acc@1 77.476 Acc@5 94.012
Accuracy of the ema network on the 50000 test images: 77.5%
--
EPOCH 79
 * Acc@1 75.340 Acc@5 92.940
Accuracy of the network on the 50000 test images: 75.3%
 * Acc@1 77.452 Acc@5 94.004
Accuracy of the ema network on the 50000 test images: 77.5%
--
EPOCH 80
 * Acc@1 75.772 Acc@5 93.198
Accuracy of the network on the 50000 test images: 75.8%
 * Acc@1 77.438 Acc@5 94.000
Accuracy of the ema network on the 50000 test images: 77.4%
--
EPOCH 81
 * Acc@1 75.848 Acc@5 93.302
Accuracy of the network on the 50000 test images: 75.8%
 * Acc@1 77.476 Acc@5 93.986
Accuracy of the ema network on the 50000 test images: 77.5%
--
EPOCH 82
 * Acc@1 75.890 Acc@5 93.386
Accuracy of the network on the 50000 test images: 75.9%
 * Acc@1 77.528 Acc@5 93.992
Accuracy of the ema network on the 50000 test images: 77.5%
--
EPOCH 83
 * Acc@1 75.878 Acc@5 93.146
Accuracy of the network on the 50000 test images: 75.9%
 * Acc@1 77.572 Acc@5 94.028
Accuracy of the ema network on the 50000 test images: 77.6%
--
EPOCH 84
 * Acc@1 75.926 Acc@5 93.348
Accuracy of the network on the 50000 test images: 75.9%
 * Acc@1 77.590 Acc@5 94.056
Accuracy of the ema network on the 50000 test images: 77.6%
--
EPOCH 85
 * Acc@1 75.920 Acc@5 93.222
Accuracy of the network on the 50000 test images: 75.9%
 * Acc@1 77.646 Acc@5 94.046
Accuracy of the ema network on the 50000 test images: 77.6%
--
EPOCH 86
 * Acc@1 75.916 Acc@5 93.254
Accuracy of the network on the 50000 test images: 75.9%
 * Acc@1 77.666 Acc@5 94.040
Accuracy of the ema network on the 50000 test images: 77.7%
--
EPOCH 87
 * Acc@1 75.808 Acc@5 93.342
Accuracy of the network on the 50000 test images: 75.8%
 * Acc@1 77.676 Acc@5 94.082
Accuracy of the ema network on the 50000 test images: 77.7%
--
EPOCH 88
 * Acc@1 76.144 Acc@5 93.464
Accuracy of the network on the 50000 test images: 76.1%
 * Acc@1 77.732 Acc@5 94.106
Accuracy of the ema network on the 50000 test images: 77.7%
--
EPOCH 89
 * Acc@1 76.142 Acc@5 93.422
Accuracy of the network on the 50000 test images: 76.1%
 * Acc@1 77.784 Acc@5 94.114
Accuracy of the ema network on the 50000 test images: 77.8%
--
EPOCH 90
 * Acc@1 76.104 Acc@5 93.348
Accuracy of the network on the 50000 test images: 76.1%
 * Acc@1 77.822 Acc@5 94.144
Accuracy of the ema network on the 50000 test images: 77.8%
--
EPOCH 91
 * Acc@1 75.914 Acc@5 93.334
Accuracy of the network on the 50000 test images: 75.9%
 * Acc@1 77.854 Acc@5 94.158
Accuracy of the ema network on the 50000 test images: 77.9%
--
EPOCH 92
 * Acc@1 76.372 Acc@5 93.474
Accuracy of the network on the 50000 test images: 76.4%
 * Acc@1 77.894 Acc@5 94.136
Accuracy of the ema network on the 50000 test images: 77.9%
--
EPOCH 93
 * Acc@1 76.252 Acc@5 93.524
Accuracy of the network on the 50000 test images: 76.3%
 * Acc@1 77.932 Acc@5 94.164
Accuracy of the ema network on the 50000 test images: 77.9%
--
EPOCH 94
 * Acc@1 76.264 Acc@5 93.372
Accuracy of the network on the 50000 test images: 76.3%
 * Acc@1 77.992 Acc@5 94.174
Accuracy of the ema network on the 50000 test images: 78.0%
--
EPOCH 95
 * Acc@1 76.218 Acc@5 93.276
Accuracy of the network on the 50000 test images: 76.2%
 * Acc@1 78.046 Acc@5 94.224
Accuracy of the ema network on the 50000 test images: 78.0%
--
EPOCH 96
 * Acc@1 76.360 Acc@5 93.526
Accuracy of the network on the 50000 test images: 76.4%
 * Acc@1 78.112 Acc@5 94.254
Accuracy of the ema network on the 50000 test images: 78.1%
--
EPOCH 97
 * Acc@1 76.630 Acc@5 93.634
Accuracy of the network on the 50000 test images: 76.6%
 * Acc@1 78.158 Acc@5 94.270
Accuracy of the ema network on the 50000 test images: 78.2%
--
EPOCH 98
 * Acc@1 76.660 Acc@5 93.590
Accuracy of the network on the 50000 test images: 76.7%
 * Acc@1 78.192 Acc@5 94.278
Accuracy of the ema network on the 50000 test images: 78.2%
--
EPOCH 99
 * Acc@1 76.804 Acc@5 93.554
Accuracy of the network on the 50000 test images: 76.8%
 * Acc@1 78.194 Acc@5 94.312
Accuracy of the ema network on the 50000 test images: 78.2%
--
EPOCH 100
 * Acc@1 76.576 Acc@5 93.524
Accuracy of the network on the 50000 test images: 76.6%
 * Acc@1 78.218 Acc@5 94.306
Accuracy of the ema network on the 50000 test images: 78.2%
--
EPOCH 101
 * Acc@1 76.230 Acc@5 93.532
Accuracy of the network on the 50000 test images: 76.2%
 * Acc@1 78.230 Acc@5 94.326
Accuracy of the ema network on the 50000 test images: 78.2%
--
EPOCH 102
 * Acc@1 76.592 Acc@5 93.674
Accuracy of the network on the 50000 test images: 76.6%
 * Acc@1 78.272 Acc@5 94.340
Accuracy of the ema network on the 50000 test images: 78.3%
--
EPOCH 103
 * Acc@1 76.672 Acc@5 93.602
Accuracy of the network on the 50000 test images: 76.7%
 * Acc@1 78.274 Acc@5 94.352
Accuracy of the ema network on the 50000 test images: 78.3%
--
EPOCH 104
 * Acc@1 76.748 Acc@5 93.580
Accuracy of the network on the 50000 test images: 76.7%
 * Acc@1 78.336 Acc@5 94.376
Accuracy of the ema network on the 50000 test images: 78.3%
--
EPOCH 105
 * Acc@1 76.338 Acc@5 93.506
Accuracy of the network on the 50000 test images: 76.3%
 * Acc@1 78.368 Acc@5 94.388
