-
Notifications
You must be signed in to change notification settings - Fork 1
/
train.py
243 lines (210 loc) · 8.96 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
import os
import argparse
import torch
import warnings
from tqdm import tqdm
from torch.utils.tensorboard import SummaryWriter
from torch.cuda.amp import autocast, GradScaler
from utils.data.dataloader import create_dataloader
from utils.misc import load_config, build_model, nms
from utils.metrics import Mean, AveragePrecision
class CheckpointManager(object):
def __init__(self, logdir, model, optim, scaler, scheduler, best_score):
self.epoch = 0
self.logdir = logdir
self.model = model
self.optim = optim
self.scaler = scaler
self.scheduler = scheduler
self.best_score = best_score
def save(self, filename):
data = {
'model_state_dict': self.model.state_dict(),
'optim_state_dict': self.optim.state_dict(),
'scaler_state_dict': self.scaler.state_dict(),
'scheduler_state_dict': self.scheduler.state_dict(),
'epoch': self.epoch,
'best_score': self.best_score,
}
torch.save(data, os.path.join(self.logdir, filename))
def restore(self, filename):
data = torch.load(os.path.join(self.logdir, filename))
self.model.load_state_dict(data['model_state_dict'])
self.optim.load_state_dict(data['optim_state_dict'])
self.scaler.load_state_dict(data['scaler_state_dict'])
self.scheduler.load_state_dict(data['scheduler_state_dict'])
self.epoch = data['epoch']
self.best_score = data['best_score']
def restore_lastest_checkpoint(self):
if os.path.exists(os.path.join(self.logdir, 'last.pth')):
self.restore('last.pth')
print("Restore the last checkpoint.")
def get_lr(optim):
for param_group in optim.param_groups:
return param_group['lr']
def train_step(images, true_boxes, true_classes, model, optim, amp, scaler,
metrics, device):
images = images.to(device)
true_boxes = [x.to(device) for x in true_boxes]
true_classes = [x.to(device) for x in true_classes]
optim.zero_grad()
with autocast(enabled=amp):
preds = model(images)
loss = model.compute_loss(preds, true_boxes, true_classes)
scaler.scale(loss).backward()
scaler.step(optim)
scaler.update()
loss = loss.item()
metrics['loss'].update(loss, images.shape[0])
def test_step(images, true_boxes, true_classes, difficulties, model, amp,
metrics, device):
images = images.to(device)
true_boxes = [x.to(device) for x in true_boxes]
true_classes = [x.to(device) for x in true_classes]
difficulties = [x.to(device) for x in difficulties]
with autocast(enabled=amp):
preds = model(images)
loss = model.compute_loss(preds, true_boxes, true_classes)
loss = loss.item()
metrics['loss'].update(loss, images.shape[0])
det_boxes, det_scores, det_classes = nms(*model.decode(preds))
metrics['APs'].update(det_boxes, det_scores, det_classes,
true_boxes, true_classes, difficulties)
def main():
parser = argparse.ArgumentParser(
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('--cfg', type=str, required=True,
help="config file")
parser.add_argument('--logdir', type=str, required=True,
help="log directory")
parser.add_argument('--workers', type=int, default=4,
help="number of dataloader workers")
parser.add_argument('--resume', action='store_true',
help="resume training")
parser.add_argument('--no_amp', action='store_true',
help="disable automatic mix precision")
parser.add_argument('--val_period', type=int, default=1,
help="number of epochs between successive validation")
args = parser.parse_args()
device = 'cuda' if torch.cuda.is_available() else 'cpu'
cfg = load_config(args.cfg)
enable_amp = (not args.no_amp)
if os.path.exists(args.logdir) and (not args.resume):
raise ValueError("Log directory %s already exists. Specify --resume "
"in command line if you want to resume the training."
