Implementation of VQ-VAE for audio as described the DeepMind's paper here.
There exists several implementations of VQ-VAE using PixelCNN as the encoder/decoder.
The goal in this repo is to have WaveNet as the encoder/decoder.
The starting code for this was adapted from
- hiwonjoon's repo using VQ-VAE with the pixel-cnn architecture https://github.com/hiwonjoon/tf-vqvae
- ibab's repo loading audio files and the implementation of the WaveNet architecture https://github.com/ibab/tensorflow-wavenet