QAnything icon indicating copy to clipboard operation
QAnything copied to clipboard

[BUG] 纯python环境安装gpu利用率默认参数为0.61导致rtx3060用不起。(手动修正代码见正文)

Open fivegg opened this issue 10 months ago • 1 comments

是否已有关于该错误的issue或讨论? | Is there an existing issue / discussion for this?

  • [X] 我已经搜索过已有的issues和讨论 | I have searched the existing issues / discussions

该问题是否在FAQ中有解答? | Is there an existing answer for this in FAQ?

  • [X] 我已经搜索过FAQ | I have searched FAQ

当前行为 | Current Behavior

gpu内存为12G,llm_engine.py报错增加gpu_memory_utilization。被迫去源码手动修改了一下 修改: anaconda3\envs\qanything-python\lib\python3.10\site-packages\vllm\engine\llm_engine.py 259行 def _init_cache(self) -> None: self.cache_config.gpu_memory_utilization=0.9

期望行为 | Expected Behavior

No response

运行环境 | Environment

- OS:
- NVIDIA Driver:
- CUDA:
- docker:
- docker-compose:
- NVIDIA GPU:
- NVIDIA GPU Memory:

QAnything日志 | QAnything logs

No response

复现方法 | Steps To Reproduce

No response

备注 | Anything else?

No response

fivegg avatar Apr 16 '24 03:04 fivegg

修改utils文件中的general_utils.py文件第199行gpu_memory_utilization = 0.8或者0.7

Hzzhang-nlp avatar Apr 16 '24 14:04 Hzzhang-nlp