-
Notifications
You must be signed in to change notification settings - Fork 42
/
Copy pathtrain.py
246 lines (199 loc) · 8.79 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
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import os
import math
import argparse
import pprint
import tqdm
from collections import defaultdict
import torch
import torch.nn.functional as F
from tensorboardX import SummaryWriter
from datasets import get_dataloader
from transforms import get_transform
from tasks import get_task
from optimizers import get_optimizer
from schedulers import get_scheduler
import utils
import utils.config
import utils.checkpoint
def evaluate_single_epoch(config, task, dataloader, epoch,
writer, postfix_dict):
task.get_model().eval()
cal_metric_once = config.eval.cal_metric_once
with torch.no_grad():
batch_size = config.eval.batch_size
total_size = len(dataloader.dataset)
total_step = math.ceil(total_size / batch_size)
tbar = tqdm.tqdm(enumerate(dataloader), total=total_step)
loss_list = []
metric_list_dict = defaultdict(list)
data_list_dict = defaultdict(list)
for i, data in tbar:
images = data['image'].cuda()
labels = data['label'].cuda()
outputs = task.forward(images=images)
loss = task.loss(outputs, labels)
loss_list.append(loss.item())
predicts = task.inference(outputs=outputs)
if cal_metric_once:
for key, value in predicts.items():
data_list_dict[key].extend(value.cpu().numpy())
for key, value in data.items():
if key == 'image':
continue
data_list_dict[key + 's'].extend(value)
else:
metric_dict = task.metrics(labels=labels, **predicts, **data)
for key, value in metric_dict.items():
metric_list_dict[key].append(value)
f_epoch = epoch + i / total_step
desc = '{:5s}'.format('dev')
desc += ', {:.2f} epoch'.format(f_epoch)
tbar.set_description(desc)
tbar.set_postfix(**postfix_dict)
if i % config.train.log_step == 0:
log_step = int(f_epoch * 10000)
if writer is not None:
annotated_images = task.annotate_to_images(images, labels, predicts)
for idx, annotated_image in enumerate(annotated_images):
writer.add_image('dev/image_{}'.format(idx), annotated_image, log_step)
log_dict = {}
log_dict['loss'] = sum(loss_list) / len(loss_list)
if cal_metric_once:
metric_dict = task.metrics(**data_list_dict)
for key, value in metric_dict.items():
log_dict[key] = value
else:
for key, values in metric_list_dict.items():
log_dict[key] = sum(values) / len(values)
for key, value in log_dict.items():
if writer is not None:
writer.add_scalar('dev/{}'.format(key), value, epoch)
if key in ['loss', 'score']:
postfix_dict['dev/{}'.format(key)] = value
return log_dict['score']
def train_single_epoch(config, task, dataloader, optimizer,
epoch, writer, postfix_dict):
task.get_model().train()
batch_size = config.train.batch_size
total_size = len(dataloader.dataset)
total_step = total_size // batch_size
log_dict = {}
tbar = tqdm.tqdm(enumerate(dataloader), total=total_step)
for i, data in tbar:
images = data['image'].cuda()
labels = data['label'].cuda()
outputs = task.forward(images=images, labels=labels)
loss = task.loss(outputs, labels)
log_dict['loss'] = loss.item()
predicts = task.inference(outputs=outputs, labels=labels)
metric_dict = task.metrics(labels=labels, **predicts)
log_dict.update(metric_dict)
loss.backward()
warmup = True
if warmup and epoch < 3:
if (i+1) % 2 == 0:
optimizer.step()
optimizer.zero_grad()
else:
if config.train.num_grad_acc is None:
optimizer.step()
optimizer.zero_grad()
elif (i+1) % config.train.num_grad_acc == 0:
optimizer.step()
optimizer.zero_grad()
f_epoch = epoch + i / total_step
log_dict['lr'] = optimizer.param_groups[0]['lr']
for key in ['lr', 'loss', 'score']:
value = log_dict[key]
postfix_dict['train/{}'.format(key)] = value
desc = '{:5s}'.format('train')
desc += ', {:.2f} epoch'.format(f_epoch)
tbar.set_description(desc)
tbar.set_postfix(**postfix_dict)
if i % config.train.log_step == 0:
log_step = int(f_epoch * 10000)
if writer is not None:
for key, value in log_dict.items():
writer.add_scalar('train/{}'.format(key), value, log_step)
annotated_images = task.annotate_to_images(images, labels, predicts)
for idx, annotated_image in enumerate(annotated_images):
writer.add_image('train/image_{}'.format(idx), annotated_image, log_step)
def train(config, task, dataloaders, optimizer, scheduler, writer, start_epoch):
scores = []
best_score = 0.0
best_score_mavg = 0.0
postfix_dict = {}
for epoch in range(start_epoch, config.train.num_epochs):
# train phase
train_single_epoch(config, task, dataloaders['train'],
optimizer, epoch, writer, postfix_dict)
# val phase
score = evaluate_single_epoch(config, task, dataloaders['dev'],
epoch, writer, postfix_dict)
scores.append(score)
if config.scheduler.name == 'reduce_lr_on_plateau':
scheduler.step(score)
elif config.scheduler.name != 'reduce_lr_on_plateau':
scheduler.step()
if epoch % config.train.save_checkpoint_epoch == 0:
utils.checkpoint.save_checkpoint(config, task.get_model(), optimizer,
epoch, 0, keep=10)
scores = scores[-20:]
score_mavg = sum(scores) / len(scores)
writer.add_scalar('dev/score_mavg', score_mavg, epoch)
if score > best_score:
best_score = score
utils.checkpoint.save_checkpoint(config, task.get_model(), optimizer,
epoch, keep=10, name='best.score')
utils.checkpoint.copy_last_n_checkpoints(config, 10, 'best.score.{:04d}.pth')
if score_mavg > best_score_mavg:
best_score_mavg = score_mavg
utils.checkpoint.save_checkpoint(config, task.get_model(), optimizer,
epoch, keep=10, name='best.score_mavg')
utils.checkpoint.copy_last_n_checkpoints(config, 10, 'best.score_mavg.{:04d}.pth')
return {'score': best_score, 'score_mavg': best_score_mavg}
def run(config):
train_dir = config.train.dir
task = get_task(config)
optimizer = get_optimizer(config, task.get_model().parameters())
checkpoint = utils.checkpoint.get_initial_checkpoint(config)
if checkpoint is not None:
last_epoch, step = utils.checkpoint.load_checkpoint(task.get_model(),
optimizer,
checkpoint)
else:
last_epoch, step = -1, -1
print('from checkpoint: {} last epoch:{}'.format(checkpoint, last_epoch))
scheduler = get_scheduler(config, optimizer, last_epoch)
preprocess_opt = task.get_preprocess_opt()
dataloaders = {split:get_dataloader(config, split,
get_transform(config, split,
**preprocess_opt))
for split in ['train', 'dev']}
writer = SummaryWriter(config.train.dir)
train(config, task, dataloaders, optimizer, scheduler,
writer, last_epoch+1)
def parse_args():
description = 'Train humpback whale identification'
parser = argparse.ArgumentParser(description=description)
parser.add_argument('--config', dest='config_file',
help='configuration filename',
default=None, type=str)
return parser.parse_args()
def main():
import warnings
warnings.filterwarnings("ignore")
print('Train humpback whale identification')
args = parse_args()
if args.config_file is None:
raise Exception('no configuration file')
config = utils.config.load(args.config_file)
pprint.PrettyPrinter(indent=2).pprint(config)
utils.prepare_train_directories(config)
run(config)
print('success!')
if __name__ == '__main__':
main()