Accuracy of the ema network on the 50000 test images: 78.4%
--
EPOCH 106
 * Acc@1 76.852 Acc@5 93.840
Accuracy of the network on the 50000 test images: 76.9%
 * Acc@1 78.408 Acc@5 94.404
Accuracy of the ema network on the 50000 test images: 78.4%
--
EPOCH 107
 * Acc@1 76.872 Acc@5 93.672
Accuracy of the network on the 50000 test images: 76.9%
 * Acc@1 78.410 Acc@5 94.406
Accuracy of the ema network on the 50000 test images: 78.4%
--
EPOCH 108
 * Acc@1 76.608 Acc@5 93.692
Accuracy of the network on the 50000 test images: 76.6%
 * Acc@1 78.442 Acc@5 94.418
Accuracy of the ema network on the 50000 test images: 78.4%
--
EPOCH 109
 * Acc@1 76.770 Acc@5 93.670
Accuracy of the network on the 50000 test images: 76.8%
 * Acc@1 78.422 Acc@5 94.414
Accuracy of the ema network on the 50000 test images: 78.4%
--
EPOCH 110
 * Acc@1 76.918 Acc@5 93.792
Accuracy of the network on the 50000 test images: 76.9%
 * Acc@1 78.500 Acc@5 94.432
Accuracy of the ema network on the 50000 test images: 78.5%
--
EPOCH 111
 * Acc@1 76.972 Acc@5 93.760
Accuracy of the network on the 50000 test images: 77.0%
 * Acc@1 78.520 Acc@5 94.416
Accuracy of the ema network on the 50000 test images: 78.5%
--
EPOCH 112
 * Acc@1 76.796 Acc@5 93.778
Accuracy of the network on the 50000 test images: 76.8%
 * Acc@1 78.524 Acc@5 94.470
Accuracy of the ema network on the 50000 test images: 78.5%
--
EPOCH 113
 * Acc@1 76.920 Acc@5 93.744
Accuracy of the network on the 50000 test images: 76.9%
 * Acc@1 78.524 Acc@5 94.470
Accuracy of the ema network on the 50000 test images: 78.5%
--
EPOCH 114
 * Acc@1 76.956 Acc@5 93.892
Accuracy of the network on the 50000 test images: 77.0%
 * Acc@1 78.524 Acc@5 94.486
Accuracy of the ema network on the 50000 test images: 78.5%
--
EPOCH 115
 * Acc@1 76.770 Acc@5 93.672
Accuracy of the network on the 50000 test images: 76.8%
 * Acc@1 78.578 Acc@5 94.528
Accuracy of the ema network on the 50000 test images: 78.6%
--
EPOCH 116
 * Acc@1 77.148 Acc@5 93.772
Accuracy of the network on the 50000 test images: 77.1%
 * Acc@1 78.598 Acc@5 94.532
Accuracy of the ema network on the 50000 test images: 78.6%
--
EPOCH 117
 * Acc@1 77.062 Acc@5 93.806
Accuracy of the network on the 50000 test images: 77.1%
 * Acc@1 78.628 Acc@5 94.506
Accuracy of the ema network on the 50000 test images: 78.6%
--
EPOCH 118
 * Acc@1 77.036 Acc@5 93.960
Accuracy of the network on the 50000 test images: 77.0%
 * Acc@1 78.650 Acc@5 94.512
Accuracy of the ema network on the 50000 test images: 78.7%
--
EPOCH 119
 * Acc@1 77.324 Acc@5 93.906
Accuracy of the network on the 50000 test images: 77.3%
 * Acc@1 78.682 Acc@5 94.518
Accuracy of the ema network on the 50000 test images: 78.7%
--
EPOCH 120
 * Acc@1 77.256 Acc@5 94.024
Accuracy of the network on the 50000 test images: 77.3%
 * Acc@1 78.746 Acc@5 94.574
Accuracy of the ema network on the 50000 test images: 78.7%
--
EPOCH 121
 * Acc@1 77.038 Acc@5 93.924
Accuracy of the network on the 50000 test images: 77.0%
 * Acc@1 78.752 Acc@5 94.600
Accuracy of the ema network on the 50000 test images: 78.8%
--
EPOCH 122
 * Acc@1 77.112 Acc@5 94.012
Accuracy of the network on the 50000 test images: 77.1%
 * Acc@1 78.744 Acc@5 94.624
Accuracy of the ema network on the 50000 test images: 78.7%
--
EPOCH 123
 * Acc@1 77.246 Acc@5 93.932
Accuracy of the network on the 50000 test images: 77.2%
 * Acc@1 78.720 Acc@5 94.662
Accuracy of the ema network on the 50000 test images: 78.7%
--
EPOCH 124
 * Acc@1 77.430 Acc@5 93.984
Accuracy of the network on the 50000 test images: 77.4%
 * Acc@1 78.786 Acc@5 94.666
Accuracy of the ema network on the 50000 test images: 78.8%
--
EPOCH 125
 * Acc@1 77.268 Acc@5 93.982
Accuracy of the network on the 50000 test images: 77.3%
 * Acc@1 78.824 Acc@5 94.694
Accuracy of the ema network on the 50000 test images: 78.8%
--
EPOCH 126
 * Acc@1 77.464 Acc@5 94.054
Accuracy of the network on the 50000 test images: 77.5%
 * Acc@1 78.810 Acc@5 94.704
Accuracy of the ema network on the 50000 test images: 78.8%
--
EPOCH 127
 * Acc@1 77.370 Acc@5 93.962
Accuracy of the network on the 50000 test images: 77.4%
 * Acc@1 78.828 Acc@5 94.690
Accuracy of the ema network on the 50000 test images: 78.8%
--
EPOCH 128
 * Acc@1 77.434 Acc@5 94.094
Accuracy of the network on the 50000 test images: 77.4%
 * Acc@1 78.868 Acc@5 94.710
Accuracy of the ema network on the 50000 test images: 78.9%
--
EPOCH 129
 * Acc@1 77.586 Acc@5 94.144
Accuracy of the network on the 50000 test images: 77.6%
 * Acc@1 78.906 Acc@5 94.698
Accuracy of the ema network on the 50000 test images: 78.9%
--
EPOCH 130
 * Acc@1 77.520 Acc@5 94.142
Accuracy of the network on the 50000 test images: 77.5%
 * Acc@1 78.930 Acc@5 94.708
Accuracy of the ema network on the 50000 test images: 78.9%
--
EPOCH 131
 * Acc@1 77.602 Acc@5 94.184
Accuracy of the network on the 50000 test images: 77.6%
 * Acc@1 78.946 Acc@5 94.718
Accuracy of the ema network on the 50000 test images: 78.9%
--
EPOCH 132
 * Acc@1 77.596 Acc@5 94.134
Accuracy of the network on the 50000 test images: 77.6%
 * Acc@1 78.966 Acc@5 94.750
Accuracy of the ema network on the 50000 test images: 79.0%
