[GH-ISSUE #71] qwen32B蒸馏模型，长度>8k时，预测一定比例乱码，出现<think><think><think><think><think><think> #60

New issue

Closed

opened 2026-10-26 21:12:39 +00:00 by bsakaguchi · 1 comment

bsakaguchi commented

2026-10-26 21:12:39 +00:00

Owner

Originally created by @daniellibin on GitHub (Jan 25, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-R1/issues/71

temperature按照0.6设置的，大家遇到过这个情况吗？
长度设置4k以内无乱码但是会截断；
长度设置8k以上大概有20%比例出现，一直重复输出

Originally created by @daniellibin on GitHub (Jan 25, 2025). Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-R1/issues/71 temperature按照0.6设置的，大家遇到过这个情况吗？长度设置4k以内无乱码但是会截断；长度设置8k以上大概有20%比例出现，一直重复输出

bsakaguchi closed this issue

2026-10-26 21:12:39 +00:00

bsakaguchi commented

2026-10-26 21:12:40 +00:00

Author

Owner

@stack-heap-overflow commented on GitHub (Jan 26, 2025):

https://huggingface.co/deepseek-ai/DeepSeek-R1/discussions/44

@stack-heap-overflow commented on GitHub (Jan 26, 2025): https://huggingface.co/deepseek-ai/DeepSeek-R1/discussions/44