EfficientNetV2-S implementation using PyTorch

Train Steps

• Modify imagenet path by changing data_dir in train.py
• python main.py --benchmark for model information
• python main.py --train for training model
• python main.py --eval for evaluation

Number of parameters: 23987826
Time per operator type:
        255.614 ms.    65.2844%. Conv
        93.8097 ms.    23.9592%. Sigmoid
        33.5431 ms.    8.56699%. Mul
        5.08615 ms.    1.29901%. ReduceMean
        3.48612 ms.   0.890364%. Add
        391.539 ms in Total
FLOP per operator type:
        17.2771 GFLOP.    99.6695%. Conv
      0.0521015 GFLOP.   0.300567%. Mul
     0.00519322 GFLOP.  0.0299591%. Add
        17.3344 GFLOP in Total
Feature Memory Read per operator type:
        336.595 MB.     54.336%. Mul
        241.329 MB.    38.9574%. Conv
        41.5457 MB.    6.70665%. Add
         619.47 MB in Total
Feature Memory Written per operator type:
        208.406 MB.    54.6548%. Mul
        152.135 MB.    39.8975%. Conv
        20.7729 MB.    5.44771%. Add
        381.313 MB in Total
Parameter Memory per operator type:
        87.9824 MB.        100%. Conv
              0 MB.          0%. Add
              0 MB.          0%. Mul
        87.9824 MB in Total