在Linux系统下，使用conda环境安装GPU版本的Paddle，安装后使用官方检测程序python -c "import paddle; paddle.utils.run_check()"检测GPU是否工作正常，出现如下报错:The third-party dynamic library (libcudnn.so) that Paddle depends on is not configured correctly.

W0322 10:44:47.675211 290774 gpu_resources.cc:61] Please NOTE: device: 0, GPU Compute Capability: 8.6, Driver API Version: 11.7, Runtime API Version: 11.6
W0322 10:44:47.675591 290774 dynamic_loader.cc:307] The third-party dynamic library (libcudnn.so) that Paddle depends on is not configured correctly. (error code is /usr/local/cuda/lib64/libcudnn.so: cannot open shared object file: No such file or directory)Suggestions:1. Check if the third-party dynamic library (e.g. CUDA, CUDNN) is installed correctly and its version is matched with paddlepaddle you installed.2. Configure third-party dynamic library environment variables as follows:- Linux: set LD_LIBRARY_PATH by `export LD_LIBRARY_PATH=...`- Windows: set PATH by `set PATH=XXX;……PreconditionNotMetError: Cannot load cudnn shared library. Cannot invoke method cudnnGetVersion.[Hint: cudnn_dso_handle should not be null.] (at /paddle/paddle/phi/backends/dynload/cudnn.cc:60)[operator < fill_constant > error]

分析报错内容，Paddle依赖的第三方链接库配置错误，根据建议，添加环境变量即可。

解决方案：

查看路径

我是用的是非root用户创建的环境，命名为paddle，环境路径为~/.conda/envs/paddle，对应的第三方动态链接库地址为~/.conda/envs/paddle/lib，根据你建立的环境名称，对应的路径为~/.conda/envs/[虚拟环境名称]/lib
若不清楚安装路径，可使用conda activate [环境名]进入环境，运行python -c "import paddle; print(paddle.__file__)"输出安装路径，我的输出结果为

/home/ubuntu/.conda/envs/paddle/lib/python3.8/site-packages/paddle/__init__.py

对应的路径为

/home/ubuntu/.conda/envs/paddle/lib
或
~/.conda/envs/paddle/lib #建议使用相对路径

添加环境变量

1. 临时方案
   每次在程序运行前设置环境变量

   export LD_LIBRARY_PATH=~/.conda/envs/paddle/lib
   python xxx.py
   

2. 永久方案
   将环境变量添加到~/.bashrc文件

   echo "export LD_LIBRARY_PATH=~/.conda/envs/paddle/lib">>~/.bashrc
   

   添加后需要关闭终端重新打开或者登录。

再次运行，错误消失。

(paddle) ubuntu@ThinkStation:~$ python -c "import paddle; paddle.utils.run_check()"
Running verify PaddlePaddle program ... 
W0322 11:10:42.037217 309060 gpu_resources.cc:61] Please NOTE: device: 0, GPU Compute Capability: 8.6, Driver API Version: 11.7, Runtime API Version: 11.6
W0322 11:10:42.042984 309060 gpu_resources.cc:91] device: 0, cuDNN Version: 8.4.
PaddlePaddle works well on 1 GPU.
W0322 11:10:47.978128 309060 fuse_all_reduce_op_pass.cc:79] Find all_reduce operators: 2. To make the speed faster, some all_reduce ops are fused during training, after fusion, the number of all_reduce ops is 2.
PaddlePaddle works well on 2 GPUs.
PaddlePaddle is installed successfully! Let's start deep learning with PaddlePaddle now.