-
Notifications
You must be signed in to change notification settings - Fork 7
/
engine.py
238 lines (189 loc) · 11.1 KB
/
engine.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
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
"""
Train and eval functions used in main.py
"""
import math
import os
import sys
from typing import Iterable
import torch
import util.misc as utils
from torchvision.utils import save_image
from datasets.demo import denorm
import shutil
import pandas as pd
# ----------------------------------------------------------------------------------
def train_one_epoch_st(model: torch.nn.Module, criterion: torch.nn.Module,
data_loader: Iterable, optimizer: torch.optim.Optimizer,lr_scheduler,
device: torch.device,logger, epoch: int, save_path:str,args,max_norm: float = 0):
model.train()
criterion.train()
metric_logger = utils.MetricLogger(delimiter=" ")
metric_logger.add_meter('lr', utils.SmoothedValue(window_size=1, fmt='{value:.6f}'))
# metric_logger.add_meter('class_error', utils.SmoothedValue(window_size=1, fmt='{value:.2f}'))
header = 'Epoch: [{}]'.format(epoch)
print_freq = 100
for it,(samples,style_images, targets) in metric_logger.log_every(data_loader, print_freq,logger, header):
#
samples = samples.to(device)
style_images = style_images.to(device)
# targets = [{k: v.to(device) for k, v in t.items()} for t in targets]
outputs = model(samples,style_images)
loss_dict = criterion(outputs, samples,style_images)
weight_dict = criterion.weight_dict
losses = sum(loss_dict[k] * weight_dict[k] for k in loss_dict.keys() if k in weight_dict)
# reduce losses over all GPUs for logging purposes
loss_dict_reduced = utils.reduce_dict(loss_dict)
loss_dict_reduced_unscaled = {f'{k}_unscaled': v
for k, v in loss_dict_reduced.items()}
loss_dict_reduced_scaled = {k: v * weight_dict[k]
for k, v in loss_dict_reduced.items() if k in weight_dict}
losses_reduced_scaled = sum(loss_dict_reduced_scaled.values())
loss_value = losses_reduced_scaled.item()
if not math.isfinite(loss_value):
logger.info("Loss is {}, stopping training".format(loss_value))
logger.info(loss_dict_reduced)
sys.exit(1)
if it % 2000==0:
if not os.path.exists(os.path.join(save_path,"train_outputs")):
os.makedirs(os.path.join(save_path,"train_outputs"))
if not os.path.exists(os.path.join(save_path,"train_content_images")):
os.makedirs(os.path.join(save_path,"train_content_images"))
if not os.path.exists(os.path.join(save_path,"train_style_images")):
os.makedirs(os.path.join(save_path,"train_style_images"))
if isinstance(outputs, tuple):
outputs,_=outputs
outputs=denorm(outputs, device)
samples.tensors=denorm(samples.tensors, device)
style_images.tensors=denorm(style_images.tensors, device)
save_image(outputs,os.path.join(save_path,"train_outputs",f'{epoch:04}_{it:08}.png') )
save_image(samples.tensors,os.path.join(save_path,"train_content_images",f'{epoch:04}_{it:08}.png' ) )
save_image(style_images.tensors,os.path.join(save_path,"train_style_images",f'{epoch:04}_{it:08}.png' ) ) # .clamp(0,1)
optimizer.zero_grad()
losses.backward()
if max_norm > 0:
torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm)
optimizer.step()
metric_logger.update(loss=loss_value, **loss_dict_reduced_scaled, **loss_dict_reduced_unscaled)
# metric_logger.update(class_error=loss_dict_reduced['class_error'])
metric_logger.update(lr=optimizer.param_groups[0]["lr"])
if it % 4000==0:
if not os.path.exists(os.path.join(save_path,"checkpoint")):
os.makedirs(os.path.join(save_path,"checkpoint"))
checkpoint_path = os.path.join(save_path,"checkpoint",f'checkpoint{epoch:04}.pth')
utils.save_on_master({
'model': model.state_dict(),
'optimizer': optimizer.state_dict(),
'lr_scheduler': lr_scheduler.state_dict(),
'epoch': epoch,
'args': args,
}, checkpoint_path)
torch.cuda.empty_cache()
# gather the stats from all processes
metric_logger.synchronize_between_processes()
logger.info("Averaged stats:{}".format(metric_logger))
return {k: meter.global_avg for k, meter in metric_logger.meters.items()}
@torch.no_grad()
def evaluate_st(model, criterion, postprocessors, data_loader, base_ds, device, logger,epoch,save_path):
model.eval()
criterion.eval()
metric_logger = utils.MetricLogger(delimiter=" ")
header = 'Test:'
iou_types = tuple(k for k in ('segm', 'bbox') if k in postprocessors.keys())
for it,(samples,style_images, targets) in metric_logger.log_every(data_loader, 100,logger, header):
samples = samples.to(device)
style_images = style_images.to(device)
outputs = model(samples,style_images)
loss_dict = criterion(outputs, samples,style_images)
weight_dict = criterion.weight_dict
# reduce losses over all GPUs for logging purposes
loss_dict_reduced = utils.reduce_dict(loss_dict)
