QAnything icon indicating copy to clipboard operation
QAnything copied to clipboard

ImportError: .../miniconda3/envs/qt/lib/python3.10/site-packages/vllm/_C.cpython-310-x86_64-linux-gnu.so: undefined symbol: _ZN2at4_ops15to_dtype_layout4callERKNS_6TensorEN3c108optionalINS5_10ScalarTypeEEENS6_INS5_6LayoutEEENS6_INS5_6DeviceEEENS6_IbEEbbNS6_INS5_12MemoryFormatEEE

Open Hzzhang-nlp opened this issue 10 months ago • 6 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

运行bash scripts/run_for_3B_in_Linux_or_WSL.sh,报错: ImportError: .../miniconda3/envs/qt/lib/python3.10/site-packages/vllm/_C.cpython-310-x86_64-linux-gnu.so: undefined symbol: _ZN2at4_ops15to_dtype_layout4callERKNS_6TensorEN3c108optionalINS5_10ScalarTypeEEENS6_INS5_6LayoutEEENS6_INS5_6DeviceEEENS6_IbEEbbNS6_INS5_12MemoryFormatEEE

期望行为 | Expected Behavior

想能帮忙解决

运行环境 | Environment

- OS:Ubuntu22.04
- NVIDIA Driver:525.147.05
- CUDA:12.0
- docker:
- docker-compose:
- NVIDIA GPU:RTX3060
- NVIDIA GPU Memory:12GB

QAnything日志 | QAnything logs

截图 2024-04-14 00-00-26

复现方法 | Steps To Reproduce

按照这种方式进行部署了: conda create -n qanything-python python=3.10 conda activate qanything-python git clone -b qanything-python-v1.3.1 https://github.com/netease-youdao/QAnything.git cd QAnything pip install -e .

然后,运行bash scripts/run_for_3B_in_Linux_or_WSL.sh,报错: ImportError: .../miniconda3/envs/qt/lib/python3.10/site-packages/vllm/_C.cpython-310-x86_64-linux-gnu.so: undefined symbol: _ZN2at4_ops15to_dtype_layout4callERKNS_6TensorEN3c108optionalINS5_10ScalarTypeEEENS6_INS5_6LayoutEEENS6_INS5_6DeviceEEENS6_IbEEbbNS6_INS5_12MemoryFormatEEE

备注 | Anything else?

截图 2024-04-14 00-00-26

Hzzhang-nlp avatar Apr 13 '24 16:04 Hzzhang-nlp

我也是遇到这个错

tangsipeng avatar Apr 15 '24 02:04 tangsipeng

装2.1.2版本的pytorch以及对应的依赖版本可以解决问题

tangsipeng avatar Apr 15 '24 03:04 tangsipeng

@tangsipeng @Hzzhang-nlp 我在测试环境中部署时默认安装的torch就是2.1.2,请问你们报错时系统的torch版本是多少?

xixihahaliu avatar Apr 15 '24 12:04 xixihahaliu

安装nvcc解决

fivegg avatar Apr 16 '24 08:04 fivegg

有解决吗,安装时也是遇到这个错误

OakLi avatar Apr 22 '24 13:04 OakLi

@xixihahaliu > 装2.1.2版本的pytorch以及对应的依赖版本可以解决问题 但是其它包依赖的又是2.2.2版本的啊,把torch改成2.1.2q其它的又报错

OakLi avatar Apr 22 '24 13:04 OakLi