[BUG] No `forward_features` for mlp_mixer

huggingface / pytorch-image-models

The largest collection of PyTorch image encoders / backbones. Including train, eval, inference, export scripts, and pretrained weights -- ResNet, ResNeXT, EfficientNet, NFNet, Vision Transformer (ViT), MobileNetV4, MobileNet-V3 & V2, RegNet, DPN, CSPNet, Swin Transformer, MaxViT, CoAtNet, ConvNeXt, and more

https://huggingface.co/docs/timm

Apache License 2.0

32.18k stars 4.76k forks source link

[BUG] No `forward_features` for mlp_mixer #694

Closed irvingzhang0512 closed 3 years ago

irvingzhang0512 commented 3 years ago

Hi, thanks for your work! It's very helpful. I have create a PR to use timm as backbones for TSN(action recognizer) open-mmlab/mmaction2#880

Describe the bug There is no forward_features in mlp_mixer.

https://github.com/rwightman/pytorch-image-models/blob/07fb05cc3d5c6f99d5b74a88e9a14db42e8e2e4f/timm/models/mlp_mixer.py#L244-L250

Expected behavior All models share the same structure, aka forward_features + head, so that we can extract features with unified interface

rwightman commented 3 years ago

@irvingzhang0512 currently have that change with some other adjustments locally, will get it up soon(ish)