PyTorch VAE algorithms

This project provides an experimentation environment for variational autoencoders. All models are implemented in PyTorch and trained with Lightning. Examples of generative modeling and representation learning are provided for MNIST and CIFAR-10.

Notebooks

• Introduction

• Bernoulli VAEs for MNIST

• Gaussian VAE for CIFAR-10

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