An unknown error occurred when using Faster Whisper

[Code-1123]

The log shows:
[2024-07-28 14:59:38,231] file_transcriber.run:61 ERROR ->
Traceback (most recent call last):
File "buzz\transcriber\file_transcriber.py", line 59, in run
File "buzz\transcriber\whisper_file_transcriber.py", line 80, in transcribe
Exception: Unknown error

[raivisdejus]

Please select a language to transcribe, I think this is an issue with "Detect language" that was recently fixed but is not yet released

[Code-1123]

Thanks for your reply! I can currently run Faster Whisper on Windows 11, Nvidia 4060-8G, but the GPU utilization fluctuates between 0 and 40. Is this normal?

[raivisdejus]

Right now it works, and I am happy that it is so :)
Next step will be to sort out GPU usage on Windows. I have some ideas of what is missing, so those fixes will come a bit later.

Essentially you need to install the pip version, the wheels and then also change the torch to CUDA version. Something similar to what was recommended https://stackoverflow.com/questions/75775272/cuda-and-openai-whisper-enforcing-gpu-instead-of-cpu-not-working or SYSTRAN/faster-whisper#140

Or use WSL and install Linux version

[Code-1123]

Got it! :D

[Code-1123]

Unfortunately the strange thing is that when the process reaches 100%, it drops back to 0%, and then still reports an error ㅠ_ㅠ：

[2024-07-31 17:06:03,980] whisper_file_transcriber.transcribe:73 DEBUG -> whisper process completed with code = 3221226505, time taken = 0:04:36.103324, number of segments = 0
[2024-07-31 17:06:03,980] file_transcriber.run:61 ERROR ->
Traceback (most recent call last):
File "buzz\transcriber\file_transcriber.py", line 59, in run
File "buzz\transcriber\whisper_file_transcriber.py", line 82, in transcribe
Exception: Unknown error

Maybe my sound sequence is too long and causing problems with the output?

[raivisdejus]

The first progress to 100% the first time you use the model will convert it to the supported format. It is ok that it drops to 0% after that.

The fact that it fails after may be an issue with model used or settings or the whole setup.

Can you share more info on your setup?

• What model size did you use
• What language was selected to transcribe to
• Was "word level timestamps" selected
• Paste this %USERPROFILE%\AppData\Local\Buzz\Buzz\Logs in the address bar of windows and check the logs file, There may be more info, like maybe some DLL was missing or some other error. If possible upload the log file somewhere
• I doubt the audio is to blame, but try some shorter part of it or shorter audio as a test.

My gut feeling tells the log file will have more info and some part of the CUDA setup for Windows is missing. Like in my tests I had some DLL missing

[Code-1123]

I've tried 3 times.
[2024-07-31 15:01:09,572]: I used the faster-whisper-large-v2 model and input a two-hour video in MP4 format. All paths were in English and the file names contained three Chinese characters.
[2024-07-31 17:00:02,040]: I imported the file in mp3 format in the same path and removed the Chinese characters.
[2024-07-31 17:01:27,877]: I guessed that large-v2 would take up too much video memory and cause overflow, so I changed to medium.

I just downloaded large-v2 from hugging face and tried to recognize a 30-minute audio using other GUI projects. It can be seen that it completed the audio recognition, but when it finally output and saved the srt file, it reported a GPUmemory overflow error. It succeeded when I changed the 10-minute audio file. Maybe the unknown error shows in the log is GPU memory overflow? I will try to recognize a shorter audio file later.

