-
Notifications
You must be signed in to change notification settings - Fork 39
/
train.py
executable file
·306 lines (243 loc) · 11.9 KB
/
train.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
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
import time
import torch.backends.cudnn as cudnn
import torch.optim
import torch.utils.data
import torchvision.transforms as transforms
from torch import nn
from torch.nn.utils.rnn import pack_padded_sequence
from models import DecoderWithAttention
from datasets import *
from utils import *
from nltk.translate.bleu_score import corpus_bleu
# Data parameters
data_folder = 'final_dataset' # folder with data files saved by create_input_files.py
data_name = 'coco_5_cap_per_img_5_min_word_freq' # base name shared by data files
# Model parameters
emb_dim = 1024 # dimension of word embeddings
attention_dim = 1024 # dimension of attention linear layers
decoder_dim = 1024 # dimension of decoder RNN
dropout = 0.5
device = torch.device("cuda" if torch.cuda.is_available() else "cpu") # sets device for model and PyTorch tensors
cudnn.benchmark = True # set to true only if inputs to model are fixed size; otherwise lot of computational overhead
# Training parameters
start_epoch = 0
epochs = 50 # number of epochs to train for (if early stopping is not triggered)
epochs_since_improvement = 0 # keeps track of number of epochs since there's been an improvement in validation BLEU
batch_size = 100
workers = 1 # for data-loading; right now, only 1 works with h5py
best_bleu4 = 0. # BLEU-4 score right now
print_freq = 100 # print training/validation stats every __ batches
checkpoint = None # path to checkpoint, None if none
def main():
"""
Training and validation.
"""
global best_bleu4, epochs_since_improvement, checkpoint, start_epoch,data_name, word_map
# Read word map
word_map_file = os.path.join(data_folder, 'WORDMAP_' + data_name + '.json')
with open(word_map_file, 'r') as j:
word_map = json.load(j)
# Initialize / load checkpoint
if checkpoint is None:
decoder = DecoderWithAttention(attention_dim=attention_dim,
embed_dim=emb_dim,
decoder_dim=decoder_dim,
vocab_size=len(word_map),
dropout=dropout)
decoder_optimizer = torch.optim.Adamax(params=filter(lambda p: p.requires_grad, decoder.parameters()))
else:
checkpoint = torch.load(checkpoint)
start_epoch = checkpoint['epoch'] + 1
epochs_since_improvement = checkpoint['epochs_since_improvement']
best_bleu4 = checkpoint['bleu-4']
decoder = checkpoint['decoder']
decoder_optimizer = checkpoint['decoder_optimizer']
# Move to GPU, if available
decoder = decoder.to(device)
# Loss functions
criterion_ce = nn.CrossEntropyLoss().to(device)
criterion_dis = nn.MultiLabelMarginLoss().to(device)
# Custom dataloaders
train_loader = torch.utils.data.DataLoader(
CaptionDataset(data_folder, data_name, 'TRAIN'),
batch_size=batch_size, shuffle=True, num_workers=workers, pin_memory=True)
val_loader = torch.utils.data.DataLoader(
CaptionDataset(data_folder, data_name, 'VAL'),
batch_size=batch_size, shuffle=True, num_workers=workers, pin_memory=True)
# Epochs
for epoch in range(start_epoch, epochs):
# Decay learning rate if there is no improvement for 8 consecutive epochs, and terminate training after 20
if epochs_since_improvement == 20:
break
if epochs_since_improvement > 0 and epochs_since_improvement % 8 == 0:
adjust_learning_rate(decoder_optimizer, 0.8)
# One epoch's training
train(train_loader=train_loader,
decoder=decoder,
criterion_ce = criterion_ce,
criterion_dis=criterion_dis,
decoder_optimizer=decoder_optimizer,
epoch=epoch)
# One epoch's validation
recent_bleu4 = validate(val_loader=val_loader,
decoder=decoder,
criterion_ce=criterion_ce,
criterion_dis=criterion_dis)
# Check if there was an improvement
is_best = recent_bleu4 > best_bleu4
best_bleu4 = max(recent_bleu4, best_bleu4)
if not is_best:
epochs_since_improvement += 1
print("\nEpochs since last improvement: %d\n" % (epochs_since_improvement,))
else:
epochs_since_improvement = 0
# Save checkpoint
save_checkpoint(data_name, epoch, epochs_since_improvement, decoder,decoder_optimizer, recent_bleu4, is_best)
def train(train_loader, decoder, criterion_ce, criterion_dis, decoder_optimizer, epoch):
"""
Performs one epoch's training.
:param train_loader: DataLoader for training data
:param decoder: decoder model
:param criterion_ce: cross entropy loss layer
:param criterion_dis : discriminative loss layer
:param decoder_optimizer: optimizer to update decoder's weights
:param epoch: epoch number
"""
decoder.train() # train mode (dropout and batchnorm is used)
batch_time = AverageMeter() # forward prop. + back prop. time
data_time = AverageMeter() # data loading time
losses = AverageMeter() # loss (per word decoded)
top5accs = AverageMeter() # top5 accuracy
start = time.time()
# Batches
for i, (imgs, caps, caplens) in enumerate(train_loader):
data_time.update(time.time() - start)
# Move to GPU, if available
imgs = imgs.to(device)
caps = caps.to(device)
caplens = caplens.to(device)
