NVIDIA-NeMo / NVIDIA-NeMo/Curator
Check Pytorch cuda context is valid across GPUs
Open
@VibhuJawa is already working on this.
Since Oct 8, 2024.
enhancement
- Dominant language
- Python
- Stars
- 1.8k
- Forks
- 328
- Avg merge
- 4d 5h
- Merged PRs (30d)
- 30
Description
Describe the bug
We have had multiple breakages of CUDA context being only used for GPU 0 in a dask+pytorch environment. Sometimes this can occur due to a library creating a cuda context with pytorch before starting the cluster.
What ends up happening is Pytorch models being deployed on GPU-0 and that issue is hard to debug .
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.