Archived: The code has moved with a cleaner implementation without pytorch lightning : https://github.com/WaveGenAI/phonira

A model that can make edit of music based on a prompt. Follow MusicGen architecture.

Installation

python -m pip install -r requirements.txt

Dataset

The dataset used is: https://huggingface.co/datasets/WaveGenAI/youtube-cc-by-music

Launch the trainig

python3 -m scripts.train --config_path config/config.yaml

Run inference

python3 -m scripts.inference --config_path config/config.yaml --checkpoint_path PATH

Then, go to http://127.0.0.1:7860

License

See the LICENSE file.