Home

Awesome

THOP: PyTorch-OpCounter

How to install

pip install thop (now continously intergrated on Github actions)

OR

pip install --upgrade git+https://github.com/Lyken17/pytorch-OpCounter.git

How to use

Results of Recent Models

The implementation are adapted from torchvision. Following results can be obtained using benchmark/evaluate_famous_models.py.

<p align="center"> <table> <tr> <td>
ModelParams(M)MACs(G)
alexnet61.100.77
vgg11132.867.74
vgg11_bn132.877.77
vgg13133.0511.44
vgg13_bn133.0511.49
vgg16138.3615.61
vgg16_bn138.3715.66
vgg19143.6719.77
vgg19_bn143.6819.83
resnet1811.691.82
resnet3421.803.68
resnet5025.564.14
resnet10144.557.87
resnet15260.1911.61
wide_resnet101_2126.8922.84
wide_resnet50_268.8811.46
</td> <td>
ModelParams(M)MACs(G)
resnext50_32x4d25.034.29
resnext101_32x8d88.7916.54
densenet1217.982.90
densenet16128.687.85
densenet16914.153.44
densenet20120.014.39
squeezenet1_01.250.82
squeezenet1_11.240.35
mnasnet0_52.220.14
mnasnet0_753.170.24
mnasnet1_04.380.34
mnasnet1_36.280.53
mobilenet_v23.500.33
shufflenet_v2_x0_51.370.05
shufflenet_v2_x1_02.280.15
shufflenet_v2_x1_53.500.31
shufflenet_v2_x2_07.390.60
inception_v327.165.75
</td> </tr> </p>