Skip to content

Latest commit

 

History

History
188 lines (140 loc) · 6.3 KB

README.md

File metadata and controls

188 lines (140 loc) · 6.3 KB
Raw

VASP Calculation Workflow Manager

This project provides a script to manage VASP calculations for optimizing atomic structures (OPT), running ab initio molecular dynamics (AIMD), and processing results into datasets for machine learning purposes. The script supports submitting VASP jobs, checking SCF convergence, running AIMD simulations, and converting data for DeePMD or Graph Neural Network (GNN) training. Additionally, it includes functionality for visualizing the occurrence of elements in the dataset.

Table of Contents

  1. Prerequisites
  2. File Structure
  3. Usage
    1. Command-Line Arguments
    2. Example Command
    3. Input File
  4. Workflow Description
  5. Detailed Instructions
    1. Prepare the Environment
    2. Set Up Input Parameters
    3. Run the Script
    4. Check Job Status
    5. Process Data
    6. Plot Element Weights
  6. Notes
  7. License
  8. Acknowledgements

Prerequisites

  • Python 3 (version > 3.10)

  • ASE (version > 3.22)

  • DeePMD-kit (version > 3.0.0a1)

  • dpdata (version > 0.2.18)

  • VASP and VASPKIT

  • SLURM (for job scheduling)

  • tqdm (version > 4.65.0，for progress bars)

    if output lmdb dataset need:

  • torch (version > 2.3.1)

  • torch_geometric

  • torch_scatter

  • lmdb (version > 1.5.1)

  • fairchem

File Structure

├── flow.py # Main script for managing VASP calculations
├── input # File containing input parameters
├── utils
│ ├── plot_weight.py # Script for plotting element weights
│ ├── INCAR_opt # INCAR file for structural optimization
│ ├── INCAR_md # INCAR file for AIMD calculations
│ └── sub.vasp # VASP submission script
└── structure_db # Directory contains POSCAR files

Usage

Command-Line Arguments

  • prefix: Prefix for files to process. Usually POSCAR
  • operation: Specify the operation to perform. Choices are:
    • opt: Submit VASP optimization jobs
    • md: Submit VASP AIMD jobs
    • optcheck: Check SCF convergence for OPT jobs
    • mdcheck: Check SCF convergence for AIMD jobs
    • dpdata: Process data into npy or lmdb format for DeePMD or GNN training
    • plot: Plot element weights in the dataset

Example Command

python flow.py PREFIX opt

Input File

The flow.py script reads parameters from an input file in the same directory. The file should contain the following parameters:

work_path: /path/to/work/
opt_INCAR_path: /path/to/work/utils/INCAR_opt
md_INCAR_path: /path/to/work/utils/INCAR_md
vasp_sub_path: /path/to/work/utils/sub.vasp
structre_db_path: /path/to/work/structure_db/
max_jobs: 25 # Max jobs in slurm queue
sleep_time: 20 # The interval time for jobs checking in slurm queue 
user_name: Your_username
step_data: 10 # Interval of data extraction by dpdata
test_size: 0.1 # Proportion of test dataset
dataset_prefix: /path/to/work/data
dataset_fmt: npy_single or npy_mix or lmdb
plot_name: element_weights.png

Workflow Description

  1. VASP OPT Batch Submission :
    • Submits VASP optimization jobs in batch.
    • The initial optimization doesn't require full structural convergence, only SCF convergence.
  2. Check SCF Convergence for OPT :
    • Checks SCF convergence for VASP optimization jobs.
  3. Run AIMD Simulations :
    • Runs ab initio molecular dynamics simulations based on SCF-converged OPT results to generate data for collection
  4. Check SCF Convergence for AIMD :
    • Checks SCF convergence for AIMD simulations and logs the number of SCF-converged steps.
  5. Process Data for DeepMD Training :
    • Converts VASP results into npy or lmdb format for training neural network potentials. Note that only AIMD data are collected by an interval
  6. Plot Element Weights :
    • Plots the occurrence of elements in the dataset as a colored periodic table.

Detailed Instructions

1. Prepare the Environment

Ensure all prerequisites are installed and properly configured.

2. Set Up Input Parameters

Create an input file in the project directory with the required parameters.

3. Run the Script

Use the command-line interface to perform various operations. For example, to submit optimization jobs, use:

python flow.py POSCAR opt

'POSCAR' is the PREFIX of the structure name in structure_db.

If the number of your POSCAR files in structure_db is large, usually you need to run this script by nohup

nohup python flow.py POSCAR opt &

After opt and optcheck, the AIMD jobs will be submited

nohup python flow.py POSCAR md &

4. Check Job Status

Use optcheck ormdcheck operations to verify SCF convergence for OPT and MD jobs, respectively.

python flow.py POSCAR optcheck
python flow.py POSCAR mdcheck

5. Process Data

After completing the VASP simulations, convert the results into a suitable format for following training:

python flow.py POSCAR dpdata

change the dataset_prefix: /path/to/work/data and dataset_fmt: npy_single or npy_mix or lmdb parameters to control which format to output.

6. Plot Element Weights

Visualize the occurrence of elements in your dataset:

python flow.py POSCAR plot

example of the plot element weight

Notes

  • Ensure the VASP, VASPKIT and DeePMD-kit executables are accessible in your environment.
  • Customize the INCAR_OPT, INCAR_MD, KPOINTS, POTCAR, and submission scripts (sub.vasp) as needed for your specific system and requirements.

License

This project is licensed under the GNU General Public License v3.0. See the LICENSE file for details.

Acknowledgements