-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathconfident_classifier.py
252 lines (211 loc) · 9.33 KB
/
confident_classifier.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
from copy import deepcopy
import logging
from typing import Any, Dict
from pytorch_lightning import LightningModule
from pytorch_lightning.utilities.types import STEP_OUTPUT
import torch
from torch import optim
import torch.nn as nn
import torch.nn.functional as tf
from cls_models import set_model_to_mode
from cls_models.base import BaseClassifier
from datasets import load_data
from eval_ood_detection import eval_classifier
class ConfidentClassifier(LightningModule):
def __init__(
self,
classifier: BaseClassifier,
generator: nn.Module,
discriminator: nn.Module,
args: Dict,
dataset: Dict,
opt: Dict = {},
) -> None:
super().__init__()
self.classifier = classifier
self.generator = generator
self.discriminator = discriminator
self.args = args
self.dataset = dataset
self.opt = opt
self.save_hyperparameters(ignore=["classifier", "generator", "discriminator"])
self.class_crit = nn.NLLLoss()
self.kl_crit = nn.KLDivLoss()
self.gan_crit = nn.BCEWithLogitsLoss()
self.console_logger = logging.getLogger("root.ConfidentClassifier")
def on_save_checkpoint(self, checkpoint: Dict[str, Any]) -> None:
for module in ("classifier", "generator", "discriminator"):
if hasattr(getattr(self, module), "_hparams_name"):
checkpoint[f"{module}_hparams_name"] = getattr(
self, module
)._hparams_name
checkpoint[f"{module}_hyper_parameters"] = getattr(self, module).hparams
def configure_optimizers(self):
disc_opt = optim.Adam(
self.discriminator.parameters(),
lr=self.opt.get("lr_disc", self.opt.get("lr", 2e-4)),
weight_decay=self.opt.get(
"weight_decay_disc", self.opt.get("weight_decay", 2e-4)
),
betas=(0, 0.9),
)
disc_sched = optim.lr_scheduler.LinearLR(
disc_opt,
start_factor=1.0,
end_factor=self.opt.get("min_lr_disc", self.opt.get("min_lr", 1e-5))
/ self.opt.get("lr_disc", self.opt.get("lr", 2e-4)),
total_iters=self.trainer.estimated_stepping_batches,
)
cls_opt = optim.Adam(
self.classifier.parameters(),
lr=self.opt.get("lr_cls", self.opt.get("lr", 2e-4)),
weight_decay=self.opt.get(
"weight_decay_cls", self.opt.get("weight_decay", 2e-4)
),
betas=(0, 0.9),
)
cls_sched = optim.lr_scheduler.LinearLR(
cls_opt,
start_factor=1.0,
end_factor=self.opt.get("min_lr_cls", self.opt.get("min_lr", 1e-5))
/ self.opt.get("lr_cls", self.opt.get("lr", 2e-4)),
total_iters=self.trainer.estimated_stepping_batches,
)
gen_opt = optim.Adam(
self.generator.parameters(),
lr=self.opt.get("lr_gen", self.opt.get("lr", 2e-4)),
weight_decay=self.opt.get(
"weight_decay_gen", self.opt.get("weight_decay", 2e-4)
),
betas=(0, 0.9),
)
gen_sched = optim.lr_scheduler.LinearLR(
gen_opt,
start_factor=1.0,
end_factor=self.opt.get("min_lr_gen", self.opt.get("min_lr", 1e-5))
/ self.opt.get("lr_gen", self.opt.get("lr", 2e-4)),
total_iters=self.trainer.estimated_stepping_batches,
)
return [disc_opt, gen_opt, cls_opt], [disc_sched, gen_sched, cls_sched]
def training_step(self, batch, batch_idx, optimizer_idx) -> STEP_OUTPUT:
x, y = batch
if optimizer_idx == 0:
############################
# Train Discriminator
############################
self.generator.eval()
set_model_to_mode(self.classifier, "eval")
with torch.no_grad():
x_tilde = self.generator(num_samples=x.shape[0])
if "image_channels" in self.dataset["cfg"]:
x_tilde = torch.sigmoid(x_tilde)
disc_x_tilde = self.discriminator(x_tilde)
disc_x = self.discriminator(x)
disc_loss = self.gan_crit(
disc_x_tilde, torch.zeros_like(disc_x_tilde)
) + self.gan_crit(
disc_x, torch.ones_like(disc_x)
