train
Description:
Performs automatic training for NEP.
Input Parameters
Important
You should use
NepTrain init <type>to generatejob.yaml, and then useNepTrain train job.yamlto start the training task.After the program runs, a
restart.yamlfile will be generated. To continue training, you can useNepTrain train restart.yaml.
Usage:
NepTrain train <config_path>
Options:
<config_path>Path to configuration file, such asjob.yaml.
Output
During training, intermediate data and the resulting potential are
written to the working directory. A restart.yaml file is created after
each iteration, allowing you to resume the workflow with
NepTrain train restart.yaml.
Example
Initialization
Run NepTrain init slurm on the command line to generate job.yaml, which contains all modifiable control parameters of the workflow.
The command produces a job.yaml file showing default execution parameters and their explanations. Set the parameters line by line on the first run, and later reuse the file as a template for future training.
Tip
If you copy a previously modified job.yaml, you can place it in the working directory and run NepTrain init slurm again to synchronize any newly added parameters during version updates.
Start Training
Run the following command on the login node:
NepTrain train job.yaml
To run in the background:
nohup NepTrain train job.yaml &