AI-Hypercomputer / AI-Hypercomputer/maxtext
GCS upload errors should not be silent
Open
- Dominant language
- Python
- Stars
- 2.4k
- Forks
- 607
- Avg merge
- 2d 19h
- Merged PRs (30d)
- 158
Description
I ran into this issue where my node didn't have sufficient ACLs to write to my GCS bucket. This didn't show up in the logs, but also inexplicably caused the job to just hang.
I'm not sure of the reason for the hang, but we GCS uploads should not silently fail. `blob.upload_from_filename` errors should be at least logged.
https://github.com/search?q=repo%3AAI-Hypercomputer%2Fmaxtext%20blob.upload_from_filename&type=code
Contributor guide
Assessment
This issue has not been assessed yet.