https://github.com/nekitmm/funnelact_pytorch
pytorch implementation of Funnel Activation (FReLU)
https://github.com/nekitmm/funnelact_pytorch
deep-learning deep-neural-networks frelu funnel-act funnel-activation pytorch resnet
Last synced: about 1 year ago
JSON representation
pytorch implementation of Funnel Activation (FReLU)
- Host: GitHub
- URL: https://github.com/nekitmm/funnelact_pytorch
- Owner: nekitmm
- License: mit
- Created: 2020-08-15T20:55:16.000Z (almost 6 years ago)
- Default Branch: master
- Last Pushed: 2020-08-16T22:57:18.000Z (almost 6 years ago)
- Last Synced: 2025-05-06T19:13:35.560Z (about 1 year ago)
- Topics: deep-learning, deep-neural-networks, frelu, funnel-act, funnel-activation, pytorch, resnet
- Language: Python
- Homepage:
- Size: 89.4 MB
- Stars: 16
- Watchers: 1
- Forks: 2
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# FunnelAct Pytorch
Pytorch implementation of Funnel Activation (FReLU): https://arxiv.org/pdf/2007.11824.pdf
Validation results are listed below:
| Model | Activation | Err@1 | Err@5 |
| :---------------------- | :--------: | :------: | :------: |
| ResNet50 | FReLU | **22.40** | **6.164** |
Note that from the file resnet_frelu.py you can call ResNet18, ResNet34, ResNet50, ResNet101 and ResNet152
but the weights in this repo only available for ResNet50 and I never tried to train other models,
so no guaranties there!
The code in this repo is based on pytorch imagenet example:
https://github.com/pytorch/examples/tree/master/imagenet
and original implementation of Funnel Activation in Megengine:
https://github.com/megvii-model/FunnelAct
Enjoy!