PyAutoencoder

A clean, modular PyTorch library for building and training autoencoders.

PyAutoencoder is an open-source Python library that gives simple access to autoencoder frameworks in PyTorch. You bring your own nn.Module encoder and decoder, and the library wraps them into ready-to-use models, each paired with its loss function.

  • Minimal, composable API: no complicated base classes and no custom training loop to learn.
  • Ready-to-use models: Autoencoder (AE), Variational Autoencoder (VAE) and Adaptive Group VAE (Ada-GVAE), with reconstruction likelihoods and KL divergence already implemented.
  • Research-oriented: lightweight and transparent, designed for quick prototyping with full control.
pip install pyautoencoder