Why does this error occur when tensorflow has no gpu, but the model can still be run normally.

W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'nvcuda.dll'; dlerror: nvcuda.dll not found
W tensorflow/stream_executor/cuda/cuda_driver.cc:269] failed call to cuInit: UNKNOWN ERROR (303)

Error:-

2022-02-26 10:47:12.726011: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library invcuda.dll'; dlerror: nvcuda.dll not found
2022-02-26 10:47:12.727262: W tensorflow/stream_executor/cuda/cuda_driver.cc:269] failed call to cuInit: UNKNOWN ERROR (303)

CodePudding user response：

Why does this error occur when Tensorflow has no GPU, but the model can still be run normally.

According to https://wikidll.com/nvidia-corporation/nvcuda-dll

"Nvcuda.dll a DLL (Dynamic Link Library) file, developed by NVIDIA Corporation, which is referred to essential system files of the Windows OS. It usually contains a set of procedures and driver functions, which may be applied by Windows."

Since you have no GPU, presumably there are no NVIDIA drivers installed, and no Nvcuda.dll. Presumably, the Tensorflow code is able to fallback to a mode where the model is run on the CPU.

You must have installed a Tensorflow for GPU version. If you want to get rid of the (noisy) warning message, force install the CPU only version, or modify the logging levels. For more details: see Tensorflow not working: Could not load dynamic library 'nvcuda.dll'; dlerror: nvcuda.dll not found