-
Notifications
You must be signed in to change notification settings - Fork 1
Expand file tree
/
Copy pathtrain_lstm.py
More file actions
164 lines (143 loc) · 6.68 KB
/
Copy pathtrain_lstm.py
File metadata and controls
164 lines (143 loc) · 6.68 KB
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
import argparse
import json
import os
import torch
from loguru import logger
from torch import nn
from torch.nn.utils.rnn import pack_padded_sequence
import trainer
from datasets import ClassifierDataset, LSTMDataset
from models import LanguageModel
def loss_fn_lstm(preds, label):
label_seq, label_seq_length = label
label_seq_length += 1
# remove <start>
label_seq = label_seq[:, 1:]
label_seq_length = label_seq_length.cpu().to(torch.int64)
label_packed = pack_padded_sequence(label_seq,
label_seq_length,
batch_first=True,
enforce_sorted=False)
return nn.CrossEntropyLoss()(preds, label_packed.data)
if __name__ == "__main__":
parser = argparse.ArgumentParser(description='Train language models.')
parser.add_argument('direction', type=str)
parser.add_argument('--model_dir', type=str, default='models')
parser.add_argument('--model_name', type=str, default='')
parser.add_argument('--data_dir', type=str, default='data')
parser.add_argument('--lstm_data_dir', type=str, default='lstm')
parser.add_argument('--classifier_data_dir',
type=str,
default='classifier')
parser.add_argument('--feature_filename',
type=str,
default='image_features.h5')
parser.add_argument('--featuremap_filename',
type=str,
default='feature_map.json')
parser.add_argument('--label_filename', type=str, default='')
parser.add_argument('--word_map_filename',
type=str,
default='word_map.json')
parser.add_argument('--label_caption_filename',
type=str,
default='label_caption.json')
parser.add_argument('--att_dim', type=int, default=1024)
parser.add_argument('--embed_dim', type=int, default=1024)
parser.add_argument('--img_embed_dim', type=int, default=1024)
parser.add_argument('--hidden_dim', type=int, default=1024)
parser.add_argument('--batch_size', type=int, default=1024)
parser.add_argument('--epoch', type=int, default=80)
parser.add_argument('--decay_every', type=int, default=20)
parser.add_argument('--decay_rate', type=float, default=0.1)
parser.add_argument('--lr', type=float, default=1e-4)
parser.add_argument('--num_workers', type=int, default=12)
parser.add_argument('--load_param', type=str, default='')
parser.add_argument('--no-logfile', dest='logfile', action='store_false')
parser.set_defaults(log_file=True)
args = parser.parse_args()
if args.direction == 'left':
model_name = 'lstm-left' if not args.model_name else args.model_name
label_filename = 'label_left.json'
elif args.direction == 'right':
model_name = 'lstm-right' if not args.model_name else args.model_name
label_filename = 'label_right.json'
else:
logger.error('direction has to be either left or right!')
exit(1)
if args.label_filename:
label_filename = args.label_filename
if not os.path.exists(args.model_dir):
os.mkdir(args.model_dir)
if args.log_file:
logger.add(os.path.join(
args.model_dir,
'{0}_log.txt',
).format(model_name),
mode='w')
model_path = os.path.join(args.model_dir, model_name)
feature_path = os.path.join(args.data_dir, args.feature_filename)
featuremap_path = os.path.join(args.data_dir, args.featuremap_filename)
label_path = os.path.join(args.data_dir, args.lstm_data_dir,
label_filename)
word_map_path = os.path.join(args.data_dir, args.lstm_data_dir,
args.word_map_filename)
cap_detection_label_path = os.path.join(args.data_dir,
args.classifier_data_dir,
args.label_caption_filename)
if args.load_param:
with open(args.load_opt, 'r') as fp:
params = json.load(fp)
else:
params = {
'batch_size': args.batch_size,
'epoch': args.epoch,
'decay_every': args.decay_every,
'decay_rate': args.decay_rate,
'lr': args.lr,
}
with open(
os.path.join(args.model_dir, '{0}_param.json'.format(model_name)),
'w') as fp:
json.dump(params, fp)
with open(os.path.join(args.model_dir, '{0}_args.json'.format(model_name)),
'w') as fp:
json.dump(vars(args), fp)
with open(word_map_path, 'r') as fp:
word_map = json.load(fp)
reversed_word_map = [(word_map[k], k) for k in word_map]
reversed_word_map = sorted(reversed_word_map)
reversed_word_map = [x[1] for x in reversed_word_map]
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
logger.info("Training {0}...".format(model_name))
train_dataset = LSTMDataset(feature_path, label_path, featuremap_path,
'train')
val_dataset = LSTMDataset(feature_path, label_path, featuremap_path, 'val')
val_cap_detection = ClassifierDataset(feature_path,
cap_detection_label_path,
featuremap_path, 'val')
model = LanguageModel(len(word_map), args.embed_dim, args.hidden_dim,
(36, 2048), args.img_embed_dim, args.att_dim, device)
if args.direction == 'left':
trainer.train_val_loss(model=model,
train_dataset=train_dataset,
val_dataset=val_dataset,
num_workers=args.num_workers,
params=params,
loss_fn=loss_fn_lstm,
model_save_path=model_path,
save_every=5,
device=device)
else:
trainer.train_val_meteor(model=model,
train_dataset=train_dataset,
val_dataset=val_dataset,
val_cap_dataset=val_cap_detection,
word_map=word_map,
reversed_word_map=reversed_word_map,
num_workers=args.num_workers,
params=params,
loss_fn=loss_fn_lstm,
model_save_path=model_path,
save_every=5,
device=device)