search

reriiasu / speech-to-text

Real-time transcription using faster-whisper
MIT License
375 stars 56 forks source link

Could not locate cudnn_ops_infer64_8.dll #12

Open Bob82u opened 5 months ago

Bob82u commented 5 months ago

Hello, I'm encountering an error "Could not locate cudnn_ops_infer64_8.dll. Please make sure it is in your library path!" when running the model with CUDA. Using CPU works fine.

Setup: • GeForce RTX 2070 + 8G • CUDA Toolkit 11.4 (downloaded from https://developer.nvidia.com/cuda-11-4-0-download-archive) • cuDNN 11.4-windows-x64-v8.2.2.26 (downloaded from https://developer.download.nvidia.com/compute/redist/cudnn/v8.2.2/ Extracted cudnn DLLs placed in: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.4\bin\

Environment path confirmed to the dll location.

  where cudnn_ops_infer64_8.dll
  C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.4\bin\cudnn_ops_infer64_8.dll

Would you have any suggestions to fix this error? In case of compatibility issues, could you recommend compatible CUDA and cuDNN versions with download links?

reriiasu commented 1 month ago

@Bob82u Hello Bob82u

Please try using CUDA 12 with cuDNN 8. Faster-Whisper does not currently support cuDNN 9.