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
    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/recattn_nearest_interp
PRINT_FREQ: 100
SAVE_FREQ: 1
SEED: 0
TAG: recattn_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, input_resolution=(64, 64), level=1, depth=2
      (blocks): ModuleList(
        (0): MLLABlock(
          dim=48, mlp_ratio=4.0, level=1, downsample=False
          (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=96, bias=True)
          (o_proj): Conv2d(48, 48, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(48, 48, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=48)
              (1): LinearAttention(
                dim=48, num_heads=2
                (qk): Conv2d(48, 96, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=48)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(48, 48, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=48)
          )
          (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=1, downsample=False
          (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=96, bias=True)
          (o_proj): Conv2d(48, 48, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(48, 48, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=48)
              (1): LinearAttention(
                dim=48, num_heads=2
                (qk): Conv2d(48, 96, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(48, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=48)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(48, 48, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=48)
          )
          (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=0, downsample=True
        (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=192, bias=True)
        (o_proj): Conv2d(96, 96, kernel_size=(1, 1), stride=(1, 1))
        (act): SiLU()
        (agg): RecAttn2d(
          (down): Sequential(
            (0): Conv2d(96, 96, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=96)
            (1): LinearAttention(
              dim=96, num_heads=2
              (qk): Conv2d(96, 192, kernel_size=(1, 1), stride=(1, 1), groups=2)
              (elu): ELU(alpha=1.0)
              (lepe): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=96)
              (rope): RoPE()
            )
            (2): Upsample(scale_factor=2.0, mode='nearest')
          )
          (conv): Conv2d(96, 96, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=96)
        )
        (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, input_resolution=(32, 32), level=0, depth=4
      (blocks): ModuleList(
        (0): MLLABlock(
          dim=96, mlp_ratio=4.0, level=0, downsample=False
          (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=192, bias=True)
          (o_proj): Conv2d(96, 96, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(96, 96, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=96)
              (1): LinearAttention(
                dim=96, num_heads=4
                (qk): Conv2d(96, 192, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=96)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(96, 96, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=96)
          )
          (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=0, downsample=False
          (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=192, bias=True)
          (o_proj): Conv2d(96, 96, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(96, 96, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=96)
              (1): LinearAttention(
                dim=96, num_heads=4
                (qk): Conv2d(96, 192, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=96)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(96, 96, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=96)
          )
          (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=0, downsample=False
          (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=192, bias=True)
          (o_proj): Conv2d(96, 96, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(96, 96, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=96)
              (1): LinearAttention(
                dim=96, num_heads=4
                (qk): Conv2d(96, 192, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=96)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(96, 96, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=96)
          )
          (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=0, downsample=False
          (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=192, bias=True)
          (o_proj): Conv2d(96, 96, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(96, 96, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=96)
              (1): LinearAttention(
                dim=96, num_heads=4
                (qk): Conv2d(96, 192, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=96)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(96, 96, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=96)
          )
          (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=-1, downsample=True
        (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=384, bias=True)
        (o_proj): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))
        (act): SiLU()
        (agg): RecAttn2d(
          (down): Sequential(
            (0): Conv2d(192, 192, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=192)
            (1): LinearAttention(
              dim=192, num_heads=4
              (qk): Conv2d(192, 384, kernel_size=(1, 1), stride=(1, 1), groups=2)
              (elu): ELU(alpha=1.0)
              (lepe): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
              (rope): RoPE()
            )
            (2): Upsample(scale_factor=2.0, mode='nearest')
          )
          (conv): Conv2d(192, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=192)
        )
        (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, input_resolution=(16, 16), level=-1, depth=8
      (blocks): ModuleList(
        (0): MLLABlock(
          dim=192, mlp_ratio=4.0, level=-1, downsample=False
          (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=384, bias=True)
          (o_proj): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(192, 192, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=192)
              (1): LinearAttention(
                dim=192, num_heads=8
                (qk): Conv2d(192, 384, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(192, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=192)
          )
          (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=-1, downsample=False
          (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=384, bias=True)
          (o_proj): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(192, 192, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=192)
              (1): LinearAttention(
                dim=192, num_heads=8
                (qk): Conv2d(192, 384, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(192, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=192)
          )
          (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=-1, downsample=False
          (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=384, bias=True)
          (o_proj): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(192, 192, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=192)
              (1): LinearAttention(
                dim=192, num_heads=8
                (qk): Conv2d(192, 384, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(192, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=192)
          )
          (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=-1, downsample=False
          (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=384, bias=True)
          (o_proj): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(192, 192, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=192)
              (1): LinearAttention(
                dim=192, num_heads=8
                (qk): Conv2d(192, 384, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(192, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=192)
          )
          (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=-1, downsample=False
          (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=384, bias=True)
          (o_proj): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(192, 192, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=192)
              (1): LinearAttention(
                dim=192, num_heads=8
