modelscope / modelscope/ms-swift
微调分类任务出现零损失
Open
Nobody has claimed this yet.
bug
- Dominant language
- Python
- Stars
- 15.7k
- Forks
- 1.7k
- Avg merge
- 1d 16h
- Merged PRs (30d)
- 136
Description
你好，
我在微调Qwen 2.5 Omni 7B模型，下面是一条训练数据：
{
"messages": [
{
"content": "<video>背景：作为一名专业的视频评价员，对这条视频进行评估。XXX这里省略",
"role": "user"
}
],
"label": 2,
"videos": [
"xxx.mp4"
]
}
标签有0，1，2。微调的过程中损失一直是0.0，是哪里出问题了吗？
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 with the provided training-data example and the Qwen 2.5 Omni 7B classification fine-tuning setup; inspect how labels and video inputs are consumed. Reproduce the reported 0.0 loss and verify whether label values 0, 1, and 2 reach the loss calculation. Done means identifying the cause and confirming non-zero loss or documenting the required input or configuration change.
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
- 25/100