-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtrainer.py
148 lines (130 loc) · 6.65 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
import torch
from torch.utils.tensorboard import SummaryWriter
from torch.nn import functional as F
import numpy as np
from sklearn.metrics import roc_auc_score
from early_stopping import EarlyStopping
from matplotlib import pyplot as plt
from sklearn.model_selection import KFold,GroupShuffleSplit
from torch.utils.data import DataLoader, ConcatDataset
from models import EEGAutoencoder
from torch import nn
import random
import os
SEED =0
torch.manual_seed(SEED)
torch.cuda.manual_seed_all(SEED)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
np.random.seed(SEED)
random.seed(SEED)
os.environ['PYTHONHASHSEED'] = str(SEED)
class Trainer():
def __init__(self, num_epochs, experiment_name,
model, rloss, closs, optimizer, scheduler,device ='cuda:0',patience=15,seed=0):
self.num_epochs =num_epochs
self.writer =SummaryWriter(experiment_name)
self.model =model
self.criterion =rloss
self.cr_loss =closs
self.optimizer =optimizer
self.scheduler =scheduler
self.early_stopping = EarlyStopping(patience=patience, verbose=True)
self.device =device
self.seed =seed
self.set_seed()
def set_seed(self):
torch.manual_seed(self.seed)
torch.cuda.manual_seed_all(self.seed)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
np.random.seed(self.seed)
random.seed(self.seed)
os.environ['PYTHONHASHSEED'] = str(self.seed)
def plot_batch_and_outputs(self, batch, outputs, show_type='img'):
Y_ticks = np.linspace(0, 128, num=98)
X_ticks = np.linspace(0, 45, num=24)
random_channel = np.random.randint(batch.shape[1])
levels = 45
if show_type == 'img':
print('True: ')
spectrum = plt.imshow(batch[0, 0, random_channel, :, :].detach().cpu().numpy(), cmap='jet')
plt.show()
print('Predicted: ')
spectrum = plt.imshow(outputs[0, 0, random_channel, :, :].detach().cpu().numpy(), cmap='jet')
plt.show()
else:
spectrum = plt.contourf(Y_ticks, X_ticks, batch[0, 0, random_channel, :, :].detach().cpu().numpy(), levels,
cmap='jet')
plt.show()
print('Predicted: ')
spectrum = plt.contourf(Y_ticks, X_ticks, outputs[0, 0, random_channel, :, :].detach().cpu().numpy(),
levels, cmap='jet')
plt.show()
def __iner_loop__(self, loader, is_Train =True):
loss,reconstr_loss,correct,total,labels_all,softmax_preds = 0,0,0,0, [],[]
for data in loader:
img, label = data
img = img.float().to(self.device)
label = label.long().to(self.device)
codes, output, preds = self.model(img)
soft_preds = F.softmax(preds.data)
_, predicted = torch.max(soft_preds, 1)
correct += (predicted == label).sum().item()
total += label.size(0)
l2_loss = self.criterion(output, img)
cr_loss1 = self.cr_loss(preds, label)
loss = l2_loss + cr_loss1
if is_Train:
self.optimizer.zero_grad()
loss.backward()
self.optimizer.step()
self.scheduler.step()
loss += loss.data
reconstr_loss +=l2_loss.data
labels_all.append(label.cpu().detach().numpy())
softmax_preds.append(soft_preds.cpu().detach().numpy()[:, 1])
labels_all = np.concatenate(labels_all)
softmax_preds = np.concatenate(softmax_preds)
return loss,reconstr_loss,correct,total,labels_all,softmax_preds,img,output
def train(self, train_loader,val_loader,vebrose=False):
for epoch in range(self.num_epochs):
train_loss,train_reconstr_loss,train_correct,train_total,_,_,_,_ =self.__iner_loop__(train_loader, is_Train=True)
if epoch % 1 == 0:
with torch.no_grad():
val_loss,val_reconstr_loss, val_correct, val_total, labels_all, softmax_preds,img,output = self.__iner_loop__(val_loader, is_Train=False)
self.writer.add_scalar('training total loss',
train_loss / train_total,
epoch)
self.writer.add_scalar('training reconstraction loss',
train_reconstr_loss / val_total,
epoch)
self.writer.add_scalar('validation total loss',
val_loss / val_total,
epoch)
self.writer.add_scalar('validation reconstraction loss',
val_reconstr_loss / val_total,
epoch)
self.writer.add_scalar('train accuracy',
train_correct / train_total,
epoch)
self.writer.add_scalar('val accuracy',
val_correct / val_total,
epoch)
self.writer.add_scalar('val roc auc',
roc_auc_score(labels_all, softmax_preds),
epoch)
if vebrose:
print('epoch [{}/{}], train loss:{:.4f}'.format(epoch + 1, self.num_epochs, train_loss / train_total))
print('epoch [{}/{}], val loss:{:.4f}'.format(epoch + 1, self.num_epochs, val_loss / val_total))
print('epoch [{}/{}], train accuracy:{:.4f}'.format(epoch + 1, self.num_epochs,
train_correct / train_total))
print('epoch [{}/{}], val accuracy:{:.4f}'.format(epoch + 1, self.num_epochs, val_correct / val_total))
print('epoch [{}/{}], val roc auc:{:.4f}'.format(epoch + 1, self.num_epochs,
roc_auc_score(labels_all, softmax_preds)))
self.plot_batch_and_outputs(img, output, show_type='c')
self.early_stopping(val_correct / val_total,val_reconstr_loss/val_total, self.model)
if self.early_stopping.early_stop:
print("Early stopping")
break
return self.early_stopping.best_score, self.early_stopping.best_reconstr_score