Examples for using AceFF neural network potential.

The ASE and batched PyTorch examples use AceFF-2-OMOL. The OpenMM examples use AceFF-2.0 until AceFF-2-OMOL support is released in OpenMM-ML.

There are 4 example notebooks They can be run in Google Colab or locally.

• Single point energy calculator with ASE.

  • ./notebooks/aceff_single_point_calculation.ipynb

• Batch minimization and molecular dynamics with PyTorch and TorchMD.

  • ./notebooks/aceff_batched_minimization_and_md.ipynb

• Running MD for a small molecule in vacuum with OpenMM.

  • ./notebooks/aceff_openmm_md_example.ipynb

• Running MD for a protein ligand complex using MM/ML mechanical embedding scheme with OpenMM.

  • ./notebooks/aceff_openmm_protein_ligand.ipynb

Python Environment for running locally

You will need to install:

• torchmd-net (3.1.0 or newer for AceFF-2-OMOL)

and depending on which example

• openmm-torch openmm-ml
• ase
• rdkit
• torchmd

For the MD examples you will need a CUDA GPU.

The AceFF model files can be downloaded from HuggingFace: https://huggingface.co/Acellera