EfficientNetV2-S implementation using PyTorch
- Configure
imagenet
path by changingdata_dir
intrain.py
python main.py --benchmark
for model informationpython main.py --train
for training modelpython main.py --eval
for evaluation
Number of parameters: 23941296
Time per operator type:
1740.04 ms. 79.8733%. Conv
246.38 ms. 11.3096%. Sigmoid
165.81 ms. 7.6112%. Mul
18.7233 ms. 0.85946%. Add
7.54697 ms. 0.346429%. ReduceMean
2178.5 ms in Total
FLOP per operator type:
17.277 GFLOP. 99.728%. Conv
0.0419269 GFLOP. 0.242015%. Mul
0.00519322 GFLOP. 0.0299768%. Add
17.3241 GFLOP in Total
Feature Memory Read per operator type:
295.89 MB. 51.1414%. Mul
241.136 MB. 41.6778%. Conv
41.5457 MB. 7.18074%. Add
578.572 MB in Total
Feature Memory Written per operator type:
167.708 MB. 49.2377%. Mul
152.127 MB. 44.6635%. Conv
20.7729 MB. 6.09876%. Add
340.608 MB in Total
Parameter Memory per operator type:
87.8034 MB. 100%. Conv
0 MB. 0%. Add
0 MB. 0%. Mul
87.8034 MB in Total