Welcome to mlcg-tk’s documentation!

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

A collection of tools for processing raw simulation data for use in training a transferable coarse grained (CG) forcefield, using the MLCG library.

Installation

Before installing mlcg-tk, you must install the `mlcg package <https://github.com/ClementiGroup/mlcg>`__.

Similar to mlcg, we encourage the use of the uv enviroment manager

Once mlcg is installed, you can install mlcg-tk as follows

git clone git@github.com:ClementiGroup/mlcg-tk.git
cd mlcg-tk
uv pip install .

For developers

Add --group dev to install additional development dependencies (black, pytest, coverage), e.g.

uv pip install . --group dev

Contents:

• Data pipeline user guide
  • About the Training Data Generation Pipeline
  • 0) Create a directory to output all the intermediate files
  • 1) Loading and processing all-atom simulation data
  • 2) Computing statistics and fitting priors
  • 3) Producing delta forces
  • 4) Package training data
  • 5) Generate simulation input
• API documentation
  • Core API
    • input_generator package
      • prior_fit package
        • dihedral module
        • fit_potentials module
        • harmonic module
        • histogram module
        • polynomial module
        • repulsion module
        • restricted_bending module
        • utils module
      • embedding_maps module
      • prior_gen module
      • prior_nls module
      • raw_data_loader module
      • raw_dataset module
      • utils module
    • prior_tools package
      • utils module
  • Scripts
    • gen_input_data.py
    • fit_priors.py
    • merge_statistics.py
    • produce_delta_forces.py
    • package_training_data.py
    • add_decoys.py
    • gen_sim_input.py