-
Notifications
You must be signed in to change notification settings - Fork 1
/
train_and_test.py
210 lines (189 loc) · 7.71 KB
/
train_and_test.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
import matplotlib.pyplot as plt
import numpy as np
import paddle
from tqdm import tqdm
def train(net, optimizer, epochs, batch_size, train_loader, val_loader, loss_function, save_path):
# 设置优化器
optim = optimizer
# 设置损失函数
loss_fn = loss_function
train_acc_list = []
train_loss_list = []
train_predicts_list = []
train_recall_list = []
val_acc_list = []
val_loss_list = []
val_predicts_list = []
val_recall_list = []
now_acc = 0
for epoch in range(epochs):
TP = 0
FN = 0
FP = 0
TN = 0
train_loss = 0
train_acc = 0
train_predicts = 0
train_recall = 0
train_nums = 0
train_loop = tqdm(train_loader, desc=f"train epoch:{epoch + 1}/{epochs}")
for idx, data in enumerate(train_loop):
x_data = data[0] # 训练数据
y_data = data[1] # 训练数据标签
predicts = net(x_data) # 预测结果
# 计算损失 等价于 prepare 中loss的设置
loss = loss_fn(predicts, y_data)
predicts_label = predicts.argmax(axis=1)
# 计算混淆矩阵
tp, fn, fp, tn = get_confusion_matrix(y_data, predicts_label)
TP += tp
FN += fn
FP += fp
TN += tn
train_loss += loss.numpy().item()
if (TP + TN) == 0:
train_acc = 0
else:
train_acc = (TP + TN) / (TP + TN + FP + FN)
if TP == 0:
train_predicts = 0
train_recall = 0
else:
train_predicts = TP / (TP + FP)
train_recall = TP / (TP + FN)
# 下面的反向传播、打印训练信息、更新参数、梯度清零都被封装到 Model.fit() 中
# 反向传播
loss.backward()
# 更新参数
optim.step()
# 梯度清零
optim.clear_grad()
train_nums += 1
train_loop.set_postfix(
{'loss': train_loss / train_nums, 'acc': train_acc, 'predicts': train_predicts, 'recall': train_recall})
# 保存训练信息
train_acc_list.append(train_acc)
train_loss_list.append(train_loss / train_nums)
train_predicts_list.append(train_predicts)
train_recall_list.append(train_recall)
val_loss = 0
val_acc = 0
val_predicts = 0
val_recall = 0
TP = 0
FN = 0
FP = 0
TN = 0
val_loop = tqdm(val_loader, desc='')
# 禁用动态图梯度计算
net.eval()
val_nums = 0
for idx, data in enumerate(val_loop):
x_data = data[0] # 测试数据
y_data = data[1] # 测试数据标签
predicts = net(x_data) # 预测结果
# 计算损失与精度
loss = loss_fn(predicts, y_data)
predicts_label = predicts.argmax(axis=1)
# 计算混淆矩阵
tp, fn, fp, tn = get_confusion_matrix(y_data, predicts_label)
TP += tp
FN += fn
FP += fp
TN += tn
val_loss += loss.numpy().item()
if (TP + TN) == 0:
val_acc = 0
else:
val_acc = (TP + TN) / (TP + TN + FP + FN)
if TP == 0:
val_predicts = 0
train_recall = 0
else:
val_predicts = TP / (TP + FP)
val_recall = TP / (TP + FN)
val_nums += 1
val_loop.set_postfix(
{'loss': val_loss / val_nums, 'acc': val_acc, 'predicts': val_predicts, 'recall': val_recall})
val_loss_list.append(val_loss / val_nums)
val_acc_list.append(val_acc)
val_predicts_list.append(val_predicts)
val_recall_list.append(val_recall)
# 保存模型
if now_acc < val_acc:
now_acc = val_acc
paddle.save(net.state_dict(),
save_path + "model/" + "acc_" + str(now_acc) + "_epoch_" + str(epoch + 1) + ".pdparams")
