-
Notifications
You must be signed in to change notification settings - Fork 4
/
train.py
283 lines (221 loc) · 12.6 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
import os
import sys
import time
import json
import shutil
import torch
import data
from model import ReLearning
from evaluation import AverageMeter, LogCollector, encode_data, do_predict
import argparse
import logging
import tensorboard_logger as tb_logger
from simpleknn.bigfile import BigFile
from utils.generic_utils import Progbar
from utils.util import read_video_set, write_csv, read_dict
from utils.common import ROOT_PATH, checkToSkip, makedirsforfile
from utils.cbvrp_eval import read_csv_to_dict, hit_k_own, recall_k_own
def main():
# Hyper Parameters
parser = argparse.ArgumentParser()
parser.add_argument("--rootpath", default=ROOT_PATH, type=str, help="rootpath (default: %s)" % ROOT_PATH)
parser.add_argument("--overwrite", default=0, type=int, help="overwrite existing file (default: 0)")
parser.add_argument('--collection', default='track_1_shows', type=str, help='collection')
parser.add_argument('--feature', default='inception-pool3', type=str, help="video feature.")
parser.add_argument('--embed_size', default=1024, type=int, help='Dimensionality of the video embedding.')
parser.add_argument('--loss', default='trl', type=str, help='loss function. (trl|netrl)')
parser.add_argument('--alpha', default=1.0, type=float, help='loss weight for irrelevant loss')
parser.add_argument("--cost_style", default='sum', type=str, help="cost_style (sum|mean)")
parser.add_argument('--max_violation', action='store_true', help='Use max instead of sum in the rank loss.')
parser.add_argument('--margin', default=0.2, type=float, help='Rank loss margin.')
parser.add_argument('--margin_irel', default=0.05, type=float, help='Irrelevant loss margin.')
parser.add_argument('--grad_clip', default=2., type=float, help='Gradient clipping threshold.')
parser.add_argument('--optimizer', default='adam', type=str, help='optimizer. (adam|rmsprop)')
parser.add_argument('--learning_rate', default=.001, type=float, help='Initial learning rate.')
parser.add_argument('--lr_decay', default=0.99, type=float, help='learning rate decay after each epoch')
parser.add_argument('--num_epochs', default=50, type=int, help='Number of training epochs.')
parser.add_argument('--batch_size', default=32, type=int, help='Size of a training mini-batch.')
parser.add_argument('--workers', default=2, type=int, help='Number of data loader workers.')
parser.add_argument('--log_step', default=100, type=int, help='Number of steps to print and record the log.')
parser.add_argument('--measure', default='cosine', help='Similarity measure used (cosine|order)')
parser.add_argument('--no_imgnorm', action='store_true', help='Do not normalize the image embeddings.')
parser.add_argument('--postfix', default='run_0', type=str, help='')
# augmentation for frame-level features
parser.add_argument('--stride', default='1', type=str, help='stride=1 means no frame-level data augmentation (default: 1)')
# augmentation for video-level features
parser.add_argument('--aug_prob', default=0.0, type=float,
help='aug_prob=0 means no frame-level data augmentation, aug_prob=0.5 means half of video use augmented features(default: 0.0)')
parser.add_argument('--perturb_intensity', default=1.0, type=float, help='perturbation intensity, epsilon in Eq.2 (default: 1.0)')
parser.add_argument('--perturb_prob', default=0.5, type=float, help='perturbation probability, p in Eq.2 (default: 0.5)')
opt = parser.parse_args()
print json.dumps(vars(opt), indent = 2)
visual_info = 'feature_%s_embed_size_%d_no_imgnorm_%s' % (opt.feature, opt.embed_size, opt.no_imgnorm)
loss_info = '%s_%s_margin_%.1f_max_violation_%s_%s' % (opt.loss, opt.measure, opt.margin, opt.max_violation, opt.cost_style)
