0.1 - Basic utilities and flows

This is a release in support of the article "Multimap targeted free energy estimation" article.

The library includes the following features.

Normalizing flows

• Masked Autoregressive Flows with a ready-to-use MADE conditioner and the following transformers:
  • Affine
  • (Circular) neural spline
  • Sum-of-squares polynomial
  • Mobius (based on that proposed in)
• Continuous normalizing flows with a dynamics layer based on Equivariant graph neural networks.
• Centroid and Oriented flows, which can be used to fix the reference system before the map.
• A PCA flow to perform the transformation in uncorrelated coordinates.

Utilities

• PyTorch wrappers for psi4 and the Atomistic Simulation Environment
  to evaluate potential energies and forces with multiple molecular simulation engines.
• A PyTorch Dataset that can wrap an MDAnalysis Universe to train on molecular simulation trajectories.
• A PyTorch-accelerated (T)FEP estimator.
• A PyTorch-accelerated bootstrap analysis utility.
• A simple storage utility class to save potential energies and log training information.