aigc-apps / aigc-apps/VideoX-Fun
是否支持模型完全参数分片(单卡batch<1)
Open
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 2.3k
- Forks
- 188
- Avg merge
- 2d 22h
- Merged PRs (30d)
- 3
Description
现在需要做一个蒸馏操作。需要放入教师模型和学生模型，单卡放不下两个模型，所以需要完全参数的并行，切分一个模型到不同GPU，每张卡上的batch size应该是小于1的，例如八张卡可能就总共batchsize=2.目前框架是否支持这个操作
Contributor guide
No contributing guide indexed for this repository
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 files, tests, or entry points. Start by locating the model-loading and distributed training entry points, then determine whether teacher and student models can be fully sharded across GPUs with a per-device batch size below one. Done means the framework's support or limitation is documented and verified with a multi-GPU distillation run.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- distributed-systems, machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100