% args.logdir)
model = build_model(cfg)
model.to(device)
train_loader = create_dataloader(cfg.train_json,
batch_size=cfg.batch_size,
image_size=cfg.input_size,
image_mean=cfg.image_mean,
image_stddev=cfg.image_stddev,
augment=True,
shuffle=True,
num_workers=args.workers)
val_loader = create_dataloader(cfg.val_json,
batch_size=cfg.batch_size,
image_size=cfg.input_size,
image_mean=cfg.image_mean,
image_stddev=cfg.image_stddev,
num_workers=args.workers)
# Criteria
optim = getattr(torch.optim, cfg.optim.pop('name'))(model.parameters(),
**cfg.optim)
scaler = GradScaler(enabled=enable_amp)
scheduler = getattr(torch.optim.lr_scheduler, cfg.scheduler.pop('name'))(
optim,
**cfg.scheduler
)
metrics = {
'loss': Mean(),
'APs': AveragePrecision(len(cfg.class_names), cfg.recall_steps)
}
# Checkpointing
ckpt = CheckpointManager(args.logdir,
model=model,
optim=optim,
scaler=scaler,
scheduler=scheduler,
best_score=0.)
ckpt.restore_lastest_checkpoint()
# TensorBoard writers
writers = {
'train': SummaryWriter(os.path.join(args.logdir, 'train')),
'val': SummaryWriter(os.path.join(args.logdir, 'val'))
}
# Kick off
for epoch in range(ckpt.epoch + 1, cfg.epochs + 1):
print("-" * 10)
print("Epoch: %d/%d" % (epoch, cfg.epochs))
# Train
model.train()
metrics['loss'].reset()
if epoch == 1:
warnings.filterwarnings(
'ignore',
".*call of `lr_scheduler.step\(\)` before `optimizer.step\(\)`.*" # noqa: W605
)
warmup_scheduler = torch.optim.lr_scheduler.LinearLR(
optim,
start_factor=0.001,
total_iters=min(1000, len(train_loader))
)
pbar = tqdm(train_loader,
bar_format="{l_bar}{bar:20}{r_bar}",
desc="Training")
for (images, true_boxes, true_classes, _) in pbar:
train_step(images,
true_boxes,
true_classes,
model=model,
optim=optim,
amp=enable_amp,
scaler=scaler,
metrics=metrics,
device=device)
loss = metrics['loss'].result
lr = get_lr(optim)
pbar.set_postfix(loss='%.5f' % metrics['loss'].result, lr=lr)
if epoch == 1:
warmup_scheduler.step()
writers['train'].add_scalar('Loss', loss, epoch)
writers['train'].add_scalar('Learning rate', get_lr(optim), epoch)
scheduler.step()
# Validation
if epoch % args.val_period == 0:
model.eval()
metrics['loss'].reset()
metrics['APs'].reset()
pbar = tqdm(val_loader,
bar_format="{l_bar}{bar:20}{r_bar}",
desc="Validation")
with torch.no_grad():
for (images, true_boxes, true_classes, difficulties) in pbar:
test_step(images,
true_boxes,
true_classes,
difficulties,
model=model,
amp=enable_amp,
metrics=metrics,
device=device)
pbar.set_postfix(loss='%.5f' % metrics['loss'].result)
APs = metrics['APs'].result
mAP50 = APs[:, 0].mean()
mAP = APs.mean()
if mAP > ckpt.best_score:
ckpt.best_score = mAP
ckpt.save('best.pth')
print("mAP@[0.5]: %.3f" % mAP50)
print("mAP@[0.5:0.95]: %.3f (best: %.3f)" % (mAP, ckpt.best_score))
writers['val'].add_scalar('Loss', metrics['loss'].result, epoch)
writers['val'].add_scalar('mAP@[0.5]', mAP50, epoch)
writers['val'].add_scalar('mAP@[0.5:0.95]', mAP, epoch)
ckpt.epoch += 1
ckpt.save('last.pth')
writers['train'].close()
writers['val'].close()
if __name__ == '__main__':
main()