For all your TensorFlow Variational Autoencoder needs.

This is the repository for the Variational Autoencoder (VAE) blogpost series from Fast Forward Labs. Start there, then check out the repo!

• Part I: Introducing Variational Autoencoders (in Prose and Code)
• Part II: Under the Hood of the Variational Autoencoder (in Prose and Code)

To train a new model, edit main.py with your desired VAE ARCHITECTURE, HYPERPARAMETERS, and paths/to/outdirs.

Then, simply:

$ python main.py

OR, restore a trained model from its saved meta_graph via:

$ python main.py <path/to/meta_graph_name>

(without the .meta suffix)