Let’s use freely available HuggingFace models to make programming more accessible.
This project is set up to use Conda for managing the Python dependency, and uses Poetry for managing Python package dependencies.
Set up the environment using Conda.
$ conda env create -f conda.yaml
$ conda activate speech-to-code
This will give you a fresh Python 3.11 environment.
Install Poetry.
If you have not done this, I reccomend looking at the official guide.
This can be done by using e.g. Poetry for installing the package at the root, using poetry install.
The most basic usage looks like the following.
from speech_to_code import SpeechToCode
converter = SpeechToCode.default()
converter.convert(file_path="path/to/audio.wav")
This will create a speech-to-code pipeline composed of Distil-Whisper (distil-small.en) and Phi-2.
For details on how to run the API, please refer to the api/README.md.