GUORUIWANG
GUORUIWANG
### Description ### Expected behavior with the suggested feature ### Other Comments
Is Ampere GPU with cuda 11.0 a necessary condition? How to solve this error? Thank you
用的llama_quant.py,未用数据集评估,量化后模型输出是乱码,请问啥原因呀
### Is there an existing issue for this? - [X] I have searched the existing issues ### Current Behavior 由于用量化的模型效果不佳,尝试用的fp16训,在8*v100出现,第一张卡报oom,但其他卡显存占用不高,有什么好的解决方式 ### Expected Behavior v100 是否上可以训 fp16 ### Steps To Reproduce...
actor、critic、sft、reward model
加速看不到效果反而更慢
尝试了chatglm 和baichuan 使用fastllm后速度反而更慢