Skip to content

Latest commit

 

History

History
34 lines (24 loc) · 652 Bytes

README.md

File metadata and controls

34 lines (24 loc) · 652 Bytes
Raw

VQVAE-WaveNet

A PyTorch implementation of VQVAE and WaveNet, forked from https://github.com/dhgrs/chainer-VQ-VAE.

The status of this repository is work in progress.

Requirements

  • Python>=3.8
  • PyTorch>=1.6.0
  • torchaudio>=0.6.0
  • homura>=2020.08

e.g.,

conda create -n ${PROJECT_NAME} python=3.8
conda activate ${PROJECT_NAME}
conda install -c pytorch pytorch torchaudio cudatoolkit==10.2
pip install git+https://github.com/moskomule/homura@v2020.08

Train and inference

Training

python train.py

Inference (generating)

python generate.py -i input_file -o output_file -m saved_model -s speaker