Hugging Face's logo

Spaces:
rayli
/
instruct-particulate
Running on Zero

App Files Community
1

Apply for a GPU community grant: Academic project

#1
by rayli - opened
Discussion
rayli
 Owner

We are building a demo for our academic research work that will be open-sourced!

The current implementation is getting a lot of torch.AcceleratorError: CUDA error: uncorrectable ECC error encountered error when starting on ZeroGPU, which seems to be caused by HF ZeroGPU assigning the request to a bad/sticky GPU worker.

I'm wondering if there is a solution to this problem and if not, could you grant us a dedicated GPU for our demo?
hysts

Hi @rayli , thanks for reporting this. I'll check with the infra team about the ECC failures. As for the grant, community GPU grants are for ZeroGPU, so it wouldn't help here.
rayli
 Owner

Thank you @hysts ! Let me know what the infra team says!

hysts

@rayli The issue should now be resolved. One of the GPU nodes had an issue, and our infra team has excluded it. The Space has also been restarted. Sorry for the disruption, and thanks again for reporting this!

Sign up or log in to comment