PyTorch模型

小猫的世界

有没有办法，可以像model.summary()方法在Keras中所做的那样，在PyTorch中打印一个模型的摘要?

1. Model Summary:
   
2. ____________________________________________________________________________________________________
   
3. Layer (type)                     Output Shape          Param #     Connected to                     
   
4. ====================================================================================================
   
5. input_1 (InputLayer)             (None, 1, 15, 27)     0                                            
   
6. ____________________________________________________________________________________________________
   
7. convolution2d_1 (Convolution2D)  (None, 8, 15, 27)     872         input_1[0][0]                    
   
8. ____________________________________________________________________________________________________
   
9. maxpooling2d_1 (MaxPooling2D)    (None, 8, 7, 27)      0           convolution2d_1[0][0]            
   
10. ____________________________________________________________________________________________________
    
11. flatten_1 (Flatten)              (None, 1512)          0           maxpooling2d_1[0][0]            
    
12. ____________________________________________________________________________________________________
    
13. dense_1 (Dense)                  (None, 1)             1513        flatten_1[0][0]                  
    
14. ====================================================================================================
    
15. Total params: 2,385
    
16. Trainable params: 2,385
    
17. Non-trainable params: 0

复制代码

神龙教

在Keras的模型中，不会得到关于模型的详细信息。简单地打印模型可以让对涉及的不同层及其规范有一些了解。比如说：

1. from torchvision import models
   
2. model = models.vgg16()
   
3. print(model)

复制代码

输出如下：

1. VGG (
   
2.   (features): Sequential (
   
3.     (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
   
4.     (1): ReLU (inplace)
   
5.     (2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
   
6.     (3): ReLU (inplace)
   
7.     (4): MaxPool2d (size=(2, 2), stride=(2, 2), dilation=(1, 1))
   
8.     (5): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
   
9.     (6): ReLU (inplace)
   
10.     (7): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    
11.     (8): ReLU (inplace)
    
12.     (9): MaxPool2d (size=(2, 2), stride=(2, 2), dilation=(1, 1))
    
13.     (10): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    
14.     (11): ReLU (inplace)
    
15.     (12): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    
16.     (13): ReLU (inplace)
    
17.     (14): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    
18.     (15): ReLU (inplace)
    
19.     (16): MaxPool2d (size=(2, 2), stride=(2, 2), dilation=(1, 1))
    
20.     (17): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    
21.     (18): ReLU (inplace)
    
22.     (19): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    
23.     (20): ReLU (inplace)
    
24.     (21): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    
25.     (22): ReLU (inplace)
    
26.     (23): MaxPool2d (size=(2, 2), stride=(2, 2), dilation=(1, 1))
    
27.     (24): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    
28.     (25): ReLU (inplace)
    
29.     (26): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    
30.     (27): ReLU (inplace)
    
31.     (28): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    
32.     (29): ReLU (inplace)
    
33.     (30): MaxPool2d (size=(2, 2), stride=(2, 2), dilation=(1, 1))
    
34.   )
    
35.   (classifier): Sequential (
    
36.     (0): Dropout (p = 0.5)
    
37.     (1): Linear (25088 -> 4096)
    
38.     (2): ReLU (inplace)
    
39.     (3): Dropout (p = 0.5)
    
40.     (4): Linear (4096 -> 4096)
    
41.     (5): ReLU (inplace)
    
42.     (6): Linear (4096 -> 1000)
    
43.   )
    
44. )

复制代码