                (qk): Conv2d(192, 384, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(192, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=192)
          )
          (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=-1, downsample=False
          (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=384, bias=True)
          (o_proj): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(192, 192, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=192)
              (1): LinearAttention(
                dim=192, num_heads=8
                (qk): Conv2d(192, 384, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(192, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=192)
          )
          (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=-1, downsample=False
          (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=384, bias=True)
          (o_proj): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(192, 192, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=192)
              (1): LinearAttention(
                dim=192, num_heads=8
                (qk): Conv2d(192, 384, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(192, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=192)
          )
          (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=-1, downsample=False
          (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=384, bias=True)
          (o_proj): Conv2d(192, 192, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(192, 192, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=192)
              (1): LinearAttention(
                dim=192, num_heads=8
                (qk): Conv2d(192, 384, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(192, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=192)
          )
          (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=-2, downsample=True
        (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=768, bias=True)
        (o_proj): Conv2d(384, 384, kernel_size=(1, 1), stride=(1, 1))
        (act): SiLU()
        (agg): RecAttn2d(
          (down): Sequential(
            (0): Conv2d(384, 384, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=384)
            (1): LinearAttention(
              dim=384, num_heads=8
              (qk): Conv2d(384, 768, kernel_size=(1, 1), stride=(1, 1), groups=2)
              (elu): ELU(alpha=1.0)
              (lepe): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=384)
              (rope): RoPE()
            )
            (2): Upsample(scale_factor=2.0, mode='nearest')
          )
          (conv): Conv2d(384, 384, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=384)
        )
        (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, input_resolution=(8, 8), level=-2, depth=4
      (blocks): ModuleList(
        (0): MLLABlock(
          dim=384, mlp_ratio=4.0, level=-2, downsample=False
          (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=768, bias=True)
          (o_proj): Conv2d(384, 384, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(384, 384, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=384)
              (1): LinearAttention(
                dim=384, num_heads=16
                (qk): Conv2d(384, 768, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=384)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(384, 384, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=384)
          )
          (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=-2, downsample=False
          (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=768, bias=True)
          (o_proj): Conv2d(384, 384, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(384, 384, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=384)
              (1): LinearAttention(
                dim=384, num_heads=16
                (qk): Conv2d(384, 768, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=384)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(384, 384, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=384)
          )
          (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=-2, downsample=False
          (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=768, bias=True)
          (o_proj): Conv2d(384, 384, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(384, 384, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=384)
              (1): LinearAttention(
                dim=384, num_heads=16
                (qk): Conv2d(384, 768, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=384)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(384, 384, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=384)
          )
          (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=-2, downsample=False
          (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=768, bias=True)
          (o_proj): Conv2d(384, 384, kernel_size=(1, 1), stride=(1, 1))
          (act): SiLU()
          (agg): RecAttn2d(
            (down): Sequential(
              (0): Conv2d(384, 384, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), groups=384)
              (1): LinearAttention(
                dim=384, num_heads=16
                (qk): Conv2d(384, 768, kernel_size=(1, 1), stride=(1, 1), groups=2)
                (elu): ELU(alpha=1.0)
                (lepe): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=384)
                (rope): RoPE()
              )
              (2): Upsample(scale_factor=2.0, mode='nearest')
            )
            (conv): Conv2d(384, 384, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=384)
          )
          (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 0.410 Acc@5 1.710
Accuracy of the network on the 50000 test images: 0.4%
 * Acc@1 0.096 Acc@5 0.500
Accuracy of the ema network on the 50000 test images: 0.1%
--
EPOCH 2
 * Acc@1 3.078 Acc@5 10.334
Accuracy of the network on the 50000 test images: 3.1%
 * Acc@1 0.104 Acc@5 0.498
Accuracy of the ema network on the 50000 test images: 0.1%
--
EPOCH 3
 * Acc@1 9.666 Acc@5 24.274
Accuracy of the network on the 50000 test images: 9.7%
 * Acc@1 0.104 Acc@5 0.532
Accuracy of the ema network on the 50000 test images: 0.1%
--
EPOCH 4
 * Acc@1 17.750 Acc@5 38.622
Accuracy of the network on the 50000 test images: 17.8%
 * Acc@1 0.100 Acc@5 0.528
Accuracy of the ema network on the 50000 test images: 0.1%
--
EPOCH 5
 * Acc@1 25.326 Acc@5 49.532
Accuracy of the network on the 50000 test images: 25.3%
 * Acc@1 0.168 Acc@5 0.534
Accuracy of the ema network on the 50000 test images: 0.2%
--
EPOCH 6
 * Acc@1 32.012 Acc@5 57.658
Accuracy of the network on the 50000 test images: 32.0%
 * Acc@1 0.130 Acc@5 0.506
Accuracy of the ema network on the 50000 test images: 0.1%
--
EPOCH 7
 * Acc@1 37.956 Acc@5 63.940
Accuracy of the network on the 50000 test images: 38.0%
 * Acc@1 0.156 Acc@5 0.632
Accuracy of the ema network on the 50000 test images: 0.2%
--
EPOCH 8
 * Acc@1 42.672 Acc@5 68.370
Accuracy of the network on the 50000 test images: 42.7%
 * Acc@1 0.128 Acc@5 0.844
Accuracy of the ema network on the 50000 test images: 0.1%
--
EPOCH 9
 * Acc@1 47.372 Acc@5 73.086
Accuracy of the network on the 50000 test images: 47.4%
 * Acc@1 0.116 Acc@5 0.778
Accuracy of the ema network on the 50000 test images: 0.1%
--
EPOCH 10
 * Acc@1 50.682 Acc@5 76.008
Accuracy of the network on the 50000 test images: 50.7%
 * Acc@1 0.184 Acc@5 0.852
Accuracy of the ema network on the 50000 test images: 0.2%
--
EPOCH 11
 * Acc@1 53.294 Acc@5 78.210
Accuracy of the network on the 50000 test images: 53.3%
 * Acc@1 0.272 Acc@5 1.036
Accuracy of the ema network on the 50000 test images: 0.3%
--
EPOCH 12
 * Acc@1 54.876 Acc@5 79.648
Accuracy of the network on the 50000 test images: 54.9%
 * Acc@1 0.302 Acc@5 1.222
Accuracy of the ema network on the 50000 test images: 0.3%
--
EPOCH 13
 * Acc@1 57.552 Acc@5 81.576
Accuracy of the network on the 50000 test images: 57.6%
 * Acc@1 0.360 Acc@5 1.774
Accuracy of the ema network on the 50000 test images: 0.4%
--
EPOCH 14
 * Acc@1 58.476 Acc@5 82.446
Accuracy of the network on the 50000 test images: 58.5%
 * Acc@1 0.536 Acc@5 2.908
Accuracy of the ema network on the 50000 test images: 0.5%
--
EPOCH 15
 * Acc@1 60.316 Acc@5 83.790
Accuracy of the network on the 50000 test images: 60.3%
 * Acc@1 1.038 Acc@5 4.864
Accuracy of the ema network on the 50000 test images: 1.0%
--
EPOCH 16
 * Acc@1 61.392 Acc@5 84.538
Accuracy of the network on the 50000 test images: 61.4%
 * Acc@1 2.200 Acc@5 8.016
Accuracy of the ema network on the 50000 test images: 2.2%
--
EPOCH 17
 * Acc@1 62.260 Acc@5 85.112
Accuracy of the network on the 50000 test images: 62.3%
 * Acc@1 4.198 Acc@5 12.704
