Active Learning of Hydrogen Combustion Reaction.

Installation and Dependencies

We recommend using conda environment to install dependencies of this library. Please install (or load) conda and then proceed with the following commands:

conda create --name torch-gpu python=3.7
conda activate torch-gpu
conda install -c conda-forge numpy scipy pandas ase pyyaml tqdm scikit-learn attrs xlsxwriter
conda install -c pytorch pytorch torchvision cudatoolkit=11.3
conda install -c conda-forge plumed py-plumed
pip install rmsd

This repository is built on top of the (NewtonNet repository)[https://github.com/THGLab/NewtonNet]. To install NewtonNet, run the following command:

cd NewtonNet
pip install -e .

The developer installation is available and for that you need to first clone H2Combustion from this repository:

git clone https://github.com/THGLab/H2Combustion_AL.git

And adding these two directories into $PYTHONPATH:

export PYTHONPATH=$PYTHONPATH:/path/to/your/directory/NewtonNet
export PYTHONPATH=$PYTHONPATH:/path/to/your/directory/H2Combustion_AL

This should take ~20 minute to install. Now, you can run combust and md modules anywhere on your computer as long as you are in the torch-gpu environment.

OS requirement

This package is supported for macOS and Linux. The package has been tested on the following systems: macOS: Big Sur (11.4) Linux: x86_64 GNU/Linux

Guidelines

• You can find several models inside the scripts directory that rely on the implemented modules in the combust and md library. The yaml file control setting of the model. Please modify parameters using the yaml files when retraining.

• The documentation of the modules are available at most cases. Please look up local classes or functions and consult with the docstrings in the code.

• Some paths are hardcoded into the code that is specific to developer. Make sure to go through and change them before using the code.

Demo

See script/demo/md_demo.py for using trained model to run md simulation. The expected output is included in script/demo/demo_output/, it takes ~10 seconds on a laptop to run this demo.

Model Training

To train a model, run this file after you've modified the settings in both the yaml file md/active_learning/config_h2_template.yml and some paths and settings in the code. Currently the code is tailored towards specific use of traing on Lawrencium HPC.

python md/active_learning/active_learning.py model_name