E2GNN/preprocess_md.py at main · Shen-Group/E2GNN · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
# %%
import os
from ase.io.vasp import read_vasp_xml
import multiprocessing as mp
import os
import lmdb
import numpy as np
import torch
from tqdm import tqdm
from pathlib import Path
import pickle
from graph_constructor import AtomsToGraphs
from sklearn.model_selection import train_test_split
import argparse
from lmdb_dataset import TrajectoryLmdbDataset
import warnings
warnings.filterwarnings("ignore")

# %%
def save_metadata(data_root):
    train_set = TrajectoryLmdbDataset({"src": os.path.join(data_root, 'train')})

    energy_list = []
    force_list = []
    for data in train_set:
        energy_list.append(data[0].y)
        force_list.append(data[0].force)

    energy = torch.cat(energy_list).numpy()
    force = torch.stack(force_list).numpy()

    norm_stats = {
        'e_mean': energy.mean(),
        'e_std': energy.std(),
        'f_mean': force.mean(),
        'f_std': force.std(),
    }
    save_path = Path(data_root)
    np.save(save_path / 'metadata', norm_stats)
    path = save_path / 'metadata.npy'
    print("norm_stats: ", np.load(path, allow_pickle=True).item())

def write_data(mp_args):
    a2g, db_path, atoms_list, atoms_indices = mp_args

    db = lmdb.open(
        db_path,
        map_size=1099511627776 * 2,
        subdir=False,
        meminit=False,
        map_async=True,
    )

    for i, atoms_index in enumerate(tqdm(atoms_indices, desc='Reading atoms objects', position=0, leave=True)):
        atoms = atoms_list[atoms_index]
        energy = atoms.get_potential_energy()
        forces = atoms.get_forces()

        data = a2g.convert(atoms)
        data.y = torch.Tensor([energy])
        data.force = torch.Tensor(forces)

        txn = db.begin(write=True)
        txn.put(f"{i}".encode("ascii"), pickle.dumps(data, protocol=-1))
        txn.commit()

    # Save count of objects in lmdb.
    txn = db.begin(write=True)
    txn.put("length".encode("ascii"), pickle.dumps(i+1, protocol=-1))
    txn.commit()

    db.sync()
    db.close()

# get_metadata()
# %%
if __name__ == '__main__':
    parser = argparse.ArgumentParser()

    # Add argument
    parser.add_argument('--data_root', type=str, default=None, help='data directory', required=True)
    parser.add_argument('--num_workers', type=int, default=1, help='number of workers')

    args = parser.parse_args()
    data_root = args.data_root
    num_workers = args.num_workers

    data_path = os.path.join(data_root, 'vasprun.xml')
    configs = read_vasp_xml(data_path, index=slice(None)) # Read xml

    atoms_list = []
    for i, atoms in enumerate(configs):
        atoms_list.append(atoms)

    data_indices = np.array(list(range(len(atoms_list))))
    train_indices, test_indices = train_test_split(data_indices, test_size=0.2, train_size=0.8, random_state=123, shuffle=True)
    train_indices, val_indices = train_test_split(train_indices, test_size=0.1, train_size=0.9, random_state=123, shuffle=True)

    print("train: ", len(train_indices))
    print("val: ", len(val_indices))
    print("test: ", len(test_indices))

    a2g = AtomsToGraphs(
        max_neigh=50,
        radius=5,
        r_energy=False,
        r_forces=False,
        r_distances=False,
        r_edges=False,
    )

    for dataset in ['train', 'val', 'test']:
        db_path = os.path.join(data_root, dataset)
        save_path = Path(db_path)
        save_path.mkdir(parents=True, exist_ok=True)
        mp_db_paths = [
            os.path.join(save_path, "data.%04d.lmdb" % i)
            for i in range(num_workers)
        ]
        if dataset == 'train':
            mp_data_indices = np.array_split(train_indices, num_workers)
        elif dataset == 'val':
            mp_data_indices = np.array_split(val_indices, num_workers)
        elif dataset == 'test':
            mp_data_indices = np.array_split(test_indices, num_workers)

        pool = mp.Pool(num_workers)
        mp_args = [
            (
                a2g,
                mp_db_paths[i],
                atoms_list,
                mp_data_indices[i]
            )
            for i in range(num_workers)
        ]

        pool.imap(write_data, mp_args)
        pool.close()
        pool.join()

    save_metadata(data_root)

# %%