search

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 4.75k forks source link

[DOCS] Define "SSL" and SWSL" in the docs(strings) #1458

Closed jpcbertoldo closed 2 years ago

jpcbertoldo commented 2 years ago

Is your feature request related to a problem? Please describe.

I was not aware of what "SSL" and "SWSL" in the model names stand for. As I was browsing some models I saw this and started to Ctrl + F for SSL and SWSL in many places but I would only find the accronym being used but never defined 🙃.

It's silly but it was not straight forward to find -- I ended up figuring it out once I read a full doc...

Describe the solution you'd like

Very simple: just define the accronym in the docs(trings?). E.g. in swsl-resnet.md:

The models in this collection utilise semi-weakly supervised learning **(SWSL)***

This would be specially useful if it was shown directly in https://paperswithcode.com/lib/timm.

I'd be happy to do it if someone would tell me where it's most appropriate so it would also update papers with code. Is it in the docs/models/.templates/models/*.md files?

jpcbertoldo commented 2 years ago

If I can add: same thing for the prefix tv_ (standing for torchvision).

It is indeed defined in a docstring (here for instance) but one cannot learn it from https://paperswithcode.com/lib/timm/resnet.

rwightman commented 2 years ago

@jpcbertoldo it is in the README.md and (out of date) https://rwightman.github.io/pytorch-image-models/models/#resnet-resnext-resnetpy ... could be better yes

I have some work on going to redo all pretrained model configs, for better metdata, multi-weight (per architecture) support, etc. That will then be used to generate model cards and Hugging Face Hub entries for each weight. This is a slow plodding process though, and interleaved with a bunch of other work so taking a while....

jpcbertoldo commented 2 years ago

Ok! Hadn't seen the summary page. I was assuming the content of those multiple places would be the same (as coming from some automatic generation).

jpcbertoldo commented 2 years ago

My button to close the issue didn't work for some reason. Feel free to close (or I'll retry later).