produce_delta_forces.py

mlcg_tk.scripts.produce_delta_forces.produce_delta_forces(dataset_name, names, tag, save_dir, prior_tag, prior_fn, device, batch_size, force_tag=None, mol_num_batches=1)[source]

Removes prior energy terms from input forces to produce delta force input for training

Parameters:

  • dataset_name (str) – Name given to specific dataset

  • names (List[str]) – List of sample names

  • tag (str) – Label given to all output files produced from dataset

  • save_dir (str) – Path to directory from which input will be loaded and to which output will be saved

  • prior_tag (str) – String identifying the specific combination of prior terms

  • prior_fn (str) – Path to filename in which prior model is saved

  • device (str) – Device on which to run delta force calculations

  • batch_size (int) – Number of frames to take per batch

  • force_tag (str) – Optional tag to identify input for a particular run of delta force calculation

  • mol_num_batches (int) – If greater than 1, will load each molecule data from the specified number of batches that were be treated as different samples

mlcg_tk.scripts.produce_delta_forces.remove_baseline_forces_collated(collated_data, model)[source]

Compute the forces on the input collated_data with the models and remove them from the reference forces contained in data_list. The computation of the forces is done on the whole data_list at once so it should not be too large.

Parameters:

  • collated_data (AtomicData) – Collated list of AtomicData instances that contain the full reference forces

  • models – SumOut object containing models that compute prior/baseline forces

Returns:

Uncollated list of AtomicData instances, where the value of the ‘forces’ field is now the delta forces (original forces minus the baseline/prior forces). An additional field ‘baseline’ forces is added, whose value is equal to the baseline/prior forces

Return type:

List[AtomicData]