Pytorch中的model.named_parameters()和model.parameters()

之前一直不清楚怎么查看模型的参数和结构，现在学习了一下。

首先搞个resnet20出来

import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.nn import init
from models.res_utils import DownsampleA, DownsampleC, DownsampleD
import math


class ResNetBasicblock(nn.Module):
    expansion = 1
    """
    RexNet basicblock (https://github.com/facebook/fb.resnet.torch/blob/master/models/resnet.lua)
    """
    def __init__(self, inplanes, planes, stride=1, downsample=None):
        super(ResNetBasicblock, self).__init__()

        self.conv_a = nn.Conv2d(inplanes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
        self.bn_a = nn.BatchNorm2d(planes)

        self.conv_b = nn.Conv2d(planes, planes, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn_b = nn.BatchNorm2d(planes)

        self.downsample = downsample

    def forward(self, x):
        residual = x

        basicblock = self.conv_a(x)
        basicblock = self.bn_a(basicblock)
        basicblock = F.relu(basicblock, inplace=True)

        basicblock = self.conv_b(basicblock)
        basicblock = self.bn_b(basicblock)

        if self.downsample is not None:
            residual = self.downsample(x)
    
        return F.relu(residual + basicblock, inplace=True)

class CifarResNet(nn.Module):
    """
    ResNet optimized for the Cifar dataset, as specified in
    https://arxiv.org/abs/1512.03385.pdf
    """
    def __init__(self, block, depth, num_classes):
        """ Constructor
        Args:
          depth: number of layers.
          num_classes: number of classes
          base_width: base width
        """
        super(CifarResNet, self).__init__()

        #Model type specifies number of layers for CIFAR-10 and CIFAR-100 model
        assert (depth - 2) % 6 == 0, 'depth should be one of 20, 32, 44, 56, 110'
        layer_blocks = (depth - 2) // 6
        print ('CifarResNet : Depth : {} , Layers for each block : {}'.format(depth, layer_blocks))

        self.num_classes = num_classes

        self.conv_1_3x3 = nn.Conv2d(3, 16, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn_1 = nn.BatchNorm2d(16)

        self.inplanes = 16
        self.stage_1 = self._make_layer(block, 16, layer_blocks, 1)
        self.stage_2 = self._make_layer(block, 32, layer_blocks, 2)
        self.stage_3 = self._make_layer(block, 64, layer_blocks, 2)
        self.avgpool = nn.AvgPool2d(8)
        self.classifier = nn.Linear(64*block.expansion, num_classes)

        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
                m.weight.data.normal_(0, math.sqrt(2. / n))
                #m.bias.data.zero_()
            elif isinstance(m, nn.BatchNorm2d):
                m.weight.data.fill_(1)
                m.bias.data.zero_()
            elif isinstance(m, nn.Linear):
                init.kaiming_normal(m.weight)
                m.bias.data.zero_()

    def _make_layer(self, block, planes, blocks, stride=1):
        downsample = None
        if stride != 1 or self.inplanes != planes * block.expansion:
            downsample = DownsampleA(self.inplanes, planes * block.expansion, stride)

        layers = []
        layers.append(block(self.inplanes, planes, stride, downsample))
        self.inplanes = planes * block.expansion
        for i in range(1, blocks):
            layers.append(block(self.inplanes, planes))

        return nn.Sequential(*layers)

    def forward(self, x):
        x = self.conv_1_3x3(x)
        x = F.relu(self.bn_1(x), inplace=True)
        x = self.stage_1(x)
        x = self.stage_2(x)
        x = self.stage_3(x)
        x = self.avgpool(x)
        x = x.view(x.size(0), -1)
        return self.classifier(x)

def resnet20(num_classes=10):
    """Constructs a ResNet-20 model for CIFAR-10 (by default)
    Args:
    num_classes (uint): number of classes
    """
    model = CifarResNet(ResNetBasicblock, 20, num_classes)
    return model

DownsampleA其实是这个东西

class DownsampleA(nn.Module):  

  def __init__(self, nIn, nOut, stride):
    super(DownsampleA, self).__init__() 
    assert stride == 2    
    self.avg = nn.AvgPool2d(kernel_size=1, stride=stride)   

  def forward(self, x):   
    x = self.avg(x)  
    return torch.cat((x, x.mul(0)), 1)  

所以最后网络结构是预处理的conv层和bn层，以及接下去的三个stage，每个stage分别是三层，最后是avgpool和全连接层

1、model.named_parameters()，迭代打印model.named_parameters()将会打印每一次迭代元素的名字和param

for name, param in net.named_parameters():
    print(name,param.requires_grad)
    param.requires_grad = False


