This Repo contains pyTorch Implementation of the paper Neural Discrete Representation Learning . As of now, VectorQuantization is done. VectorQuantizedEMA method is not yet Implemented but will be done soon. And also,will try to pair the Network with an Autoregressive Model soon.
If you want to Train the model,plot Loss and view the Test reconstructions,then use command
python test.py
If you just want to train the model and view train Loss plots, then use
python train.py
The hyperparameters of the Model and Network Architecture is taken from author's Implementation.
FID Score on reconstructed test images: 21.55