====================================================
[2024-07-31 15:01:09,572] _api.release:286 DEBUG -> Attempting to release lock 2133363470672 on C:\Users\xxxx\AppData\Local\Buzz\Buzz\Cache\models.locks\models--guillaumekln--faster-whisper-large-v2\bf2a9746382e1aa7ffff6b3a0d137ed9edbd9670c3b87e5d35f5e85e70d0333a.lock
[2024-07-31 15:01:09,572] _api.release:289 DEBUG -> Lock 2133363470672 released on C:\Users\xxxx\AppData\Local\Buzz\Buzz\Cache\models.locks\models--guillaumekln--faster-whisper-large-v2\bf2a9746382e1aa7ffff6b3a0d137ed9edbd9670c3b87e5d35f5e85e70d0333a.lock
[2024-07-31 15:01:11,088] file_transcriber_queue_worker.run:56 DEBUG -> Starting next transcription task
[2024-07-31 15:01:11,100] whisper_file_transcriber.transcribe:50 DEBUG -> Starting whisper file transcription, task = FileTranscriptionTask(transcription_options=TranscriptionOptions(language='ko', task=<Task.TRANSCRIBE: 'transcribe'>, model=TranscriptionModel(), word_level_timings=False, temperature=(0.0, 0.2, 0.4, 0.6, 0.8, 1.0), initial_prompt='', openai_access_token='', enable_llm_translation=False, llm_prompt='', llm_model=''), file_transcription_options=FileTranscriptionOptions(file_paths=['E:/xxxx/xxxx/xxxx/xxxx-xxx/9 .mp4'], url=None, output_formats={<OutputFormat.SRT: 'srt'>}), model_path='C:\Users\xxxx\AppData\Local\Buzz\Buzz\Cache\models\models--guillaumekln--faster-whisper-large-v2\snapshots\f541c54c566e32dc1fbce16f98df699208837e8b', id=31820218, uid=UUID('73d00e30-3c85-4bd8-86d3-f3e292107c67'), segments=[], status=None, error=None, queued_at=None, started_at=None, completed_at=None, output_directory=None, source=<Source.FILE_IMPORT: 'file_import'>, file_path='E:/xxxx/xxxx/xxxx/xxxx/9==some chinese chars==.mp4', url=None, fraction_downloaded=0.0)
[2024-07-31 16:14:40,739] whisper_file_transcriber.transcribe:73 DEBUG -> whisper process completed with code = 3221226505, time taken = 1:13:29.639518, number of segments = 0
[2024-07-31 16:14:40,755] file_transcriber.run:61 ERROR ->
Traceback (most recent call last):
File "buzz\transcriber\file_transcriber.py", line 59, in run
File "buzz\transcriber\whisper_file_transcriber.py", line 82, in transcribe
Exception: Unknown error
[2024-07-31 16:14:40,761] file_transcriber_queue_worker.run:40 DEBUG -> Waiting for next transcription task
[2024-07-31 16:59:57,661] settings.init:14 DEBUG -> Settings filename: \HKEY_CURRENT_USER\Software\Buzz\OrganizationDefaults
[2024-07-31 16:59:57,676] settings.init:14 DEBUG -> Settings filename: \HKEY_CURRENT_USER\Software\Buzz\OrganizationDefaults
[2024-07-31 17:00:02,026] file_transcriber_queue_worker.run:56 DEBUG -> Starting next transcription task
[2024-07-31 17:00:02,040] whisper_file_transcriber.transcribe:50 DEBUG -> Starting whisper file transcription, task = FileTranscriptionTask(transcription_options=TranscriptionOptions(language='ko', task=<Task.TRANSCRIBE: 'transcribe'>, model=TranscriptionModel(), word_level_timings=False, temperature=(0.0, 0.2, 0.4, 0.6, 0.8, 1.0), initial_prompt='', openai_access_token='', enable_llm_translation=False, llm_prompt='', llm_model=''), file_transcription_options=FileTranscriptionOptions(file_paths=['E:/xxxx/xxxx/xxxx/xxxx-xxx/1/1.MP3'], url=None, output_formats={<OutputFormat.SRT: 'srt'>}), model_path='C:\Users\xxxx\AppData\Local\Buzz\Buzz\Cache\models\models--guillaumekln--faster-whisper-large-v2\snapshots\f541c54c566e32dc1fbce16f98df699208837e8b', id=93709178, uid=UUID('e9b7589d-2eeb-40d4-93e2-c5f25278d287'), segments=[], status=None, error=None, queued_at=None, started_at=None, completed_at=None, output_directory=None, source=<Source.FILE_IMPORT: 'file_import'>, file_path='E:/xxxx/xxxx/xxxx/xxxx/1/1.MP3', url=None, fraction_downloaded=0.0)
[2024-07-31 17:00:20,864] whisper_file_transcriber.transcribe:73 DEBUG -> whisper process completed with code = 3221226505, time taken = 0:00:18.824139, number of segments = 0
[2024-07-31 17:00:20,864] file_transcriber.run:61 ERROR ->
Traceback (most recent call last):
File "buzz\transcriber\file_transcriber.py", line 59, in run
File "buzz\transcriber\whisper_file_transcriber.py", line 82, in transcribe
Exception: Unknown error
[2024-07-31 17:00:20,864] file_transcriber_queue_worker.run:40 DEBUG -> Waiting for next transcription task
[2024-07-31 17:00:48,856] settings.init:14 DEBUG -> Settings filename: \HKEY_CURRENT_USER\Software\Buzz\OrganizationDefaults
[2024-07-31 17:00:48,872] settings.init:14 DEBUG -> Settings filename: \HKEY_CURRENT_USER\Software\Buzz\OrganizationDefaults
[2024-07-31 17:00:48,935] settings.init:14 DEBUG -> Settings filename: \HKEY_CURRENT_USER\Software\Buzz\OrganizationDefaults
[2024-07-31 17:00:48,935] settings.init:14 DEBUG -> Settings filename: \HKEY_CURRENT_USER\Software\Buzz\OrganizationDefaults
[2024-07-31 17:01:04,030] settings.init:14 DEBUG -> Settings filename: \HKEY_CURRENT_USER\Software\Buzz\OrganizationDefaults
[2024-07-31 17:01:04,030] settings.init:14 DEBUG -> Settings filename: \HKEY_CURRENT_USER\Software\Buzz\OrganizationDefaults
[2024-07-31 17:01:27,871] file_transcriber_queue_worker.run:56 DEBUG -> Starting next transcription task
[2024-07-31 17:01:27,877] whisper_file_transcriber.transcribe:50 DEBUG -> Starting whisper file transcription, task = FileTranscriptionTask(transcription_options=TranscriptionOptions(language='ko', task=<Task.TRANSCRIBE: 'transcribe'>, model=TranscriptionModel(), word_level_timings=False, temperature=(0.0, 0.2, 0.4, 0.6, 0.8, 1.0), initial_prompt='', openai_access_token='', enable_llm_translation=False, llm_prompt='', llm_model=''), file_transcription_options=FileTranscriptionOptions(file_paths=['E:/xxxx/xxxx/xxxx/xxxx-xxx/1/1.MP3'], url=None, output_formats={<OutputFormat.SRT: 'srt'>}), model_path='C:\Users\xxxx\AppData\Local\Buzz\Buzz\Cache\models\models--guillaumekln--faster-whisper-medium\snapshots\8701f851d407f3f47e091bb13b8dac5290c7f7fb', id=73272940, uid=UUID('72f43d8e-eee5-44bb-8d99-2674d3b77a11'), segments=[], status=None, error=None, queued_at=None, started_at=None, completed_at=None, output_directory=None, source=<Source.FILE_IMPORT: 'file_import'>, file_path='E:/xxxx/xxxx/xxxx/xxxx/1/1.MP3', url=None, fraction_downloaded=0.0)
[2024-07-31 17:06:03,980] whisper_file_transcriber.transcribe:73 DEBUG -> whisper process completed with code = 3221226505, time taken = 0:04:36.103324, number of segments = 0
[2024-07-31 17:06:03,980] file_transcriber.run:61 ERROR ->
Traceback (most recent call last):
File "buzz\transcriber\file_transcriber.py", line 59, in run
File "buzz\transcriber\whisper_file_transcriber.py", line 82, in transcribe
Exception: Unknown error

