Shadows comments

Repositories
Issues
Comments

Results 2 comments of


                                            Shadows

如何使用v2.13.0的自定义模型功能（how to use 2.13.0 multi models)

我在使用该项目调用自己微调的大模型接口时，输出的token被限制在了100，请问我应该如何修改。

GRPO with reward model. CUDA out of memory. How to fix? Thank you very much.

> > Thank you [@LWShowTime](https://github.com/LWShowTime) . Setting `gradient_accumulation_steps: 1` does not solve the OOM. > > Maybe you can add some information of your gpu devices. For GRPOTrainer, I can...