--
EPOCH 133
 * Acc@1 77.494 Acc@5 94.128
Accuracy of the network on the 50000 test images: 77.5%
 * Acc@1 79.012 Acc@5 94.766
Accuracy of the ema network on the 50000 test images: 79.0%
--
EPOCH 134
 * Acc@1 77.828 Acc@5 94.162
Accuracy of the network on the 50000 test images: 77.8%
 * Acc@1 79.066 Acc@5 94.770
Accuracy of the ema network on the 50000 test images: 79.1%
--
EPOCH 135
 * Acc@1 77.300 Acc@5 94.040
Accuracy of the network on the 50000 test images: 77.3%
 * Acc@1 79.060 Acc@5 94.796
Accuracy of the ema network on the 50000 test images: 79.1%
--
EPOCH 136
 * Acc@1 77.696 Acc@5 94.128
Accuracy of the network on the 50000 test images: 77.7%
 * Acc@1 79.092 Acc@5 94.778
Accuracy of the ema network on the 50000 test images: 79.1%
--
EPOCH 137
 * Acc@1 78.020 Acc@5 94.092
Accuracy of the network on the 50000 test images: 78.0%
 * Acc@1 79.142 Acc@5 94.788
Accuracy of the ema network on the 50000 test images: 79.1%
--
EPOCH 138
 * Acc@1 77.840 Acc@5 94.238
Accuracy of the network on the 50000 test images: 77.8%
 * Acc@1 79.174 Acc@5 94.806
Accuracy of the ema network on the 50000 test images: 79.2%
--
EPOCH 139
 * Acc@1 77.834 Acc@5 94.226
Accuracy of the network on the 50000 test images: 77.8%
 * Acc@1 79.162 Acc@5 94.826
Accuracy of the ema network on the 50000 test images: 79.2%
--
EPOCH 140
 * Acc@1 77.918 Acc@5 94.230
Accuracy of the network on the 50000 test images: 77.9%
 * Acc@1 79.194 Acc@5 94.840
Accuracy of the ema network on the 50000 test images: 79.2%
--
EPOCH 141
 * Acc@1 77.808 Acc@5 94.258
Accuracy of the network on the 50000 test images: 77.8%
 * Acc@1 79.258 Acc@5 94.874
Accuracy of the ema network on the 50000 test images: 79.3%
--
EPOCH 142
 * Acc@1 77.988 Acc@5 94.226
Accuracy of the network on the 50000 test images: 78.0%
 * Acc@1 79.258 Acc@5 94.864
Accuracy of the ema network on the 50000 test images: 79.3%
--
EPOCH 143
 * Acc@1 78.068 Acc@5 94.246
Accuracy of the network on the 50000 test images: 78.1%
 * Acc@1 79.310 Acc@5 94.882
Accuracy of the ema network on the 50000 test images: 79.3%
--
EPOCH 144
 * Acc@1 78.026 Acc@5 94.190
Accuracy of the network on the 50000 test images: 78.0%
 * Acc@1 79.328 Acc@5 94.882
Accuracy of the ema network on the 50000 test images: 79.3%
--
EPOCH 145
 * Acc@1 78.088 Acc@5 94.314
Accuracy of the network on the 50000 test images: 78.1%
 * Acc@1 79.346 Acc@5 94.904
Accuracy of the ema network on the 50000 test images: 79.3%
--
EPOCH 146
 * Acc@1 77.900 Acc@5 94.304
Accuracy of the network on the 50000 test images: 77.9%
 * Acc@1 79.354 Acc@5 94.930
Accuracy of the ema network on the 50000 test images: 79.4%
--
EPOCH 147
 * Acc@1 77.972 Acc@5 94.278
Accuracy of the network on the 50000 test images: 78.0%
 * Acc@1 79.370 Acc@5 94.942
Accuracy of the ema network on the 50000 test images: 79.4%
--
EPOCH 148
 * Acc@1 77.954 Acc@5 94.300
Accuracy of the network on the 50000 test images: 78.0%
 * Acc@1 79.396 Acc@5 94.952
Accuracy of the ema network on the 50000 test images: 79.4%
--
EPOCH 149
 * Acc@1 78.094 Acc@5 94.374
Accuracy of the network on the 50000 test images: 78.1%
 * Acc@1 79.404 Acc@5 94.956
Accuracy of the ema network on the 50000 test images: 79.4%
--
EPOCH 150
 * Acc@1 78.370 Acc@5 94.336
Accuracy of the network on the 50000 test images: 78.4%
 * Acc@1 79.374 Acc@5 94.974
Accuracy of the ema network on the 50000 test images: 79.4%
--
EPOCH 151
 * Acc@1 78.194 Acc@5 94.310
Accuracy of the network on the 50000 test images: 78.2%
 * Acc@1 79.372 Acc@5 94.978
Accuracy of the ema network on the 50000 test images: 79.4%
--
EPOCH 152
 * Acc@1 78.062 Acc@5 94.374
Accuracy of the network on the 50000 test images: 78.1%
 * Acc@1 79.416 Acc@5 94.970
Accuracy of the ema network on the 50000 test images: 79.4%
--
EPOCH 153
 * Acc@1 78.072 Acc@5 94.412
Accuracy of the network on the 50000 test images: 78.1%
 * Acc@1 79.428 Acc@5 94.972
Accuracy of the ema network on the 50000 test images: 79.4%
--
EPOCH 154
 * Acc@1 78.256 Acc@5 94.524
Accuracy of the network on the 50000 test images: 78.3%
 * Acc@1 79.464 Acc@5 95.002
Accuracy of the ema network on the 50000 test images: 79.5%
--
EPOCH 155
 * Acc@1 77.982 Acc@5 94.460
Accuracy of the network on the 50000 test images: 78.0%
 * Acc@1 79.482 Acc@5 95.030
Accuracy of the ema network on the 50000 test images: 79.5%
--
EPOCH 156
 * Acc@1 78.436 Acc@5 94.458
Accuracy of the network on the 50000 test images: 78.4%
 * Acc@1 79.490 Acc@5 95.020
Accuracy of the ema network on the 50000 test images: 79.5%
--
EPOCH 157
 * Acc@1 78.588 Acc@5 94.562
Accuracy of the network on the 50000 test images: 78.6%
 * Acc@1 79.512 Acc@5 95.044
Accuracy of the ema network on the 50000 test images: 79.5%
--
EPOCH 158
 * Acc@1 78.274 Acc@5 94.462
Accuracy of the network on the 50000 test images: 78.3%
 * Acc@1 79.574 Acc@5 95.062
Accuracy of the ema network on the 50000 test images: 79.6%
--
EPOCH 159
 * Acc@1 78.460 Acc@5 94.530
Accuracy of the network on the 50000 test images: 78.5%
 * Acc@1 79.594 Acc@5 95.106
Accuracy of the ema network on the 50000 test images: 79.6%
--
EPOCH 160
