-
Notifications
You must be signed in to change notification settings - Fork 1
/
alchemy_data.py
163 lines (135 loc) · 5.56 KB
/
alchemy_data.py
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
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
"""
Generate pytorch ready .pt files, and as database parser
Usage:
alchemy_data.py [--task=<str>]
Options:
--task=<str> Task (dev, valid_00, test, ...) to generate .pt files [default: all]
"""
from docopt import docopt
import torch
from torch_geometric.data import InMemoryDataset, Data
import pandas as pd
import numpy as np
from pathlib import Path
import pickle
def atom_dat_reader(atm_dict, target):
# vertex
atom_vertex = np.hstack([
# atm_dict["atm_coord"],
atm_dict["atm_charge"][:, None],
atm_dict["atm_symbol_onehot"],
atm_dict["atm_addcharge"][:, None],
atm_dict["atm_aromatic"][:, None],
atm_dict["atm_hybrid"],
])
# edge index
natm = atom_vertex.shape[0]
natm_rangerep = np.arange(natm)[:, None].repeat(natm, axis=1)
atom_edgeidx = np.array([natm_rangerep.flatten(), natm_rangerep.T.flatten()])
# edge
atom_edge = np.hstack([
atm_dict["atm_dist"].reshape(-1)[:, None],
atm_dict["atm_nuceng_adaj"].reshape(-1)[:, None],
atm_dict["atm_edge_type"].reshape(5, -1).T,
])
# construct atom-network data
atom_data = Data(
x=torch.as_tensor(atom_vertex, dtype=torch.float32),
edge_index=torch.as_tensor(atom_edgeidx, dtype=torch.long),
edge_attr=torch.as_tensor(atom_edge, dtype=torch.float32),
y=torch.as_tensor(target, dtype=torch.float32)
)
return atom_data
def orbital_dat_reader(ao_dict, target):
# vertex index (to its atom)
orbital_vertindex = ao_dict["ao_idx"]
# vertex
orbital_vertex = np.hstack([
# ao_dict["ao_coord"],
ao_dict["ao_atomchg"][:, None],
ao_dict["ao_atomhot"],
ao_dict["ao_zeta"][:, None],
ao_dict["ao_valence"][:, None],
ao_dict["ao_spacial_x"][:, None],
ao_dict["ao_spacial_y"][:, None],
ao_dict["ao_spacial_z"][:, None],
])
# edge index
nao = orbital_vertex.shape[0]
nao_rangerep = np.arange(nao)[:, None].repeat(nao, axis=1)
orbital_edgeidx = np.array([nao_rangerep.flatten(), nao_rangerep.T.flatten()])
# edge
orbital_edge = np.hstack([
ao_dict["ao_dist"].reshape(-1)[:, None],
ao_dict["int1e_ovlp"].reshape(-1)[:, None],
ao_dict["int1e_kin"].reshape(-1)[:, None],
ao_dict["int1e_nuc"].reshape(-1)[:, None],
ao_dict["int1e_r"].reshape(3, -1).T,
ao_dict["rdm1e"].reshape(-1)[:, None],
])
# construct atom-network data
orbital_data = Data(
x=torch.as_tensor(orbital_vertex, dtype=torch.float32),
edge_index=torch.as_tensor(orbital_edgeidx, dtype=torch.long),
edge_attr=torch.as_tensor(orbital_edge, dtype=torch.float32),
y=torch.as_tensor(target, dtype=torch.float32),
atom_idx=torch.as_tensor(orbital_vertindex, dtype=torch.long)
)
return orbital_data
class AlchemyData(InMemoryDataset):
def __init__(self, mode="dev", net_type="atom", root_path=".", train_csv_path=None, transform=None, pre_transform=None):
self.mode = mode
self.net = net_type
self.root_path = Path(root_path)
self.train_csv_path = train_csv_path # type: str or None
self.target = NotImplemented
super(AlchemyData, self).__init__(root_path, transform=transform, pre_transform=pre_transform)
self.data, self.slices = torch.load(self.processed_paths[0])
@property
def raw_file_names(self):
return self.mode + ".dat"
@property
def processed_file_names(self):
return self.net + "-" + self.mode + ".pt"
def _download(self):
pass
def download(self):
pass
def process(self):
process_count = 0
if self.train_csv_path is not None:
self.target = pd.read_csv(self.train_csv_path, index_col=0, usecols=['gdb_idx', ] + ['property_{}'.format(x) for x in range(12)])
self.target = self.target[['property_{}'.format(x) for x in range(12)]]
dat_file = self.raw_paths[0]
with open(dat_file, "rb") as dat:
dat_dict = pickle.load(dat)
data_atom_list = []
data_orbital_list = []
for entry in dat_dict:
# process_count += 1
# if process_count % 10 == 0:
# print("processed " + str(process_count))
target = torch.as_tensor(self.target.loc[entry].tolist(), dtype=torch.float32) \
if self.target is not NotImplemented \
else torch.as_tensor([entry], dtype=torch.float32)
dat_atom, dat_orbital = dat_dict[entry]
data_atom_list.append(atom_dat_reader(dat_atom, torch.as_tensor(target, dtype=torch.float32).unsqueeze(0)))
data_orbital_list.append(orbital_dat_reader(dat_orbital, torch.as_tensor(target, dtype=torch.float32).unsqueeze(0)))
data_atom, slices_atom = self.collate(data_atom_list)
torch.save((data_atom, slices_atom), self.processed_dir + "/atom-" + self.mode + ".pt")
data_orbital, slices_orbital = self.collate(data_orbital_list)
torch.save((data_orbital, slices_orbital), self.processed_dir + "/orbital-" + self.mode + ".pt")
if __name__ == '__main__':
arguments = docopt(__doc__)
print(arguments)
TASK = arguments["--task"]
tasks = [TASK]
if TASK == "all":
tasks = ["dev", "test", "valid_00", "valid_01", "valid_02", "valid_03", "valid_04"]
for t in tasks:
print("processing task " + t + "...")
if t == "test":
train_csv_path = None
else:
train_csv_path = "./raw/train.csv"
AlchemyData(mode=t, net_type="atom", train_csv_path=train_csv_path)