deepspeedai / deepspeedai/DeepSpeed
[REQUEST]Support Galore
Open
Nobody has claimed this yet.
enhancement
- Dominant language
- Python
- Stars
- 43.1k
- Forks
- 5k
- Avg merge
- 4d 15h
- Merged PRs (30d)
- 112
Description
Galore(https://arxiv.org/abs/2403.03507) is a latest training method, which can greatly save VRAM. Transformers already support it, but there are issues when working with deepspeed(https://github.com/hiyouga/LLaMA-Factory/issues/2929). It would be better if any support could be provided.
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
The issue names no file, test, or entry point. Start by reviewing the reported DeepSpeed/Transformers interaction and the linked LLaMA-Factory issue; done would require a concrete Galore support scope and validation for DeepSpeed training.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 20/100