named_parameters function

• Returns... As an iterator model All the parameters in , The return value is one Dictionaries ： contain Name of parameter and Value size ;
• Used in internal implementation A recursive algorithm , So for nested network parameters , Meeting Recursive traversal , Output Bottom layer parameters , See the following example ;

named_children() function

• This function is used to output the data in the network Layer 1 module name and instance object , Only the top module name will be displayed ;

named_modules() function

• This function is used to recursively output the data in the network Module name and instance object of each layer , The first mock exam will display the name of each module in all layers.

Example

import torch
import torch.nn as nn

class TestModel(nn.Module):
    def __init__(self):
        super(TestModel,self).__init__()
        #  Conventional convolution layer , The input channel is 3, The output channel is 12
        self.conv1 = nn.Conv2d(3, 12, kernel_size=3, padding=1)

        self.layer1 = nn.Sequential(
            nn.Conv2d(12, 6, kernel_size=3, padding=1),
            nn.Conv2d(6, 6, kernel_size=3, padding=1),
            nn.ReLU(inplace=True)
        )

    def forward(self, x):
        x = self.conv1(x)
        x = self.layer1(x)
 
model = TestModel()

named_parameters = model.named_parameters()
print('------------parameters-----------------')
for name, parameter in named_parameters:
    print(name)

named_children = model.named_children()
print('------------parameters-----------------')
for name, children in named_children:
    print(name)

named_modules = model.named_modules()
print('------------modules-----------------')
for name, module in named_modules:
    print(name)


------------ Output --------------------------
------------parameters-----------------
conv1.weight
conv1.bias
layer1.0.weight   #  You can see that for sequential The modules in the module are given a numeric number 
layer1.0.bias
layer1.1.weight
layer1.1.bias
------------parameters-----------------
conv1
layer1
------------modules-----------------
conv1
layer1
layer1.0
layer1.1
layer1.2