-
Notifications
You must be signed in to change notification settings - Fork 0
/
trainer.py
193 lines (161 loc) · 5.17 KB
/
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
import os
from typing import Callable, Optional
from torch import nn
import numpy as np
import torch
import torch.nn.parallel
import torch.utils.data.distributed
from torch.utils.data import DataLoader
from torch.optim import Optimizer
from monai.data import decollate_batch
import visualize
from utils import utils
from logger import WandBLogger
def train_epoch(
model: nn.Module,
loader: DataLoader,
optimizer: Optimizer,
epoch: int,
loss_func: Callable,
logger: WandBLogger,
device: str or torch.device,
) -> float or torch.Tensor:
model.train()
epoch_loss = 0.
for idx, batch_data in enumerate(loader):
data, target = batch_data["image"].to(device), batch_data["label"].to(device)
utils.zero_grad(model)
logits = model(data)
loss = loss_func(logits, target)
loss.backward()
optimizer.step()
loss_value = loss.item()
epoch_loss += loss_value / len(loader)
print(
"Epoch {} {}/{}".format(epoch, idx, len(loader)),
"loss: {:.4f}".format(loss_value),
)
logger("train/loss", loss_value)
utils.zero_grad(model)
return epoch_loss
def val_epoch(
model: nn.Module,
loader: DataLoader,
epoch: int,
acc_func: Callable,
logger: WandBLogger,
device: str or torch.device,
model_inferer=None,
post_label=None,
post_pred=None
):
if model_inferer is None:
model_inferer = lambda x: model(x)
model.eval()
with torch.no_grad():
avg_acc = 0.
images = []
for idx, batch_data in enumerate(loader):
data, target = batch_data["image"].to(device), batch_data["label"].to(device)
logits = model_inferer(data)
if not logits.is_cuda:
target = target.cpu()
val_labels_list = decollate_batch(target)
if post_label is not None:
val_labels_convert = [post_label(val_label_tensor) for val_label_tensor in val_labels_list]
val_outputs_list = decollate_batch(logits)
if post_pred is not None:
val_output_convert = [post_pred(val_pred_tensor) for val_pred_tensor in val_outputs_list]
acc = acc_func(y_pred=val_output_convert, y=val_labels_convert)
acc_list = acc.detach().cpu().numpy()
val_acc = np.mean([np.nanmean(l) for l in acc_list])
avg_acc += val_acc / len(loader)
if idx < 5:
image = visualize.create_image_visual(
source=data.cpu().numpy()[0, 0],
target=target.cpu().numpy().squeeze(),
output=logits.cpu().numpy().argmax(1).squeeze(),
)
images.append(image)
logger("val/acc", avg_acc)
acc_by_class = acc_func.aggregate()[0].mean(0)
logger.log_val_dice(acc_by_class)
logger.log_image(np.vstack(images), "val")
print(
"validation {}".format(epoch),
"acc/val",
avg_acc,
)
print(acc_by_class)
def save_checkpoint(
model: nn.Module,
epoch: int,
log_dir: str,
optimizer: Optional[Optimizer] = None,
scheduler=None,
):
state_dict = model.state_dict()
save_dict = {"epoch": epoch, "state_dict": state_dict}
snapshot_path = os.path.join(log_dir, 'snapshots')
if not os.path.exists(snapshot_path):
os.makedirs(snapshot_path)
filename = os.path.join(snapshot_path, f'chk_{epoch}.pt')
torch.save(save_dict, filename)
print("Saving checkpoint", filename)
def run_training(
config: dict,
model: nn.Module,
train_loader: DataLoader,
val_loader: DataLoader,
optimizer: Optimizer,
loss_func: Callable,
acc_func: Callable,
log_dir: str,
device: str or torch.device,
val_every: int,
save_every: int,
model_inferer=None,
scheduler=None,
start_epoch: int = 0,
post_label=None,
post_pred=None,
):
logger = WandBLogger(
config=config,
model=model,
)
epoch = start_epoch
while True:
print("Epoch:", epoch)
train_loss = train_epoch(
model,
train_loader,
optimizer,
epoch=epoch,
loss_func=loss_func,
device=device,
logger=logger,
)
print(
"training {}".format(epoch),
"loss: {:.4f}".format(train_loss),
)
if (epoch + 1) % val_every == 0:
val_epoch(
model,
val_loader,
epoch=epoch,
acc_func=acc_func,
device=device,
logger=logger,
model_inferer=model_inferer,
post_label=post_label,
post_pred=post_pred,
)
if (epoch + 1) % save_every == 0:
save_checkpoint(
model, epoch, log_dir, optimizer=optimizer, scheduler=scheduler
)
if scheduler is not None:
scheduler.step()
epoch += 1