 * Acc@1 78.642 Acc@5 94.622
Accuracy of the network on the 50000 test images: 78.6%
 * Acc@1 79.632 Acc@5 95.102
Accuracy of the ema network on the 50000 test images: 79.6%
--
EPOCH 161
 * Acc@1 78.524 Acc@5 94.444
Accuracy of the network on the 50000 test images: 78.5%
 * Acc@1 79.702 Acc@5 95.090
Accuracy of the ema network on the 50000 test images: 79.7%
--
EPOCH 162
 * Acc@1 78.666 Acc@5 94.588
Accuracy of the network on the 50000 test images: 78.7%
 * Acc@1 79.686 Acc@5 95.098
Accuracy of the ema network on the 50000 test images: 79.7%
--
EPOCH 163
 * Acc@1 78.646 Acc@5 94.622
Accuracy of the network on the 50000 test images: 78.6%
 * Acc@1 79.762 Acc@5 95.104
Accuracy of the ema network on the 50000 test images: 79.8%
--
EPOCH 164
 * Acc@1 78.560 Acc@5 94.590
Accuracy of the network on the 50000 test images: 78.6%
 * Acc@1 79.768 Acc@5 95.116
Accuracy of the ema network on the 50000 test images: 79.8%
--
EPOCH 165
 * Acc@1 78.458 Acc@5 94.610
Accuracy of the network on the 50000 test images: 78.5%
 * Acc@1 79.796 Acc@5 95.126
Accuracy of the ema network on the 50000 test images: 79.8%
--
EPOCH 166
 * Acc@1 78.618 Acc@5 94.664
Accuracy of the network on the 50000 test images: 78.6%
 * Acc@1 79.764 Acc@5 95.148
Accuracy of the ema network on the 50000 test images: 79.8%
--
EPOCH 167
 * Acc@1 78.670 Acc@5 94.646
Accuracy of the network on the 50000 test images: 78.7%
 * Acc@1 79.770 Acc@5 95.148
Accuracy of the ema network on the 50000 test images: 79.8%
--
EPOCH 168
 * Acc@1 78.838 Acc@5 94.656
Accuracy of the network on the 50000 test images: 78.8%
 * Acc@1 79.816 Acc@5 95.172
Accuracy of the ema network on the 50000 test images: 79.8%
--
EPOCH 169
 * Acc@1 78.740 Acc@5 94.688
Accuracy of the network on the 50000 test images: 78.7%
 * Acc@1 79.822 Acc@5 95.186
Accuracy of the ema network on the 50000 test images: 79.8%
--
EPOCH 170
 * Acc@1 78.950 Acc@5 94.676
Accuracy of the network on the 50000 test images: 78.9%
 * Acc@1 79.868 Acc@5 95.184
Accuracy of the ema network on the 50000 test images: 79.9%
--
EPOCH 171
 * Acc@1 78.974 Acc@5 94.722
Accuracy of the network on the 50000 test images: 79.0%
 * Acc@1 79.930 Acc@5 95.184
Accuracy of the ema network on the 50000 test images: 79.9%
--
EPOCH 172
 * Acc@1 78.972 Acc@5 94.746
Accuracy of the network on the 50000 test images: 79.0%
 * Acc@1 79.940 Acc@5 95.206
Accuracy of the ema network on the 50000 test images: 79.9%
--
EPOCH 173
 * Acc@1 78.762 Acc@5 94.710
Accuracy of the network on the 50000 test images: 78.8%
 * Acc@1 79.926 Acc@5 95.206
Accuracy of the ema network on the 50000 test images: 79.9%
--
EPOCH 174
 * Acc@1 79.056 Acc@5 94.752
Accuracy of the network on the 50000 test images: 79.1%
 * Acc@1 79.968 Acc@5 95.240
Accuracy of the ema network on the 50000 test images: 80.0%
--
EPOCH 175
 * Acc@1 79.034 Acc@5 94.868
Accuracy of the network on the 50000 test images: 79.0%
 * Acc@1 79.972 Acc@5 95.254
Accuracy of the ema network on the 50000 test images: 80.0%
--
EPOCH 176
 * Acc@1 79.030 Acc@5 94.828
Accuracy of the network on the 50000 test images: 79.0%
 * Acc@1 80.016 Acc@5 95.274
Accuracy of the ema network on the 50000 test images: 80.0%
--
EPOCH 177
 * Acc@1 79.058 Acc@5 94.798
Accuracy of the network on the 50000 test images: 79.1%
 * Acc@1 80.022 Acc@5 95.308
Accuracy of the ema network on the 50000 test images: 80.0%
--
EPOCH 178
 * Acc@1 79.230 Acc@5 94.860
Accuracy of the network on the 50000 test images: 79.2%
 * Acc@1 80.038 Acc@5 95.310
Accuracy of the ema network on the 50000 test images: 80.0%
--
EPOCH 179
 * Acc@1 79.226 Acc@5 94.912
Accuracy of the network on the 50000 test images: 79.2%
 * Acc@1 80.068 Acc@5 95.314
Accuracy of the ema network on the 50000 test images: 80.1%
--
EPOCH 180
 * Acc@1 79.170 Acc@5 94.800
Accuracy of the network on the 50000 test images: 79.2%
 * Acc@1 80.150 Acc@5 95.348
Accuracy of the ema network on the 50000 test images: 80.2%
--
EPOCH 181
 * Acc@1 79.250 Acc@5 94.896
Accuracy of the network on the 50000 test images: 79.2%
 * Acc@1 80.164 Acc@5 95.332
Accuracy of the ema network on the 50000 test images: 80.2%
--
EPOCH 182
 * Acc@1 79.202 Acc@5 94.852
Accuracy of the network on the 50000 test images: 79.2%
 * Acc@1 80.256 Acc@5 95.348
Accuracy of the ema network on the 50000 test images: 80.3%
--
EPOCH 183
 * Acc@1 79.334 Acc@5 94.926
Accuracy of the network on the 50000 test images: 79.3%
 * Acc@1 80.254 Acc@5 95.354
Accuracy of the ema network on the 50000 test images: 80.3%
--
EPOCH 184
 * Acc@1 79.310 Acc@5 94.926
Accuracy of the network on the 50000 test images: 79.3%
 * Acc@1 80.304 Acc@5 95.366
Accuracy of the ema network on the 50000 test images: 80.3%
--
EPOCH 185
 * Acc@1 79.110 Acc@5 94.960
Accuracy of the network on the 50000 test images: 79.1%
 * Acc@1 80.372 Acc@5 95.388
Accuracy of the ema network on the 50000 test images: 80.4%
--
EPOCH 186
 * Acc@1 79.464 Acc@5 95.086
Accuracy of the network on the 50000 test images: 79.5%
 * Acc@1 80.386 Acc@5 95.396
Accuracy of the ema network on the 50000 test images: 80.4%
--
EPOCH 187
 * Acc@1 79.540 Acc@5 95.066
