执行如下命令后报错pip3 install paddlepaddle-gpu --upgrade报错日志(PaddleVideo) E:\PaddleVideopython Python 3.7.12 | packaged by conda-forge | (default, Oct 26 2021, 05:35:01) [MSC v.1916 64 bit (AMD64)] on win32 Type help, copyright, credits or license for more information. import paddle paddle.utils.run_check() Running verify PaddlePaddle program ... I0706 15:51:38.917606 1980 interpretercore.cc:237] New Executor is Running. W0706 15:51:38.918606 1980 gpu_resources.cc:119] Please NOTE: device: 0, GPU Compute Capability: 8.6, Driver API Version: 12.9, Runtime API Version: 11.8 W0706 15:51:38.920606 1980 dynamic_loader.cc:274] Note: [Recommend] copy cudnn into CUDA installation directory. For instance, download cudnn-10.0-windows10-x64-v7.6.5.32.zip from NVIDIAs official website, then, unzip it and copy it into C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0 You should do this according to your CUDA installation directory and CUDNN version. Traceback (most recent call last): File stdin, line 1, in module File D:\Develop\anaconda3\2025\envs\PaddleVideo\lib\site-packages\paddle\utils\install_check.py, line 249, in run_check _run_static_single(use_cuda, use_xpu) File D:\Develop\anaconda3\2025\envs\PaddleVideo\lib\site-packages\paddle\utils\install_check.py, line 146, in _run_static_single exe.run(startup_prog) File D:\Develop\anaconda3\2025\envs\PaddleVideo\lib\site-packages\paddle\fluid\executor.py, line 1401, in run use_pruneuse_prune, File D:\Develop\anaconda3\2025\envs\PaddleVideo\lib\site-packages\paddle\fluid\executor.py, line 1619, in _run_impl scope, list(feed.keys()), fetch_list, return_numpy File D:\Develop\anaconda3\2025\envs\PaddleVideo\lib\site-packages\paddle\fluid\executor.py, line 655, in run scope, feed_names, fetch_list RuntimeError: In user code: File stdin, line 1, in module File D:\Develop\anaconda3\2025\envs\PaddleVideo\lib\site-packages\paddle\utils\install_check.py, line 249, in run_check _run_static_single(use_cuda, use_xpu) File D:\Develop\anaconda3\2025\envs\PaddleVideo\lib\site-packages\paddle\utils\install_check.py, line 133, in _run_static_single input, out, weight _simple_network() File D:\Develop\anaconda3\2025\envs\PaddleVideo\lib\site-packages\paddle\utils\install_check.py, line 34, in _simple_network attrpaddle.ParamAttr(initializerpaddle.nn.initializer.Constant(0.1)), File D:\Develop\anaconda3\2025\envs\PaddleVideo\lib\site-packages\paddle\tensor\creation.py, line 223, in create_parameter attr, shape, convert_dtype(dtype), is_bias, default_initializer File D:\Develop\anaconda3\2025\envs\PaddleVideo\lib\site-packages\paddle\fluid\layer_helper_base.py, line 436, in create_parameter **attr._to_kwargs(with_initializerTrue) File D:\Develop\anaconda3\2025\envs\PaddleVideo\lib\site-packages\paddle\fluid\framework.py, line 3935, in create_parameter initializer(param, self) File D:\Develop\anaconda3\2025\envs\PaddleVideo\lib\site-packages\paddle\nn\initializer\initializer.py, line 40, in __call__ return self.forward(param, block) File D:\Develop\anaconda3\2025\envs\PaddleVideo\lib\site-packages\paddle\nn\initializer\constant.py, line 75, in forward stop_gradientTrue, File D:\Develop\anaconda3\2025\envs\PaddleVideo\lib\site-packages\paddle\fluid\framework.py, line 4019, in append_op attrskwargs.get(attrs, None), File D:\Develop\anaconda3\2025\envs\PaddleVideo\lib\site-packages\paddle\fluid\framework.py, line 2781, in __init__ for frame in traceback.extract_stack(): PreconditionNotMetError: The third-party dynamic library (cudnn64_8.dll) that Paddle depends on is not configured correctly. (error code is 126) 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 PATHXXX; (at ..\paddle\phi\backends\dynload\dynamic_loader.cc:301) [operator fill_constant error] 该错误很明确PaddlePaddle 找不到cudnn64_8.dll错误代码 126 表示 DLL 找不到。环境是CUDA 11.8Runtime APIGPU Compute Capability 8.6可能是 RTX 30/40 系列缺少 cuDNN v8 的 DLL解决方法1. 检查当前 PaddlePaddle 版本和 CUDA 版本匹配先确认你安装的 PaddlePaddle 版本pip show paddlepaddle-gpu2. 下载并安装 cuDNN 8 for CUDA 11.x由于你的 CUDA Runtime 是 11.8需要 cuDNN 8.x 版本前往 NVIDIA cuDNN 下载页需要注册 NVIDIA 账号下载cuDNN Library for Windows (x64)版本选择8.9.x或8.6.xforCUDA 11.x解压后将文件复制到 CUDA 安装目录解压后你会看到bin/、include/、lib/三个文件夹将这些文件分别复制到C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\ ├── bin\ ← 复制 cudnn64_8.dll 等到此目录 ├── include\ ← 复制 cudnn*.h 等到此目录 └── lib\x64\ ← 复制 cudnn*.lib 等到此目录3. 确保 PATH 包含 CUDA bin 目录# 检查 PATH 中是否有 CUDA binecho$env:PATH# 如果没有临时添加或永久添加到系统环境变量$env:PATH C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\bin;$env:PATH4. 验证复制完 DLL 后重启终端重新测试importpaddle paddle.utils.run_check()如果 PaddlePaddle 版本是 CUDA 10.x 编译的你还需要重新安装匹配 CUDA 11.x 的版本# 卸载旧版本pip uninstall paddlepaddle-gpu# 安装 CUDA 11.8 版本python-m pip install paddlepaddle-gpu2.6.1-i https://pypi.tuna.tsinghua.edu.cn/simple核心问题就是cuDNN DLL 缺失安装后即可解决。