if opt.loss == 'netrl':
loss_info += '_alpha_%.1f_margin_irel_%.2f' % (opt.alpha, opt.margin_irel)
optimizer_info = '%s_lr_%.5f_%.2f_bs_%d' % ( opt.optimizer, opt.learning_rate, opt.lr_decay, opt.batch_size)
data_argumentation_info = 'frame_stride_%s_video_prob_%.1f_perturb_intensity_%.5f_perturb_prob_%.2f' % (opt.stride, opt.aug_prob, opt.perturb_intensity, opt.perturb_prob)
opt.logger_name = os.path.join(opt.rootpath, opt.collection, 'cv', 'ReLearning', visual_info, loss_info, optimizer_info, data_argumentation_info, opt.postfix)
if checkToSkip(os.path.join(opt.logger_name,'model_best.pth.tar'), opt.overwrite):
sys.exit(0)
if checkToSkip(os.path.join(opt.logger_name,'val_perf.txt'), opt.overwrite):
sys.exit(0)
makedirsforfile(os.path.join(opt.logger_name,'model_best.pth.tar'))
logging.basicConfig(format='%(asctime)s %(message)s', level=logging.INFO)
tb_logger.configure(opt.logger_name, flush_secs=5)
# reading data
train_video_set_file = os.path.join(opt.rootpath, opt.collection, 'split', 'train.csv' )
val_video_set_file = os.path.join(opt.rootpath, opt.collection, 'split', 'val.csv' )
train_video_list = read_video_set(train_video_set_file)
val_video_list = read_video_set(val_video_set_file)
train_rootpath = os.path.join(opt.rootpath, opt.collection, 'relevance_train.csv')
val_rootpath = os.path.join(opt.rootpath, opt.collection, 'relevance_val.csv')
val_video2gtrank = read_csv_to_dict(val_rootpath)
stride_list = map(int, opt.stride.strip().split('-'))
opt.sum_subs = sum(stride_list)
if opt.aug_prob <= 0:
opt.feature = "avg-" + opt.feature + "-stride%s" % opt.stride
video_feat_path = os.path.join(opt.rootpath, opt.collection, 'FeatureData', opt.feature)
video_feats = BigFile(video_feat_path)
opt.feature_dim = video_feats.ndims
# Load data loaders
if opt.sum_subs > 1:
video2subvideo_path = os.path.join(video_feat_path, 'video2subvideo.txt')
video2subvideo = read_dict(video2subvideo_path)
train_loader = data.get_video_da_loader(train_rootpath, video_feats, opt, opt.batch_size, True, opt.workers,
video2subvideo, opt.sum_subs, feat_path=video_feat_path)
else:
train_loader = data.get_video_da_loader(train_rootpath, video_feats, opt, opt.batch_size, True, opt.workers, feat_path=video_feat_path)
val_feat_loader = data.get_feat_loader(val_video_list, video_feats, opt.batch_size, False, 1)
cand_feat_loader = data.get_feat_loader(train_video_list + val_video_list, video_feats, opt.batch_size, False, 1)
# Construct the model
model = ReLearning(opt)
# Train the Model
best_rsum = 0
best_hit_k_scores = 0
best_recall_K_scoress = 0
no_impr_counter = 0
lr_counter = 0
fout_val_perf_hist = open(os.path.join(opt.logger_name,'val_perf_hist.txt'), 'w')
for epoch in range(opt.num_epochs):
# train for one epoch
print "\nEpoch: ", epoch + 1
print "learning rate: ", get_learning_rate(model.optimizer)
train(opt, train_loader, model, epoch)
# evaluate on validation set
rsum, hit_k_scores, recall_K_scores = validate(val_feat_loader, cand_feat_loader, model, val_video2gtrank, log_step=opt.log_step, opt=opt)
# remember best R@ sum and save checkpoint
is_best = rsum > best_rsum
best_rsum = max(rsum, best_rsum)
if is_best:
best_hit_k_scores = hit_k_scores
best_recall_K_scoress = recall_K_scores
print 'current perf: ', rsum
print 'best perf: ', best_rsum
print 'current hit_top_k: ', [round(x,3) for x in hit_k_scores]
print 'current recall_top_k: ', [round(x,3) for x in recall_K_scores]
fout_val_perf_hist.write("epoch_%d %f\n" % (epoch, rsum))
fout_val_perf_hist.flush()
save_checkpoint({
'epoch': epoch + 1,
'model': model.state_dict(),
'best_rsum': best_rsum,
'opt': opt,
'Eiters': model.Eiters,
}, is_best, filename='checkpoint_epoch_%s.pth.tar' % epoch, prefix=opt.logger_name + '/')
lr_counter += 1
decay_learning_rate(opt, model.optimizer, opt.lr_decay)
if not is_best:
