huggingface / huggingface/datasets

Comment/feature request: Huggingface 502s from GHA

Open
#7,759 0 comments 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
22k
Forks
3.4k
Avg merge
5d 7h
Merged PRs (30d)
17

Description

This is no longer a pressing issue, but for completeness I am reporting that in August 26th, GET requests to `https://datasets-server.huggingface.co/info\?dataset\=livebench/math` were returning 502s when invoked from [github actions](https://github.com/UKGovernmentBEIS/inspect_evals/actions/runs/17241892475/job/48921123754) (that link will expire eventually, [here are the logs](https://github.com/user-attachments/files/22233578/logs_44225296943.zip)).

When invoked from actions, it appeared to be consistently failing for ~6 hours. However, these 502s never occurred when the request was invoked from my local machine in that same time period.

I suspect that this is related to how the requests are routed with github actions versus locally.

Its not clear to me if the request even reached huggingface servers or if its the github proxy that stopped it from going through, but I wanted to report it nonetheless in case this is helpful information. I'm curious if huggingface can do anything on their end to confirm cause.

And a feature request for if this happens in the future (assuming huggingface has visibilty on it): A "datasets status" page highlighting if 502s occur for specific individual datasets could be useful for people debugging on the other end of this!

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.