Accuracy of the ema network on the 50000 test images: 4.2%
--
EPOCH 18
 * Acc@1 63.284 Acc@5 85.830
Accuracy of the network on the 50000 test images: 63.3%
 * Acc@1 7.234 Acc@5 19.414
Accuracy of the ema network on the 50000 test images: 7.2%
--
EPOCH 19
 * Acc@1 63.594 Acc@5 86.036
Accuracy of the network on the 50000 test images: 63.6%
 * Acc@1 11.174 Acc@5 27.254
Accuracy of the ema network on the 50000 test images: 11.2%
--
EPOCH 20
 * Acc@1 64.862 Acc@5 86.798
Accuracy of the network on the 50000 test images: 64.9%
 * Acc@1 16.126 Acc@5 35.440
Accuracy of the ema network on the 50000 test images: 16.1%
--
EPOCH 21
 * Acc@1 64.840 Acc@5 87.056
Accuracy of the network on the 50000 test images: 64.8%
 * Acc@1 21.788 Acc@5 43.736
Accuracy of the ema network on the 50000 test images: 21.8%
--
EPOCH 22
 * Acc@1 66.044 Acc@5 87.690
Accuracy of the network on the 50000 test images: 66.0%
 * Acc@1 27.620 Acc@5 51.316
Accuracy of the ema network on the 50000 test images: 27.6%
--
EPOCH 23
 * Acc@1 66.860 Acc@5 88.170
Accuracy of the network on the 50000 test images: 66.9%
 * Acc@1 33.338 Acc@5 58.136
Accuracy of the ema network on the 50000 test images: 33.3%
--
EPOCH 24
 * Acc@1 67.806 Acc@5 88.666
Accuracy of the network on the 50000 test images: 67.8%
 * Acc@1 38.686 Acc@5 64.296
Accuracy of the ema network on the 50000 test images: 38.7%
--
EPOCH 25
 * Acc@1 68.098 Acc@5 88.830
Accuracy of the network on the 50000 test images: 68.1%
 * Acc@1 43.584 Acc@5 69.082
Accuracy of the ema network on the 50000 test images: 43.6%
--
EPOCH 26
 * Acc@1 67.970 Acc@5 88.816
Accuracy of the network on the 50000 test images: 68.0%
 * Acc@1 48.102 Acc@5 73.278
Accuracy of the ema network on the 50000 test images: 48.1%
--
EPOCH 27
 * Acc@1 69.038 Acc@5 89.332
Accuracy of the network on the 50000 test images: 69.0%
 * Acc@1 51.928 Acc@5 76.808
Accuracy of the ema network on the 50000 test images: 51.9%
--
EPOCH 28
 * Acc@1 69.384 Acc@5 89.524
Accuracy of the network on the 50000 test images: 69.4%
 * Acc@1 55.344 Acc@5 79.726
Accuracy of the ema network on the 50000 test images: 55.3%
--
EPOCH 29
 * Acc@1 69.308 Acc@5 89.718
Accuracy of the network on the 50000 test images: 69.3%
 * Acc@1 58.526 Acc@5 82.092
Accuracy of the ema network on the 50000 test images: 58.5%
--
EPOCH 30
 * Acc@1 69.322 Acc@5 89.614
Accuracy of the network on the 50000 test images: 69.3%
 * Acc@1 60.962 Acc@5 83.964
Accuracy of the ema network on the 50000 test images: 61.0%
--
EPOCH 31
 * Acc@1 69.866 Acc@5 90.072
Accuracy of the network on the 50000 test images: 69.9%
 * Acc@1 63.092 Acc@5 85.478
Accuracy of the ema network on the 50000 test images: 63.1%
--
EPOCH 32
 * Acc@1 70.216 Acc@5 90.290
Accuracy of the network on the 50000 test images: 70.2%
 * Acc@1 64.906 Acc@5 86.716
Accuracy of the ema network on the 50000 test images: 64.9%
--
EPOCH 33
 * Acc@1 70.640 Acc@5 90.654
Accuracy of the network on the 50000 test images: 70.6%
 * Acc@1 66.396 Acc@5 87.638
Accuracy of the ema network on the 50000 test images: 66.4%
--
EPOCH 34
 * Acc@1 70.918 Acc@5 90.604
Accuracy of the network on the 50000 test images: 70.9%
 * Acc@1 67.604 Acc@5 88.452
Accuracy of the ema network on the 50000 test images: 67.6%
--
EPOCH 35
 * Acc@1 71.026 Acc@5 90.520
Accuracy of the network on the 50000 test images: 71.0%
 * Acc@1 68.776 Acc@5 89.172
Accuracy of the ema network on the 50000 test images: 68.8%
--
EPOCH 36
 * Acc@1 71.236 Acc@5 90.772
Accuracy of the network on the 50000 test images: 71.2%
 * Acc@1 69.730 Acc@5 89.726
Accuracy of the ema network on the 50000 test images: 69.7%
--
EPOCH 37
 * Acc@1 71.746 Acc@5 91.068
Accuracy of the network on the 50000 test images: 71.7%
 * Acc@1 70.526 Acc@5 90.134
Accuracy of the ema network on the 50000 test images: 70.5%
--
EPOCH 38
 * Acc@1 71.468 Acc@5 90.820
Accuracy of the network on the 50000 test images: 71.5%
 * Acc@1 71.134 Acc@5 90.512
Accuracy of the ema network on the 50000 test images: 71.1%
--
EPOCH 39
 * Acc@1 72.124 Acc@5 91.276
Accuracy of the network on the 50000 test images: 72.1%
 * Acc@1 71.702 Acc@5 90.822
Accuracy of the ema network on the 50000 test images: 71.7%
--
EPOCH 40
 * Acc@1 72.100 Acc@5 91.270
Accuracy of the network on the 50000 test images: 72.1%
 * Acc@1 72.170 Acc@5 91.138
Accuracy of the ema network on the 50000 test images: 72.2%
--
EPOCH 41
 * Acc@1 72.336 Acc@5 91.378
Accuracy of the network on the 50000 test images: 72.3%
 * Acc@1 72.596 Acc@5 91.390
Accuracy of the ema network on the 50000 test images: 72.6%
--
EPOCH 42
 * Acc@1 72.810 Acc@5 91.564
Accuracy of the network on the 50000 test images: 72.8%
 * Acc@1 73.044 Acc@5 91.594
Accuracy of the ema network on the 50000 test images: 73.0%
--
EPOCH 43
 * Acc@1 72.384 Acc@5 91.542
Accuracy of the network on the 50000 test images: 72.4%
 * Acc@1 73.430 Acc@5 91.808
Accuracy of the ema network on the 50000 test images: 73.4%
--
EPOCH 44
 * Acc@1 72.290 Acc@5 91.340
Accuracy of the network on the 50000 test images: 72.3%
 * Acc@1 73.808 Acc@5 91.970
Accuracy of the ema network on the 50000 test images: 73.8%
--
EPOCH 45
 * Acc@1 72.950 Acc@5 91.642
Accuracy of the network on the 50000 test images: 72.9%
 * Acc@1 74.136 Acc@5 92.126
Accuracy of the ema network on the 50000 test images: 74.1%
--
EPOCH 46
 * Acc@1 72.780 Acc@5 91.698
Accuracy of the network on the 50000 test images: 72.8%
 * Acc@1 74.422 Acc@5 92.308
Accuracy of the ema network on the 50000 test images: 74.4%
--
EPOCH 47
 * Acc@1 73.136 Acc@5 91.874
Accuracy of the network on the 50000 test images: 73.1%
 * Acc@1 74.700 Acc@5 92.436
Accuracy of the ema network on the 50000 test images: 74.7%
--
EPOCH 48
 * Acc@1 73.128 Acc@5 91.778
Accuracy of the network on the 50000 test images: 73.1%
 * Acc@1 74.874 Acc@5 92.542
Accuracy of the ema network on the 50000 test images: 74.9%
--
EPOCH 49
 * Acc@1 73.450 Acc@5 91.854
Accuracy of the network on the 50000 test images: 73.4%
 * Acc@1 75.102 Acc@5 92.634
Accuracy of the ema network on the 50000 test images: 75.1%
--
EPOCH 50
 * Acc@1 73.164 Acc@5 91.744
Accuracy of the network on the 50000 test images: 73.2%
 * Acc@1 75.242 Acc@5 92.766
Accuracy of the ema network on the 50000 test images: 75.2%
--
EPOCH 51
 * Acc@1 73.340 Acc@5 92.026
Accuracy of the network on the 50000 test images: 73.3%
 * Acc@1 75.366 Acc@5 92.852
Accuracy of the ema network on the 50000 test images: 75.4%
--
EPOCH 52
 * Acc@1 73.204 Acc@5 92.034
Accuracy of the network on the 50000 test images: 73.2%
 * Acc@1 75.562 Acc@5 92.944
Accuracy of the ema network on the 50000 test images: 75.6%
--
EPOCH 53
 * Acc@1 73.512 Acc@5 92.022
Accuracy of the network on the 50000 test images: 73.5%
 * Acc@1 75.716 Acc@5 93.014
Accuracy of the ema network on the 50000 test images: 75.7%
--
EPOCH 54
 * Acc@1 73.588 Acc@5 91.974
Accuracy of the network on the 50000 test images: 73.6%
 * Acc@1 75.810 Acc@5 93.082
Accuracy of the ema network on the 50000 test images: 75.8%
--
EPOCH 55
 * Acc@1 73.524 Acc@5 92.038
Accuracy of the network on the 50000 test images: 73.5%
 * Acc@1 75.962 Acc@5 93.150
Accuracy of the ema network on the 50000 test images: 76.0%
--
EPOCH 56
 * Acc@1 73.518 Acc@5 92.166
Accuracy of the network on the 50000 test images: 73.5%
 * Acc@1 76.068 Acc@5 93.202
Accuracy of the ema network on the 50000 test images: 76.1%
--
EPOCH 57
 * Acc@1 73.776 Acc@5 92.346
Accuracy of the network on the 50000 test images: 73.8%
 * Acc@1 76.178 Acc@5 93.272
Accuracy of the ema network on the 50000 test images: 76.2%
--
EPOCH 58
 * Acc@1 73.926 Acc@5 92.326
Accuracy of the network on the 50000 test images: 73.9%
 * Acc@1 76.282 Acc@5 93.326
Accuracy of the ema network on the 50000 test images: 76.3%
--
EPOCH 59
 * Acc@1 73.972 Acc@5 92.350
Accuracy of the network on the 50000 test images: 74.0%
 * Acc@1 76.402 Acc@5 93.394
Accuracy of the ema network on the 50000 test images: 76.4%
--
EPOCH 60
 * Acc@1 74.034 Acc@5 92.134
Accuracy of the network on the 50000 test images: 74.0%
 * Acc@1 76.504 Acc@5 93.448
Accuracy of the ema network on the 50000 test images: 76.5%
--
EPOCH 61
 * Acc@1 74.572 Acc@5 92.492
Accuracy of the network on the 50000 test images: 74.6%
 * Acc@1 76.596 Acc@5 93.482
Accuracy of the ema network on the 50000 test images: 76.6%
--
EPOCH 62
 * Acc@1 73.922 Acc@5 92.470
Accuracy of the network on the 50000 test images: 73.9%
 * Acc@1 76.712 Acc@5 93.536
Accuracy of the ema network on the 50000 test images: 76.7%
--
EPOCH 63
 * Acc@1 74.180 Acc@5 92.466
Accuracy of the network on the 50000 test images: 74.2%
 * Acc@1 76.818 Acc@5 93.572
Accuracy of the ema network on the 50000 test images: 76.8%
--
EPOCH 64
 * Acc@1 74.574 Acc@5 92.502
Accuracy of the network on the 50000 test images: 74.6%