Accuracy of the network on the 50000 test images: 79.5%
 * Acc@1 80.412 Acc@5 95.402
Accuracy of the ema network on the 50000 test images: 80.4%
--
EPOCH 188
 * Acc@1 79.392 Acc@5 95.010
Accuracy of the network on the 50000 test images: 79.4%
 * Acc@1 80.460 Acc@5 95.418
Accuracy of the ema network on the 50000 test images: 80.5%
--
EPOCH 189
 * Acc@1 79.440 Acc@5 94.964
Accuracy of the network on the 50000 test images: 79.4%
 * Acc@1 80.482 Acc@5 95.444
Accuracy of the ema network on the 50000 test images: 80.5%
--
EPOCH 190
 * Acc@1 79.460 Acc@5 95.046
Accuracy of the network on the 50000 test images: 79.5%
 * Acc@1 80.500 Acc@5 95.456
Accuracy of the ema network on the 50000 test images: 80.5%
--
EPOCH 191
 * Acc@1 79.698 Acc@5 95.124
Accuracy of the network on the 50000 test images: 79.7%
 * Acc@1 80.522 Acc@5 95.458
Accuracy of the ema network on the 50000 test images: 80.5%
--
EPOCH 192
 * Acc@1 79.664 Acc@5 94.950
Accuracy of the network on the 50000 test images: 79.7%
 * Acc@1 80.542 Acc@5 95.458
Accuracy of the ema network on the 50000 test images: 80.5%
--
EPOCH 193
 * Acc@1 79.464 Acc@5 95.116
Accuracy of the network on the 50000 test images: 79.5%
 * Acc@1 80.556 Acc@5 95.460
Accuracy of the ema network on the 50000 test images: 80.6%
--
EPOCH 194
 * Acc@1 79.744 Acc@5 95.122
Accuracy of the network on the 50000 test images: 79.7%
 * Acc@1 80.568 Acc@5 95.446
Accuracy of the ema network on the 50000 test images: 80.6%
--
EPOCH 195
 * Acc@1 79.768 Acc@5 95.072
Accuracy of the network on the 50000 test images: 79.8%
 * Acc@1 80.592 Acc@5 95.460
Accuracy of the ema network on the 50000 test images: 80.6%
--
EPOCH 196
 * Acc@1 79.734 Acc@5 95.124
Accuracy of the network on the 50000 test images: 79.7%
 * Acc@1 80.612 Acc@5 95.440
Accuracy of the ema network on the 50000 test images: 80.6%
--
EPOCH 197
 * Acc@1 79.672 Acc@5 95.128
Accuracy of the network on the 50000 test images: 79.7%
 * Acc@1 80.618 Acc@5 95.438
Accuracy of the ema network on the 50000 test images: 80.6%
--
EPOCH 198
 * Acc@1 79.936 Acc@5 95.116
Accuracy of the network on the 50000 test images: 79.9%
 * Acc@1 80.640 Acc@5 95.464
Accuracy of the ema network on the 50000 test images: 80.6%
--
EPOCH 199
 * Acc@1 79.902 Acc@5 95.154
Accuracy of the network on the 50000 test images: 79.9%
 * Acc@1 80.642 Acc@5 95.480
Accuracy of the ema network on the 50000 test images: 80.6%
--
EPOCH 200
 * Acc@1 79.780 Acc@5 95.122
Accuracy of the network on the 50000 test images: 79.8%
 * Acc@1 80.648 Acc@5 95.508
Accuracy of the ema network on the 50000 test images: 80.6%
--
EPOCH 201
 * Acc@1 79.714 Acc@5 95.068
Accuracy of the network on the 50000 test images: 79.7%
 * Acc@1 80.670 Acc@5 95.528
Accuracy of the ema network on the 50000 test images: 80.7%
--
EPOCH 202
 * Acc@1 79.952 Acc@5 95.178
Accuracy of the network on the 50000 test images: 80.0%
 * Acc@1 80.700 Acc@5 95.512
Accuracy of the ema network on the 50000 test images: 80.7%
--
EPOCH 203
 * Acc@1 80.014 Acc@5 95.252
Accuracy of the network on the 50000 test images: 80.0%
 * Acc@1 80.698 Acc@5 95.540
Accuracy of the ema network on the 50000 test images: 80.7%
--
EPOCH 204
 * Acc@1 79.988 Acc@5 95.320
Accuracy of the network on the 50000 test images: 80.0%
 * Acc@1 80.764 Acc@5 95.572
Accuracy of the ema network on the 50000 test images: 80.8%
--
EPOCH 205
 * Acc@1 80.104 Acc@5 95.152
Accuracy of the network on the 50000 test images: 80.1%
 * Acc@1 80.784 Acc@5 95.568
Accuracy of the ema network on the 50000 test images: 80.8%
--
EPOCH 206
 * Acc@1 80.092 Acc@5 95.254
Accuracy of the network on the 50000 test images: 80.1%
 * Acc@1 80.814 Acc@5 95.590
Accuracy of the ema network on the 50000 test images: 80.8%
--
EPOCH 207
 * Acc@1 80.312 Acc@5 95.280
Accuracy of the network on the 50000 test images: 80.3%
 * Acc@1 80.820 Acc@5 95.572
Accuracy of the ema network on the 50000 test images: 80.8%
--
EPOCH 208
 * Acc@1 80.060 Acc@5 95.308
Accuracy of the network on the 50000 test images: 80.1%
 * Acc@1 80.810 Acc@5 95.592
Accuracy of the ema network on the 50000 test images: 80.8%
--
EPOCH 209
 * Acc@1 80.072 Acc@5 95.312
Accuracy of the network on the 50000 test images: 80.1%
 * Acc@1 80.808 Acc@5 95.584
Accuracy of the ema network on the 50000 test images: 80.8%
--
EPOCH 210
 * Acc@1 80.124 Acc@5 95.304
Accuracy of the network on the 50000 test images: 80.1%
 * Acc@1 80.842 Acc@5 95.582
Accuracy of the ema network on the 50000 test images: 80.8%
--
EPOCH 211
 * Acc@1 80.254 Acc@5 95.432
Accuracy of the network on the 50000 test images: 80.3%
 * Acc@1 80.866 Acc@5 95.622
Accuracy of the ema network on the 50000 test images: 80.9%
--
EPOCH 212
 * Acc@1 80.172 Acc@5 95.292
Accuracy of the network on the 50000 test images: 80.2%
 * Acc@1 80.862 Acc@5 95.632
Accuracy of the ema network on the 50000 test images: 80.9%
--
EPOCH 213
 * Acc@1 80.138 Acc@5 95.300
Accuracy of the network on the 50000 test images: 80.1%
 * Acc@1 80.872 Acc@5 95.640
Accuracy of the ema network on the 50000 test images: 80.9%
--
EPOCH 214
 * Acc@1 80.234 Acc@5 95.294
Accuracy of the network on the 50000 test images: 80.2%