loss_dict_reduced_scaled = {k: v * weight_dict[k]
for k, v in loss_dict_reduced.items() if k in weight_dict}
loss_dict_reduced_unscaled = {f'{k}_unscaled': v
for k, v in loss_dict_reduced.items()}
metric_logger.update(loss=sum(loss_dict_reduced_scaled.values()),
**loss_dict_reduced_scaled,
**loss_dict_reduced_unscaled)
if it % 200==0:
if not os.path.exists(os.path.join(save_path,"eval_outputs")):
os.makedirs(os.path.join(save_path,"eval_outputs"))
if not os.path.exists(os.path.join(save_path,"eval_content_images")):
os.makedirs(os.path.join(save_path,"eval_content_images"))
if not os.path.exists(os.path.join(save_path,"eval_style_images")):
os.makedirs(os.path.join(save_path,"eval_style_images"))
if isinstance(outputs, tuple):
outputs,_=outputs
outputs=denorm(outputs, device)
samples.tensors=denorm(samples.tensors, device)
style_images.tensors=denorm(style_images.tensors, device)
save_image(outputs,os.path.join(save_path,"eval_outputs",f'{epoch:04}_{it:08}.png' ) )
save_image(samples.tensors,os.path.join(save_path,"eval_content_images",f'{epoch:04}_{it:08}.png' ) )
save_image(style_images.tensors,os.path.join(save_path,"eval_style_images",f'{epoch:04}_{it:08}.png' ) ) # .clamp(0,1)
# gather the stats from all processes
metric_logger.synchronize_between_processes()
print("Averaged stats:", metric_logger)
stats = {k: meter.global_avg for k, meter in metric_logger.meters.items()}
return stats
def save_batch_images(save_path,outputs,samples,style_images,epoch,it,device):
outputs=denorm(outputs, device)
samples.tensors=denorm(samples.tensors, device)
style_images.tensors=denorm(style_images.tensors, device)
# print("outputs.shape:",outputs.shape)
save_image(outputs,os.path.join(save_path,"test_outputs",f'{epoch:04}',f'{epoch:04}_{it:08}.png' ) )
save_image(samples.tensors,os.path.join(save_path,"test_content_images",f'{epoch:04}',f'{epoch:04}_{it:08}.png' ) )
save_image(style_images.tensors,os.path.join(save_path,"test_style_images",f'{epoch:04}',f'{epoch:04}_{it:08}.png' ) )
def test_st(model, criterion, postprocessors, data_loader, device, logger,epoch,save_path):
model.eval()
criterion.eval()
metric_logger = utils.MetricLogger(delimiter=" ")
# metric_logger.add_meter('class_error', utils.SmoothedValue(window_size=1, fmt='{value:.2f}'))
header = 'Test:'
iou_types = tuple(k for k in ('segm', 'bbox') if k in postprocessors.keys())
print("len(data_loader):",len(data_loader))
if os.path.exists(os.path.join(save_path,"test_outputs",f'{epoch:04}')):
shutil.rmtree(os.path.join(save_path,"test_outputs",f'{epoch:04}'))
os.makedirs(os.path.join(save_path,"test_outputs",f'{epoch:04}'))
tmp_out=[]
for it,(samples,style_images, targets) in metric_logger.log_every(data_loader, 100,logger, header):
# if it<=3:
# continue
samples = samples.to(device)
style_images = style_images.to(device)
# targets = [{k: v.to(device) for k, v in t.items()} for t in targets]
outputs = model(samples,style_images)
loss_dict = criterion(outputs, samples,style_images)
weight_dict = criterion.weight_dict
# reduce losses over all GPUs for logging purposes
loss_dict_reduced = utils.reduce_dict(loss_dict)
loss_dict_reduced_scaled = {k: v * weight_dict[k]
for k, v in loss_dict_reduced.items() if k in weight_dict}
loss_dict_reduced_unscaled = {f'{k}_unscaled': v
for k, v in loss_dict_reduced.items()}
metric_logger.update(loss=sum(loss_dict_reduced_scaled.values()),
**loss_dict_reduced_scaled,
**loss_dict_reduced_unscaled)
if it % 1==0:
if not os.path.exists(os.path.join(save_path,"test_content_images",f'{epoch:04}')):
os.makedirs(os.path.join(save_path,"test_content_images",f'{epoch:04}'))
if not os.path.exists(os.path.join(save_path,"test_style_images",f'{epoch:04}')):
os.makedirs(os.path.join(save_path,"test_style_images",f'{epoch:04}'))
if isinstance(outputs, tuple):
outputs,_=outputs
# if False:
if "content_image_name" in targets[0]:
for i in range(len(outputs)):
c_name=targets[i]["content_image_name"]
s_name=targets[i]["style_image_name"]
save_name="{}_{}".format(c_name,s_name)
output_i=denorm(outputs[i], device)
save_image(output_i,os.path.join(save_path,"test_outputs",f'{epoch:04}',f'{epoch:04}_{save_name}.png' ) )
sample_i=denorm(samples.tensors[i], device)
save_image(sample_i,os.path.join(save_path,"test_content_images",f'{epoch:04}',f'{epoch:04}_{save_name}.png' ) )
else:
save_batch_images(save_path,outputs,samples,style_images,epoch,it,device)
# gather the stats from all processes
metric_logger.synchronize_between_processes()
print("Averaged stats:", metric_logger)
stats = {k: meter.global_avg for k, meter in metric_logger.meters.items()}
return stats