RevGraphVampNet

This repo contains the code for RevGraphVAMPNet

REVGraphVAMPNet code

Usage

training

python train.py --pre-train-epoch 300 --epochs 1000  --batch-size 500 --lr 0.0005 
--hidden 16  --num-atoms 42 --num-classes 4 --num_neighbors 10 --conv_type SchNet 
--dmin 0 --dmax 8. --step 0.5 --dist-data ../intermediate/red_5nbrs_1ns_dist_min.npy 
--nbr-data ../intermediate/red_5nbrs_1ns_inds_min.npy  
--data-info ../intermediate/red_5nbrs_1ns_datainfo_min.npy --residual  
--train  --score-method VAMPCE --save-folder ab_sch4_1 

testing

python train.py --pre-train-epoch 300 --epochs 1000  --batch-size 500 --lr 0.0005 
--hidden 16  --num-atoms 42 --num-classes 4 --num_neighbors 10 --conv_type SchNet 
--dmin 0 --dmax 8. --step 0.5 --dist-data ../intermediate/red_5nbrs_1ns_dist_min.npy 
--nbr-data ../intermediate/red_5nbrs_1ns_inds_min.npy  
--data-info ../intermediate/red_5nbrs_1ns_datainfo_min.npy 
--residual  --score-method VAMPCE --return-emb --return-attn --score-method VAMPCE
 --save-folder abred_all_1 --trained-model ab_sch4_1/best_allnet.pt

Requirements

• pytorch
• deeptime
• torch_scatter

Sources:

• VAMPNet code is based on deeptime package deeptime
• GraphNet code is based on the GraphVampNet
• SchNet code is based on the cgnet
• physical constraint code is based on the RevNet

Cite

If you use this code please cite the following paper:

Ying Huang ...