Official paper: beta-VAE: Learning Basic Visual Concepts with a Constrained Variational Framework
This repository's main notebook is results.ipynb
.
Links to download the datasets:
To train the models run train.py
.
Feel free to play with these parameters:
dataset_name
= CelebA
/Chairs
/DSprites
/FERG
num_epochs = 10
batch_size= 64
trainset_percentage = 0.8
betas = [1, 5, 25, 50, 100, 250, 500]