Add more model resources

[stevhliu]

A continuation of #19767 to add existing official resources (blog posts, notebooks, scripts, etc.) directly to the model docs for 20 of the most popular architectures based on last month's pageviews. I'm not sure whether there are existing resources for all of these models, but I'll check it out, and if not we can either:

• Move to the next most popular model
• Or it could be a good opportunity to create some resources for it

Tracking model progress and updates:

• BERT (see Add BERT resources #19852)
• T5 (see Add T5 resources #19878)
• RoBERTa (see Add RoBERTa resources #19911)
• GPT2 (see Add GPT2 resources #19879)
• BLOOM (see Add BLOOM resources #19881)
• BART (see Add BART resources #19928)
• ViT (assigned to @stanleycai95)
• DistilBERT (see Add DistilBERT resources #19930)
• Wav2Vec2 (see Add wav2vec2 resources #19931)
• LayoutLMV3 (see Add LayoutLMv3 resource #19932)
• CLIP (assigned to @ambujpawar, see Add clip resources to the transformers documentation #20190)
• LayoutLM (assigned to @avisinghal6)
• GPT-J (assigned to @adit299)
• TrOCR (assigned to @huangperry)
• LayoutLMV2 (assigned to @y3sar)
• ALBERT (assigned to @ENate)
• OPT (assigned to @alissadb)
• DeBERTa (assigned to @Saad135, see Add to DeBERTa resources #20155)
• OpenAI GPT (assigned to @shogohida, see [Docs] Add resources of OpenAI GPT #20084)
• XLM-RoBERTa (assigned to @hazrulakmal)

[sayakpaul]

@stevhliu @NielsRogge is there a plan to add the additional vision tasks we currently support to this page? For example object detection, segmentation, video classification. Then there seems to be an entire section missing on multimodal models.

I understand for these tasks, we don't have pipelines yet. But I believe we can still enlist these tasks with our existing resources (notebooks, blog posts, scripts, etc.).

Another suggestion is to provide consolidated model links for each of the tasks enlisted in that page. For example, for image classification, it could be https://huggingface.co/models?pipeline_tag=image-classification.

WDYT?

Cc: @osanseviero @nateraw

[NielsRogge]

@stevhliu @NielsRogge is there a plan to add the additional vision tasks we currently support to this page?

Yes I've pinged @stevhliu yesterday about this ;) he's working on it

[sayakpaul]

Amazing! @stevhliu, if possible could tag me in the PR when you open it? Would love to take a look.

[stevhliu]

For sure @sayakpaul! I'm working on a proposal to reorganize/update the tasks summary docs a bit to include the additional vision/multimodal tasks you mentioned.

Also great suggestion to link to the models; we can use these recently added icons in the docs :)

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.

[avisinghal6]

Hi, I am a newbie to open source and would like to contribute. @NielsRogge can I contribute to this issue?

[stevhliu]

Welcome @avisinghal6, we'd be more than happy for you to make a contribution! 🤗

The remaining models available are LayoutLM, LayoutLMV2, TrOCR, and OPT. Let me know which model you'd like to take on, and any questions you might have!

[avisinghal6]

@stevhliu, I can start working on LayoutLM model. So I need to search for existing official resources on hugging face website and attach relevant ones to the documentation of LayoutLM model ?

[stevhliu]

Yup, check out issue #20055 for a list of resources to search from and the BERT resources section for an example of what it should look like!

[avisinghal6]

@stevhliu, I have added the resources for LayoutLM. Commit : d360021

[stevhliu]

Hi @avisinghal6, would you mind opening a pull request on the Transformers repository for your contribution? You can check out this guide for more details how. Thanks! 😄

[avisinghal6]

Hi @stevhliu, I have created the PR#21377

[SarangShrivastava]

Hi @stevhliu - I want to add resources for LayoutLMv2. Will submit a PR soon.

[stevhliu]

Looking forward to your contribution @SarangShrivastava!

[rajveer43]

hey, @stevhliu I want to work on any of the model if its available

[stevhliu]

Hey @SarangShrivastava , just checking to see if you're still interested in making a model contribution. Totally cool if you aren't available anymore, I'll unassign you from the model you claimed and let someone else take a stab at it. Thanks!

@rajveer43 thanks for your interest! If any of the models free up, I'll let you know 🤗

[y3sar]

Hello @stevhliu I would like to make my first contribution to transformers and this looks like a great place to start😄. Is there any way I can help? I can see all the models have been assigned is it possible to work on a model already assigned? Thanks in advance

[stevhliu]

Thanks for the interest; TrOCR, LayoutLMV2, and ALBERT are now available!

[y3sar]

@stevhliu I would like to start working on LayoutLMV2.

[y3sar]

@stevhliu I have a question regarding 🌎 emoji. Exactly which resources should be marked by it?
Thanks in advance

[stevhliu]

🌎 is for unofficial Hugging Face resources, such as community-created ones. For example, if you look at the GPT2 resources, there are links to notebooks for generating lyrics and tweets from contributors :)

[y3sar]

@stevhliu thanks for the response. I noticed in BERT resources under text-classification a notebook by NielsRogge is not marked 🌎. Are these notebooks considered official?

[unitinguncle]

Hello @stevhliu I would like to make my first contribution to open source transformers and this looks like a great place to start. Is there any way I can help? Thanks in advance

[stevhliu]

Hi, thanks for your interest @unitinguncle! Currently, all the models are taken but I'll let you know if something is available. In the meantime, feel free to also take a look at some of the Good First Issue's to see if there is anything else you may be interested in!

[raj-pandey55]

Hello @stevhliu I would like to make my first contribution to transformers and this looks like a great place to start😄. Is there any way I can make a contribution here? Thanks in advance

[stevhliu]

Thanks for your interest @raj-pandey55! It looks like ALBERT is the only model left to complete. @ENate, are you still interested in making a contribution? If not, I'll reassign it to @raj-pandey55 🙂

[ENate]

@stevhliu I am still interested and hope to finish up the doc and open a pull request soon. Sorry for the delay (I am available again after some pause). I actually noticed that there is an issue with building the doc and commented already. Will do so soon.

[ENate]

I commented about the issue under the link

https://github.com/PyAV-Org/PyAV/issues/1140
and will hope to at least the get the doc build. Thanks

[raj-pandey55]

Thanks for letting me know @ENate . @stevhliu if any new opportunity comes up, please let me know🤗. Thanks in advance

[divyam-prajapati]

Hi there! 👋

Hello @stevhliu, I'm new to open source contributions, and I'm excited to start my journey by helping out here. I'm eager to learn and collaborate with the community. If there are any specific areas in the documentation that need attention, please let me know, and I'd be happy to take them on.

[stevhliu]

All done! 🥳