 * Acc@1 76.870 Acc@5 93.638
Accuracy of the ema network on the 50000 test images: 76.9%
--
EPOCH 65
 * Acc@1 74.038 Acc@5 92.292
Accuracy of the network on the 50000 test images: 74.0%
 * Acc@1 76.934 Acc@5 93.658
Accuracy of the ema network on the 50000 test images: 76.9%
--
EPOCH 66
 * Acc@1 74.114 Acc@5 92.612
Accuracy of the network on the 50000 test images: 74.1%
 * Acc@1 76.910 Acc@5 93.714
Accuracy of the ema network on the 50000 test images: 76.9%
--
EPOCH 67
 * Acc@1 74.504 Acc@5 92.490
Accuracy of the network on the 50000 test images: 74.5%
 * Acc@1 76.984 Acc@5 93.780
Accuracy of the ema network on the 50000 test images: 77.0%
--
EPOCH 68
 * Acc@1 74.620 Acc@5 92.706
Accuracy of the network on the 50000 test images: 74.6%
 * Acc@1 77.042 Acc@5 93.832
Accuracy of the ema network on the 50000 test images: 77.0%
--
EPOCH 69
 * Acc@1 73.894 Acc@5 92.280
Accuracy of the network on the 50000 test images: 73.9%
 * Acc@1 77.116 Acc@5 93.858
Accuracy of the ema network on the 50000 test images: 77.1%
--
EPOCH 70
 * Acc@1 74.588 Acc@5 92.634
Accuracy of the network on the 50000 test images: 74.6%
 * Acc@1 77.242 Acc@5 93.854
Accuracy of the ema network on the 50000 test images: 77.2%
--
EPOCH 71
 * Acc@1 74.502 Acc@5 92.516
Accuracy of the network on the 50000 test images: 74.5%
 * Acc@1 77.358 Acc@5 93.890
Accuracy of the ema network on the 50000 test images: 77.4%
--
EPOCH 72
 * Acc@1 74.866 Acc@5 92.782
Accuracy of the network on the 50000 test images: 74.9%
 * Acc@1 77.404 Acc@5 93.960
Accuracy of the ema network on the 50000 test images: 77.4%
--
EPOCH 73
 * Acc@1 74.818 Acc@5 92.788
Accuracy of the network on the 50000 test images: 74.8%
 * Acc@1 77.424 Acc@5 93.966
Accuracy of the ema network on the 50000 test images: 77.4%
--
EPOCH 74
 * Acc@1 74.776 Acc@5 92.616
Accuracy of the network on the 50000 test images: 74.8%
 * Acc@1 77.474 Acc@5 94.004
Accuracy of the ema network on the 50000 test images: 77.5%
--
EPOCH 75
 * Acc@1 74.756 Acc@5 92.720
Accuracy of the network on the 50000 test images: 74.8%
 * Acc@1 77.542 Acc@5 94.014
Accuracy of the ema network on the 50000 test images: 77.5%
--
EPOCH 76
 * Acc@1 75.518 Acc@5 93.086
Accuracy of the network on the 50000 test images: 75.5%
 * Acc@1 77.602 Acc@5 94.036
Accuracy of the ema network on the 50000 test images: 77.6%
--
EPOCH 77
 * Acc@1 75.548 Acc@5 93.088
Accuracy of the network on the 50000 test images: 75.5%
 * Acc@1 77.686 Acc@5 94.074
Accuracy of the ema network on the 50000 test images: 77.7%
--
EPOCH 78
 * Acc@1 75.838 Acc@5 93.310
Accuracy of the network on the 50000 test images: 75.8%
 * Acc@1 77.714 Acc@5 94.102
Accuracy of the ema network on the 50000 test images: 77.7%
--
EPOCH 79
 * Acc@1 75.696 Acc@5 93.332
Accuracy of the network on the 50000 test images: 75.7%
 * Acc@1 77.750 Acc@5 94.114
Accuracy of the ema network on the 50000 test images: 77.8%
--
EPOCH 80
 * Acc@1 75.994 Acc@5 93.354
Accuracy of the network on the 50000 test images: 76.0%
 * Acc@1 77.736 Acc@5 94.112
Accuracy of the ema network on the 50000 test images: 77.7%
--
EPOCH 81
 * Acc@1 75.966 Acc@5 93.294
Accuracy of the network on the 50000 test images: 76.0%
 * Acc@1 77.742 Acc@5 94.154
Accuracy of the ema network on the 50000 test images: 77.7%
--
EPOCH 82
 * Acc@1 76.096 Acc@5 93.202
Accuracy of the network on the 50000 test images: 76.1%
 * Acc@1 77.728 Acc@5 94.160
Accuracy of the ema network on the 50000 test images: 77.7%
--
EPOCH 83
 * Acc@1 76.038 Acc@5 93.340
Accuracy of the network on the 50000 test images: 76.0%
 * Acc@1 77.792 Acc@5 94.160
Accuracy of the ema network on the 50000 test images: 77.8%
--
EPOCH 84
 * Acc@1 75.938 Acc@5 93.314
Accuracy of the network on the 50000 test images: 75.9%
 * Acc@1 77.858 Acc@5 94.162
Accuracy of the ema network on the 50000 test images: 77.9%
--
EPOCH 85
 * Acc@1 76.082 Acc@5 93.346
Accuracy of the network on the 50000 test images: 76.1%
 * Acc@1 77.890 Acc@5 94.208
Accuracy of the ema network on the 50000 test images: 77.9%
--
EPOCH 86
 * Acc@1 76.172 Acc@5 93.392
Accuracy of the network on the 50000 test images: 76.2%
 * Acc@1 77.890 Acc@5 94.234
Accuracy of the ema network on the 50000 test images: 77.9%
--
EPOCH 87
 * Acc@1 76.124 Acc@5 93.398
Accuracy of the network on the 50000 test images: 76.1%
 * Acc@1 77.882 Acc@5 94.272
Accuracy of the ema network on the 50000 test images: 77.9%
--
EPOCH 88
 * Acc@1 76.350 Acc@5 93.532
Accuracy of the network on the 50000 test images: 76.3%
 * Acc@1 77.952 Acc@5 94.264
Accuracy of the ema network on the 50000 test images: 78.0%
--
EPOCH 89
 * Acc@1 76.320 Acc@5 93.474
Accuracy of the network on the 50000 test images: 76.3%
 * Acc@1 77.944 Acc@5 94.270
Accuracy of the ema network on the 50000 test images: 77.9%
--
EPOCH 90
 * Acc@1 76.420 Acc@5 93.532
Accuracy of the network on the 50000 test images: 76.4%
 * Acc@1 77.956 Acc@5 94.282
Accuracy of the ema network on the 50000 test images: 78.0%
--
EPOCH 91
 * Acc@1 76.508 Acc@5 93.584
Accuracy of the network on the 50000 test images: 76.5%
 * Acc@1 77.996 Acc@5 94.268
Accuracy of the ema network on the 50000 test images: 78.0%
--
EPOCH 92
 * Acc@1 76.324 Acc@5 93.520
Accuracy of the network on the 50000 test images: 76.3%
 * Acc@1 78.056 Acc@5 94.276
Accuracy of the ema network on the 50000 test images: 78.1%
--
EPOCH 93
 * Acc@1 76.468 Acc@5 93.586
Accuracy of the network on the 50000 test images: 76.5%
 * Acc@1 78.050 Acc@5 94.292
Accuracy of the ema network on the 50000 test images: 78.1%
--
EPOCH 94
 * Acc@1 76.516 Acc@5 93.628
Accuracy of the network on the 50000 test images: 76.5%
 * Acc@1 78.076 Acc@5 94.292
Accuracy of the ema network on the 50000 test images: 78.1%
--
EPOCH 95
 * Acc@1 76.430 Acc@5 93.536
Accuracy of the network on the 50000 test images: 76.4%
 * Acc@1 78.124 Acc@5 94.332
Accuracy of the ema network on the 50000 test images: 78.1%
--
EPOCH 96
 * Acc@1 76.274 Acc@5 93.526
Accuracy of the network on the 50000 test images: 76.3%
 * Acc@1 78.146 Acc@5 94.336
Accuracy of the ema network on the 50000 test images: 78.1%
--
EPOCH 97
 * Acc@1 76.396 Acc@5 93.622
Accuracy of the network on the 50000 test images: 76.4%
 * Acc@1 78.146 Acc@5 94.372
Accuracy of the ema network on the 50000 test images: 78.1%
--
EPOCH 98
 * Acc@1 76.606 Acc@5 93.650
Accuracy of the network on the 50000 test images: 76.6%
 * Acc@1 78.166 Acc@5 94.358
Accuracy of the ema network on the 50000 test images: 78.2%
--
EPOCH 99
 * Acc@1 76.748 Acc@5 93.768
Accuracy of the network on the 50000 test images: 76.7%
 * Acc@1 78.220 Acc@5 94.354
Accuracy of the ema network on the 50000 test images: 78.2%
--
EPOCH 100
 * Acc@1 76.790 Acc@5 93.678
Accuracy of the network on the 50000 test images: 76.8%
 * Acc@1 78.260 Acc@5 94.376
Accuracy of the ema network on the 50000 test images: 78.3%
--
EPOCH 101
 * Acc@1 76.450 Acc@5 93.642
Accuracy of the network on the 50000 test images: 76.4%
 * Acc@1 78.262 Acc@5 94.402
Accuracy of the ema network on the 50000 test images: 78.3%
--
EPOCH 102
 * Acc@1 76.820 Acc@5 93.790
Accuracy of the network on the 50000 test images: 76.8%
 * Acc@1 78.272 Acc@5 94.418
Accuracy of the ema network on the 50000 test images: 78.3%
--
EPOCH 103
 * Acc@1 76.640 Acc@5 93.676
Accuracy of the network on the 50000 test images: 76.6%
 * Acc@1 78.298 Acc@5 94.426
Accuracy of the ema network on the 50000 test images: 78.3%
--
EPOCH 104
 * Acc@1 76.806 Acc@5 93.778
Accuracy of the network on the 50000 test images: 76.8%
 * Acc@1 78.388 Acc@5 94.430
Accuracy of the ema network on the 50000 test images: 78.4%
--
EPOCH 105
 * Acc@1 76.780 Acc@5 93.712
Accuracy of the network on the 50000 test images: 76.8%
 * Acc@1 78.374 Acc@5 94.436
Accuracy of the ema network on the 50000 test images: 78.4%
--
EPOCH 106
 * Acc@1 76.828 Acc@5 93.804
Accuracy of the network on the 50000 test images: 76.8%
 * Acc@1 78.426 Acc@5 94.458
Accuracy of the ema network on the 50000 test images: 78.4%
--
EPOCH 107
 * Acc@1 76.978 Acc@5 93.832
Accuracy of the network on the 50000 test images: 77.0%
 * Acc@1 78.500 Acc@5 94.460
Accuracy of the ema network on the 50000 test images: 78.5%
--
EPOCH 108
 * Acc@1 76.912 Acc@5 93.774
Accuracy of the network on the 50000 test images: 76.9%
 * Acc@1 78.546 Acc@5 94.478
Accuracy of the ema network on the 50000 test images: 78.5%
--
EPOCH 109
 * Acc@1 76.930 Acc@5 93.822
Accuracy of the network on the 50000 test images: 76.9%
 * Acc@1 78.548 Acc@5 94.520
Accuracy of the ema network on the 50000 test images: 78.5%
--
EPOCH 110
 * Acc@1 77.060 Acc@5 93.886
Accuracy of the network on the 50000 test images: 77.1%
 * Acc@1 78.588 Acc@5 94.546
Accuracy of the ema network on the 50000 test images: 78.6%
--
EPOCH 111