# Forward prop.
scores, scores_d,caps_sorted, decode_lengths, sort_ind = decoder(imgs, caps, caplens)
#Max-pooling across predicted words across time steps for discriminative supervision
scores_d = scores_d.max(1)[0]
# Since we decoded starting with <start>, the targets are all words after <start>, up to <end>
targets = caps_sorted[:, 1:]
targets_d = torch.zeros(scores_d.size(0),scores_d.size(1)).to(device)
targets_d.fill_(-1)
for length in decode_lengths:
targets_d[:,:length-1] = targets[:,:length-1]
# Remove timesteps that we didn't decode at, or are pads
# pack_padded_sequence is an easy trick to do this
scores, _ = pack_padded_sequence(scores, decode_lengths, batch_first=True)
targets, _ = pack_padded_sequence(targets, decode_lengths, batch_first=True)
# Calculate loss
loss_d = criterion_dis(scores_d,targets_d.long())
loss_g = criterion_ce(scores, targets)
loss = loss_g + (10 * loss_d)
# Back prop.
decoder_optimizer.zero_grad()
loss.backward()
# Clip gradients when they are getting too large
torch.nn.utils.clip_grad_norm_(filter(lambda p: p.requires_grad, decoder.parameters()), 0.25)
# Update weights
decoder_optimizer.step()
# Keep track of metrics
top5 = accuracy(scores, targets, 5)
losses.update(loss.item(), sum(decode_lengths))
top5accs.update(top5, sum(decode_lengths))
batch_time.update(time.time() - start)
start = time.time()
# Print status
if i % print_freq == 0:
print('Epoch: [{0}][{1}/{2}]\t'
'Batch Time {batch_time.val:.3f} ({batch_time.avg:.3f})\t'
'Data Load Time {data_time.val:.3f} ({data_time.avg:.3f})\t'
'Loss {loss.val:.4f} ({loss.avg:.4f})\t'
'Top-5 Accuracy {top5.val:.3f} ({top5.avg:.3f})'.format(epoch, i, len(train_loader),
batch_time=batch_time,
data_time=data_time, loss=losses,
top5=top5accs))
def validate(val_loader, decoder, criterion_ce, criterion_dis):
"""
Performs one epoch's validation.
:param val_loader: DataLoader for validation data.
:param decoder: decoder model
:param criterion_ce: cross entropy loss layer
:param criterion_dis : discriminative loss layer
:return: BLEU-4 score
"""
decoder.eval() # eval mode (no dropout or batchnorm)
batch_time = AverageMeter()
losses = AverageMeter()
top5accs = AverageMeter()
start = time.time()
references = list() # references (true captions) for calculating BLEU-4 score
hypotheses = list() # hypotheses (predictions)
# Batches
with torch.no_grad():
for i, (imgs, caps, caplens,allcaps) in enumerate(val_loader):
# Move to device, if available
imgs = imgs.to(device)
caps = caps.to(device)
caplens = caplens.to(device)
scores, scores_d, caps_sorted, decode_lengths, sort_ind = decoder(imgs, caps, caplens)
#Max-pooling across predicted words across time steps for discriminative supervision
scores_d = scores_d.max(1)[0]
# Since we decoded starting with <start>, the targets are all words after <start>, up to <end>
targets = caps_sorted[:, 1:]
targets_d = torch.zeros(scores_d.size(0),scores_d.size(1)).to(device)
targets_d.fill_(-1)
for length in decode_lengths:
targets_d[:,:length-1] = targets[:,:length-1]
# Remove timesteps that we didn't decode at, or are pads
# pack_padded_sequence is an easy trick to do this
scores_copy = scores.clone()
scores, _ = pack_padded_sequence(scores, decode_lengths, batch_first=True)
targets, _ = pack_padded_sequence(targets, decode_lengths, batch_first=True)
# Calculate loss
loss_d = criterion_dis(scores_d,targets_d.long())
loss_g = criterion_ce(scores, targets)
loss = loss_g + (10 * loss_d)
# Keep track of metrics
losses.update(loss.item(), sum(decode_lengths))
top5 = accuracy(scores, targets, 5)
top5accs.update(top5, sum(decode_lengths))
batch_time.update(time.time() - start)
start = time.time()
if i % print_freq == 0:
print('Validation: [{0}/{1}]\t'
'Batch Time {batch_time.val:.3f} ({batch_time.avg:.3f})\t'
'Loss {loss.val:.4f} ({loss.avg:.4f})\t'
'Top-5 Accuracy {top5.val:.3f} ({top5.avg:.3f})\t'.format(i, len(val_loader), batch_time=batch_time,
loss=losses, top5=top5accs))
# Store references (true captions), and hypothesis (prediction) for each image
# If for n images, we have n hypotheses, and references a, b, c... for each image, we need -
# references = [[ref1a, ref1b, ref1c], [ref2a, ref2b], ...], hypotheses = [hyp1, hyp2, ...]
# References
allcaps = allcaps[sort_ind] # because images were sorted in the decoder
for j in range(allcaps.shape[0]):
img_caps = allcaps[j].tolist()
img_captions = list(
map(lambda c: [w for w in c if w not in {word_map['<start>'], word_map['<pad>']}],
img_caps)) # remove <start> and pads
references.append(img_captions)
# Hypotheses
_, preds = torch.max(scores_copy, dim=2)
preds = preds.tolist()
temp_preds = list()
for j, p in enumerate(preds):
temp_preds.append(preds[j][:decode_lengths[j]]) # remove pads
preds = temp_preds
hypotheses.extend(preds)
assert len(references) == len(hypotheses)
# Calculate BLEU-4 scores
bleu4 = corpus_bleu(references, hypotheses)
bleu4 = round(bleu4,4)
print(
'\n * LOSS - {loss.avg:.3f}, TOP-5 ACCURACY - {top5.avg:.3f}, BLEU-4 - {bleu}\n'.format(
loss=losses,
top5=top5accs,
bleu=bleu4))
return bleu4
if __name__ == '__main__':
main()