hoshi-hiyouga

ping @mlinmg

> No problem, I'll do it tomorrow. Alse please send the error it gives you, I can't manage to reproduce it script: ``` CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \ --stage sft \...

too many dummy commit in this pr, open another one should be more efficient btw, qwen1.5 models do not work for me in both non-FA2 and FA2 paths

It may be because the embed tokens and lm head are saved in 32-bit precision, leading to an increase in the file size. You can merge the lora adapter using...

更新代码，选择导出设备为 cuda

in 3 days

now we provide the `empty` template that allows training on the text without a specific format, you can pre-process your dataset and template them, and use `empty` template to fine-tune...

sorry it is not supported yet

Llama-3 model can generate texts in our Linux environment. I think it is likely an issue with your hardware and environment. 

@JieShenAI 还没支持。