search

KoljaB / RealtimeSTT

A robust, efficient, low-latency speech-to-text library with advanced voice activity detection, wake word activation and instant transcription.
MIT License
2.09k stars 190 forks source link

How to choose the CUDA version？ #67

Closed jacobtang closed 5 months ago

jacobtang commented 5 months ago

hello,@KoljaB I update the latest RealtimeSTT,but something wrong with the CUDA version for the ctranslate2 in the linux env(ubuntu).

RealTimeSTT: root - ERROR - Error initializing main faster_whisper transcription model: CUDA failed with error the operation cannot be performed in the present state Traceback (most recent call last): File "/home/ubuntu/stt_translate/audio_recorder.py", line 629, in _transcription_worker model = faster_whisper.WhisperModel( File "/usr/local/lib/python3.10/dist-packages/faster_whisper/transcribe.py", line 144, in init self.model = ctranslate2.models.Whisper( RuntimeError: CUDA failed with error the operation cannot be performed in the present state

How to choose the CUDA version with the latest RealtimeSTT？should I download cuDNN v8.7.0 (November 28th, 2022), for CUDA 11.x"? thanks!

KoljaB commented 5 months ago

Ctranslate2 stopped support for CUDA 11. Please upgrade to CUDA 12. See here.

jacobtang commented 5 months ago

[2024-06-12 16:16:38.255] [ctranslate2] [thread 30905] [warning] The compute type inferred from the saved model is float16, but the target device or backend do not support efficient float16 computation. The model weights have been automatically converted to use the float32 compute type instead the CUDA info： vcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2023 NVIDIA Corporation Built on Fri_Jan__6_16:45:21_PST_2023 Cuda compilation tools, release 12.0, V12.0.140 Build cuda_12.0.r12.0/compiler.32267302_0