Do you have any feedback about Discussions and PRs on the Hub?
Can't wait to get some community feedback!
Hellooo world! 🤗
G'day g'day 🦘👋
Hello there!
Guten Tag
Hola! 👋
Hey! 🤗
hallo wie geht's?
Saletti zäme!
Nice =)
Hi everybody 🤗
Bonjour !
From what I've been able to test so far, it's soooo cool.
A few remarks:
- I had to type julien-c name by hand to tag him in a PR (the first Hub PR ever to what I have been told 😎). It didn't offer it automatically like on GitHub for example. Would it be possible to add this feature?
- In the same logic I would have added an emoji "🎉" to his comment to celebrate this first PR but obviously it is not possible to react. Would it be possible to add this feature?
- I tried to propose a PR of the PIAF readme.md file but it returns me on a 404 error. Any idea why that is? (I see this on commonvoice too)
- I wanted to put a screen of the 404 error from the previous point to illustrate this but it didn't copy/paste me the image. Would it be possible to add this feature?
It is an extraordinary work that you have just delivered, a huge thank you to you 🙏
Ah and it seems you can't edit your message either
Hi everyone
Thanks for your feedback and suggestions we aim to iterate swiftly on those 🔥
🔥 yeay 🎉 just tested it out making an easy PR https://huggingface.co/spaces/osanseviero/hugging-gallery/discussions/1 everything went smoothly.
It would be nice to autocomplete/suggest emojis -> :fire: in comments.
How do I even create PR using local machine or API?
Thinking about this, there is a lot of potential in that feature... in-between wikipedia discussion page and github code PR.
Agree with @nazneen that auto-complete to tag people would be super nice
An exciting step toward a more open & ethical ML future. 🙌 🎉 Keep all the great feedback coming!
Hi @lbourdois those are very good suggestions (we'll probably add all of them 🎉)
Hi
@victor
,
A new suggestion that I'm about to find.
I opened a PR to add tags related to languages (https://huggingface.co/Geotrend/distilbert-base-en-ar-cased/discussions/1). I just realized that I made a mistake: the different languages should not be separated by a comma but should be returned to the line each time with a new dash.
I would like to change my PR but can't find a way to do other than closing my current PR and re-opening another one.
It would be friendlier to have the possibility to add changes to a PR as long as it is not merge like it is on GitHub.
It would be great to have an overview of all PRs that I have opened and to see their status
💥
Can we have notifications on the hub instead of mail?
@lbourdois I agree, we should be able to do that on the website directly. (you can always pull your pr branch locally and do your change in another commit then push but it's a bit complex)
How about PR to add a model/space to an organization? @osanseviero @victor
Ah and it seems you can't edit your message either
It's shipped @lbourdois and reactions + images upload are coming soon.
Sorry folks, I've clicked on the wrong button!
Hey all! Now you can react to comments!
It would be very helpful to add thumbs-down emoji 👎 to comments. On GitHub, I'll often land on an issues page if I'm having trouble, and there's a long thread of proposed solutions. Seeing comments with many 👎 emojis lets you know that that solution did not actually work.
Wow, I’m looking forward to dive into this!
Thus far I wasn’t sure whether this is possible. Thank you for your amazing work!
@osanseviero
I’ve recently re-visited kaggle and got reminded of their feature to attach notebooks / community code to some of their hosted datasets. I imagine that a standardized hf notebook that uses a particular dataset could be fruitful and enrich the community section.
For example this notebook (https://huggingface.co/blog/fine-tune-segformer#fine-tune-a-semantic-segmentation-model-with-a-custom-dataset) would be attached in the dataset csection of this ds-repo:
Ups, that wasn’t meant to be that long 😂
Thanks again :)
Best,
Mike
Feature Request - Sort/Filter Models by Size
Hello,
First of all, I would like to thank you for HuggingFaceHub, which allows me to discover and learn so much about the world of AI.
It's Great.
I would like to propose a new feature that could be useful to other users on the platform:
On the Models Page: Add Sorting or Filtering by Model Size
1. Sort by LLM size: Add an option to sort models by size in the existing list.
Or alternatively,
2. Filter by size: Integrate one or two buttons or an option in the "OTHERS" menu on the models page that would allow specifying a maximum or minimum size.
This would help users quickly find models suited to their specific needs, especially based on available hardware resources.
Best regards,
Hello I believe that the BertPooler is a mistaken concept and may lead to sub-optimal performance in sequence classification. I've verified this on sequence classification tasks only and simply removing it does far better often. Kindly check the file transformers/models/bert/modeling_bert.py.
Thanks!



