python应用运行报错，部分报错信息如下：
7920475 [E:onnxruntime:Default, provider_bridge_ort.cc:1744 onnxruntime::TryGetProviderInfo_CUDA] C:\a_work\1\s\onnxruntime\core\session\provider_bridge_ort.cc:1426 onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : LoadLibrary failed with error 126 “” when trying to load “lib\site-packages\onnxruntime\capi\onnxruntime_providers_cuda.dll”
EP Error C:\a_work\1\s\onnxruntime\python\onnxruntime_pybind_state.cc:866 onnxruntime::python::CreateExecutionProviderInstance CUDA_PATH is set but CUDA wasnt able to be loaded. Please install the correct version of CUDA andcuDNN as mentioned in the GPU requirements page (https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements), make sure they’re in the PATH, and that your GPU is supported.
when using [‘CUDAExecutionProvider’, ‘CPUExecutionProvider’]
Falling back to [‘CUDAExecutionProvider’, ‘CPUExecutionProvider’] and retrying.
RuntimeError: C:\a_work\1\s\onnxruntime\python\onnxruntime_pybind_state.cc:866 onnxruntime::python::CreateExecutionProviderInstance CUDA_PATH is set but CUDA wasnt able to be loaded. Please install the correct version of CUDA andcuDNN as mentioned in the GPU requirements page (https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements), make sure they’re in the PATH, and that your GPU is supported.
这是onnxruntime调用CUDA失败导致的错误，主要是onnxruntime的问题，解决方法如下：
首先检查你的电脑上是否安装CUDA和cuDNN，如果没安装的话可以看这个教程安装《Windows电脑安装CUDA及配置CUDNN环境详细教程》
你的python环境里应该也没有安装optimum，按下面步骤操作即可
先运行命令卸载onnxruntime-gpu
pip uninstall onnxruntime-gpu如果你安装的是onnxruntime的话，那就把onnxruntime卸载，然后运行下面命令安装optimum
pip install optimum[onnxruntime-gpu]这个过程会同时将evaluate，onnxruntime-gpu，optimum安装上去，重新运行python应用的话就没有这个CUDA报错了。
onnxruntime_providers_cuda.dll相关报错解决方法看另一篇文章《python报错[ONNXRuntimeError] : FAIL : LoadLibrary failed with error “\capi\onnxruntime_providers_cuda.dll”解决方法》