 * Acc@1 76.858 Acc@5 93.826
Accuracy of the network on the 50000 test images: 76.9%
 * Acc@1 78.604 Acc@5 94.560
Accuracy of the ema network on the 50000 test images: 78.6%
--
EPOCH 112
 * Acc@1 77.086 Acc@5 93.924
Accuracy of the network on the 50000 test images: 77.1%
 * Acc@1 78.626 Acc@5 94.588
Accuracy of the ema network on the 50000 test images: 78.6%
--
EPOCH 113
 * Acc@1 77.128 Acc@5 93.914
Accuracy of the network on the 50000 test images: 77.1%
 * Acc@1 78.654 Acc@5 94.594
Accuracy of the ema network on the 50000 test images: 78.7%
--
EPOCH 114
 * Acc@1 77.232 Acc@5 94.126
Accuracy of the network on the 50000 test images: 77.2%
 * Acc@1 78.638 Acc@5 94.604
Accuracy of the ema network on the 50000 test images: 78.6%
--
EPOCH 115
 * Acc@1 77.014 Acc@5 93.940
Accuracy of the network on the 50000 test images: 77.0%
 * Acc@1 78.646 Acc@5 94.626
Accuracy of the ema network on the 50000 test images: 78.6%
--
EPOCH 116
 * Acc@1 77.322 Acc@5 93.952
Accuracy of the network on the 50000 test images: 77.3%
 * Acc@1 78.636 Acc@5 94.622
Accuracy of the ema network on the 50000 test images: 78.6%
--
EPOCH 117
 * Acc@1 77.222 Acc@5 93.870
Accuracy of the network on the 50000 test images: 77.2%
 * Acc@1 78.646 Acc@5 94.606
Accuracy of the ema network on the 50000 test images: 78.6%
--
EPOCH 118
 * Acc@1 77.180 Acc@5 93.906
Accuracy of the network on the 50000 test images: 77.2%
 * Acc@1 78.710 Acc@5 94.644
Accuracy of the ema network on the 50000 test images: 78.7%
--
EPOCH 119
 * Acc@1 77.392 Acc@5 94.020
Accuracy of the network on the 50000 test images: 77.4%
 * Acc@1 78.714 Acc@5 94.666
Accuracy of the ema network on the 50000 test images: 78.7%
--
EPOCH 120
 * Acc@1 77.460 Acc@5 94.046
Accuracy of the network on the 50000 test images: 77.5%
 * Acc@1 78.774 Acc@5 94.694
Accuracy of the ema network on the 50000 test images: 78.8%
--
EPOCH 121
 * Acc@1 77.260 Acc@5 94.064
Accuracy of the network on the 50000 test images: 77.3%
 * Acc@1 78.826 Acc@5 94.684
Accuracy of the ema network on the 50000 test images: 78.8%
--
EPOCH 122
 * Acc@1 77.386 Acc@5 94.040
Accuracy of the network on the 50000 test images: 77.4%
 * Acc@1 78.874 Acc@5 94.692
Accuracy of the ema network on the 50000 test images: 78.9%
--
EPOCH 123
 * Acc@1 77.568 Acc@5 94.070
Accuracy of the network on the 50000 test images: 77.6%
 * Acc@1 78.894 Acc@5 94.722
Accuracy of the ema network on the 50000 test images: 78.9%
--
EPOCH 124
 * Acc@1 77.498 Acc@5 94.172
Accuracy of the network on the 50000 test images: 77.5%
 * Acc@1 78.908 Acc@5 94.730
Accuracy of the ema network on the 50000 test images: 78.9%
--
EPOCH 125
 * Acc@1 77.538 Acc@5 94.148
Accuracy of the network on the 50000 test images: 77.5%
 * Acc@1 78.936 Acc@5 94.736
Accuracy of the ema network on the 50000 test images: 78.9%
--
EPOCH 126
 * Acc@1 77.526 Acc@5 94.142
Accuracy of the network on the 50000 test images: 77.5%
 * Acc@1 78.990 Acc@5 94.734
Accuracy of the ema network on the 50000 test images: 79.0%
--
EPOCH 127
 * Acc@1 77.504 Acc@5 94.050
Accuracy of the network on the 50000 test images: 77.5%
 * Acc@1 79.018 Acc@5 94.756
Accuracy of the ema network on the 50000 test images: 79.0%
--
EPOCH 128
 * Acc@1 77.518 Acc@5 94.104
Accuracy of the network on the 50000 test images: 77.5%
 * Acc@1 79.060 Acc@5 94.768
Accuracy of the ema network on the 50000 test images: 79.1%
--
EPOCH 129
 * Acc@1 77.854 Acc@5 94.138
Accuracy of the network on the 50000 test images: 77.9%
 * Acc@1 79.040 Acc@5 94.780
Accuracy of the ema network on the 50000 test images: 79.0%
--
EPOCH 130
 * Acc@1 77.580 Acc@5 94.192
Accuracy of the network on the 50000 test images: 77.6%
 * Acc@1 79.090 Acc@5 94.804
Accuracy of the ema network on the 50000 test images: 79.1%
--
EPOCH 131
 * Acc@1 77.692 Acc@5 94.118
Accuracy of the network on the 50000 test images: 77.7%
 * Acc@1 79.138 Acc@5 94.806
Accuracy of the ema network on the 50000 test images: 79.1%
--
EPOCH 132
 * Acc@1 77.674 Acc@5 94.142
Accuracy of the network on the 50000 test images: 77.7%
 * Acc@1 79.188 Acc@5 94.800
Accuracy of the ema network on the 50000 test images: 79.2%
--
EPOCH 133
 * Acc@1 77.770 Acc@5 94.248
Accuracy of the network on the 50000 test images: 77.8%
 * Acc@1 79.176 Acc@5 94.812
Accuracy of the ema network on the 50000 test images: 79.2%
--
EPOCH 134
 * Acc@1 77.958 Acc@5 94.280
Accuracy of the network on the 50000 test images: 78.0%
 * Acc@1 79.212 Acc@5 94.856
Accuracy of the ema network on the 50000 test images: 79.2%
--
EPOCH 135
 * Acc@1 77.736 Acc@5 94.120
Accuracy of the network on the 50000 test images: 77.7%
 * Acc@1 79.234 Acc@5 94.852
Accuracy of the ema network on the 50000 test images: 79.2%
--
EPOCH 136
 * Acc@1 77.924 Acc@5 94.218
Accuracy of the network on the 50000 test images: 77.9%
 * Acc@1 79.244 Acc@5 94.886
Accuracy of the ema network on the 50000 test images: 79.2%
--
EPOCH 137
 * Acc@1 77.814 Acc@5 94.248
Accuracy of the network on the 50000 test images: 77.8%
 * Acc@1 79.314 Acc@5 94.934
Accuracy of the ema network on the 50000 test images: 79.3%
--
EPOCH 138
 * Acc@1 77.968 Acc@5 94.282
Accuracy of the network on the 50000 test images: 78.0%
 * Acc@1 79.348 Acc@5 94.934
Accuracy of the ema network on the 50000 test images: 79.3%
--
EPOCH 139
 * Acc@1 78.016 Acc@5 94.466
Accuracy of the network on the 50000 test images: 78.0%
 * Acc@1 79.336 Acc@5 94.936
Accuracy of the ema network on the 50000 test images: 79.3%
--
EPOCH 140
 * Acc@1 78.056 Acc@5 94.398
Accuracy of the network on the 50000 test images: 78.1%
 * Acc@1 79.336 Acc@5 94.942
Accuracy of the ema network on the 50000 test images: 79.3%
--
EPOCH 141
 * Acc@1 77.690 Acc@5 94.284
Accuracy of the network on the 50000 test images: 77.7%
 * Acc@1 79.332 Acc@5 94.956
Accuracy of the ema network on the 50000 test images: 79.3%
--
EPOCH 142
 * Acc@1 78.050 Acc@5 94.344
Accuracy of the network on the 50000 test images: 78.0%
 * Acc@1 79.360 Acc@5 94.974
Accuracy of the ema network on the 50000 test images: 79.4%
--
EPOCH 143
 * Acc@1 78.178 Acc@5 94.366
Accuracy of the network on the 50000 test images: 78.2%
 * Acc@1 79.332 Acc@5 95.022
Accuracy of the ema network on the 50000 test images: 79.3%
--
EPOCH 144
 * Acc@1 78.268 Acc@5 94.410
Accuracy of the network on the 50000 test images: 78.3%
 * Acc@1 79.384 Acc@5 95.024
Accuracy of the ema network on the 50000 test images: 79.4%
--
EPOCH 145
 * Acc@1 78.282 Acc@5 94.400
Accuracy of the network on the 50000 test images: 78.3%
 * Acc@1 79.424 Acc@5 95.016
Accuracy of the ema network on the 50000 test images: 79.4%
--
EPOCH 146
 * Acc@1 78.326 Acc@5 94.418
Accuracy of the network on the 50000 test images: 78.3%
 * Acc@1 79.430 Acc@5 95.050
Accuracy of the ema network on the 50000 test images: 79.4%
--
EPOCH 147
 * Acc@1 78.154 Acc@5 94.388
Accuracy of the network on the 50000 test images: 78.2%
 * Acc@1 79.444 Acc@5 95.038
Accuracy of the ema network on the 50000 test images: 79.4%
--
EPOCH 148
 * Acc@1 78.034 Acc@5 94.396
Accuracy of the network on the 50000 test images: 78.0%
 * Acc@1 79.470 Acc@5 95.066
Accuracy of the ema network on the 50000 test images: 79.5%
--
EPOCH 149
 * Acc@1 78.382 Acc@5 94.460
Accuracy of the network on the 50000 test images: 78.4%
 * Acc@1 79.534 Acc@5 95.072
Accuracy of the ema network on the 50000 test images: 79.5%
--
EPOCH 150
 * Acc@1 78.196 Acc@5 94.454
Accuracy of the network on the 50000 test images: 78.2%
 * Acc@1 79.582 Acc@5 95.070
Accuracy of the ema network on the 50000 test images: 79.6%
--
EPOCH 151
 * Acc@1 78.392 Acc@5 94.462
Accuracy of the network on the 50000 test images: 78.4%
 * Acc@1 79.572 Acc@5 95.050
Accuracy of the ema network on the 50000 test images: 79.6%
--
EPOCH 152
 * Acc@1 78.014 Acc@5 94.356
Accuracy of the network on the 50000 test images: 78.0%
 * Acc@1 79.586 Acc@5 95.068
Accuracy of the ema network on the 50000 test images: 79.6%
--
EPOCH 153
 * Acc@1 78.410 Acc@5 94.566
Accuracy of the network on the 50000 test images: 78.4%
 * Acc@1 79.616 Acc@5 95.060
Accuracy of the ema network on the 50000 test images: 79.6%
--
EPOCH 154
 * Acc@1 78.288 Acc@5 94.492
Accuracy of the network on the 50000 test images: 78.3%
 * Acc@1 79.686 Acc@5 95.102
Accuracy of the ema network on the 50000 test images: 79.7%
--
EPOCH 155
 * Acc@1 78.412 Acc@5 94.500
Accuracy of the network on the 50000 test images: 78.4%
 * Acc@1 79.688 Acc@5 95.116
Accuracy of the ema network on the 50000 test images: 79.7%
--
EPOCH 156
 * Acc@1 78.488 Acc@5 94.566
Accuracy of the network on the 50000 test images: 78.5%
 * Acc@1 79.710 Acc@5 95.132
Accuracy of the ema network on the 50000 test images: 79.7%
--
EPOCH 157
 * Acc@1 78.648 Acc@5 94.654
Accuracy of the network on the 50000 test images: 78.6%
 * Acc@1 79.692 Acc@5 95.128
