-
Notifications
You must be signed in to change notification settings - Fork 1
/
engine_for_finetuning.py
260 lines (220 loc) · 10.9 KB
/
engine_for_finetuning.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
255
256
257
258
259
260
import math
import sys
from typing import Iterable, Optional
import torch
import torch.nn.functional as F
from timm.data import Mixup
from timm.utils import accuracy, ModelEma
import utils
from consistency import consistency_loss
from third_party.clip_loss import CLIPLoss, DenoisingCLIPLoss, GaussianCLIPLoss
def train_class_batch(model, samples, target, criterion):
if isinstance(criterion, CLIPLoss):
outputs = model(samples, return_feature=True)
loss = criterion(outputs, target)
elif isinstance(criterion, DenoisingCLIPLoss):
outputs, features = model(samples, return_feature=True, return_output=True)
noisy_features, clean_features = features.chunk(2, dim=0)
_, outputs = outputs.chunk(2, dim=0)
loss = criterion(outputs, noisy_features, clean_features, target)
elif isinstance(criterion, GaussianCLIPLoss):
outputs, features = model(samples, return_feature=True, return_output=True)
view1, view2 = features.chunk(2, dim=0)
loss = criterion(outputs, view1, view2, target)
else:
outputs = model(samples)
loss = criterion(outputs, target)
return loss, outputs
def get_loss_scale_for_deepspeed(model):
optimizer = model.optimizer
return optimizer.loss_scale if hasattr(optimizer, "loss_scale") else optimizer.cur_scale
def train_one_epoch(model: torch.nn.Module, criterion: torch.nn.Module,
data_loader: Iterable, optimizer: torch.optim.Optimizer,
device: torch.device, epoch: int, loss_scaler, max_norm: float = 0,
model_ema: Optional[ModelEma] = None, model_emas = None,
mixup_fn: Optional[Mixup] = None, log_writer=None,
start_steps=None, lr_schedule_values=None, wd_schedule_values=None,
num_training_steps_per_epoch=None, update_freq=None,
post_transform=None, consistency_lbd: float = 0., con_loss='default',
jsd_lbd: float = 0., denoise=None):
model.train(True)
metric_logger = utils.MetricLogger(delimiter=" ")
metric_logger.add_meter('lr', utils.SmoothedValue(window_size=1, fmt='{value:.6f}'))
metric_logger.add_meter('min_lr', utils.SmoothedValue(window_size=1, fmt='{value:.6f}'))
header = 'Epoch: [{}]'.format(epoch)
if loss_scaler is None:
model.zero_grad()
model.micro_steps = 0
else:
optimizer.zero_grad()
print_freq = 100
for data_iter_step, (samples, targets) in enumerate(metric_logger.log_every(data_loader, print_freq, header)):
step = data_iter_step // update_freq
if step >= num_training_steps_per_epoch:
continue
it = start_steps + step # global training iteration
# Update LR & WD for the first acc
if lr_schedule_values is not None or wd_schedule_values is not None and data_iter_step % update_freq == 0:
for i, param_group in enumerate(optimizer.param_groups):
if lr_schedule_values is not None:
param_group["lr"] = lr_schedule_values[it] * param_group["lr_scale"]
if wd_schedule_values is not None and param_group["weight_decay"] > 0:
param_group["weight_decay"] = wd_schedule_values[it]
samples = samples.to(device, non_blocking=True)
targets = targets.to(device, non_blocking=True)
if denoise is not None:
with torch.no_grad():
with torch.cuda.amp.autocast():
samples = denoise(samples)
if mixup_fn is not None:
samples, targets = mixup_fn(samples, targets)
if consistency_lbd > 0:
samples = torch.cat([samples, samples])
targets = torch.cat([targets, targets])
if jsd_lbd > 0:
samples0 = samples
samples = torch.cat([samples, samples])
if isinstance(criterion, DenoisingCLIPLoss):
samples0 = samples
if isinstance(criterion, GaussianCLIPLoss):
samples = torch.cat([samples, samples])
targets = torch.cat([targets, targets])
with torch.no_grad():
if post_transform is not None:
samples = post_transform(samples)
if isinstance(criterion, DenoisingCLIPLoss):
for t in post_transform.transforms[1:]: # Ignore Gaussian noise
samples0 = t(samples0)
samples = torch.cat([samples, samples0])
if jsd_lbd > 0:
for t in post_transform.transforms[1:]: # Ignore Gaussian noise
samples0 = t(samples0)
samples1, samples2 = samples.chunk(2, dim=0)
lam1 = torch.rand(samples0.size(0), device=device)
lam2 = torch.rand(samples0.size(0), device=device)
samples1 = (1 - lam1).view(-1, 1, 1, 1) * samples0 + lam1.view(-1, 1, 1, 1) * samples1
samples2 = (1 - lam2).view(-1, 1, 1, 1) * samples0 + lam2.view(-1, 1, 1, 1) * samples2
samples = samples0
if loss_scaler is None:
samples = samples.half()
loss, output = train_class_batch(
model, samples, targets, criterion)
else:
with torch.cuda.amp.autocast():
loss, output = train_class_batch(
model, samples, targets, criterion)
if consistency_lbd > 0:
logits = torch.chunk(output, 2, dim=0)
loss_con = consistency_loss(logits, lbd=consistency_lbd, loss=con_loss)
loss = loss + loss_con
if jsd_lbd > 0:
with torch.cuda.amp.autocast():
output1 = model(samples1)
output2 = model(samples2)
p_clean, p_aug1, p_aug2 = torch.softmax(output, dim=1), \
torch.softmax(output1, dim=1), \
torch.softmax(output2, dim=1)
# Clamp mixture distribution to avoid exploding KL divergence
p_mixture = torch.clamp((p_clean + p_aug1 + p_aug2) / 3., 1e-7, 1).log()
loss_jsd = jsd_lbd * (F.kl_div(p_mixture, p_clean, reduction='batchmean') +
F.kl_div(p_mixture, p_aug1, reduction='batchmean') +
F.kl_div(p_mixture, p_aug2, reduction='batchmean')) / 3.