# Early stop occurs if the validation performance
# does not improve in ten consecutive epochs.
no_impr_counter += 1
if no_impr_counter > 10:
print ("Early stopping happened")
break
# when the validation performance has decreased after an epoch,
# we divide the learning rate by 2 and continue training;
# but we use each learning rate for at least 3 epochs
if lr_counter > 2:
decay_learning_rate(opt, model.optimizer, 0.5)
lr_counter = 0
else:
# lr_counter = 0
no_impr_counter = 0
fout_val_perf_hist.close()
# output val performance
print json.dumps(vars(opt), indent = 2)
print '\nbest performance on validation:'
print 'hit_top_k', [round(x,3) for x in best_hit_k_scores]
print 'recall_top_k', [round(x,3) for x in best_recall_K_scoress]
with open(os.path.join(opt.logger_name,'val_perf.txt'), 'w') as fout:
fout.write('best performance on validation:')
fout.write('\nhit_top_k: ' + ", ".join(map(str, [round(x,3) for x in best_hit_k_scores])))
fout.write('\necall_top_k: ' + ", ".join(map(str, [round(x,3) for x in best_recall_K_scoress])))
# generate and run the shell script for test
templete = ''.join(open( 'TEMPLATE_eval.sh' ).readlines())
striptStr = templete.replace('@@@rootpath@@@', opt.rootpath)
striptStr = striptStr.replace('@@@collection@@@', opt.collection)
striptStr = striptStr.replace('@@@overwrite@@@', str(opt.overwrite))
striptStr = striptStr.replace('@@@model_path@@@', opt.logger_name)
runfile = 'do_eval_%s.sh' % opt.collection
open( runfile, 'w' ).write(striptStr+'\n')
os.system('chmod +x %s' % runfile)
os.system('./%s' % runfile)
def train(opt, train_loader, model, epoch):
# average meters to record the training statistics
batch_time = AverageMeter()
data_time = AverageMeter()
train_logger = LogCollector()
# switch to train mode
model.train_start()
progbar = Progbar(train_loader.dataset.length)
end = time.time()
for i, train_data in enumerate(train_loader):
# measure data loading time
data_time.update(time.time() - end)
# make sure train logger is used
model.logger = train_logger
# Update the model
b_size, loss = model.train_emb(*train_data)
# print loss
progbar.add(b_size, values=[("loss", loss)])
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
# Record logs in tensorboard
tb_logger.log_value('epoch', epoch, step=model.Eiters)
tb_logger.log_value('step', i, step=model.Eiters)
tb_logger.log_value('batch_time', batch_time.val, step=model.Eiters)
tb_logger.log_value('data_time', data_time.val, step=model.Eiters)
model.logger.tb_log(tb_logger, step=model.Eiters)
def validate(val_feat_loader, cand_feat_loader, model, video2gtrank, log_step=100, opt=None):
# compute the encoding for all the validation images and captions
val_video_embs, val_video_ids_list = encode_data(model, val_feat_loader, log_step, logging.info)
cand_video_embs, cand_video_ids_list = encode_data(model, cand_feat_loader, log_step, logging.info)
video2predrank = do_predict(val_video_embs, val_video_ids_list, cand_video_embs, cand_video_ids_list, output_dir=None, overwrite=0, no_imgnorm=opt.no_imgnorm)
hit_top_k = [5, 10, 20, 30]
recall_top_k = [50, 100, 200, 300]
hit_k_scores = hit_k_own(video2gtrank, video2predrank, top_k=hit_top_k)
recall_K_scores = recall_k_own(video2gtrank, video2predrank, top_k=recall_top_k)
for i, k in enumerate(hit_top_k):
tb_logger.log_value('hit_%d' % k, hit_k_scores[i], step=model.Eiters)
for i, k in enumerate(recall_top_k):
tb_logger.log_value('recall_%d' % k, recall_K_scores[i], step=model.Eiters)
currscore = recall_K_scores[1]
return currscore, hit_k_scores, recall_K_scores
def save_checkpoint(state, is_best, filename='checkpoint.pth.tar', prefix=''):
torch.save(state, prefix + filename)
if is_best:
shutil.copyfile(prefix + filename, prefix + 'model_best.pth.tar')
def decay_learning_rate(opt, optimizer, decay):
"""decay learning rate to the last LR"""
for param_group in optimizer.param_groups:
param_group['lr'] = param_group['lr']*decay
def get_learning_rate(optimizer):
"""decay learning rate to the last LR"""
lr_list = []
for param_group in optimizer.param_groups:
lr_list.append(param_group['lr'])
return lr_list
if __name__ == '__main__':
main()