-
Notifications
You must be signed in to change notification settings - Fork 5
/
Copy pathrun_symnet_multi.py
159 lines (106 loc) · 5.2 KB
/
run_symnet_multi.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
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from utils.base_solver import BaseSolver
import os, logging, importlib, re, copy, random, tqdm, argparse
import os.path as osp
import cPickle as pickle
import numpy as np
from collections import defaultdict
import tensorflow as tf
import tensorflow.contrib.slim as slim
import torch
from utils import config as cfg
from utils import dataset, utils
from utils.evaluator import Multi_Evaluator
from run_symnet import make_parser as make_basic_parser
from run_symnet import SolverWrapper as CZSLSolverWrapper
def make_parser():
parser = make_basic_parser()
parser.add_argument("--lambda_atten", type=float, default=0)
parser.add_argument("--lambda_multi_rmd", type=float, default=0)
parser.add_argument("--sample_bar", default=0.1, type=float)
parser.add_argument("--bce_neg_weight", default=0., type=float)
return parser
def main():
logger = logging.getLogger('MAIN')
parser = make_parser()
args = parser.parse_args()
args.network = 'symnet_multi'
utils.duplication_check(args)
utils.display_args(args, logger)
logger.info("Loading dataset")
train_dataloader = dataset.get_dataloader(args.data, 'train',
batchsize=args.bz,args=args)
test_dataloader = dataset.get_dataloader(args.data, 'test',
batchsize=args.test_bz, args=args)
logger.info("Loading network and solver")
network = importlib.import_module('network.'+args.network)
net = network.Network(train_dataloader, args)
with utils.create_session() as sess:
sw = SolverWrapper(net, train_dataloader, test_dataloader, args)
sw.trainval_model(sess, args.epoch)
class SolverWrapper(CZSLSolverWrapper):
def trainval_model(self, sess, max_epoch):
self.criterion = 'mAUC'
logger = self.logger('train_model')
logger.info('Begin training')
lr, score_op, train_op, train_summary_op = self.construct_graph(sess)
self.initialize(sess)
sess.graph.finalize()
writer = tf.summary.FileWriter(self.log_dir, sess.graph)
evaluator = Multi_Evaluator()
best_report = defaultdict(dict)
for epoch in range(self.epoch_num+1, max_epoch+1):
################################## train ####################################
summary = None
for batch_ind, batch in tqdm.tqdm(enumerate(self.train_dataloader),
total=len(self.train_dataloader),
postfix='epoch %d/%d'%(epoch, max_epoch)):
if isinstance(lr, float):
eval_lr = lr
else:
eval_lr = lr.eval()
summary = self.network.train_step(sess, batch, eval_lr, train_op, train_summary_op)
writer.add_summary(summary, float(epoch))
if self.args.test_freq>0 and epoch % self.args.test_freq == 0:
############################## test ################################
all_attr = []
all_pred = defaultdict(list)
for image_ind, batch in tqdm.tqdm(enumerate(self.test_dataloader), total=len(self.test_dataloader), postfix='test'):
predictions = self.network.test_step(sess, batch, score_op)
all_attr.append(torch.from_numpy(batch[1]))
for key in score_op.keys():
all_pred[key].append(predictions[key])
all_report = {'Current':{}, 'Best':{}}
for name, pred in all_pred.items():
mAP, mAUC = evaluator(
np.concatenate(pred,0),
np.concatenate(all_attr,0)
)
# save to tensorboard
report_dict = {
'mAP': mAP,
'mAUC': mAUC,
'name': self.args.name,
'epoch': epoch,
}
if self.criterion not in best_report[name] or report_dict[self.criterion] > best_report[name][self.criterion]:
best_report[name] = report_dict
# print test results
all_report['Current'][name] = report_dict
all_report['Best'][name] = best_report[name]
for key in all_report.keys():
print("%s %s: "%(key, name) + utils.formated_multi_result(all_report[key][name]))
# save to tensorboard
summary = tf.Summary()
for key, value in report_dict.items():
if key not in ['name', 'epoch']:
summary.value.add(tag="%s/%s"%(name,key), simple_value=value)
writer.add_summary(summary, epoch)
if epoch % self.args.snapshot_freq == 0:
self.snapshot(sess, epoch)
writer.close()
logger.info('Finished.')
if __name__=="__main__":
main()