plt.title("acc") # 设置图形标题
plt.plot(range(len(train_acc_list)), train_acc_list, color='blue', linewidth=2.0,
label='train_acc') # 绘制曲线，设置颜色、线宽和线型
plt.plot(range(len(val_acc_list)), val_acc_list, color='red', linewidth=2.0, label='val_acc') # 绘制曲线，设置颜色、线宽和线型
plt.legend(loc=0, frameon=True, facecolor='white')
plt.savefig(save_path + "acc.png")
plt.show() # 显示图像
plt.title("loss") # 设置图形标题
plt.plot(range(len(train_loss_list)), train_loss_list, color='blue', linewidth=2.0,
label='train_loss') # 绘制曲线，设置颜色、线宽和线型
plt.plot(range(len(val_loss_list)), val_loss_list, color='red', linewidth=2.0, label='val_loss') # 绘制曲线，设置颜色、线宽和线型
plt.legend(loc=0, frameon=True, facecolor='white')
plt.savefig(save_path + "loss.png")
plt.show() # 显示图像
plt.title("predicts") # 设置图形标题
plt.plot(range(len(train_predicts_list)), train_predicts_list, color='blue', linewidth=2.0,
label='train_predicts') # 绘制曲线，设置颜色、线宽和线型
plt.plot(range(len(val_predicts_list)), val_predicts_list, color='red', linewidth=2.0, label='val_predicts')
plt.legend(loc=0, frameon=True, facecolor='white')
plt.savefig(save_path + "predicts.png")
plt.show() # 显示图像
plt.title("recall") # 设置图形标题
plt.plot(range(len(train_recall_list)), train_recall_list, color='blue', linewidth=2.0,
label='train_recall') # 绘制曲线，设置颜色、线宽和线型
plt.plot(range(len(val_recall_list)), val_recall_list, color='red', linewidth=2.0,
label='val_recall') # 绘制曲线，设置颜色、线宽和线型
plt.legend(loc=0, frameon=True, facecolor='white')
plt.savefig(save_path + "recall.png")
plt.show() # 显示图像
# 保存训练信息
np.savetxt(save_path + 'train_acc.txt', train_acc_list)
np.savetxt(save_path + 'train_loss.txt', train_loss_list)
np.savetxt(save_path + 'val_acc.txt', val_acc_list)
np.savetxt(save_path + 'val_loss.txt', val_loss_list)
np.savetxt(save_path + 'train_predicts.txt', train_predicts_list)
np.savetxt(save_path + 'train_recall.txt', train_recall_list)
np.savetxt(save_path + 'val_predicts.txt', val_predicts_list)
np.savetxt(save_path + 'val_recall.txt', val_recall_list)
def get_confusion_matrix(true_labels, pred_labels, num_classes=2):
cm = [[0 for _ in range(num_classes)] for _ in range(num_classes)]
for true, pred in zip(true_labels, pred_labels):
cm[true][pred] += 1
return cm[0][0], cm[0][1], cm[1][0], cm[1][1]
def test(dataloader, model):
TP = 0
FN = 0
FP = 0
TN = 0
test_acc = 0
test_predicts = 0
test_recall = 0
for batch_id, data in enumerate(dataloader):
x_data = data[0] # 测试数据
y_data = data[1] # 测试数据标签
predicts = model(x_data) # 预测结果
predicts_label = predicts.argmax(axis=1)
# 计算混淆矩阵
tp, fn, fp, tn = get_confusion_matrix(y_data, predicts_label)
TP += tp
FN += fn
FP += fp
TN += tn
if (TP + TN) == 0:
test_acc = 0
else:
test_acc = (TP + TN) / (TP + TN + FP + FN)
if TP == 0:
test_predicts = 0
test_recall = 0
else:
test_predicts = TP / (TP + FP)
test_recall = TP / (TP + FN)
return test_acc, test_predicts, test_recall, 2 * test_predicts * test_recall / (test_predicts + test_recall)