-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathLaserMix_Trainer.py
224 lines (167 loc) · 8.21 KB
/
LaserMix_Trainer.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
# Common
import os
import wandb
import numpy as np
import time
# torch
import torch
import torch.nn as nn
import torch.nn.functional as F
import MinkowskiEngine as ME
from network.lr_adjust import adjust_learning_rate
from utils import common as com
from domain_mix.laserMix import LaserMix
from trainer_base import Base_Trainer
def my_worker_init_fn(worker_id):
np.random.seed(np.random.get_state()[1][0] + worker_id)
class Trainer(Base_Trainer):
def __init__(self,
cfg,
net_G, ema_G,
G_optim,
logger, tf_writer, device):
super().__init__(cfg, logger, tf_writer, device)
print("This is a laserMix trainer.")
self.net_G = net_G
self.G_optim = G_optim
""" Define Loss Function """
self.criterion = nn.CrossEntropyLoss(ignore_index=0) # seg loss
if self.cfg.SOURCE_LOSS.lambda_lov > 0.:
from network.lov_loss import Lovasz_loss
self.lov_criterion = Lovasz_loss(ignore=0)
if cfg.MEAN_TEACHER.use_mt:
self.ema_G = ema_G
self.create_ema_model(self.ema_G, self.net_G)
self.laserMix = LaserMix(self.cfg)
print("Init LaserMix Trainer Done.")
def train(self):
for epoch in range(self.cfg.TRAIN.MAX_EPOCHS):
print("This is epoch: {}".format(epoch))
self.train_one_epoch()
def train_one_epoch(self):
for tgt_BData in self.tgt_train_loader:
self.wb_dict = {}
self.c_iter += 1
start_t = time.time()
self.set_lr()
self.set_zero_grad()
# send data to GPU
src_BData = self.src_TraDL.next()
self.src_BData = self.send_data2GPU(src_BData)
self.tgt_BData = self.send_data2GPU(tgt_BData)
""" tea-model forward the raw target scans """
self.tgt_BData['pseudo_label'] = self.get_pseudo_label()
# laserMix scans
self.masked_batch = self.laserMix.mix(self.src_BData, self.tgt_BData)
# update G
src_loss = self.train_source()
tgt_loss = self.train_target()
(src_loss + tgt_loss).backward()
self.G_optim.step()
if self.cfg.MEAN_TEACHER.use_mt and \
self.cfg.MEAN_TEACHER.alpha_ema > 0 and \
self.c_iter % self.cfg.MEAN_TEACHER.update_every == 0:
self.update_ema_variables(self.ema_G, self.net_G)
if self.c_iter % self.cfg.TRAIN.LOG_PERIOD == 0:
print('iter:{0:6d}, '
'src_Ls:{1:.4f}, '
'tgt_Ls:{2:.4f}, '
'itr:{3:.3f}, '
'Exp:{4}'.format(self.c_iter,
self.wb_dict['netG/all_src_loss'],
self.wb_dict['netG/all_tgt_loss'],
time.time() - start_t,
self.cfg.TRAIN.EXP_NAME))
self.save_log() # save logs
if self.c_iter % self.t_val_iter == 0: # Traget domain val.
self.valid_and_save()
if self.c_iter % self.s_val_iter == 0: # Source domain val.
_ = self.src_valer.rolling_predict(self.net_G, self.ema_G, self.c_iter, domain='src')
if self.c_iter % 10 == 0:
torch.cuda.empty_cache()
if self.c_iter == self.cfg.TRAIN.MAX_ITERS:
if self.c_iter % self.t_val_iter != 0:
self.valid_and_save()
print("Finish training, this is max iter: {}".format(self.c_iter))
quit()
torch.cuda.empty_cache()
def train_source(self):# ===========train G ================
# Train with Source. compute source seg loss
if self.cfg.DATASET_SOURCE.USE_DGT:
src_G_in = ME.SparseTensor(self.src_BData['aug_feats_mink'],
self.src_BData['aug_coords_mink'])
src_labels = self.src_BData['aug_labels_mink']
else:
src_G_in = ME.SparseTensor(self.src_BData['feats_mink'], self.src_BData['coords_mink'])
src_labels = self.src_BData['labels_mink']
src_G_out = self.net_G(src_G_in)
src_stu_logit = src_G_out['sp_out']
all_src_loss = 0.