Accuracy of the ema network on the 50000 test images: 79.7%
--
EPOCH 158
 * Acc@1 78.482 Acc@5 94.588
Accuracy of the network on the 50000 test images: 78.5%
 * Acc@1 79.746 Acc@5 95.162
Accuracy of the ema network on the 50000 test images: 79.7%
--
EPOCH 159
 * Acc@1 78.392 Acc@5 94.622
Accuracy of the network on the 50000 test images: 78.4%
 * Acc@1 79.756 Acc@5 95.188
Accuracy of the ema network on the 50000 test images: 79.8%
--
EPOCH 160
 * Acc@1 78.756 Acc@5 94.644
Accuracy of the network on the 50000 test images: 78.8%
 * Acc@1 79.810 Acc@5 95.192
Accuracy of the ema network on the 50000 test images: 79.8%
--
EPOCH 161
 * Acc@1 78.590 Acc@5 94.550
Accuracy of the network on the 50000 test images: 78.6%
 * Acc@1 79.828 Acc@5 95.196
Accuracy of the ema network on the 50000 test images: 79.8%
--
EPOCH 162
 * Acc@1 78.608 Acc@5 94.624
Accuracy of the network on the 50000 test images: 78.6%
 * Acc@1 79.830 Acc@5 95.184
Accuracy of the ema network on the 50000 test images: 79.8%
--
EPOCH 163
 * Acc@1 78.942 Acc@5 94.728
Accuracy of the network on the 50000 test images: 78.9%
 * Acc@1 79.858 Acc@5 95.210
Accuracy of the ema network on the 50000 test images: 79.9%
--
EPOCH 164
 * Acc@1 78.534 Acc@5 94.662
Accuracy of the network on the 50000 test images: 78.5%
 * Acc@1 79.870 Acc@5 95.222
Accuracy of the ema network on the 50000 test images: 79.9%
--
EPOCH 165
 * Acc@1 78.572 Acc@5 94.714
Accuracy of the network on the 50000 test images: 78.6%
 * Acc@1 79.892 Acc@5 95.250
Accuracy of the ema network on the 50000 test images: 79.9%
--
EPOCH 166
 * Acc@1 78.692 Acc@5 94.718
Accuracy of the network on the 50000 test images: 78.7%
 * Acc@1 79.916 Acc@5 95.256
Accuracy of the ema network on the 50000 test images: 79.9%
--
EPOCH 167
 * Acc@1 78.858 Acc@5 94.748
Accuracy of the network on the 50000 test images: 78.9%
 * Acc@1 79.982 Acc@5 95.274
Accuracy of the ema network on the 50000 test images: 80.0%
--
EPOCH 168
 * Acc@1 78.964 Acc@5 94.670
Accuracy of the network on the 50000 test images: 79.0%
 * Acc@1 80.028 Acc@5 95.304
Accuracy of the ema network on the 50000 test images: 80.0%
--
EPOCH 169
 * Acc@1 78.944 Acc@5 94.772
Accuracy of the network on the 50000 test images: 78.9%
 * Acc@1 80.018 Acc@5 95.322
Accuracy of the ema network on the 50000 test images: 80.0%
--
EPOCH 170
 * Acc@1 78.914 Acc@5 94.778
Accuracy of the network on the 50000 test images: 78.9%
 * Acc@1 80.040 Acc@5 95.356
Accuracy of the ema network on the 50000 test images: 80.0%
--
EPOCH 171
 * Acc@1 78.944 Acc@5 94.782
Accuracy of the network on the 50000 test images: 78.9%
 * Acc@1 80.082 Acc@5 95.346
Accuracy of the ema network on the 50000 test images: 80.1%
--
EPOCH 172
 * Acc@1 78.904 Acc@5 94.746
Accuracy of the network on the 50000 test images: 78.9%
 * Acc@1 80.114 Acc@5 95.336
Accuracy of the ema network on the 50000 test images: 80.1%
--
EPOCH 173
 * Acc@1 79.334 Acc@5 94.852
Accuracy of the network on the 50000 test images: 79.3%
 * Acc@1 80.112 Acc@5 95.324
Accuracy of the ema network on the 50000 test images: 80.1%
--
EPOCH 174
 * Acc@1 79.136 Acc@5 94.930
Accuracy of the network on the 50000 test images: 79.1%
 * Acc@1 80.138 Acc@5 95.334
Accuracy of the ema network on the 50000 test images: 80.1%
--
EPOCH 175
 * Acc@1 79.074 Acc@5 94.876
Accuracy of the network on the 50000 test images: 79.1%
 * Acc@1 80.170 Acc@5 95.340
Accuracy of the ema network on the 50000 test images: 80.2%
--
EPOCH 176
 * Acc@1 79.046 Acc@5 94.866
Accuracy of the network on the 50000 test images: 79.0%
 * Acc@1 80.242 Acc@5 95.364
Accuracy of the ema network on the 50000 test images: 80.2%
--
EPOCH 177
 * Acc@1 79.314 Acc@5 94.930
Accuracy of the network on the 50000 test images: 79.3%
 * Acc@1 80.244 Acc@5 95.378
Accuracy of the ema network on the 50000 test images: 80.2%
--
EPOCH 178
 * Acc@1 79.222 Acc@5 94.858
Accuracy of the network on the 50000 test images: 79.2%
 * Acc@1 80.266 Acc@5 95.386
Accuracy of the ema network on the 50000 test images: 80.3%
--
EPOCH 179
 * Acc@1 79.156 Acc@5 94.912
Accuracy of the network on the 50000 test images: 79.2%
 * Acc@1 80.236 Acc@5 95.380
Accuracy of the ema network on the 50000 test images: 80.2%
--
EPOCH 180
 * Acc@1 79.364 Acc@5 94.896
Accuracy of the network on the 50000 test images: 79.4%
 * Acc@1 80.278 Acc@5 95.380
Accuracy of the ema network on the 50000 test images: 80.3%
--
EPOCH 181
 * Acc@1 79.292 Acc@5 94.944
Accuracy of the network on the 50000 test images: 79.3%
 * Acc@1 80.286 Acc@5 95.354
Accuracy of the ema network on the 50000 test images: 80.3%
--
EPOCH 182
 * Acc@1 79.442 Acc@5 94.870
Accuracy of the network on the 50000 test images: 79.4%
 * Acc@1 80.326 Acc@5 95.380
Accuracy of the ema network on the 50000 test images: 80.3%
--
EPOCH 183
 * Acc@1 79.326 Acc@5 95.008
Accuracy of the network on the 50000 test images: 79.3%
 * Acc@1 80.364 Acc@5 95.414
Accuracy of the ema network on the 50000 test images: 80.4%
--
EPOCH 184
 * Acc@1 79.334 Acc@5 94.908
Accuracy of the network on the 50000 test images: 79.3%
 * Acc@1 80.406 Acc@5 95.412
Accuracy of the ema network on the 50000 test images: 80.4%
--
EPOCH 185
 * Acc@1 79.624 Acc@5 95.138
Accuracy of the network on the 50000 test images: 79.6%
 * Acc@1 80.388 Acc@5 95.430
Accuracy of the ema network on the 50000 test images: 80.4%
--
EPOCH 186
 * Acc@1 79.696 Acc@5 95.026
Accuracy of the network on the 50000 test images: 79.7%
 * Acc@1 80.432 Acc@5 95.446
Accuracy of the ema network on the 50000 test images: 80.4%
--
EPOCH 187
 * Acc@1 79.676 Acc@5 95.114
Accuracy of the network on the 50000 test images: 79.7%
 * Acc@1 80.484 Acc@5 95.438
Accuracy of the ema network on the 50000 test images: 80.5%
--
EPOCH 188
 * Acc@1 79.564 Acc@5 94.996
Accuracy of the network on the 50000 test images: 79.6%
 * Acc@1 80.494 Acc@5 95.450
Accuracy of the ema network on the 50000 test images: 80.5%
--
EPOCH 189
 * Acc@1 79.554 Acc@5 94.996
Accuracy of the network on the 50000 test images: 79.6%
 * Acc@1 80.522 Acc@5 95.484
Accuracy of the ema network on the 50000 test images: 80.5%
--
EPOCH 190
 * Acc@1 79.614 Acc@5 95.094
Accuracy of the network on the 50000 test images: 79.6%
 * Acc@1 80.538 Acc@5 95.502
Accuracy of the ema network on the 50000 test images: 80.5%
--
EPOCH 191
 * Acc@1 79.678 Acc@5 95.140
Accuracy of the network on the 50000 test images: 79.7%
 * Acc@1 80.544 Acc@5 95.518
Accuracy of the ema network on the 50000 test images: 80.5%
--
EPOCH 192
 * Acc@1 79.602 Acc@5 95.030
Accuracy of the network on the 50000 test images: 79.6%
 * Acc@1 80.576 Acc@5 95.510
Accuracy of the ema network on the 50000 test images: 80.6%
--
EPOCH 193
 * Acc@1 79.670 Acc@5 95.118
Accuracy of the network on the 50000 test images: 79.7%
 * Acc@1 80.582 Acc@5 95.516
Accuracy of the ema network on the 50000 test images: 80.6%
--
EPOCH 194
 * Acc@1 79.822 Acc@5 95.246
Accuracy of the network on the 50000 test images: 79.8%
 * Acc@1 80.576 Acc@5 95.542
Accuracy of the ema network on the 50000 test images: 80.6%
--
EPOCH 195
 * Acc@1 79.760 Acc@5 95.180
Accuracy of the network on the 50000 test images: 79.8%
 * Acc@1 80.606 Acc@5 95.536
Accuracy of the ema network on the 50000 test images: 80.6%
--
EPOCH 196
 * Acc@1 79.710 Acc@5 95.248
Accuracy of the network on the 50000 test images: 79.7%
 * Acc@1 80.662 Acc@5 95.550
Accuracy of the ema network on the 50000 test images: 80.7%
--
EPOCH 197
 * Acc@1 79.816 Acc@5 95.172
Accuracy of the network on the 50000 test images: 79.8%
 * Acc@1 80.686 Acc@5 95.566
Accuracy of the ema network on the 50000 test images: 80.7%
--
EPOCH 198
 * Acc@1 79.816 Acc@5 95.220
Accuracy of the network on the 50000 test images: 79.8%
 * Acc@1 80.740 Acc@5 95.586
Accuracy of the ema network on the 50000 test images: 80.7%
--
EPOCH 199
 * Acc@1 79.886 Acc@5 95.196
Accuracy of the network on the 50000 test images: 79.9%
 * Acc@1 80.762 Acc@5 95.614
Accuracy of the ema network on the 50000 test images: 80.8%
--
EPOCH 200
 * Acc@1 79.876 Acc@5 95.262
Accuracy of the network on the 50000 test images: 79.9%
 * Acc@1 80.782 Acc@5 95.606
Accuracy of the ema network on the 50000 test images: 80.8%
--
EPOCH 201
 * Acc@1 80.010 Acc@5 95.312
Accuracy of the network on the 50000 test images: 80.0%
 * Acc@1 80.808 Acc@5 95.594
Accuracy of the ema network on the 50000 test images: 80.8%
--
EPOCH 202
 * Acc@1 80.042 Acc@5 95.274
Accuracy of the network on the 50000 test images: 80.0%
 * Acc@1 80.830 Acc@5 95.580
Accuracy of the ema network on the 50000 test images: 80.8%
--
EPOCH 203
 * Acc@1 79.756 Acc@5 95.164
Accuracy of the network on the 50000 test images: 79.8%
 * Acc@1 80.818 Acc@5 95.582
Accuracy of the ema network on the 50000 test images: 80.8%
--
EPOCH 204
 * Acc@1 80.216 Acc@5 95.232
