[GH-ISSUE #505] sft cause repeat sentence issue #402

Closed
opened 2026-10-26 21:20:41 +00:00 by bsakaguchi · 3 comments
Owner

Originally created by @Auyin111 on GitHub (Feb 26, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-R1/issues/505

After I used llama factory to sft the 14b distilled qwen model using personal dataset, the model repeat same sentence again and again for a long time.
My training dataset format is xxx answer and has 380 training dataset.

Any idea on it? Thanks

Originally created by @Auyin111 on GitHub (Feb 26, 2025). Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-R1/issues/505 After I used llama factory to sft the 14b distilled qwen model using personal dataset, the model repeat same sentence again and again for a long time. My training dataset format is <think> xxx </think> answer and has 380 training dataset. Any idea on it? Thanks
bsakaguchi 2026-10-26 21:20:41 +00:00
Author
Owner

@MSS444 commented on GitHub (Mar 18, 2025):

did you get an answer if yes then please tell i got a similar pattern on phi4 as well, i also used llamafactory

<!-- gh-comment-id:2731762528 --> @MSS444 commented on GitHub (Mar 18, 2025): did you get an answer if yes then please tell i got a similar pattern on phi4 as well, i also used llamafactory
Author
Owner

@github-actions[bot] commented on GitHub (Apr 19, 2025):

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. If you believe this issue is still relevant, please leave a comment to keep it open. Thank you for your contributions!

<!-- gh-comment-id:2816380074 --> @github-actions[bot] commented on GitHub (Apr 19, 2025): This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. If you believe this issue is still relevant, please leave a comment to keep it open. Thank you for your contributions!
Author
Owner

@github-actions[bot] commented on GitHub (May 23, 2025):

false

<!-- gh-comment-id:2902917328 --> @github-actions[bot] commented on GitHub (May 23, 2025): false
Sign in to join this conversation.
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
deepseek-ai/DeepSeek-R1#402
No description provided.