Add support for AMD GPUs (ROCm backend) or non-NVIDIA hardware
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 18
- Forks
- 28
- Avg merge
- 2d 2h
- Merged PRs (30d)
- 7
Description
This came up while attempting to run training jobs on MI100 node on Delta, where CUDA-based containers do not work. Currently, we assume NVIDIA environment such as use of --nv in containers, CUDA-specific dependencies here. This limits the use of the AMD GPUs (MI100) using --rocm.
Proposed changes
- Add support for ROCm-enabled PyTorch builds
- Update container workflow to support --rocm (Apptainer)
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.
Research direction
Start with projects/train/pyproject.toml at the CUDA-specific dependencies, then trace the container workflow that currently uses Apptainer's --nv option. Check how PyTorch builds and container options are selected for training jobs. Done means ROCm-enabled PyTorch dependencies and --rocm container execution support are documented and usable on AMD MI100 hardware.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- devops, machine-learning
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 48/100