hiyouga / LLaMA-Factory

Unify Efficient Fine-Tuning of 100+ LLMs
Apache License 2.0
25.52k stars 3.16k forks source link

Qwen dpo训练卡住 #4631

Closed yxk9810 closed 2 days ago

yxk9810 commented 2 days ago

Reminder

System Info

使用: "dpo_qwen": { "file_name": "qwen_dpo.json", "ranking": true }, 数据格式参考: pair的格式,cutofflen 设置2048

Reproduction

image 模型训练一直卡再这里,尝试调整了输入的长度还是没用?请问这个问题有遇到过么?

Expected behavior

No response

Others

No response

hiyouga commented 2 days ago

你没有安装 CUDA 版本的 pytorch