How do I use this with an existing Whisper installation?

#1
by athu16 - opened

Since this is a custom Whisper model, how do I use it in OpenAI whisper? The whisper.load_model() function only accepts strings like "base", "small", "medium", etc.

Hi, while I have not personally used a custom Whisper model in OpenAI Whisper, my guess is that you can do so by following these steps:

Step 1: Install git-lfs and clone the Hugging Face repository. Once this is done, the model will be saved as a directory. (observe the folder located on the left-hand side in Colab.)
!sudo apt-get install git-lfs
!git clone https://huggingface.co/DrishtiSharma/whisper-large-v2-marathi

Step 2: Install the dependencies related to openai.

Step 3: import openai.whisper as whisper
model_path = "/content/whisper-large-v2-marathi"
model = whisper.load_model(model_path)

this didn't work at all.. It's giving this error:
RuntimeError: Model /san_test/whisper-medium-ar not found; available models = ['tiny.en', 'tiny', 'base.en', 'base', 'small.en', 'small', 'medium.en', 'medium', 'large-v1', 'large-v2', 'large']

Sign up or log in to comment