-
Notifications
You must be signed in to change notification settings - Fork 30
/
main.py
185 lines (164 loc) · 8.1 KB
/
main.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
175
176
177
178
179
180
181
182
183
184
185
import argparse
import os
import time
from datetime import datetime
import logging
import torch.nn as nn
import torch.backends.cudnn as cudnn
from torch.autograd import Variable
import math
from torch.nn.utils import clip_grad_norm
from torch.nn.utils.rnn import pack_padded_sequence
from data import build_vocab, get_coco_data, get_iterator
from utils import setup_logging, adjust_optimizer, AverageMeter, select_optimizer
from model import CaptionModel
from torchvision.models import resnet
model_names = sorted(name for name in resnet.__dict__
if name.islower() and not name.startswith("__")
and callable(resnet.__dict__[name]))
parser = argparse.ArgumentParser(description='COCO caption genration training')
parser.add_argument('--results_dir', metavar='RESULTS_DIR', default='./results',
help='results dir')
parser.add_argument('--save', metavar='SAVE', default='',
help='saved folder')
parser.add_argument('--cnn', '-a', metavar='CNN', default='resnet50',
choices=model_names,
help='cnn feature extraction architecture: ' +
' | '.join(model_names) +
' (default: resnet50)')
parser.add_argument('--embedding_size', default=256, type=int,
help='size of word embedding used')
parser.add_argument('--rnn_size', default=256, type=int,
help='size of rnn hidden layer')
parser.add_argument('--num_layers', default=2, type=int,
help='number of rnn layers to use')
parser.add_argument('--max_length', default=30, type=int,
help='maximum time length to feed')
parser.add_argument('--type', default='torch.cuda.FloatTensor',
help='type of tensor - e.g torch.cuda.HalfTensor')
parser.add_argument('-j', '--workers', default=8, type=int, metavar='N',
help='number of data loading workers (default: 8)')
parser.add_argument('--epochs', default=10, type=int, metavar='N',
help='number of total epochs to run')
parser.add_argument('--finetune_epoch', default=3, type=int,
help='epoch to start cnn finetune')
parser.add_argument('--start_epoch', default=0, type=int, metavar='N',
help='manual epoch number (useful on restarts)')
parser.add_argument('-b', '--batch-size', default=32, type=int,
metavar='N', help='mini-batch size (default: 32)')
parser.add_argument('-eb', '--eval_batch_size', default=128, type=int,
metavar='N', help='mini-batch size (default: 128)')
parser.add_argument('--optimizer', default='SGD', type=str, metavar='OPT',
help='optimizer function used')
parser.add_argument('--grad_clip', default=5., type=float,
help='gradient max norm')
parser.add_argument('--lr', '--learning_rate', default=0.1, type=float,
metavar='LR', help='initial learning rate')
parser.add_argument('--lr_decay', '--learning_rate_decay', default=0.8, type=float,
metavar='LR', help='learning rate decay')
parser.add_argument('--momentum', default=0.9, type=float, metavar='M',
help='momentum')
parser.add_argument('--weight-decay', '--wd', default=1e-4, type=float,
metavar='W', help='weight decay (default: 1e-4)')
parser.add_argument('--share_weights', default=False, type=bool,
help='share embedder and classifier weights')
parser.add_argument('--print_freq', '-p', default=10, type=int,
metavar='N', help='print frequency (default: 10)')
parser.add_argument('-e', '--evaluate', type=str, metavar='FILE',
help='evaluate model FILE on validation set')
def main():
global args
args = parser.parse_args()
if args.save is '':
args.save = datetime.now().strftime('%Y-%m-%d_%H-%M-%S')
save_path = os.path.join(args.results_dir, args.save)
if not os.path.exists(save_path):
os.makedirs(save_path)
setup_logging(os.path.join(save_path, 'log.txt'))
checkpoint_file = os.path.join(save_path, 'checkpoint_epoch_%s.pth.tar')
logging.debug("run arguments: %s", args)
logging.info("using pretrained cnn %s", args.cnn)
cnn = resnet.__dict__[args.cnn](pretrained=True)
vocab = build_vocab()
model = CaptionModel(cnn, vocab,
embedding_size=args.embedding_size,
rnn_size=args.rnn_size,
num_layers=args.num_layers,
share_embedding_weights=args.share_weights)
train_data = get_iterator(get_coco_data(vocab, train=True),
batch_size=args.batch_size,
max_length=args.max_length,
shuffle=True,
num_workers=args.workers)
val_data = get_iterator(get_coco_data(vocab, train=False),
batch_size=args.eval_batch_size,
max_length=args.max_length,
shuffle=False,
num_workers=args.workers)
if 'cuda' in args.type:
cudnn.benchmark = True
model.cuda()
optimizer = select_optimizer(
args.optimizer, params=model.parameters(), lr=args.lr)
regime = lambda e: {'lr': args.lr * (args.lr_decay ** e),
'momentum': args.momentum,
'weight_decay': args.weight_decay}
model.finetune_cnn(False)
def forward(model, data, training=True, optimizer=None):
use_cuda = 'cuda' in args.type
loss = nn.CrossEntropyLoss()
perplexity = AverageMeter()
batch_time = AverageMeter()
data_time = AverageMeter()
if training:
model.train()
else:
model.eval()
end = time.time()
for i, (imgs, (captions, lengths)) in enumerate(data):
data_time.update(time.time() - end)
if use_cuda:
imgs = imgs.cuda()
captions = captions.cuda(async=True)
imgs = Variable(imgs, volatile=not training)
captions = Variable(captions, volatile=not training)
input_captions = captions[:-1]
target_captions = pack_padded_sequence(captions, lengths)[0]
pred, _ = model(imgs, input_captions, lengths)
err = loss(pred, target_captions)
perplexity.update(math.exp(err.data[0]))
if training:
optimizer.zero_grad()
err.backward()
clip_grad_norm(model.rnn.parameters(), args.grad_clip)
optimizer.step()
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
if i % args.print_freq == 0:
logging.info('{phase} - Epoch: [{0}][{1}/{2}]\t'
'Time {batch_time.val:.3f} ({batch_time.avg:.3f})\t'
'Data {data_time.val:.3f} ({data_time.avg:.3f})\t'
'Perplexity {perp.val:.4f} ({perp.avg:.4f})'.format(
epoch, i, len(data),
phase='TRAINING' if training else 'EVALUATING',
batch_time=batch_time,
data_time=data_time, perp=perplexity))
return perplexity.avg
for epoch in range(args.start_epoch, args.epochs):
if epoch >= args.finetune_epoch:
model.finetune_cnn(True)
optimizer = adjust_optimizer(
optimizer, epoch, regime)
# Train
train_perp = forward(
model, train_data, training=True, optimizer=optimizer)
# Evaluate
val_perp = forward(model, val_data, training=False)
logging.info('\n Epoch: {0}\t'
'Training Perplexity {train_perp:.4f} \t'
'Validation Perplexity {val_perp:.4f} \n'
.format(epoch + 1, train_perp=train_perp, val_perp=val_perp))
model.save_checkpoint(checkpoint_file % (epoch + 1))
if __name__ == '__main__':
main()