 * Acc@1 80.918 Acc@5 95.620
Accuracy of the ema network on the 50000 test images: 80.9%
--
EPOCH 215
 * Acc@1 80.502 Acc@5 95.436
Accuracy of the network on the 50000 test images: 80.5%
 * Acc@1 80.962 Acc@5 95.618
Accuracy of the ema network on the 50000 test images: 81.0%
--
EPOCH 216
 * Acc@1 80.298 Acc@5 95.478
Accuracy of the network on the 50000 test images: 80.3%
 * Acc@1 80.982 Acc@5 95.644
Accuracy of the ema network on the 50000 test images: 81.0%
--
EPOCH 217
 * Acc@1 80.312 Acc@5 95.450
Accuracy of the network on the 50000 test images: 80.3%
 * Acc@1 81.020 Acc@5 95.670
Accuracy of the ema network on the 50000 test images: 81.0%
--
EPOCH 218
 * Acc@1 80.302 Acc@5 95.414
Accuracy of the network on the 50000 test images: 80.3%
 * Acc@1 81.034 Acc@5 95.674
Accuracy of the ema network on the 50000 test images: 81.0%
--
EPOCH 219
 * Acc@1 80.604 Acc@5 95.490
Accuracy of the network on the 50000 test images: 80.6%
 * Acc@1 81.062 Acc@5 95.684
Accuracy of the ema network on the 50000 test images: 81.1%
--
EPOCH 220
 * Acc@1 80.516 Acc@5 95.454
Accuracy of the network on the 50000 test images: 80.5%
 * Acc@1 81.048 Acc@5 95.694
Accuracy of the ema network on the 50000 test images: 81.0%
--
EPOCH 221
 * Acc@1 80.450 Acc@5 95.516
Accuracy of the network on the 50000 test images: 80.5%
 * Acc@1 81.082 Acc@5 95.668
Accuracy of the ema network on the 50000 test images: 81.1%
--
EPOCH 222
 * Acc@1 80.652 Acc@5 95.552
Accuracy of the network on the 50000 test images: 80.7%
 * Acc@1 81.082 Acc@5 95.676
Accuracy of the ema network on the 50000 test images: 81.1%
--
EPOCH 223
 * Acc@1 80.468 Acc@5 95.500
Accuracy of the network on the 50000 test images: 80.5%
 * Acc@1 81.074 Acc@5 95.694
Accuracy of the ema network on the 50000 test images: 81.1%
--
EPOCH 224
 * Acc@1 80.620 Acc@5 95.488
Accuracy of the network on the 50000 test images: 80.6%
 * Acc@1 81.124 Acc@5 95.714
Accuracy of the ema network on the 50000 test images: 81.1%
--
EPOCH 225
 * Acc@1 80.602 Acc@5 95.566
Accuracy of the network on the 50000 test images: 80.6%
 * Acc@1 81.118 Acc@5 95.738
Accuracy of the ema network on the 50000 test images: 81.1%
--
EPOCH 226
 * Acc@1 80.598 Acc@5 95.558
Accuracy of the network on the 50000 test images: 80.6%
 * Acc@1 81.138 Acc@5 95.742
Accuracy of the ema network on the 50000 test images: 81.1%
--
EPOCH 227
 * Acc@1 80.636 Acc@5 95.642
Accuracy of the network on the 50000 test images: 80.6%
 * Acc@1 81.200 Acc@5 95.744
Accuracy of the ema network on the 50000 test images: 81.2%
--
EPOCH 228
 * Acc@1 80.676 Acc@5 95.568
Accuracy of the network on the 50000 test images: 80.7%
 * Acc@1 81.196 Acc@5 95.750
Accuracy of the ema network on the 50000 test images: 81.2%
--
EPOCH 229
 * Acc@1 80.676 Acc@5 95.594
Accuracy of the network on the 50000 test images: 80.7%
 * Acc@1 81.202 Acc@5 95.760
Accuracy of the ema network on the 50000 test images: 81.2%
--
EPOCH 230
 * Acc@1 80.886 Acc@5 95.648
Accuracy of the network on the 50000 test images: 80.9%
 * Acc@1 81.246 Acc@5 95.782
Accuracy of the ema network on the 50000 test images: 81.2%
--
EPOCH 231
 * Acc@1 80.816 Acc@5 95.634
Accuracy of the network on the 50000 test images: 80.8%
 * Acc@1 81.278 Acc@5 95.784
Accuracy of the ema network on the 50000 test images: 81.3%
--
EPOCH 232
 * Acc@1 80.820 Acc@5 95.616
Accuracy of the network on the 50000 test images: 80.8%
 * Acc@1 81.316 Acc@5 95.778
Accuracy of the ema network on the 50000 test images: 81.3%
--
EPOCH 233
 * Acc@1 80.832 Acc@5 95.674
Accuracy of the network on the 50000 test images: 80.8%
 * Acc@1 81.282 Acc@5 95.792
Accuracy of the ema network on the 50000 test images: 81.3%
--
EPOCH 234
 * Acc@1 81.138 Acc@5 95.664
Accuracy of the network on the 50000 test images: 81.1%
 * Acc@1 81.298 Acc@5 95.814
Accuracy of the ema network on the 50000 test images: 81.3%
--
EPOCH 235
 * Acc@1 80.934 Acc@5 95.566
Accuracy of the network on the 50000 test images: 80.9%
 * Acc@1 81.306 Acc@5 95.838
Accuracy of the ema network on the 50000 test images: 81.3%
--
EPOCH 236
 * Acc@1 80.882 Acc@5 95.716
Accuracy of the network on the 50000 test images: 80.9%
 * Acc@1 81.360 Acc@5 95.852
Accuracy of the ema network on the 50000 test images: 81.4%
--
EPOCH 237
 * Acc@1 80.990 Acc@5 95.716
Accuracy of the network on the 50000 test images: 81.0%
 * Acc@1 81.382 Acc@5 95.860
Accuracy of the ema network on the 50000 test images: 81.4%
--
EPOCH 238
 * Acc@1 81.024 Acc@5 95.692
Accuracy of the network on the 50000 test images: 81.0%
 * Acc@1 81.408 Acc@5 95.892
Accuracy of the ema network on the 50000 test images: 81.4%
--
EPOCH 239
 * Acc@1 81.094 Acc@5 95.712
Accuracy of the network on the 50000 test images: 81.1%
 * Acc@1 81.426 Acc@5 95.872
Accuracy of the ema network on the 50000 test images: 81.4%
--
EPOCH 240
 * Acc@1 81.106 Acc@5 95.706
Accuracy of the network on the 50000 test images: 81.1%
 * Acc@1 81.440 Acc@5 95.888
Accuracy of the ema network on the 50000 test images: 81.4%
--
EPOCH 241
 * Acc@1 81.192 Acc@5 95.778
Accuracy of the network on the 50000 test images: 81.2%
 * Acc@1 81.460 Acc@5 95.890
