Giters

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), MobileNet-V3/V2, RegNet, DPN, CSPNet, Swin Transformer, MaxViT, CoAtNet, ConvNeXt, and more

Home Page:https://huggingface.co/docs/timm

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

[BUG] unknown model

yolandazhou1222 opened this issue · comments

commented

These three models can be recognized in timm 0.8.19.dev0, but report unknown model in timm 0.8.21.dev0 and 0.9.12.
['gluon_xception65', 'tv_densenet121', 'xcit_small_24_p16_384_dist'])
'xcit_small_24_p16_384_dist' seemed to be renamed to 'xcit_small_24_p16_384.fb_dist_in1k', because their pth file url are the same.
Did the other two models also change their names? Or they are not supported now?

commented

@yolandazhou1222 the densenet and xcit were renamed. I actually have a mapping mechanism for that but looks like I missed the densenet ones and one and only xcit was missed. I can add remapping for those so old names work..

However, I did remove the gluon_xception alltogether. I think there was something not quite right about that iteration of the model compared to the original TF version it was based on. So I did my own aligned xception impl, converted tensorflow weights, and trained my own weights as well.