https://github.com/clcarwin/convert_torch_to_pytorch

Convert torch t7 model to pytorch model and source.
https://github.com/clcarwin/convert_torch_to_pytorch

pytorch

Last synced: about 1 month ago
JSON representation

Convert torch t7 model to pytorch model and source.

• Host: GitHub
• URL: https://github.com/clcarwin/convert_torch_to_pytorch
• Owner: clcarwin
• License: mit
• Created: 2017-02-16T06:23:52.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2023-07-10T23:21:03.000Z (11 months ago)
• Last Synced: 2024-01-25T18:07:55.851Z (4 months ago)
• Topics: pytorch
• Language: Python
• Size: 19.5 KB
• Stars: 533
• Watchers: 16
• Forks: 161
• Open Issues: 28
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Lists

• Awesome-pytorch-list - convert_torch_to_pytorch
• Awesome-pytorch-list-CNVersion - convert_torch_to_pytorch

README

        
# Convert torch to pytorch

Convert torch t7 model to pytorch model and source.

## Convert

```bash

python convert_torch.py -m vgg16.t7

```

Two file will be created ```vgg16.py``` ```vgg16.pth```

## Example

```python

import vgg16

model = vgg16.vgg16

model.load_state_dict(torch.load('vgg16.pth'))

model.eval()

...

```

## Validated

All the models in this table can be converted and the results have been validated.

| Network             | Download |

| ------------------- | -------- |

| AlexNet | [cnn-benchmarks](https://github.com/jcjohnson/cnn-benchmarks) |

| Inception-V1 | [cnn-benchmarks](https://github.com/jcjohnson/cnn-benchmarks) |

| VGG-16 | [cnn-benchmarks](https://github.com/jcjohnson/cnn-benchmarks) |

| VGG-19 | [cnn-benchmarks](https://github.com/jcjohnson/cnn-benchmarks) |

| ResNet-18 | [cnn-benchmarks](https://github.com/jcjohnson/cnn-benchmarks) |

| ResNet-200 | [cnn-benchmarks](https://github.com/jcjohnson/cnn-benchmarks) |

| ResNeXt-50 (32x4d) | [ResNeXt](https://github.com/facebookresearch/ResNeXt) |

| ResNeXt-101 (32x4d) | [ResNeXt](https://github.com/facebookresearch/ResNeXt) |

| ResNeXt-101 (64x4d) | [ResNeXt](https://github.com/facebookresearch/ResNeXt) |

| DenseNet-264 (k=32) | [DenseNet](https://github.com/liuzhuang13/DenseNet#results-on-imagenet-and-pretrained-models) |

| DenseNet-264 (k=48) | [DenseNet](https://github.com/liuzhuang13/DenseNet#results-on-imagenet-and-pretrained-models) |