This project is to reduce the number of samples required for training a large atomic model (LAM) developed based on the cloud-native workflow framework dflow. So far, DeePMD, OCP and Nequip are supported.
Install this project by
pip install .
Configure workflow server required by dflow before submit a workflow by command line
dpclean submit input.json
Some example input files can be found in examples.