Open patrick91 opened 10 months ago
I actually specified --device mps and still get an error?
insanely-fast-whisper --file-name live-stream-jimmy.mp4 --device-id "mps"
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
Traceback (most recent call last):
File "/Users/jimmylv/.local/bin/insanely-fast-whisper", line 8, in <module>
sys.exit(main())
File "/Users/jimmylv/.local/pipx/venvs/insanely-fast-whisper/lib/python3.9/site-packages/insanely_fast_whisper/cli.py", line 86, in main
pipe = pipeline(
File "/Users/jimmylv/.local/pipx/venvs/insanely-fast-whisper/lib/python3.9/site-packages/transformers/pipelines/__init__.py", line 1070, in pipeline
return pipeline_class(model=model, framework=framework, task=task, **kwargs)
File "/Users/jimmylv/.local/pipx/venvs/insanely-fast-whisper/lib/python3.9/site-packages/transformers/pipelines/automatic_speech_recognition.py", line 239, in __init__
self.model.to(device)
File "/Users/jimmylv/.local/pipx/venvs/insanely-fast-whisper/lib/python3.9/site-packages/transformers/modeling_utils.py", line 2271, in to
return super().to(*args, **kwargs)
File "/Users/jimmylv/.local/pipx/venvs/insanely-fast-whisper/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1141, in to
device, dtype, non_blocking, convert_to_format = torch._C._nn._parse_to(*args, **kwargs)
RuntimeError: Invalid device string: 'cuda:mps'
So, you can pass
--device mps
to make it work on mX macs, but I wonder if it wouldn't be nicer to automatically detect that? :)