) # type: torch.Tensor
self.log("disc_loss", disc_loss.item(), on_epoch=True, on_step=False)
return disc_loss
elif optimizer_idx == 1:
############################
# Train Generator
############################
self.discriminator.eval()
set_model_to_mode(self.classifier, "eval")
x_tilde = self.generator(num_samples=x.shape[0])
if "image_channels" in self.dataset["cfg"]:
x_tilde = torch.sigmoid(x_tilde)
disc_x_tilde = self.discriminator(x_tilde)
class_x_tilde = self.classifier(x_tilde)[0]
class_loss_x_tilde = self.kl_crit(
tf.log_softmax(class_x_tilde, dim=1),
torch.full_like(class_x_tilde, 1.0 / class_x_tilde.shape[1]),
)
gen_loss = (
-self.gan_crit(disc_x_tilde, torch.zeros_like(disc_x_tilde))
+ self.args["beta"] * class_loss_x_tilde
)
self.log("gen_loss", gen_loss.item(), on_epoch=True, on_step=False)
return gen_loss
elif optimizer_idx == 2:
############################
# Train Classifier
############################
self.generator.eval()
self.discriminator.eval()
with torch.no_grad():
x_tilde = self.generator(num_samples=x.shape[0])
if "image_channels" in self.dataset["cfg"]:
x_tilde = torch.sigmoid(x_tilde)
class_x = self.classifier(x)[0]
class_x_tilde = self.classifier(x_tilde)[0]
class_loss_x_tilde = self.kl_crit(
tf.log_softmax(class_x_tilde, dim=1),
torch.full_like(class_x_tilde, 1.0 / class_x_tilde.shape[1]),
)
class_loss_x = self.class_crit(torch.log(class_x + 1e-16), y)
class_loss = class_loss_x + self.args["beta"] * class_loss_x_tilde
self.log("class_loss", class_loss.item(), on_epoch=True, on_step=False)
return class_loss
def validation_step(self, batch, batch_idx, **kwargs) -> None:
set_model_to_mode(self.classifier, "eval")
x, y = batch
y_hat = self.classifier(x)[0].argmax(1)
val_acc = (y_hat == y).float().mean().item()
self.log("val_acc", val_acc, on_epoch=True)
def on_train_epoch_end(self) -> None:
if getattr(self, "max_acc", 0) < self.trainer.logged_metrics.get(
"val_acc", 0
) and self.args.get("ood_datasets", ""):
self.console_logger.info("Evaluating OOD-Detection...")
self.max_acc = self.trainer.logged_metrics["val_acc"]
eval_dataset_config = deepcopy(self.dataset["cfg"])
eval_dataset_config["mode"] = "eval"
eval_dataset_config["static"] = True
indist_dataset = load_data(**eval_dataset_config)[0]
self.console_logger.debug(
f"{len(indist_dataset)} In-Dist Samples of "
f"{eval_dataset_config['name']}"
)
ood_datasets = []
if self.args["ood_datasets"] is None:
raise ValueError("No OOD dataset specified")
else:
for d in self.args["ood_datasets"].split(","):
ood_cfg = dict(
name=d,
mode=eval_dataset_config["mode"],
static=True,
image_channels=eval_dataset_config["image_channels"],
)
if "image_size" in eval_dataset_config:
ood_cfg["image_size"] = eval_dataset_config["image_size"]
ood_datasets.append(load_data(**ood_cfg)[0])
self.console_logger.debug(
f"{len(ood_datasets[-1])} OOD-Samples of '{d}'"
)
set_model_to_mode(self.classifier, "eval")
result = eval_classifier(
classifier=self.classifier,
indist_dataset=indist_dataset,
ood_datasets=ood_datasets,
args=self.args,
log=self.console_logger,
)[0]
for ood_index, ood_name in enumerate(self.args["ood_datasets"].split(",")):
for metric_index, metric_name in enumerate(
["auroc", "aupr-in", "aupr-out", "fpr95tpr"]
):
self.log(
f"{metric_name}-{ood_name}", result[ood_index, metric_index]
)
self.log("auroc-all", result[-2, 0])
self.log("aupr-in-all", result[-2, 1])
self.log("aupr-out-all", result[-2, 2])
self.log("fpr95tpr-all", result[-2, 3])
self.log("auroc-succ/fail", result[-1, 0])
self.log("aupr-s-succ/fail", result[-1, 1])
self.log("aupr-f-succ/fail", result[-1, 2])
self.log("fpr95tpr-succ/fail", result[-1, 3])