modelscope / modelscope/ms-swift
Qwen2.5VL resume_from_checkpoint key mismatch
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 15.7k
- Forks
- 1.7k
- Avg merge
- 1d 16h
- Merged PRs (30d)
- 136
Description
Checklist / 检查清单
- I have searched existing issues, and this is a new bug report. / 我已经搜索过现有的 issues，确认这是一个新的 bug report。
Bug Description / Bug 描述
当transformers版本大于4.52的时候，训练Qwen2.5VL并保存ckpt会保存为self.visual和self.model，而resume的时候模型会初始化为self.model.visual和self.model.language_model，从而导致resume_from_checkpoint会报提示key mismatch，无法加载前一次训练的ckpt模型
[补充]：开启deepspeed之后不会报key mismatch的信息，不确定是否正确加载了
How to Reproduce / 如何复现
transformers>=4.52，Qwen2.5VL训练时resume_from_checkpoint
Additional Information / 补充信息
部分mismatch信息：
No response
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 by reproducing Qwen2.5VL training and resume_from_checkpoint with transformers >=4.52, then compare the checkpoint keys saved before and after model initialization. Verify whether the reported self.visual versus self.model.visual and self.model.language_model mismatch prevents correct loading, including the deepspeed case; done means resuming loads the prior checkpoint without mismatches and preserves training state.
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
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 38/100