-
Notifications
You must be signed in to change notification settings - Fork 2
/
train_amp.py
254 lines (209 loc) · 10.6 KB
/
train_amp.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
import os
import numpy as np
import sys
import logging
from time import time
from tensorboardX import SummaryWriter
import argparse
import torch
from apex import amp
from apex.parallel import DistributedDataParallel
from torch.optim.lr_scheduler import StepLR
import torch.distributed as dist
# from torch.nn.parallel import DistributedDataParallel
from loss import SimpleLoss, DiscriminativeLoss
from data.dataset import semantic_dataset, semantic_dataset_dist
from data.const import NUM_CLASSES
from evaluation.iou import get_batch_iou
from evaluation.angle_diff import calc_angle_diff
from model import get_model
from evaluate import onehot_encoding, eval_iou
def write_log(writer, ious, title, counter):
writer.add_scalar(f'{title}/iou', torch.mean(ious[1:]), counter)
for i, iou in enumerate(ious):
writer.add_scalar(f'{title}/class_{i}/iou', iou, counter)
# def setup():
# dist.init_process_group('nccl')
# def cleanup():
# dist.destroy_process_group()
def train(args):
rank = 0
if args.multi_gpu:
dist.init_process_group('nccl')
rank = dist.get_rank()
pid = os.getpid()
print(f'current pid:{pid}')
print(f'current rank:{rank}')
device_id = rank % torch.cuda.device_count()
torch.cuda.set_device(rank)
torch.cuda.empty_cache()
if rank == 0:
if not os.path.exists(args.logdir):
os.mkdir(args.logdir)
logging.basicConfig(filename=os.path.join(args.logdir, "results.log"),
filemode='w',
format='%(asctime)s: %(message)s',
datefmt='%Y-%m-%d %H:%M:%S',
level=logging.INFO)
logging.getLogger('shapely.geos').setLevel(logging.CRITICAL)
logger = logging.getLogger()
logger.addHandler(logging.StreamHandler(sys.stdout))
data_conf = {
'num_channels': NUM_CLASSES + 1,
'image_size': args.image_size,
'xbound': args.xbound,
'ybound': args.ybound,
'zbound': args.zbound,
'dbound': args.dbound,
'thickness': args.thickness,
'angle_class': args.angle_class,
}
if args.multi_gpu:
train_loader, val_loader, train_sampler, val_sampler = semantic_dataset_dist(args.version, args.dataroot, args.prior_map_root, data_conf, args.bsz, args.nworkers, args.multi_gpu)
else:
train_loader, val_loader = semantic_dataset(args.version, args.dataroot, args.prior_map_root, data_conf, args.bsz,
args.nworkers, args.multi_gpu)
model = get_model(args.model, data_conf, args, args.instance_seg, args.embedding_dim, args.direction_pred, args.angle_class)
if args.multi_gpu:
model = model.to(device_id)
opt = torch.optim.Adam(model.parameters(), lr=args.lr, weight_decay=args.weight_decay)
sched = StepLR(opt, 10, 0.1)
model, opt = amp.initialize(model, opt, opt_level='O2')
if args.model_root != '':
model.load_state_dict(torch.load(args.model_root))
if args.finetune:
model.load_state_dict(torch.load(args.modelf), strict=False)
for name, param in model.named_parameters():
if 'bevencode.up' in name or 'bevencode.layer3' in name:
param.requires_grad = True
else:
param.requires_grad = False
if args.multi_gpu:
model = DistributedDataParallel(model)
else:
model.cuda()
if rank==0:
writer = SummaryWriter(logdir=args.logdir)
loss_fn = SimpleLoss(args.pos_weight).to(device_id)
embedded_loss_fn = DiscriminativeLoss(args.embedding_dim, args.delta_v, args.delta_d).to(device_id)
direction_loss_fn = torch.nn.BCELoss(reduction='none').to(device_id)
model.train()
counter = 0
last_idx = len(train_loader) - 1
for epoch in range(args.nepochs):
if args.multi_gpu:
train_sampler.set_epoch(epoch)
val_sampler.set_epoch(epoch)
for batchi, (imgs, trans, rots, intrins, post_trans, post_rots, lidar_data, lidar_mask, car_trans,
yaw_pitch_roll, semantic_gt, instance_gt, direction_gt, prior_map) in enumerate(train_loader):
t0 = time()
opt.zero_grad()
semantic, embedding, direction = model(imgs.to(device_id), trans.to(device_id), rots.to(device_id), intrins.to(device_id),
post_trans.to(device_id), post_rots.to(device_id), lidar_data.to(device_id),
lidar_mask.to(device_id), car_trans.to(device_id), yaw_pitch_roll.to(device_id),
prior_map.to(device_id))
semantic_gt = semantic_gt.cuda().float()
instance_gt = instance_gt.cuda()
seg_loss = loss_fn(semantic, semantic_gt)
if args.instance_seg:
var_loss, dist_loss, reg_loss = embedded_loss_fn(embedding, instance_gt)
else:
var_loss = 0
dist_loss = 0
reg_loss = 0
if args.direction_pred:
direction_gt = direction_gt.cuda()
lane_mask = (1 - direction_gt[:, 0]).unsqueeze(1)
direction_loss = direction_loss_fn(torch.softmax(direction, 1), direction_gt)