loss = loss + loss_jsd
loss_value = loss.item()
restart = torch.zeros(1).cuda()
if not math.isfinite(loss_value):
print("Loss is {}, stopping training".format(loss_value), force=True)
restart = torch.ones(1).cuda()
#sys.exit(1)
restarts = torch.sum(utils.all_gather_batch([restart])[0])
if restarts > 0:
sys.exit(1)
if loss_scaler is None:
loss /= update_freq
model.backward(loss)
model.step()
if (data_iter_step + 1) % update_freq == 0:
# model.zero_grad()
# Deepspeed will call step() & model.zero_grad() automatic
if model_ema is not None:
model_ema.update(model)
if model_emas is not None:
for model_ema in model_emas:
model_ema.update(model)
grad_norm = None
loss_scale_value = get_loss_scale_for_deepspeed(model)
else:
# this attribute is added by timm on one optimizer (adahessian)
is_second_order = hasattr(optimizer, 'is_second_order') and optimizer.is_second_order
loss /= update_freq
grad_norm = loss_scaler(loss, optimizer, clip_grad=max_norm,
parameters=model.parameters(), create_graph=is_second_order,
update_grad=(data_iter_step + 1) % update_freq == 0)
if (data_iter_step + 1) % update_freq == 0:
optimizer.zero_grad()
if model_ema is not None:
model_ema.update(model)
if model_emas is not None:
for model_ema in model_emas:
model_ema.update(model)
loss_scale_value = loss_scaler.state_dict()["scale"]
torch.cuda.synchronize()
if mixup_fn is None:
class_acc = (output.max(-1)[-1] == targets).float().mean()
else:
class_acc = None
metric_logger.update(loss=loss_value)
metric_logger.update(class_acc=class_acc)
metric_logger.update(loss_scale=loss_scale_value)
min_lr = 10.
max_lr = 0.
for group in optimizer.param_groups:
min_lr = min(min_lr, group["lr"])
max_lr = max(max_lr, group["lr"])
metric_logger.update(lr=max_lr)
metric_logger.update(min_lr=min_lr)
weight_decay_value = None
for group in optimizer.param_groups:
if group["weight_decay"] > 0:
weight_decay_value = group["weight_decay"]
metric_logger.update(weight_decay=weight_decay_value)
metric_logger.update(grad_norm=grad_norm)
if log_writer is not None:
log_writer.update(loss=loss_value, head="loss")
log_writer.update(class_acc=class_acc, head="loss")
log_writer.update(loss_scale=loss_scale_value, head="opt")
log_writer.update(lr=max_lr, head="opt")
log_writer.update(min_lr=min_lr, head="opt")
log_writer.update(weight_decay=weight_decay_value, head="opt")
log_writer.update(grad_norm=grad_norm, head="opt")
log_writer.set_step()
# gather the stats from all processes
metric_logger.synchronize_between_processes()
print("Averaged stats:", metric_logger)
return {k: meter.global_avg for k, meter in metric_logger.meters.items()}
@torch.no_grad()
def evaluate(data_loader, model, device):
criterion = torch.nn.CrossEntropyLoss()
metric_logger = utils.MetricLogger(delimiter=" ")
header = 'Test:'
# switch to evaluation mode
model.eval()
for batch in metric_logger.log_every(data_loader, 200, header):
images = batch[0]
target = batch[-1]
images = images.to(device, non_blocking=True)
target = target.to(device, non_blocking=True)
# compute output
with torch.cuda.amp.autocast():
output = model(images)
loss = criterion(output, target)
acc1, acc5 = accuracy(output, target, topk=(1, 5))
batch_size = images.shape[0]
metric_logger.update(loss=loss.item())
metric_logger.meters['acc1'].update(acc1.item(), n=batch_size)
metric_logger.meters['acc5'].update(acc5.item(), n=batch_size)
# gather the stats from all processes
metric_logger.synchronize_between_processes()
print('* Acc@1 {top1.global_avg:.3f} Acc@5 {top5.global_avg:.3f} loss {losses.global_avg:.3f}'
.format(top1=metric_logger.acc1, top5=metric_logger.acc5, losses=metric_logger.loss))
return {k: meter.global_avg for k, meter in metric_logger.meters.items()}