modelscope / modelscope/ms-swift
Megatron-sft全参数微调Qwen3VL-30B-A3B时，16卡效率正常，32卡、64卡训练异常缓慢
Open
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 15.7k
- Forks
- 1.7k
- Avg merge
- 1d 16h
- Merged PRs (30d)
- 136
Description
/// Shell ///
16-A800：16.19/its
32-A800/64-A800：835.43/its
torch2.6.0，transformer_engine2.8.0，cuda12.4.1
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
No source file, test, or entry point is identified. Start by reproducing full-parameter Megatron SFT for Qwen3VL-30B-A3B with the reported torch, Transformer Engine, and CUDA versions at 16, 32, and 64 A800s. Done means identifying the cause of the scaling slowdown, applying a fix, and verifying the multi-node performance.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- distributed-systems, machine-learning, performance
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100