[Code-1123]

My configuration is Windows 11, i5-13500HX, 16.0 GB, RTX4060-8G

[Code-1123]

I just tried to use a 5-minute audio, but unfortunately it still Unknown error, and unlike other models, when an error occurs, a command line window will not pop up after execution.

[2024-07-31 23:48:38,898] settings.init:14 DEBUG -> Settings filename: \HKEY_CURRENT_USER\Software\Buzz\OrganizationDefaults
[2024-07-31 23:48:38,898] settings.init:14 DEBUG -> Settings filename: \HKEY_CURRENT_USER\Software\Buzz\OrganizationDefaults
[2024-07-31 23:48:41,934] file_transcriber_queue_worker.run:56 DEBUG -> Starting next transcription task
[2024-07-31 23:48:41,940] whisper_file_transcriber.transcribe:50 DEBUG -> Starting whisper file transcription, task = FileTranscriptionTask(transcription_options=TranscriptionOptions(language='ko', task=<Task.TRANSCRIBE: 'transcribe'>, model=TranscriptionModel(), word_level_timings=False, temperature=(0.0, 0.2, 0.4, 0.6, 0.8, 1.0), initial_prompt='', openai_access_token='', enable_llm_translation=False, llm_prompt='', llm_model=''), file_transcription_options=FileTranscriptionOptions(file_paths=['E:/XXX/XXX/XXX/XXXX-XXX/1/3.MP3'], url=None, output_formats={<OutputFormat.SRT: 'srt'>}), model_path='C:\Users\XXX\AppData\Local\Buzz\Buzz\Cache\models\models--guillaumekln--faster-whisper-large-v2\snapshots\f541c54c566e32dc1fbce16f98df699208837e8b', id=6128112, uid=UUID('7ad53a8d-20c3-4c13-88d8-af0c5cfbd44d'), segments=[], status=None, error=None, queued_at=None, started_at=None, completed_at=None, output_directory=None, source=<Source.FILE_IMPORT: 'file_import'>, file_path='E:/XXX/XXX/XXX/XXX/1/3.MP3', url=None, fraction_downloaded=0.0)
[2024-07-31 23:48:53,137] whisper_file_transcriber.transcribe:73 DEBUG -> whisper process completed with code = 3221226505, time taken = 0:00:11.196818, number of segments = 0
[2024-07-31 23:48:53,137] file_transcriber.run:61 ERROR ->
Traceback (most recent call last):
File "buzz\transcriber\file_transcriber.py", line 59, in run
File "buzz\transcriber\whisper_file_transcriber.py", line 82, in transcribe
Exception: Unknown error
[2024-07-31 23:48:53,137] file_transcriber_queue_worker.run:40 DEBUG -> Waiting for next transcription task

[raivisdejus]

I was able to run Whisper and Huggingface whisper with CUDA GPU support on Windows. See note here #860 (comment)

[raivisdejus]

Faster whisper requires CUDA 12, on systems with older CUDA version CPU will be used