sample

Sample programs

Outline

You have the following directory structure.

- alkane
  - train.py
  - evaluate.py
  - visualize.ipynb
- QM5
- acess_heptane
- acess_octane

Run sample code

Training & Evaluation

Ase database is located at ex_matgl/DBs. First, you make the symbolic link for the database

cd alkane
ln -s ../../DBs/alkane.db .
CUDA_VISIBLE_DEVICES=0 python train.py
CUDA_VISIBLE_DEVICES=0 python evaluate.py

After you perform training, you get the trained model saved_models/best_model.ckpt. After evaluation, you can get the npz file of evalueate_answer.npz and evaluate_prediction.npz. Jupyter notebook visualize.ipynb show you the results.

Test for Heptane and Octane

After leaning neural network potential of alkane following the above instructions, run the below codes.

cd acess_heptane
ln -s ../../DBs/heptane.db .
CUDA_VISIBLE_DEVICES=0 python evaluate.py

You can get the npz file of heptane_evalueate_answer.npz and heptane_evaluate_prediction.npz. Jupyter notebook acess_heptane.ipynb show you the results.

Structual Optimization of heptane about S1

After leaning neural network potential of alkane following the above instructions, run the below codes.

cd relax_heptane
ln -s ../../DBs/heptane.db .
python relax.py

You can find optimize_sturctures/bfgs_heptane.traj and optimized structure of heptane about S1 state in it.