search

rhasspy / wyoming-faster-whisper

Wyoming protocol server for faster whisper speech to text system
MIT License
91 stars 33 forks source link

RuntimeError: Library libcublas.so.12 is not found #30

Closed griffindodd closed 3 weeks ago

griffindodd commented 7 months ago

Ubuntu 20.84 CUDA 11.8 Nvidia Driver 525.147.05 Python 3.8 Pytorch 2.2.1+cu118 cuBLAS 11.11.3.6 cuDNN 8.7.0.84

The service works well enough when using CPU, but I cannot get it to work with CUDA. Every guide I have have followed so far has used CUDA 11.8 but it seems that the version of cTranslate2 used in faster-whisper 1.0.0 now requires CUDA12 libraries.

On trying to use CUDA I get the Error: RuntimeError: Library libcublas.so.12 is not found

rufinus commented 2 months ago

see https://github.com/rhasspy/wyoming-faster-whisper/issues/35#issuecomment-2305540613 this is working currently.