modelscope / modelscope/ms-swift

Finetuned OSS through swift but deployed through native vllm produces only the reasoning content with actual content as empty ' '

Open
#6,839 12 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

stale
Dominant language
Python
Stars
15.7k
Forks
1.7k
Avg merge
1d 16h
Merged PRs (30d)
136

Description

I have finetuned the gpt-oss 20b model for 1 epoch, with a good balance of low medium and high reasoning datapoints but when deployed through vllm it produces "" in actual content and produces the correct reasoning.

Is there something that is being missed or how to retain the base model behaviour.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by comparing the Swift finetuning setup with native vLLM deployment for the gpt-oss 20b model, focusing on the response where reasoning is present but actual content is empty. The issue names no files, tests, commands, or entry points, so a minimal reproduction and the relevant training and serving configuration are needed. Done means identifying the cause and documenting a reproducible configuration or correction.

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

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.