What is WaveNet?

WaveNet is a machine learning architecture used for audio generation. Instead of utilizing RNNs, WaveNet uses dilated convolutions to train.

This project reimplements the paper in Tensorflow with Keras backend.

See paper and blog for more information: https://deepmind.com/blog/article/wavenet-generative-model-raw-audio https://arxiv.org/pdf/1609.03499.pdf

Necessary Tools:

1. Python 3
2. Docker (Docker Engine API v1.40 for gpu)

Works on all platforms but tested on Ubuntu 18

Building and Running

Pull docker image

docker pull tensorflow/tensorflow:2.1.0-gpu-py3

Build Code

docker build -t wavenet/latest .

Run code

docker run -v $(pwd)/saved_data:/saved_data:rw --gpus all -it --rm --name wavenetbox wavenet/latest