想确定一个SFT的一个细节

QwenLM / Qwen2.5

Qwen2.5 is the large language model series developed by Qwen team, Alibaba Cloud.

9.56k stars 586 forks source link

Closed xiaomao19970819 closed 2 months ago

xiaomao19970819 commented 3 months ago

若使用官方给定的transformers的trainner进行训练，在计算ce loss时，prompt对应的token的loss没有被mask掉，我的理解是对的吗？

jklj077 commented 2 months ago

This has been raised mutiple times in the issues; please refer to those issues first.

We advise you to use training frameworks, including Axolotl, Llama-Factory, Swift, etc., to finetune your models with SFT, DPO, PPO, etc.