This repository is forked from the great EnCodec repository associated with the High Fidelity Neural Audio Compression paper that aims to compress audio. We have added some functionality to train/finetune the EnCodec model for its use in the Neuralink Compression Challenge.
Python 3.8 is required and a reasonably recent version of PyTorch (1.11.0 ideally).
To install (Neuro-)EnCodec, you can run from this repository:
- Clone repo locally
pip install .The train script does not support distributed training, yet. Soon to come...
We want to thank the following open-source repositories on which we based this project:
The code in this repository is released under the MIT license as found in the LICENSE file.