modelscope / modelscope/ms-swift
开启packing后训练seq_acc计算偏低
Open
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 15.7k
- Forks
- 1.7k
- Avg merge
- 1d 16h
- Merged PRs (30d)
- 136
Description
开启packing参数后， 训练时train和eval的loss/acc趋势正常，但seq_acc数值都特别低，0.35左右。vllm对eval数据推理的acc是这个的两倍之多 0.85
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 file, test, or entry point is identified in the report. Reproduce training with packing enabled, compare train/eval seq_acc near 0.35 with vLLM eval accuracy near 0.85, and trace the packing-related metric calculation; done means the reported sequence accuracy agrees with the evaluation result.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 35/100