#
conv_1_3x3.weight False
bn_1.weight False
bn_1.bias False
stage_1.0.conv_a.weight False
stage_1.0.bn_a.weight False
stage_1.0.bn_a.bias False
stage_1.0.conv_b.weight False
stage_1.0.bn_b.weight False
stage_1.0.bn_b.bias False
stage_1.1.conv_a.weight False
stage_1.1.bn_a.weight False
stage_1.1.bn_a.bias False
stage_1.1.conv_b.weight False
stage_1.1.bn_b.weight False
stage_1.1.bn_b.bias False
stage_1.2.conv_a.weight False
stage_1.2.bn_a.weight False
stage_1.2.bn_a.bias False
stage_1.2.conv_b.weight False
stage_1.2.bn_b.weight False
stage_1.2.bn_b.bias False
stage_2.0.conv_a.weight False
stage_2.0.bn_a.weight False
stage_2.0.bn_a.bias False
stage_2.0.conv_b.weight False
stage_2.0.bn_b.weight False
stage_2.0.bn_b.bias False
stage_2.1.conv_a.weight False
stage_2.1.bn_a.weight False
stage_2.1.bn_a.bias False
stage_2.1.conv_b.weight False
stage_2.1.bn_b.weight False
stage_2.1.bn_b.bias False
stage_2.2.conv_a.weight False
stage_2.2.bn_a.weight False
stage_2.2.bn_a.bias False
stage_2.2.conv_b.weight False
stage_2.2.bn_b.weight False
stage_2.2.bn_b.bias False
stage_3.0.conv_a.weight False
stage_3.0.bn_a.weight False
stage_3.0.bn_a.bias False
stage_3.0.conv_b.weight False
stage_3.0.bn_b.weight False
stage_3.0.bn_b.bias False
stage_3.1.conv_a.weight False
stage_3.1.bn_a.weight False
stage_3.1.bn_a.bias False
stage_3.1.conv_b.weight False
stage_3.1.bn_b.weight False
stage_3.1.bn_b.bias False
stage_3.2.conv_a.weight False
stage_3.2.bn_a.weight False
stage_3.2.bn_a.bias False
stage_3.2.conv_b.weight False
stage_3.2.bn_b.weight False
stage_3.2.bn_b.bias False
classifier.weight False
classifier.bias False

并且可以更改参数的可训练属性，第一次打印是True，这是第二次，就是False了

2、model.parameters()，迭代打印model.parameters()将会打印每一次迭代元素的param而不会打印名字，这是他和named_parameters的区别，两者都可以用来改变requires_grad的属性

for index, param in enumerate(net.parameters()):
    print(param.shape)

#
torch.Size([16, 3, 3, 3])
torch.Size([16])
torch.Size([16])
torch.Size([16, 16, 3, 3])
torch.Size([16])
torch.Size([16])
torch.Size([16, 16, 3, 3])
torch.Size([16])
torch.Size([16])
torch.Size([16, 16, 3, 3])
torch.Size([16])
torch.Size([16])
torch.Size([16, 16, 3, 3])
torch.Size([16])
torch.Size([16])
torch.Size([16, 16, 3, 3])
torch.Size([16])
torch.Size([16])
torch.Size([16, 16, 3, 3])
torch.Size([16])
torch.Size([16])
torch.Size([32, 16, 3, 3])
torch.Size([32])
torch.Size([32])
torch.Size([32, 32, 3, 3])
torch.Size([32])
torch.Size([32])
torch.Size([32, 32, 3, 3])
torch.Size([32])
torch.Size([32])
torch.Size([32, 32, 3, 3])
torch.Size([32])
torch.Size([32])
torch.Size([32, 32, 3, 3])
torch.Size([32])
torch.Size([32])
torch.Size([32, 32, 3, 3])
torch.Size([32])
torch.Size([32])
torch.Size([64, 32, 3, 3])
torch.Size([64])
torch.Size([64])
torch.Size([64, 64, 3, 3])
torch.Size([64])
torch.Size([64])
torch.Size([64, 64, 3, 3])
torch.Size([64])
torch.Size([64])
torch.Size([64, 64, 3, 3])
torch.Size([64])
torch.Size([64])
torch.Size([64, 64, 3, 3])
torch.Size([64])
torch.Size([64])
torch.Size([64, 64, 3, 3])
torch.Size([64])
torch.Size([64])
torch.Size([10, 64])
torch.Size([10])

可以看出这些参数的尺寸