问题描述

环境：ubuntu，gpu：2080ti
这是报错信息： 

[WARNING] ME(764200,7f3e5617a280,python):2024-03-16-22:49:25.731.518 [mindspore/ccsrc/runtime/hardware/device_context_manager.cc:608] SelectGpuPlugin] Env CUDA_HOME is /usr/local/cuda, but dlopen file_name failed, reason: Load dynamic library: libmindspore_ascend.so.2 failed. libge_runner.so: cannot open shared object file: No such file or directory 

Load dynamic library: libmindspore_ascend.so.1 failed. libge_runner.so: cannot open shared object file: No such file or directory 

Load dynamic library: libmindspore_gpu.so.11.6 failed. libcudnn.so.8: cannot open shared object file: No such file or directory 

 

[ERROR] ME(764200:139905709089408,MainProcess):2024-03-16-22:49:26.143.621 [mindspore/run_check/_check_version.py:219] libcuda.so (need by mindspore-gpu) is not found. Please confirm that libmindspore_gpu.so is in directory:/home/ubuntu/anaconda3/envs/MPCT_mindspore/lib/python3.7/site-packages/mindspore/run_check/../lib/plugin and the correct cuda version has been installed, you can refer to the installation guidelines: https://www.mindspore.cn/install 

[ERROR] ME(764200:139905709089408,MainProcess):2024-03-16-22:49:26.143.771 [mindspore/run_check/_check_version.py:219] libcudnn.so (need by mindspore-gpu) is not found. Please confirm that libmindspore_gpu.so is in directory:/home/ubuntu/anaconda3/envs/MPCT_mindspore/lib/python3.7/site-packages/mindspore/run_check/../lib/plugin and the correct cuda version has been installed, you can refer to the installation guidelines: https://www.mindspore.cn/install 
...
 

RuntimeError: Unsupported device target GPU. This process only supports one of the ['CPU']. Please check whether the GPU environment is installed and configured correctly, and check whether current mindspore wheel package was built with "-e GPU". For details, please refer to "Device load error message". 

 

---------------------------------------------------- 

- Device load error message: 

---------------------------------------------------- 

Load dynamic library: libmindspore_ascend.so.2 failed. libge_runner.so: cannot open shared object file: No such file or directory 

Load dynamic library: libmindspore_ascend.so.1 failed. libge_runner.so: cannot open shared object file: No such file or directory 

Load dynamic library: libmindspore_gpu.so.11.6 failed. libcudnn.so.8: cannot open shared object file: No such file or directory 

是我的环境变量没有设置好吗？

解决方案

libcuda.so (need by mindspore-gpu) is not found

libcudnn.so (need by mindspore-gpu) is not found

先确定cuda和cudnn是否正确安装

确定cuda 执行如下命令：

nvcc -V

确定cudnn 执行如下命令：

cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2

cat /usr/local/cuda/include/cudnn_version.h | grep CUDNN_MAJOR -A 2