Accuracy of the ema network on the 50000 test images: 81.5%
--
EPOCH 242
 * Acc@1 81.172 Acc@5 95.794
Accuracy of the network on the 50000 test images: 81.2%
 * Acc@1 81.470 Acc@5 95.896
Accuracy of the ema network on the 50000 test images: 81.5%
--
EPOCH 243
 * Acc@1 81.186 Acc@5 95.744
Accuracy of the network on the 50000 test images: 81.2%
 * Acc@1 81.516 Acc@5 95.912
Accuracy of the ema network on the 50000 test images: 81.5%
--
EPOCH 244
 * Acc@1 80.974 Acc@5 95.740
Accuracy of the network on the 50000 test images: 81.0%
 * Acc@1 81.514 Acc@5 95.924
Accuracy of the ema network on the 50000 test images: 81.5%
--
EPOCH 245
 * Acc@1 81.228 Acc@5 95.764
Accuracy of the network on the 50000 test images: 81.2%
 * Acc@1 81.524 Acc@5 95.932
Accuracy of the ema network on the 50000 test images: 81.5%
--
EPOCH 246
 * Acc@1 81.174 Acc@5 95.796
Accuracy of the network on the 50000 test images: 81.2%
 * Acc@1 81.536 Acc@5 95.932
Accuracy of the ema network on the 50000 test images: 81.5%
--
EPOCH 247
 * Acc@1 81.360 Acc@5 95.852
Accuracy of the network on the 50000 test images: 81.4%
 * Acc@1 81.558 Acc@5 95.932
Accuracy of the ema network on the 50000 test images: 81.6%
--
EPOCH 248
 * Acc@1 81.382 Acc@5 95.892
Accuracy of the network on the 50000 test images: 81.4%
 * Acc@1 81.566 Acc@5 95.926
Accuracy of the ema network on the 50000 test images: 81.6%
--
EPOCH 249
 * Acc@1 81.250 Acc@5 95.770
Accuracy of the network on the 50000 test images: 81.2%
 * Acc@1 81.582 Acc@5 95.940
Accuracy of the ema network on the 50000 test images: 81.6%
--
EPOCH 250
 * Acc@1 81.218 Acc@5 95.838
Accuracy of the network on the 50000 test images: 81.2%
 * Acc@1 81.586 Acc@5 95.958
Accuracy of the ema network on the 50000 test images: 81.6%
--
EPOCH 251
 * Acc@1 81.462 Acc@5 95.884
Accuracy of the network on the 50000 test images: 81.5%
 * Acc@1 81.596 Acc@5 95.958
Accuracy of the ema network on the 50000 test images: 81.6%
--
EPOCH 252
 * Acc@1 81.436 Acc@5 95.814
Accuracy of the network on the 50000 test images: 81.4%
 * Acc@1 81.606 Acc@5 95.972
Accuracy of the ema network on the 50000 test images: 81.6%
--
EPOCH 253
 * Acc@1 81.498 Acc@5 95.842
Accuracy of the network on the 50000 test images: 81.5%
 * Acc@1 81.634 Acc@5 95.990
Accuracy of the ema network on the 50000 test images: 81.6%
--
EPOCH 254
 * Acc@1 81.454 Acc@5 95.928
Accuracy of the network on the 50000 test images: 81.5%
 * Acc@1 81.646 Acc@5 96.006
Accuracy of the ema network on the 50000 test images: 81.6%
--
EPOCH 255
 * Acc@1 81.428 Acc@5 95.892
Accuracy of the network on the 50000 test images: 81.4%
 * Acc@1 81.688 Acc@5 96.008
Accuracy of the ema network on the 50000 test images: 81.7%
--
EPOCH 256
 * Acc@1 81.452 Acc@5 95.890
Accuracy of the network on the 50000 test images: 81.5%
 * Acc@1 81.690 Acc@5 96.020
Accuracy of the ema network on the 50000 test images: 81.7%
--
EPOCH 257
 * Acc@1 81.400 Acc@5 95.904
Accuracy of the network on the 50000 test images: 81.4%
 * Acc@1 81.702 Acc@5 96.032
Accuracy of the ema network on the 50000 test images: 81.7%
--
EPOCH 258
 * Acc@1 81.556 Acc@5 95.918
Accuracy of the network on the 50000 test images: 81.6%
 * Acc@1 81.730 Acc@5 96.036
Accuracy of the ema network on the 50000 test images: 81.7%
--
EPOCH 259
 * Acc@1 81.572 Acc@5 95.924
Accuracy of the network on the 50000 test images: 81.6%
 * Acc@1 81.750 Acc@5 96.040
Accuracy of the ema network on the 50000 test images: 81.7%
--
EPOCH 260
 * Acc@1 81.616 Acc@5 95.952
Accuracy of the network on the 50000 test images: 81.6%
 * Acc@1 81.760 Acc@5 96.026
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 261
 * Acc@1 81.678 Acc@5 95.960
Accuracy of the network on the 50000 test images: 81.7%
 * Acc@1 81.748 Acc@5 96.042
Accuracy of the ema network on the 50000 test images: 81.7%
--
EPOCH 262
 * Acc@1 81.674 Acc@5 96.014
Accuracy of the network on the 50000 test images: 81.7%
 * Acc@1 81.790 Acc@5 96.046
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 263
 * Acc@1 81.542 Acc@5 95.964
Accuracy of the network on the 50000 test images: 81.5%
 * Acc@1 81.810 Acc@5 96.056
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 264
 * Acc@1 81.630 Acc@5 95.940
Accuracy of the network on the 50000 test images: 81.6%
 * Acc@1 81.788 Acc@5 96.054
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 265
 * Acc@1 81.800 Acc@5 96.008
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.794 Acc@5 96.050
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 266
 * Acc@1 81.602 Acc@5 96.018
Accuracy of the network on the 50000 test images: 81.6%
 * Acc@1 81.804 Acc@5 96.052
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 267
 * Acc@1 81.682 Acc@5 95.972
Accuracy of the network on the 50000 test images: 81.7%
 * Acc@1 81.800 Acc@5 96.060
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 268
 * Acc@1 81.804 Acc@5 96.028
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.802 Acc@5 96.050
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 269
 * Acc@1 81.654 Acc@5 95.998
Accuracy of the network on the 50000 test images: 81.7%
