-
Notifications
You must be signed in to change notification settings - Fork 0
/
fitter.py
196 lines (171 loc) · 6.45 KB
/
fitter.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
import datetime
from pathlib import Path
import torch
from torch.utils.data import DataLoader
import time
import os
import glob
from typing import Dict, Any, Callable
class Fitter:
def __init__(
self,
WORK_DIR: str,
INPUT_DIR: str,
model: torch.nn.Module,
device: torch.device,
n_epochs: int,
lr: float,
loss_fn: Any,
step_scheduler: bool,
validation_scheduler: bool,
scheduler_class: Callable,
scheduler_params: Dict[str, Any],
verbose: bool = True,
verbose_step: int = 10,
):
self.epoch: int = 0
self.WORK_DIR: str = WORK_DIR
self.INPUT_DIR: str = INPUT_DIR
self.start_time: str = datetime.datetime.now().isoformat()
self.log_path: str = f"{self.WORK_DIR}/output/{self.start_time}"
Path(self.log_path).mkdir(parents=True, exist_ok=True)
self.best_summary_loss: float = 10 ** 5
self.model: torch.nn.Module = model
self.device: torch.device = device
self.optimizer: torch.optim.Optimizer = torch.optim.AdamW(
self.model.parameters(), lr=lr
)
self.scheduler = scheduler_class(self.optimizer, **scheduler_params)
self.n_epochs: int = n_epochs
self.loss_fn = loss_fn
self.step_scheduler: bool = step_scheduler
self.validation_scheduler: bool = validation_scheduler
self.verbose: bool = verbose
self.verbose_step: int = verbose_step
self.log(f"Fitter prepared. Device is {self.device}")
def fit(self, train_loader: DataLoader, valid_loader: DataLoader) -> None:
for e in range(self.n_epochs):
if self.verbose:
lr = self.optimizer.param_groups[0]["lr"]
timestamp = datetime.datetime.now().isoformat()
self.log(f"\n{timestamp}\nLR: {lr}")
start = time.time()
summary_loss = self._train_one_epoch(train_loader)
self.log(
f"[RESULT]: Train. Epoch: {self.epoch}, summary_loss: {summary_loss.avg:.5f}, time: {(time.time() - start):.5f}"
)
self.save(f"{self.log_path}/last-checkpoint.bin")
start = time.time()
summary_loss = self._validation(valid_loader)
self.log(
f"[RESULT]: Val. Epoch: {self.epoch}, summary_loss: {summary_loss.avg:.5f}, time: {(time.time() - start):.5f}"
)
if summary_loss.avg < self.best_summary_loss:
self.best_summary_loss = summary_loss.avg
self.model.eval()
self.save(
f"{self.log_path}/best-checkpoint-{str(self.epoch).zfill(3)}epoch.bin"
)
for path in sorted(
glob.glob(f"{self.log_path}/best-checkpoint-*epoch.bin")
)[:-3]:
os.remove(path)
if self.validation_scheduler:
self.scheduler.step(metrics=summary_loss.avg)
self.epoch += 1
def _train_one_epoch(self, train_loader: DataLoader):
self.model.train()
summary_loss = self.loss_fn
start = time.time()
for step, (images, targets, iamge_ids) in enumerate(train_loader):
if self.verbose:
print(
f"Train Step {step}/{len(train_loader)}, "
+ f"summary_loss: {summary_loss.avg:.5f}, "
+ f"time: {(time.time() - start):.5f}",
end="\r",
)
images = torch.stack(images)
images = images.to(self.device).float()
batch_size = images.shape[0]
bboxes = [target["bboxes"].to(self.device) for target in targets]
labels = [target["labels"].to(self.device) for target in targets]
self.optimizer.zero_grad()
loss, _, _ = self.model(images, bboxes, labels)
loss.backward()
summary_loss.update(loss.detach().item(), batch_size)
self.optimizer.step()
if self.step_scheduler:
self.scheduler.step()
return summary_loss
def _validation(self, valid_loader: DataLoader):
self.model.eval()
summary_loss = self.loss_fn
start = time.time()
for step, (images, targets, image_ids) in enumerate(valid_loader):
if self.verbose:
if step % self.verbose_step == 0:
print(
f"Val Step {step}/{len(valid_loader)}, "
+ f"summary_loss: {summary_loss.avg:.5f}, "
+ f"time: {(time.time() - start):.5f}",
end="\r",
)
with torch.no_grad():
images = torch.stack(images)
batch_size = images.shape[0]
images = images.to(self.device).float()
boxes = [target["boxes"].to(self.device).float() for target in targets]
labels = [
target["labels"].to(self.device).float() for target in targets
]
loss, _, _ = self.model(images, boxes, labels)
summary_loss.update(loss.detach().item(), batch_size)
return summary_loss
def save(self, path: str):
self.model.eval()
torch.save(
{
"model_state_dict": self.model.model.state_dict(),
"optimizer_state_dict": self.optimizer.state_dict(),
"scheduler_state_dict": self.scheduler.state_dict(),
"best_summary_loss": self.best_summary_loss,
"epoch": self.epoch,
},
path,
)
def log(self, message: str):
if self.verbose:
print(message)
with open(f"{self.log_path}/log.txt", "a+") as logger:
logger.write(f"{message}\n")
def get_fitter(
WORK_DIR: str,
INPUT_DIR: str,
model: torch.nn.Module,
device: torch.device,
n_epochs: int,
lr: float,
loss_fn: Any,
step_scheduler: bool,
validation_scheduler: bool,
scheduler_class: Any,
scheduler_params: Dict[str, Any],
verbose: bool = True,
verbose_step: int = 10,
) -> Fitter:
return Fitter(
WORK_DIR,
INPUT_DIR,
model,
device,
n_epochs,
lr,
loss_fn,
step_scheduler,
validation_scheduler,
scheduler_class,
scheduler_params,
verbose,
verbose_step,
)