-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathtrain_cpt.py
163 lines (146 loc) · 6.14 KB
/
train_cpt.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
# coding:utf8
import tqdm
import os
from copy import deepcopy
import time
import json
import sys
import pdb
import traceback
from bdb import BdbQuit
import torch
from opts import parse_opt
from models.concept_detector import ConceptDetector
from dataloader import get_concept_dataloader
def clip_gradient(optimizer, grad_clip):
for group in optimizer.param_groups:
for param in group['params']:
param.grad.data.clamp_(-grad_clip, grad_clip)
def train():
dataset_name = opt.dataset_name
idx2concept = json.load(open(os.path.join(opt.captions_dir, dataset_name, 'idx2concept.json'), 'r'))
img_concepts = json.load(open(os.path.join(opt.captions_dir, dataset_name, 'img_concepts.json'), 'r'))
cpt_detector = ConceptDetector(idx2concept, opt.settings)
cpt_detector.to(opt.device)
lr = opt.concept_lr
optimizer, criterion = cpt_detector.get_optim_criterion(lr)
if opt.concept_resume:
print("====> loading checkpoint '{}'".format(opt.concept_resume))
chkpoint = torch.load(opt.concept_resume, map_location=lambda s, l: s)
assert opt.settings == chkpoint['settings'], \
'opt.settings and resume model settings are different'
assert idx2concept == chkpoint['idx2concept'], \
'idx2concept and resume model idx2concept are different'
assert dataset_name == chkpoint['dataset_name'], \
'dataset_name and resume model dataset_name are different'
cpt_detector.load_state_dict(chkpoint['model'])
optimizer.load_state_dict(chkpoint['optimizer'])
lr = optimizer.param_groups[0]['lr']
print("====> loaded checkpoint '{}', epoch: {}"
.format(opt.concept_resume, chkpoint['epoch']))
concept2idx = {}
for i, w in enumerate(idx2concept):
concept2idx[w] = i
ground_truth = deepcopy(img_concepts['test'])
print('====> process image concepts begin')
img_concepts_id = {}
for split, concepts in img_concepts.items():
print('convert %s concepts to index' % split)
img_concepts_id[split] = {}
for fn, cpts in tqdm.tqdm(concepts.items()):
cpts = [concept2idx[c] for c in cpts if c in concept2idx]
img_concepts_id[split][fn] = cpts
img_concepts = img_concepts_id
print('====> process image concepts end')
f_fc = os.path.join(opt.feats_dir, dataset_name, '%s_fc.h5' % dataset_name)
train_data = get_concept_dataloader(
f_fc, img_concepts['train'], len(idx2concept),
opt.concept_bs, opt.concept_num_works)
val_data = get_concept_dataloader(
f_fc, img_concepts['val'], len(idx2concept),
opt.concept_bs, opt.concept_num_works, shuffle=False)
test_data = get_concept_dataloader(
f_fc, img_concepts['test'], len(idx2concept),
opt.concept_bs, opt.concept_num_works, shuffle=False)
def forward(data, training=True):
cpt_detector.train(training)
loss_val = 0.0
for _, fc_feats, cpts_tensors in tqdm.tqdm(data):
fc_feats = fc_feats.to(opt.device)
cpts_tensors = cpts_tensors.to(opt.device)
pred = cpt_detector(fc_feats)
loss = criterion(pred, cpts_tensors)
loss_val += loss.item()
if training:
optimizer.zero_grad()
loss.backward()
clip_gradient(optimizer, opt.grad_clip)
optimizer.step()
return loss_val / len(data)
checkpoint = os.path.join(opt.checkpoint, 'concept', dataset_name)
if not os.path.exists(checkpoint):
os.makedirs(checkpoint)
previous_loss = None
for epoch in range(opt.concept_epochs):
print('--------------------epoch: %d' % epoch)
train_loss = forward(train_data)
with torch.no_grad():
val_loss = forward(val_data, training=False)
# test
test_loss = 0.0
pre = 0.0
recall = 0.0
last_score = 0.0
for fns, fc_feats, cpts_tensors in tqdm.tqdm(test_data):
fc_feats = fc_feats.to(opt.device)
cpts_tensors = cpts_tensors.to(opt.device)
pred, concepts, scores = cpt_detector.sample(fc_feats, num=opt.num_concepts)
loss = criterion(pred, cpts_tensors)
test_loss += loss.item()
tmp_pre = 0.0
tmp_rec = 0.0
for i, fn in enumerate(fns):
cpts = concepts[i]
grdt = ground_truth[fn]
jiaoji = len(set(grdt) - (set(grdt) - set(cpts)))
tmp_pre += jiaoji / len(cpts)
tmp_rec += jiaoji / len(grdt)
pre += tmp_pre / len(fns)
recall += tmp_rec / len(fns)
last_score += float(scores[:, -1].mean())
data_len = len(test_data)
test_loss = test_loss / data_len
pre = pre / data_len
recall = recall / data_len
last_score = last_score / data_len
if previous_loss is not None and val_loss > previous_loss:
lr = lr * 0.5
for param_group in optimizer.param_groups:
param_group['lr'] = lr
previous_loss = val_loss
print('train_loss: %.4f, val_loss: %.4f, test_loss: %.4f, '
'precision: %.4f, recall: %.4f, last_score: %.4f' %
(train_loss, val_loss, test_loss, pre, recall, last_score))
if epoch > -1:
chkpoint = {
'epoch': epoch,
'model': cpt_detector.state_dict(),
'optimizer': optimizer.state_dict(),
'settings': opt.settings,
'idx2concept': idx2concept,
'dataset_name': dataset_name,
}
checkpoint_path = os.path.join(checkpoint, 'model_%d_%.4f_%.4f_%s.pth' % (
epoch, train_loss, val_loss, time.strftime('%m%d-%H%M')))
torch.save(chkpoint, checkpoint_path)
if __name__ == '__main__':
try:
opt = parse_opt()
train()
except BdbQuit:
sys.exit(1)
except Exception:
traceback.print_exc()
print('')
pdb.post_mortem()
sys.exit(1)