Accuracy of the network on the 50000 test images: 80.2%
 * Acc@1 80.844 Acc@5 95.604
Accuracy of the ema network on the 50000 test images: 80.8%
--
EPOCH 205
 * Acc@1 80.248 Acc@5 95.362
Accuracy of the network on the 50000 test images: 80.2%
 * Acc@1 80.876 Acc@5 95.610
Accuracy of the ema network on the 50000 test images: 80.9%
--
EPOCH 206
 * Acc@1 80.186 Acc@5 95.400
Accuracy of the network on the 50000 test images: 80.2%
 * Acc@1 80.910 Acc@5 95.642
Accuracy of the ema network on the 50000 test images: 80.9%
--
EPOCH 207
 * Acc@1 80.306 Acc@5 95.386
Accuracy of the network on the 50000 test images: 80.3%
 * Acc@1 80.918 Acc@5 95.668
Accuracy of the ema network on the 50000 test images: 80.9%
--
EPOCH 208
 * Acc@1 80.118 Acc@5 95.214
Accuracy of the network on the 50000 test images: 80.1%
 * Acc@1 80.930 Acc@5 95.670
Accuracy of the ema network on the 50000 test images: 80.9%
--
EPOCH 209
 * Acc@1 80.264 Acc@5 95.344
Accuracy of the network on the 50000 test images: 80.3%
 * Acc@1 80.942 Acc@5 95.652
Accuracy of the ema network on the 50000 test images: 80.9%
--
EPOCH 210
 * Acc@1 80.312 Acc@5 95.306
Accuracy of the network on the 50000 test images: 80.3%
 * Acc@1 80.988 Acc@5 95.654
Accuracy of the ema network on the 50000 test images: 81.0%
--
EPOCH 211
 * Acc@1 80.398 Acc@5 95.386
Accuracy of the network on the 50000 test images: 80.4%
 * Acc@1 81.014 Acc@5 95.650
Accuracy of the ema network on the 50000 test images: 81.0%
--
EPOCH 212
 * Acc@1 80.500 Acc@5 95.298
Accuracy of the network on the 50000 test images: 80.5%
 * Acc@1 81.062 Acc@5 95.650
Accuracy of the ema network on the 50000 test images: 81.1%
--
EPOCH 213
 * Acc@1 80.434 Acc@5 95.442
Accuracy of the network on the 50000 test images: 80.4%
 * Acc@1 81.078 Acc@5 95.678
Accuracy of the ema network on the 50000 test images: 81.1%
--
EPOCH 214
 * Acc@1 80.394 Acc@5 95.446
Accuracy of the network on the 50000 test images: 80.4%
 * Acc@1 81.092 Acc@5 95.696
Accuracy of the ema network on the 50000 test images: 81.1%
--
EPOCH 215
 * Acc@1 80.494 Acc@5 95.502
Accuracy of the network on the 50000 test images: 80.5%
 * Acc@1 81.112 Acc@5 95.702
Accuracy of the ema network on the 50000 test images: 81.1%
--
EPOCH 216
 * Acc@1 80.488 Acc@5 95.586
Accuracy of the network on the 50000 test images: 80.5%
 * Acc@1 81.130 Acc@5 95.708
Accuracy of the ema network on the 50000 test images: 81.1%
--
EPOCH 217
 * Acc@1 80.550 Acc@5 95.500
Accuracy of the network on the 50000 test images: 80.5%
 * Acc@1 81.158 Acc@5 95.730
Accuracy of the ema network on the 50000 test images: 81.2%
--
EPOCH 218
 * Acc@1 80.530 Acc@5 95.490
Accuracy of the network on the 50000 test images: 80.5%
 * Acc@1 81.170 Acc@5 95.738
Accuracy of the ema network on the 50000 test images: 81.2%
--
EPOCH 219
 * Acc@1 80.602 Acc@5 95.484
Accuracy of the network on the 50000 test images: 80.6%
 * Acc@1 81.204 Acc@5 95.750
Accuracy of the ema network on the 50000 test images: 81.2%
--
EPOCH 220
 * Acc@1 80.500 Acc@5 95.476
Accuracy of the network on the 50000 test images: 80.5%
 * Acc@1 81.222 Acc@5 95.744
Accuracy of the ema network on the 50000 test images: 81.2%
--
EPOCH 221
 * Acc@1 80.646 Acc@5 95.510
Accuracy of the network on the 50000 test images: 80.6%
 * Acc@1 81.210 Acc@5 95.742
Accuracy of the ema network on the 50000 test images: 81.2%
--
EPOCH 222
 * Acc@1 80.536 Acc@5 95.496
Accuracy of the network on the 50000 test images: 80.5%
 * Acc@1 81.250 Acc@5 95.766
Accuracy of the ema network on the 50000 test images: 81.2%
--
EPOCH 223
 * Acc@1 80.382 Acc@5 95.418
Accuracy of the network on the 50000 test images: 80.4%
 * Acc@1 81.242 Acc@5 95.780
Accuracy of the ema network on the 50000 test images: 81.2%
--
EPOCH 224
 * Acc@1 80.794 Acc@5 95.592
Accuracy of the network on the 50000 test images: 80.8%
 * Acc@1 81.270 Acc@5 95.768
Accuracy of the ema network on the 50000 test images: 81.3%
--
EPOCH 225
 * Acc@1 80.562 Acc@5 95.564
Accuracy of the network on the 50000 test images: 80.6%
 * Acc@1 81.270 Acc@5 95.766
Accuracy of the ema network on the 50000 test images: 81.3%
--
EPOCH 226
 * Acc@1 80.556 Acc@5 95.522
Accuracy of the network on the 50000 test images: 80.6%
 * Acc@1 81.236 Acc@5 95.780
Accuracy of the ema network on the 50000 test images: 81.2%
--
EPOCH 227
 * Acc@1 80.808 Acc@5 95.622
Accuracy of the network on the 50000 test images: 80.8%
 * Acc@1 81.236 Acc@5 95.772
Accuracy of the ema network on the 50000 test images: 81.2%
--
EPOCH 228
 * Acc@1 80.832 Acc@5 95.674
Accuracy of the network on the 50000 test images: 80.8%
 * Acc@1 81.272 Acc@5 95.762
Accuracy of the ema network on the 50000 test images: 81.3%
--
EPOCH 229
 * Acc@1 80.836 Acc@5 95.574
Accuracy of the network on the 50000 test images: 80.8%
 * Acc@1 81.296 Acc@5 95.800
Accuracy of the ema network on the 50000 test images: 81.3%
--
EPOCH 230
 * Acc@1 80.866 Acc@5 95.574
Accuracy of the network on the 50000 test images: 80.9%
 * Acc@1 81.312 Acc@5 95.798
Accuracy of the ema network on the 50000 test images: 81.3%
--
EPOCH 231
 * Acc@1 80.848 Acc@5 95.626
Accuracy of the network on the 50000 test images: 80.8%
 * Acc@1 81.370 Acc@5 95.798
Accuracy of the ema network on the 50000 test images: 81.4%
--
EPOCH 232
 * Acc@1 80.966 Acc@5 95.636
Accuracy of the network on the 50000 test images: 81.0%
 * Acc@1 81.396 Acc@5 95.812
Accuracy of the ema network on the 50000 test images: 81.4%
--
EPOCH 233
 * Acc@1 80.926 Acc@5 95.656
Accuracy of the network on the 50000 test images: 80.9%
 * Acc@1 81.432 Acc@5 95.836
Accuracy of the ema network on the 50000 test images: 81.4%
--
EPOCH 234
 * Acc@1 80.954 Acc@5 95.702
Accuracy of the network on the 50000 test images: 81.0%
 * Acc@1 81.420 Acc@5 95.854
Accuracy of the ema network on the 50000 test images: 81.4%
--
EPOCH 235
 * Acc@1 80.892 Acc@5 95.680
Accuracy of the network on the 50000 test images: 80.9%
 * Acc@1 81.450 Acc@5 95.876
Accuracy of the ema network on the 50000 test images: 81.5%
--
EPOCH 236
 * Acc@1 81.064 Acc@5 95.678
Accuracy of the network on the 50000 test images: 81.1%
 * Acc@1 81.482 Acc@5 95.874
Accuracy of the ema network on the 50000 test images: 81.5%
--
EPOCH 237
 * Acc@1 81.106 Acc@5 95.670
Accuracy of the network on the 50000 test images: 81.1%
 * Acc@1 81.500 Acc@5 95.874
Accuracy of the ema network on the 50000 test images: 81.5%
--
EPOCH 238
 * Acc@1 81.244 Acc@5 95.778
Accuracy of the network on the 50000 test images: 81.2%
 * Acc@1 81.494 Acc@5 95.888
Accuracy of the ema network on the 50000 test images: 81.5%
--
EPOCH 239
 * Acc@1 81.078 Acc@5 95.690
Accuracy of the network on the 50000 test images: 81.1%
 * Acc@1 81.514 Acc@5 95.894
Accuracy of the ema network on the 50000 test images: 81.5%
--
EPOCH 240
 * Acc@1 81.108 Acc@5 95.722
Accuracy of the network on the 50000 test images: 81.1%
 * Acc@1 81.538 Acc@5 95.898
Accuracy of the ema network on the 50000 test images: 81.5%
--
EPOCH 241
 * Acc@1 81.220 Acc@5 95.772
Accuracy of the network on the 50000 test images: 81.2%
 * Acc@1 81.548 Acc@5 95.906
Accuracy of the ema network on the 50000 test images: 81.5%
--
EPOCH 242
 * Acc@1 81.334 Acc@5 95.796
Accuracy of the network on the 50000 test images: 81.3%
 * Acc@1 81.566 Acc@5 95.920
Accuracy of the ema network on the 50000 test images: 81.6%
--
EPOCH 243
 * Acc@1 81.354 Acc@5 95.746
Accuracy of the network on the 50000 test images: 81.4%
 * Acc@1 81.602 Acc@5 95.922
Accuracy of the ema network on the 50000 test images: 81.6%
--
EPOCH 244
 * Acc@1 81.248 Acc@5 95.778
Accuracy of the network on the 50000 test images: 81.2%
 * Acc@1 81.616 Acc@5 95.920
Accuracy of the ema network on the 50000 test images: 81.6%
--
EPOCH 245
 * Acc@1 81.312 Acc@5 95.770
Accuracy of the network on the 50000 test images: 81.3%
 * Acc@1 81.598 Acc@5 95.912
Accuracy of the ema network on the 50000 test images: 81.6%
--
EPOCH 246
 * Acc@1 81.430 Acc@5 95.816
Accuracy of the network on the 50000 test images: 81.4%
 * Acc@1 81.636 Acc@5 95.908
Accuracy of the ema network on the 50000 test images: 81.6%
--
EPOCH 247
 * Acc@1 81.330 Acc@5 95.786
Accuracy of the network on the 50000 test images: 81.3%
 * Acc@1 81.650 Acc@5 95.920
Accuracy of the ema network on the 50000 test images: 81.6%
--
EPOCH 248
 * Acc@1 81.542 Acc@5 95.930
Accuracy of the network on the 50000 test images: 81.5%
 * Acc@1 81.634 Acc@5 95.934
Accuracy of the ema network on the 50000 test images: 81.6%
--
EPOCH 249
 * Acc@1 81.426 Acc@5 95.908
Accuracy of the network on the 50000 test images: 81.4%
 * Acc@1 81.662 Acc@5 95.948
Accuracy of the ema network on the 50000 test images: 81.7%
--
EPOCH 250
 * Acc@1 81.532 Acc@5 95.844
Accuracy of the network on the 50000 test images: 81.5%
 * Acc@1 81.726 Acc@5 95.966
