pesmaker train

train continues from a training YAML. It uses the same smart workflow driver as next, but is easier to remember after dataset collection.

Use

pesmaker validate train.yaml
pesmaker train train.yaml

For a training YAML, this prepares model-training inputs and then prints the next submit step when a submit script is ready.

Example

project: train_initial_structure

training:
  model: nep
  output_dir: training
  dataset: train.xyz
  command: nep

For two-step NEP training, add:

training:
  model: nep
  output_dir: training
  dataset: train.xyz
  two_step: true

Run pesmaker next train.yaml after the first training job produces training/step1/nep.txt; PESMaker then creates training/step2 from step1 and prepares the second-step input.

Plot training results

After NEP writes loss.out, energy_train.out, force_train.out, and stress_train.out, run:

pesmaker plot train

PESMaker writes figures under plot/, auto-detecting outputs in the current directory, training/step2, training/step1, or training.