-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathgen.py
174 lines (142 loc) · 6.38 KB
/
gen.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
164
165
166
167
168
169
170
171
172
173
174
import argparse
import time
import math
import torch
from torch import nn
import torch.optim as optim
from torch.optim.lr_scheduler import ReduceLROnPlateau
import torch.nn.functional as F
from torch.utils.data import DataLoader
from model import *
from rdkit import Chem
def evaluate(data_iter):
model.eval()
total_loss = 0
with torch.no_grad():
for data, label in data_iter:
targets = data[:, 1:]
inputs = data[:, :-1]
if torch.cuda.is_available()==True:
targets=targets.cuda()
inputs=inputs.cuda()
outputs = model(inputs)
final_output = outputs.contiguous().view(-1, n_words)
final_target = targets.contiguous().view(-1)
loss = criterion(final_output, final_target)
total_loss += loss.item()
return total_loss / len(data_iter)
def sample(idx2word, set_smi, num_samples):
model.eval()
n_words = len(idx2word)
set_mols = [Chem.MolToInchiKey(Chem.MolFromSmiles(smi)) for smi in set_smi]
n = 0
new_mols = []
new_smiles = []
lss = 0
for i in range(num_samples):
input = torch.ones(1, 1, dtype=torch.long)
if torch.cuda.is_available()==True:
input=input.cuda()
word = '&'
while word[-1] != '\n':
output = model(input)
final_output = output.contiguous().view(-1, n_words)
word_id = torch.multinomial(F.softmax(final_output[-1, :], dim=-1), num_samples=1).unsqueeze(0)
input = torch.cat((input, word_id), dim=1)
word += idx2word[word_id.item()]
if bool(Chem.MolFromSmiles(word[1:])):
n += 1
mol = Chem.MolToInchiKey(Chem.MolFromSmiles(word[1:]))
if mol not in set_mols and mol not in new_mols:
new_mols += [mol]
new_smiles += [word[1:]]
if i != 0 and i % 10000 == 0:
print(len(new_smiles) - lss)
lss = len(new_smiles)
print(n / num_samples)
return new_smiles
if __name__ == "__main__":
parser = argparse.ArgumentParser(description='Generative Modeling')
parser.add_argument('--batch_size', type=int, default=32,
metavar='N', help='batch size (default: 32)')
parser.add_argument('--dropout', type=float, default=0.2,
help='dropout applied to layers (default: 0.2)')
parser.add_argument('--emb_dropout', type=float, default=0.1,
help='dropout applied to the embedded layer (default: 0.1)')
parser.add_argument('--epochs', type=int, default=200,
help='upper epoch limit (default: 200)')
parser.add_argument('--ksize', type=int, default=3,
help='kernel size (default: 3)')
parser.add_argument('--emsize', type=int, default=32,
help='size of word embeddings (default: 32)')
parser.add_argument('--levels', type=int, default=4,
help='levels (default: 4)')
parser.add_argument('--lr', type=float, default=0.001,
help='initial learning rate (default: 0.001)')
parser.add_argument('--nhid', type=int, default=256,
help='number of hidden units per layer (default: 256)')
parser.add_argument('--optim', type=str, default='Adam',
help='optimizer type (default: Adam)')
parser.add_argument('--save_name', type=str, default='gen.pt',
help='the name of save model')
args = parser.parse_args()
print(args)
torch.manual_seed(1024)
word2idx, idx2word = torch.load("data/opv_dic.pt")
train_data, val_data, test_data = torch.load("data/opv_data.pt")
train_iter = DataLoader(train_data, args.batch_size, shuffle=True)
val_iter = DataLoader(val_data, args.batch_size, shuffle=False)
test_iter = DataLoader(test_data, args.batch_size, shuffle=False)
n_words = len(word2idx)
model = GEN(args.emsize, n_words, hid_size=args.nhid, n_levels=args.levels,
kernel_size=args.ksize, emb_dropout=args.emb_dropout, dropout=args.dropout )
if torch.cuda.is_available()==True:
model.cuda()
criterion = nn.CrossEntropyLoss()
optimizer = getattr(optim, args.optim)(model.parameters(), lr=args.lr)
scheduler = ReduceLROnPlateau(optimizer, 'min')
best_vloss = 100
try:
for epoch in range(1, args.epochs + 1):
start_time = time.time()
model.train()
total_loss = 0
for data, label in train_iter:
targets = data[:, 1:]
inputs = data[:, :-1]
if torch.cuda.is_available()==True:
targets=targets.cuda()
inputs=inputs.cuda()
optimizer.zero_grad()
outputs = model(inputs)
final_output = outputs.contiguous().view(-1, n_words)
final_target = targets.contiguous().view(-1)
loss = criterion(final_output, final_target)
loss.backward()
optimizer.step()
total_loss += loss.item()
print('| epoch: {:3d} | train loss: {:5.6f} '.format
(epoch, total_loss / len(train_iter)))
val_loss = evaluate(val_iter)
scheduler.step(val_loss)
print('-' * 89)
print('| time: {:5.4f}s | valid loss: {:5.6f} | valid ppl: {:8.4f}'.format
((time.time() - start_time), val_loss, math.exp(val_loss)))
print('-' * 89)
if val_loss < best_vloss:
print('Save model!\n')
torch.save(model.state_dict(), "results/saved_models/" + str(args.levels) + args.save_name)
best_vloss = val_loss
except KeyboardInterrupt:
print('-' * 89)
print('Exiting from training early')
model.load_state_dict(torch.load("results/saved_models/" + str(args.levels) + args.save_name), strict=True)
test_loss = evaluate(test_iter)
print('=' * 89)
print('| End of training | test loss {:5.4f} | test ppl {:8.4f}'.format(test_loss, math.exp(test_loss)))
print('=' * 89)
with open('data/smi_c.txt', 'r') as smi:
set_smiles = smi.readlines()
new_smiles = sample(idx2word, set_smiles, num_samples=100000)
with open("results/" + str(args.levels) + 'sample.txt', 'w') as f:
f.writelines(new_smiles)