MLX verisons of the models
#1
by
fuzzbin
- opened
Hello!
First, thank you for providing these models. We ar testing different use cases for local transcription.
So to the question: Does anyone know how to convert this model so it can be used with MLX?
Hi,
Glad to read you are finding use cases for it!
Unfortunately, it's not very straightforward. First, you need to convert the HF model to a PyTorch .ptmodel file. We've used this convert_hf_whisper() function with some success in the past. And once you have the .ptfile, you might be abe to convert it to MLX by following the mlx-examples convert.py script, but it hasn't been updated in a while, so I'm not sure if it will work.
Good luck :)