# loss 1. main classifier CE loss
src_seg_loss = self.criterion(src_stu_logit.F, src_labels)
all_src_loss = all_src_loss + src_seg_loss
self.wb_dict['netG/seg_Loss'] = src_seg_loss.mean()
# loss 2. lov loss
if self.cfg.SOURCE_LOSS.lambda_lov > 0.:
lovasz_loss = self.lov_criterion(F.softmax(src_stu_logit.F, dim=1), src_labels)
all_src_loss = all_src_loss + lovasz_loss
self.wb_dict['netG/lov_Loss'] = lovasz_loss.mean()
self.wb_dict['netG/all_src_loss'] = all_src_loss.mean()
return all_src_loss
def train_target(self):
all_tgt_loss = 0.
# LaserMix
t2s_stensor = ME.SparseTensor(coordinates=self.masked_batch["mixed_coors_1"].int(),
features=self.masked_batch["mixed_feats_1"])
lmix_outDict = self.net_G(t2s_stensor)
lmix_logit = lmix_outDict['sp_out']
t2s_labels = self.masked_batch["mixed_lbls_1"].cuda()
t2s_loss = self.criterion(lmix_logit.F, t2s_labels.long())
all_tgt_loss = all_tgt_loss + t2s_loss
self.wb_dict['netG/pse_seg_loss'] = t2s_loss.mean()
self.wb_dict['netG/all_tgt_loss'] = all_tgt_loss.mean()
return all_tgt_loss
def valid_and_save(self):
cp_fn = os.path.join(self.cfg.TRAIN.MODEL_DIR, 'cp_current.tar')
self.fast_save_CP(cp_fn)
# If you want save model checkpoint, set cfg.TRAIN.SAVE_MORE_ITER = True
if self.c_iter > self.cfg.TRAIN.SAVE_ITER and self.cfg.TRAIN.SAVE_MORE_ITER:
cp_fn = os.path.join(self.cfg.TRAIN.MODEL_DIR, 'cp_{}_iter.tar'.format(self.c_iter))
self.fast_save_CP(cp_fn)
tgt_sp_iou = self.tgt_valer.rolling_predict(self.net_G, self.ema_G, self.c_iter, domain='tgt')
if (tgt_sp_iou > self.best_IoU_after_saveIter and self.c_iter > self.cfg.TRAIN.SAVE_ITER) or \
tgt_sp_iou > self.ml_info['bt_tgt_spIoU']:
s_name = 'target_Sp'
if (tgt_sp_iou > self.best_IoU_after_saveIter and self.c_iter > self.cfg.TRAIN.SAVE_ITER):
# 由于点云GAN不稳定，有时候好的结果在最开始出现，所以添加这个if
self.best_IoU_after_saveIter = tgt_sp_iou
s_name = 'target_Sp_After'
self.best_IoU_iter = self.c_iter
self.ml_info['bt_tgt_spIoU'] = tgt_sp_iou
wandb.run.summary["bt_tgt_spIoU"] = tgt_sp_iou
com.save_best_check(self.net_G,
self.G_optim, None,
self.c_iter, self.logger,
self.cfg.TRAIN.MODEL_DIR, name=s_name,
iou=tgt_sp_iou)
torch.cuda.empty_cache()
def save_log(self):
self.wb_dict['lr/lr_G'] = self.G_optim.state_dict()['param_groups'][0]['lr']
for k, v in self.wb_dict.items():
self.tf_writer.add_scalar(k, v, self.c_iter)
wandb.log({k: v}, step=self.c_iter)
def set_zero_grad(self):
self.net_G.train() # set model to training mode
self.G_optim.zero_grad()
def set_lr(self):
current_lr_G = adjust_learning_rate(self.cfg.OPTIMIZER.LEARNING_RATE_G,
self.c_iter, self.cfg.TRAIN.MAX_ITERS,
self.cfg.TRAIN.PREHEAT_STEPS)
for index in range(len(self.G_optim.param_groups)):
self.G_optim.param_groups[index]['lr'] = current_lr_G
def fast_save_CP(self, checkpoint_file):
com.save_checkpoint(checkpoint_file,
self.net_G,
self.G_optim,
None,
self.c_iter)