-
Notifications
You must be signed in to change notification settings - Fork 40
/
Copy pathostia_trainner.py
191 lines (160 loc) · 7.32 KB
/
ostia_trainner.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
# -*- coding: UTF-8 -*-
# @Time : 14/05/2020 16:48
# @Author : BubblyYi
# @FileName: ostia_trainner.py
# @Software: PyCharm
import os
import matplotlib
matplotlib.use('AGG')
import matplotlib.pyplot as plt
import torch
from torch.utils.data import DataLoader
from time import time
import sys
from datetime import datetime
class Trainer(object):
def __init__(self, batch_size, num_workers, train_dataset, val_dataset, model, model_name, optimizer, criterion, save_num = 0,start_epoch=0, max_epoch=1000, initial_lr=0.01, checkpoint_path=None):
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
self.batch_size = batch_size
self.num_workers = num_workers
self.train_dataset = train_dataset
self.val_dataset = val_dataset
self.model = model
self.model_name = model_name
self.optimizer = optimizer
self.initial_lr = initial_lr
self.criterion = criterion
self.all_tr_loss = []
self.all_val_loss = []
self.all_tr_direction_loss = []
self.all_val_direction_loss = []
self.all_tr_radius_loss = []
self.all_val_radius_loss = []
self.all_tr_err = []
self.all_val_err = []
self.best_test_loss = 2**31
self.log_file = None
self.start_epoch = start_epoch
self.max_epoch = max_epoch
self.train_loader = DataLoader(self.train_dataset, batch_size=self.batch_size, shuffle=True, num_workers=self.num_workers)
self.val_loader = DataLoader(self.val_dataset, batch_size=self.batch_size, shuffle=True, num_workers=self.num_workers)
self.checkpoint_path = checkpoint_path
self.output_folder = "logs"
if not os.path.exists(self.output_folder):
os.makedirs(self.output_folder)
self.save_num = save_num
def train_step(self, epoch):
self.print_to_log_file("\nEpoch: ", epoch + 1)
self.model.train()
train_loss = 0.
total = 0
for idx, (inputs, labels) in enumerate(self.train_loader):
inputs, labels = inputs.to(self.device), labels.to(self.device)
outputs = self.model(inputs)
outputs = outputs.view((len(labels)))
loss = self.criterion(outputs.float(), labels.float())
self.optimizer.zero_grad()
loss.backward()
self.optimizer.step()
train_loss += loss.item()
total += labels.size(0)
print_str = "Train Loss:{:.5f} Total :{:}".format(train_loss / len(self.train_loader),total)
self.print_to_log_file(print_str)
return train_loss / len(self.train_loader)
def val_step(self, epoch):
self.model.eval()
test_loss = 0.
total = 0
if True:
for idx, (inputs, labels) in enumerate(self.val_loader):
inputs, labels = inputs.to(self.device), labels.to(self.device)
outputs = self.model(inputs)
outputs = outputs.view((len(labels)))
loss = self.criterion(outputs.float(), labels.float())
test_loss += loss.item()
total += labels.size(0)
print_str = "Val Loss:{:.5f} Total:{:}".format(test_loss / len(self.val_loader),
total)
self.print_to_log_file(print_str)
print("test loss",test_loss/len(self.val_loader))
print("best test loss", self.best_test_loss)
if (test_loss/len(self.val_loader)) < self.best_test_loss:
print("saving model")
self.best_test_loss = test_loss/len(self.val_loader)
save_fold = "../checkpoint/ostia_checkpoints/"
if not os.path.exists(save_fold):
os.makedirs(save_fold)
model_save_path = save_fold + "/" + self.model_name + "_model_s"+str(self.save_num)+".pkl"
self.save_best_checkpoint(model_save_path, test_loss, epoch)
print_str = "Saving parameters to " + model_save_path
self.print_to_log_file(print_str)
return test_loss / len(self.val_loader)
def poly_lr(self, epoch, max_epochs, initial_lr, exponent=0.9):
return initial_lr * (1 - epoch / max_epochs) ** exponent
def lr_decay(self, epoch, max_epochs, initial_lr):
for params in self.optimizer.param_groups:
params['lr'] = self.poly_lr(epoch, max_epochs, initial_lr, exponent=1.5)
lr = params['lr']
print_str = "Learning rate adjusted to {}".format(lr)
self.print_to_log_file(print_str)
def plot_progress(self, epoch):
x_epoch = list(range(len(self.all_tr_loss)))
plt.plot(x_epoch, self.all_tr_loss, color="b", linestyle="--", marker="*", label='train')
plt.plot(x_epoch, self.all_val_loss, color="r", linestyle="--", marker="*", label='val')
plt.legend()
plt.rcParams['savefig.dpi'] = 300
plt.rcParams['figure.dpi'] = 300
plt.savefig("Total_loss_ostia_model_s"+str(self.save_num)+".jpg")
plt.close()
def save_best_checkpoint(self, model_save_path, curr_loss, epoch):
checkpoint = {
'net_dict': self.model.state_dict(),
'curr_loss': curr_loss,
'epoch': epoch,
'optimizer_state_dict': self.optimizer.state_dict(),
'batch_size': self.batch_size,
'train_loss': self.all_tr_loss,
'val_loss': self.all_val_loss,
'initial_lr': self.initial_lr
}
torch.save(checkpoint, model_save_path)
def print_to_log_file(self, *args, also_print_to_console=True, add_timestamp=True):
timestamp = time()
dt_object = datetime.fromtimestamp(timestamp)
if add_timestamp:
args = ("%s:" % dt_object, *args)
if self.log_file is None:
if not os.path.isdir(self.output_folder):
os.mkdir(self.output_folder)
timestamp = datetime.now()
self.log_file = os.path.join(self.output_folder, "training_log_%d_%d_%d_%02.0d_%02.0d_%02.0d.txt" %
(timestamp.year, timestamp.month, timestamp.day, timestamp.hour, timestamp.minute,
timestamp.second))
with open(self.log_file, 'w') as f:
f.write("Starting... \n")
successful = False
max_attempts = 5
ctr = 0
while not successful and ctr < max_attempts:
try:
with open(self.log_file, 'a+') as f:
for a in args:
f.write(str(a))
f.write(" ")
f.write("\n")
successful = True
except IOError:
print("%s: failed to log: " % datetime.fromtimestamp(timestamp), sys.exc_info())
ctr += 1
if also_print_to_console:
print(*args)
def run_train(self):
print("Start training")
self.model.to(self.device)
for epoch in range(self.start_epoch, self.max_epoch):
train_loss = self.train_step(epoch)
val_loss = self.val_step(epoch)
self.all_tr_loss.append(train_loss)
self.all_val_loss.append(val_loss)
self.plot_progress(epoch)
self.lr_decay(epoch, self.max_epoch, self.initial_lr)