Spaces:
Running
on
CPU Upgrade
About New Rate Limit System
Implementing a rate limit is a good idea, but it's important to note that submitting five 180B or 70B models doesn't equate to submitting five 7B or 13B models. What are your thoughts on this? Thanks.
You can use Gradio Client which used to work but is now broken. Or you can access the Result datasets but you won't be getting the view of the leaderboard, only the raw results.
I will open an new issue about the Gradio API being not working .
@mhemetfaik
This is totally correct, and we might make the rate limit more important for bigger models in the future if needed.
However, this is both a matter of compute (as you mentioned) and of model diversity: we want many people to be able to submit, not just a handful of labs and users, which is why we do not do the difference for now.
@clefourrier I'm trying the notebook available on the page:
https://huggingface.co/blog/graphml-classification
gives me the following error:
Is there a solution?
Thanks in advance,
Cordially,
Mario Manzo
@mariomanzo
This issue is completely unrelated to the leaderboard. Please try to only comment here for leaderboard related issues to not spam.
In this case, I suspect the machine you are trying to run the notebook on does not have enough RAM to process the graph. If you need more support on this issue, please open an issue on the blog's github.