How are you running Simpaudio?
Running from source (Python)
Which feature was affected?
Speech-to-Text (Whisper)
What happened?
Bug Description
Running the self-test (python app.py --selftest) or attempting audio transcription on certain NVIDIA GPU / driver configurations causes faster-whisper to fail with an unhandled cuBLAS exception instead of falling back gracefully to CPU execution.
Error Output / Traceback
[FAIL] whisper-stt: RuntimeError: cuBLAS failed with status CUBLAS_STATUS_NOT_SUPPORTED
Traceback (most recent call last):
File "..\simpaudio\selftest.py", line 28, in check
fn()
File "..\simpaudio\selftest.py", line 80, in test_whisper
result = stt.transcribe(wav, status_callback=lambda m: None)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "..\simpaudio\whisper_stt.py", line 21, in transcribe
segments, info = self._model.transcribe(
^^^^^^^^^^^^^^^^^^^^^^^
File "..\simpaudio\venv\Lib\site-packages\faster_whisper\transcribe.py", line 942, in transcribe
) = self.detect_language(
^^^^^^^^^^^^^^^^^^^^^
File "..\simpaudio\venv\Lib\site-packages\faster_whisper\transcribe.py", line 1819, in detect_language
encoder_output = self.encode(
^^^^^^^^^^^^
File "..\simpaudio\venv\Lib\site-packages\faster_whisper\transcribe.py", line 1400, in encode
return self.model.encode(features, to_cpu=to_cpu)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: cuBLAS failed with status CUBLAS_STATUS_NOT_SUPPORTED
[FAIL] overall: FAILURES DETECTED
What did you expect to happen?
When CUDA/cuBLAS acceleration is unsupported or fails on the host machine's GPU setup, whisper_stt.py should catch the runtime error and fall back gracefully to CPU execution (device="cpu"), or display a clear notification rather than crashing the process/failing the selftest suite.
Operating system
Microsoft Windows 11 Version 25H2 (OS Build 26200.7840)
Steps to reproduce
No response
Checklist
How are you running Simpaudio?
Running from source (Python)
Which feature was affected?
Speech-to-Text (Whisper)
What happened?
Bug Description
Running the self-test (
python app.py --selftest) or attempting audio transcription on certain NVIDIA GPU / driver configurations causesfaster-whisperto fail with an unhandled cuBLAS exception instead of falling back gracefully to CPU execution.Error Output / Traceback
What did you expect to happen?
When CUDA/cuBLAS acceleration is unsupported or fails on the host machine's GPU setup, whisper_stt.py should catch the runtime error and fall back gracefully to CPU execution (device="cpu"), or display a clear notification rather than crashing the process/failing the selftest suite.
Operating system
Microsoft Windows 11 Version 25H2 (OS Build 26200.7840)
Steps to reproduce
No response
Checklist