 * Acc@1 81.826 Acc@5 96.070
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 270
 * Acc@1 81.714 Acc@5 96.068
Accuracy of the network on the 50000 test images: 81.7%
 * Acc@1 81.812 Acc@5 96.060
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 271
 * Acc@1 81.652 Acc@5 96.016
Accuracy of the network on the 50000 test images: 81.7%
 * Acc@1 81.804 Acc@5 96.056
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 272
 * Acc@1 81.628 Acc@5 96.014
Accuracy of the network on the 50000 test images: 81.6%
 * Acc@1 81.828 Acc@5 96.056
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 273
 * Acc@1 81.636 Acc@5 95.978
Accuracy of the network on the 50000 test images: 81.6%
 * Acc@1 81.840 Acc@5 96.058
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 274
 * Acc@1 81.746 Acc@5 96.040
Accuracy of the network on the 50000 test images: 81.7%
 * Acc@1 81.838 Acc@5 96.054
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 275
 * Acc@1 81.692 Acc@5 96.006
Accuracy of the network on the 50000 test images: 81.7%
 * Acc@1 81.834 Acc@5 96.056
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 276
 * Acc@1 81.782 Acc@5 96.068
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.834 Acc@5 96.050
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 277
 * Acc@1 81.862 Acc@5 96.024
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.854 Acc@5 96.052
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 278
 * Acc@1 81.838 Acc@5 96.064
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.846 Acc@5 96.072
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 279
 * Acc@1 81.806 Acc@5 96.090
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.816 Acc@5 96.094
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 280
 * Acc@1 81.780 Acc@5 96.084
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.796 Acc@5 96.094
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 281
 * Acc@1 81.814 Acc@5 96.048
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.836 Acc@5 96.096
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 282
 * Acc@1 81.724 Acc@5 96.056
Accuracy of the network on the 50000 test images: 81.7%
 * Acc@1 81.838 Acc@5 96.096
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 283
 * Acc@1 81.830 Acc@5 96.094
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.830 Acc@5 96.096
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 284
 * Acc@1 81.844 Acc@5 96.038
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.828 Acc@5 96.086
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 285
 * Acc@1 81.804 Acc@5 96.062
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.854 Acc@5 96.084
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 286
 * Acc@1 81.794 Acc@5 96.080
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.868 Acc@5 96.084
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 287
 * Acc@1 81.800 Acc@5 96.054
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.870 Acc@5 96.088
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 288
 * Acc@1 81.826 Acc@5 96.068
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.850 Acc@5 96.088
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 289
 * Acc@1 81.858 Acc@5 96.070
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.866 Acc@5 96.096
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 290
 * Acc@1 81.924 Acc@5 96.118
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.882 Acc@5 96.102
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 291
 * Acc@1 81.880 Acc@5 96.142
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.864 Acc@5 96.098
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 292
 * Acc@1 81.920 Acc@5 96.092
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.868 Acc@5 96.106
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 293
 * Acc@1 81.864 Acc@5 96.044
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.852 Acc@5 96.118
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 294
 * Acc@1 81.882 Acc@5 96.118
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.864 Acc@5 96.118
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 295
 * Acc@1 81.894 Acc@5 96.100
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.878 Acc@5 96.132
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 296
 * Acc@1 81.826 Acc@5 96.076
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.878 Acc@5 96.128
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 297
 * Acc@1 81.942 Acc@5 96.096
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.872 Acc@5 96.124
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 298
 * Acc@1 81.896 Acc@5 96.110
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.868 Acc@5 96.136
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 299
 * Acc@1 81.780 Acc@5 96.118
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.882 Acc@5 96.140
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 300
 * Acc@1 81.930 Acc@5 96.124
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.886 Acc@5 96.150
Accuracy of the ema network on the 50000 test images: 81.9%