Replies: 3 comments 5 replies
-
CUDA 12.1 or higher needed. https://github.com/chidiwilliams/buzz/blob/main/CONTRIBUTING.md#gpu-support |
Beta Was this translation helpful? Give feedback.
-
I've updated to 12.1 but the speed of transcribing seems no different and my CPU is nearly at full peak with my GPU idle. |
Beta Was this translation helpful? Give feedback.
-
Faster whisper needs cuDNN and cuBLAS (https://developer.nvidia.com/cublas) that right now does not have the installer for Windows on nvidia site, but regular Whisper should use GPU on any CUDA version. With GPU the speed is +/- similar. Next release will have necessary GPU support on Windows included. To get it right now you can download latest development build from here https://github.com/chidiwilliams/buzz/actions/workflows/ci.yml?query=branch%3Amain |
Beta Was this translation helpful? Give feedback.
-
I have an RTX 3090 and it's simply not recognizing it for some reason and using my CPU instead.
Beta Was this translation helpful? Give feedback.
All reactions