3) Producing delta forces

Command:

mlcg-tk-produce_delta_forces produce_delta_forces --config configuration_files/trpcage_delta_forces.yaml

This procedure will load the prior model specified by prior_fn and then once again loop over all sample names provided. It will then calculate and remove the baseline forces using the coordinates, forces, embeddings, and neighbourlists created in the previous step. It will then save the delta forces which can then be used for training.

The following example script shows how delta forces can be computed on a computing cluster using GPU acceleration:

#!/bin/bash
#SBATCH --ntasks-per-node=1
#SBATCH --nodes=1
#SBATCH --mem=24G
#SBATCH --time=4-00:00:00
#SBATCH --gres=gpu:1
#SBATCH --partition=gpu
#SBATCH --output=trpcage_delta_forces_gpu.log
#SBATCH --job-name=test_job

mlcg-tk-produce_delta_forces produce_delta_forces --config configuration_files/trpcage_delta_forces.yaml

Here, make sure to specify cuda for the device option in the configuration file. Note that depending on the GPU being used and its available memory, it may be necessary to adjust the batch_size.