classification-cifar10-pytorch:通过PyTorch在cifar10数据集中训练几个经典分类网络

# Classification-cifar10-pytorch I am testing several classical classification networks performance on cifar10 dataset by PyTorch! [Chinese blog](https://blog.csdn.net/laizi_laizi/article/details/103006497) # Requirements - pytorch - torchsummary - python3.x # Results | Model | My Acc. | Total params | Estimated Total Size (MB) | Trainable params | Params size (MB) |Saved model size (MB)|GPU memory usage(MB) | ----------------- | ----------- | ------ | ---|--- | --- | --- |--- | | [MobileNetV2](https://arxiv.org/abs/1801.04381) | 92.64% | 2,296,922 | 36.14 | 2,296,922 | 8.76 | 8.96 | 3107 | | [VGG16](https://arxiv.org/abs/1409.1556) | 94.27% | 14,728,266 | 62.77 |14,728,266 |56.18 |59.0 | 1229 | | [PreActResNet18](https://arxiv.org/abs/1603.05027) | 94.70% | 11,171,146 | 53.38 | 11,171,146 | 42.61 | 44.7 | 1665 | | [ResNeXt29(2x64d)](https://arxiv.org/abs/1611.05431) | 95.09% | 9,128,778 | 99.84 | 9,128,778 | 34.82 | 36.7 | 5779 | | [ResNet50](https://arxiv.org/abs/1512.03385) | 95.22% | 23,520,842 | 155.86 | 23,520,842 | 89.72 | 94.4 | 5723 | | [DPN92](https://arxiv.org/abs/1707.01629) | 95.42% | 34,236,634 | 243.50 | 34,236,634 | 130.60 | 137.5 | 10535 | | [ResNeXt29(32x4d)](https://arxiv.org/abs/1611.05431) | 95.49% | 4,774,218 | 83.22 | 4,774,218 | 18.21 | 19.2 | 5817 | | [DenseNet121](https://arxiv.org/abs/1608.06993) | 95.55% | 6,956,298 | 105.05 | 6,956,298 | 26.54 | 28.3 | 8203 | | [ResNet18](https://arxiv.org/abs/1512.03385) | 95.59% | 11,173,962 | 53.89 | 11,173,962 | 42.63 | 44.8 | 1615 | | [ResNet101](https://arxiv.org/abs/1512.03385) | 95.62% | 42,512,970 | 262.31 | 42,512,970 | 162.17 | 170.6 | 8857 | **Note**: 1. Above GPU memory usage(MB) was observed with batch size=128. 2. For PreActResNet18, I set initial learning rate=0.1, but it can't converge, so I set it's initial lr=0.01. 3. I firstly train **VGG16**, **ResNet18** and **ResNet50** with total epochs=400. But I want to get results earlier, so for remaining networks, I set total epochs=300 (besides, afterwards it just improve a little). 4. Run the scripts: `python main.py --resume --lr 0.1 --trainbs 128 --testbs 100` # Pre-trained models You can obtain pre-traind models(as above list) from here: [[Baidu Drive](https://pan.baidu.com/s/1oUfaxFnghIdClCFMf3A11Q)] [[Google Drive](https://drive.google.com/open?id=1PLwxkczvKq86ATRD7SB-5w31omuORUNV)]