Accuracy of the ema network on the 50000 test images: 81.7%
--
EPOCH 251
 * Acc@1 81.498 Acc@5 95.856
Accuracy of the network on the 50000 test images: 81.5%
 * Acc@1 81.704 Acc@5 95.984
Accuracy of the ema network on the 50000 test images: 81.7%
--
EPOCH 252
 * Acc@1 81.398 Acc@5 95.838
Accuracy of the network on the 50000 test images: 81.4%
 * Acc@1 81.700 Acc@5 96.004
Accuracy of the ema network on the 50000 test images: 81.7%
--
EPOCH 253
 * Acc@1 81.418 Acc@5 95.886
Accuracy of the network on the 50000 test images: 81.4%
 * Acc@1 81.692 Acc@5 96.008
Accuracy of the ema network on the 50000 test images: 81.7%
--
EPOCH 254
 * Acc@1 81.642 Acc@5 95.894
Accuracy of the network on the 50000 test images: 81.6%
 * Acc@1 81.704 Acc@5 96.002
Accuracy of the ema network on the 50000 test images: 81.7%
--
EPOCH 255
 * Acc@1 81.444 Acc@5 95.906
Accuracy of the network on the 50000 test images: 81.4%
 * Acc@1 81.690 Acc@5 96.008
Accuracy of the ema network on the 50000 test images: 81.7%
--
EPOCH 256
 * Acc@1 81.590 Acc@5 95.942
Accuracy of the network on the 50000 test images: 81.6%
 * Acc@1 81.700 Acc@5 96.014
Accuracy of the ema network on the 50000 test images: 81.7%
--
EPOCH 257
 * Acc@1 81.550 Acc@5 95.920
Accuracy of the network on the 50000 test images: 81.6%
 * Acc@1 81.742 Acc@5 96.014
Accuracy of the ema network on the 50000 test images: 81.7%
--
EPOCH 258
 * Acc@1 81.554 Acc@5 95.924
Accuracy of the network on the 50000 test images: 81.6%
 * Acc@1 81.758 Acc@5 96.012
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 259
 * Acc@1 81.576 Acc@5 95.896
Accuracy of the network on the 50000 test images: 81.6%
 * Acc@1 81.766 Acc@5 96.016
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 260
 * Acc@1 81.662 Acc@5 95.966
Accuracy of the network on the 50000 test images: 81.7%
 * Acc@1 81.790 Acc@5 96.006
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 261
 * Acc@1 81.754 Acc@5 95.950
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.800 Acc@5 96.016
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 262
 * Acc@1 81.732 Acc@5 96.002
Accuracy of the network on the 50000 test images: 81.7%
 * Acc@1 81.788 Acc@5 96.012
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 263
 * Acc@1 81.714 Acc@5 95.988
Accuracy of the network on the 50000 test images: 81.7%
 * Acc@1 81.826 Acc@5 96.034
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 264
 * Acc@1 81.750 Acc@5 95.986
Accuracy of the network on the 50000 test images: 81.7%
 * Acc@1 81.832 Acc@5 96.044
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 265
 * Acc@1 81.784 Acc@5 95.960
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.840 Acc@5 96.028
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 266
 * Acc@1 81.662 Acc@5 95.972
Accuracy of the network on the 50000 test images: 81.7%
 * Acc@1 81.832 Acc@5 96.018
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 267
 * Acc@1 81.750 Acc@5 95.980
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.838 Acc@5 96.032
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 268
 * Acc@1 81.860 Acc@5 96.036
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.820 Acc@5 96.044
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 269
 * Acc@1 81.742 Acc@5 96.016
Accuracy of the network on the 50000 test images: 81.7%
 * Acc@1 81.798 Acc@5 96.056
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 270
 * Acc@1 81.868 Acc@5 96.048
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.816 Acc@5 96.054
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 271
 * Acc@1 81.828 Acc@5 95.950
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.848 Acc@5 96.056
Accuracy of the ema network on the 50000 test images: 81.8%
--
EPOCH 272
 * Acc@1 81.836 Acc@5 96.000
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.876 Acc@5 96.052
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 273
 * Acc@1 81.826 Acc@5 95.956
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.898 Acc@5 96.060
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 274
 * Acc@1 81.848 Acc@5 96.026
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.928 Acc@5 96.086
Accuracy of the ema network on the 50000 test images: 81.9%
--
EPOCH 275
 * Acc@1 81.856 Acc@5 96.006
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.968 Acc@5 96.096
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 276
 * Acc@1 81.862 Acc@5 96.038
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.968 Acc@5 96.086
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 277
 * Acc@1 81.884 Acc@5 96.030
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.962 Acc@5 96.090
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 278
 * Acc@1 81.988 Acc@5 96.082
Accuracy of the network on the 50000 test images: 82.0%
 * Acc@1 81.950 Acc@5 96.086
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 279
 * Acc@1 81.842 Acc@5 96.000
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 81.978 Acc@5 96.076
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 280
 * Acc@1 81.940 Acc@5 96.068
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.970 Acc@5 96.086
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 281
 * Acc@1 81.938 Acc@5 96.084
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.960 Acc@5 96.080
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 282
 * Acc@1 81.990 Acc@5 96.120
Accuracy of the network on the 50000 test images: 82.0%
 * Acc@1 81.980 Acc@5 96.076
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 283
 * Acc@1 81.948 Acc@5 96.144
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 81.990 Acc@5 96.076
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 284
 * Acc@1 81.994 Acc@5 96.110
Accuracy of the network on the 50000 test images: 82.0%
 * Acc@1 82.000 Acc@5 96.078
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 285
 * Acc@1 81.848 Acc@5 96.040
Accuracy of the network on the 50000 test images: 81.8%
 * Acc@1 82.012 Acc@5 96.088
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 286
 * Acc@1 82.038 Acc@5 96.060
Accuracy of the network on the 50000 test images: 82.0%
 * Acc@1 82.018 Acc@5 96.088
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 287
 * Acc@1 81.984 Acc@5 96.056
Accuracy of the network on the 50000 test images: 82.0%
 * Acc@1 82.022 Acc@5 96.084
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 288
 * Acc@1 81.926 Acc@5 95.982
Accuracy of the network on the 50000 test images: 81.9%
 * Acc@1 82.018 Acc@5 96.098
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 289
 * Acc@1 82.008 Acc@5 96.030
Accuracy of the network on the 50000 test images: 82.0%
 * Acc@1 82.028 Acc@5 96.092
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 290
 * Acc@1 82.038 Acc@5 96.078
Accuracy of the network on the 50000 test images: 82.0%
 * Acc@1 82.048 Acc@5 96.102
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 291
 * Acc@1 82.078 Acc@5 96.110
Accuracy of the network on the 50000 test images: 82.1%
 * Acc@1 82.052 Acc@5 96.100
Accuracy of the ema network on the 50000 test images: 82.1%
--
EPOCH 292
 * Acc@1 82.072 Acc@5 96.104
Accuracy of the network on the 50000 test images: 82.1%
 * Acc@1 82.052 Acc@5 96.096
Accuracy of the ema network on the 50000 test images: 82.1%
--
EPOCH 293
 * Acc@1 81.950 Acc@5 96.074
Accuracy of the network on the 50000 test images: 82.0%
 * Acc@1 82.034 Acc@5 96.106
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 294
 * Acc@1 82.032 Acc@5 96.108
Accuracy of the network on the 50000 test images: 82.0%
 * Acc@1 82.056 Acc@5 96.108
Accuracy of the ema network on the 50000 test images: 82.1%
--
EPOCH 295
 * Acc@1 82.032 Acc@5 96.104
Accuracy of the network on the 50000 test images: 82.0%
 * Acc@1 82.048 Acc@5 96.100
Accuracy of the ema network on the 50000 test images: 82.0%
--
EPOCH 296
 * Acc@1 82.024 Acc@5 96.104
Accuracy of the network on the 50000 test images: 82.0%
 * Acc@1 82.052 Acc@5 96.100
Accuracy of the ema network on the 50000 test images: 82.1%
--
EPOCH 297
 * Acc@1 82.050 Acc@5 96.088
Accuracy of the network on the 50000 test images: 82.1%
 * Acc@1 82.058 Acc@5 96.102
Accuracy of the ema network on the 50000 test images: 82.1%
--
EPOCH 298
 * Acc@1 82.072 Acc@5 96.084
Accuracy of the network on the 50000 test images: 82.1%
 * Acc@1 82.062 Acc@5 96.096
Accuracy of the ema network on the 50000 test images: 82.1%
--
EPOCH 299
 * Acc@1 81.998 Acc@5 96.088
Accuracy of the network on the 50000 test images: 82.0%
 * Acc@1 82.066 Acc@5 96.104
Accuracy of the ema network on the 50000 test images: 82.1%
--
EPOCH 300
 * Acc@1 82.052 Acc@5 96.064
Accuracy of the network on the 50000 test images: 82.1%
 * Acc@1 82.076 Acc@5 96.108
Accuracy of the ema network on the 50000 test images: 82.1%