direction_loss = (direction_loss * lane_mask).sum() / (lane_mask.sum() * direction_loss.shape[1] + 1e-6)
angle_diff = calc_angle_diff(direction, direction_gt, args.angle_class)
else:
direction_loss = 0
angle_diff = 0
final_loss = seg_loss * args.scale_seg + var_loss * args.scale_var + dist_loss * args.scale_dist + direction_loss * args.scale_direction
with amp.scale_loss(final_loss, opt) as scaled_loss:
scaled_loss.backward()
torch.nn.utils.clip_grad_norm_(model.parameters(), args.max_grad_norm)
opt.step()
counter += 1
t1 = time()
if counter % 10 == 0 and rank == 0:
intersects, union = get_batch_iou(onehot_encoding(semantic), semantic_gt)
iou = intersects / (union + 1e-7)
logger.info(f"TRAIN[{epoch:>3d}]: [{batchi:>4d}/{last_idx}] "
f"Time: {t1-t0:>7.4f} "
f"Loss: {final_loss.item():>7.4f} "
f"IOU: {np.array2string(iou[1:].numpy(), precision=3, floatmode='fixed')}")
write_log(writer, iou, 'train', counter)
writer.add_scalar('train/step_time', t1 - t0, counter)
writer.add_scalar('train/seg_loss', seg_loss, counter)
writer.add_scalar('train/var_loss', var_loss, counter)
writer.add_scalar('train/dist_loss', dist_loss, counter)
writer.add_scalar('train/reg_loss', reg_loss, counter)
writer.add_scalar('train/direction_loss', direction_loss, counter)
writer.add_scalar('train/final_loss', final_loss, counter)
writer.add_scalar('train/angle_diff', angle_diff, counter)
iou = eval_iou(model, val_loader)
if rank==0:
logger.info(f"EVAL[{epoch:>2d}]: "
f"IOU: {np.array2string(iou[1:].numpy(), precision=3, floatmode='fixed')}")
write_log(writer, iou, 'eval', counter)
model_name = os.path.join(args.logdir, f"model{epoch}.pt")
torch.save(model.module.state_dict(), model_name)
logger.info(f"{model_name} saved")
model.train()
sched.step()
# cleanup()
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='HDMapNet training.')
# multi_gpu config
parser.add_argument("--multi_gpu", type=bool, default=True)
parser.add_argument("--fusion_mode", type=str, default="concat", choices=['concat', 'swin-atten', 'atten', 'deform-atten', 'masked-atten', 'masked-atten-seg'])
parser.add_argument('--align_fusion', action='store_true')
parser.add_argument("--local_rank", type=int, default=0)
# logging config
parser.add_argument("--logdir", type=str, default='./runs_detrans')
# nuScenes config
# parser.add_argument('--dataroot', type=str, default='/data3/nuscenes')
parser.add_argument('--dataroot', type=str, default='/opt/data/private/nuScenes_trainval')
parser.add_argument('--prior_map_root', type=str, default='/opt/data/private/prior_map_dataset/prior_map_trainval')
parser.add_argument('--version', type=str, default='v1.0-trainval', choices=['v1.0-trainval', 'v1.0-mini'])
# model config
parser.add_argument("--model", type=str, default='HDMapNet_cam')
# training config
parser.add_argument("--nepochs", type=int, default=30)
parser.add_argument("--max_grad_norm", type=float, default=5.0)
parser.add_argument("--pos_weight", type=float, default=2.13)
parser.add_argument("--bsz", type=int, default=24)
parser.add_argument("--nworkers", type=int, default=16)
parser.add_argument("--lr", type=float, default=1e-3)
parser.add_argument("--weight_decay", type=float, default=1e-7)
# finetune config
parser.add_argument('--finetune', action='store_true')
parser.add_argument('--modelf', type=str, default=None)
# checkpoint config
parser.add_argument('--model_root', type=str, default='')
# data config
parser.add_argument("--thickness", type=int, default=5)
parser.add_argument("--image_size", nargs=2, type=int, default=[128, 352])
parser.add_argument("--xbound", nargs=3, type=float, default=[-30.0, 30.0, 0.15])
parser.add_argument("--ybound", nargs=3, type=float, default=[-15.0, 15.0, 0.15])
parser.add_argument("--zbound", nargs=3, type=float, default=[-10.0, 10.0, 20.0])
parser.add_argument("--dbound", nargs=3, type=float, default=[4.0, 45.0, 1.0])
# embedding config
parser.add_argument('--instance_seg', action='store_true')
parser.add_argument("--embedding_dim", type=int, default=16)
parser.add_argument("--delta_v", type=float, default=0.5)
parser.add_argument("--delta_d", type=float, default=3.0)
# direction config
parser.add_argument('--direction_pred', action='store_true')
parser.add_argument('--angle_class', type=int, default=36)
# map prior config
parser.add_argument('--map_prior', action='store_true')
# loss config
parser.add_argument("--scale_seg", type=float, default=1.0)
parser.add_argument("--scale_var", type=float, default=1.0)
parser.add_argument("--scale_dist", type=float, default=1.0)
parser.add_argument("--scale_direction", type=float, default=0.2)
args = parser.parse_args()
train(args)