Introduction to variational autoencoders

This project provides an experimentation environment for VAE algorithms. They can be used for generative modeling and representation learning. The models are implemented in PyTorch and trained with Lightning. Application examples based on MNIST and CIFAR-10 are provided.

Installation

pip install -e .

Training

python scripts/main.py fit --config config/mnist_dense.yaml

python scripts/main.py fit --config config/mnist_conv.yaml

python scripts/main.py fit --config config/cifar10.yaml

Notebooks

• Introduction

• Bernoulli VAEs for MNIST

• Gaussian VAE for CIFAR-10