Welcome to mlcg’s documentation!

This is the base code to create models such as the ones described in [TransCGSchnet] and [CGSchnet].

This repository collects a set of tools to apply machine learning techniques to coarse grain atomic systems.

Installation

The dependencies are defined in requirements.txt but some packages are not well handled by pip. So start by installing pytorch and pytorch-geometric with conda, e.g.:

conda install pytorch cudatoolkit=11.3 -c pytorch
conda install pyg -c pyg -c conda-forge

Support for the MACE model can be enabled with:

pip install git+https://github.com/felixmusil/mace.git@develop

Support for the TorchMD-Net models can be enabled with:

pip install git+https://github.com/felixmusil/torchmd-net.git

CLI

The models defined in this library can be convinietly trained using the pytorch-lightning CLI utilities.

Examples

Please take a look into the examples folder to see how to use this code to train a model over an existing dataset.

Contents

• Data Representation
  • AtomicData
• Geometry
  • Internal Coordinate Functions
  • Topology Utilities
  • Statistics
• Neighbor Lists
  • Neighbor List Utilities
  • Torch Implementation
  • ASE Implementation
• Coarse Graining
  • build_cg_topology()
  • build_cg_matrix()
• Datasets
  • Chignolin Dataset
  • Alanine Dipeptide Dataset
  • Custom H5 Dataset
• Priors
  • Harmonic
  • HarmonicBonds
  • HarmonicAngles
  • Dihedral
  • Repulsion
• Models
  • SchNet Utilities
  • MACE Utilities
  • Radial Basis Functions
  • Cutoff Functions
  • Model Building Utilities
  • Loss Functions
• Training
  • Scripts
  • Utils for using Pytorch Lightning
• Simulations
  • NVT ensemble
  • Utilities
• Developer Documentation
  • Documentation
• Bibliography
  • Dataset & Simulation References
  • CGnet References
  • Neural Network Architectures & Implementations

Indices and tables

• Index

• Module Index

• Search Page