pytorch学习笔记十一：查看模型的层和参数信息的几种方式

首先，我们先创建一个只有两层的简单的模型，第一层是没有参数的FlattenLayer,第二层是有参数的Linear.

import torch
from torch import nn

num_inputs = 784
num_outputs = 10

class FlattenLayer(nn.Module):
    def __init__(self):
        super(FlattenLayer, self).__init__()
    def forward(self, x): # x shape: (batch, *, *, ...)
        return x.view(x.shape[0], -1)

from collections import OrderedDict
net = nn.Sequential(
        # FlattenLayer(),
        # nn.Linear(num_inputs, num_outputs)
        OrderedDict([
          ('flatten', FlattenLayer()),
          ('linear', nn.Linear(num_inputs, num_outputs))])
        )

一、.state_dict()

print(net.state_dict())
for name in net.state_dict():
    print(name)

输出：

二、.named_parameters()

print(list(net.named_parameters()))
for name,param in net.named_parameters():
    print(name)
    print(param)

如果想要只取出numpy数据，我们不能直接使用param.numpy()，应该使用param.detach().numpy()

我们还可以通过下面这种方式取出参数名和数值：

params = list(net.named_parameters())
print(params[0][0])#name
print(params[0][1].data)#data

三、.modules()

print(list(net.modules()))
for layer in net.modules():
    print(layer)

输出：

[Sequential(
  (flatten): FlattenLayer()
  (linear): Linear(in_features=784, out_features=10, bias=True)
), FlattenLayer(), Linear(in_features=784, out_features=10, bias=True)]
Sequential(
  (flatten): FlattenLayer()
  (linear): Linear(in_features=784, out_features=10, bias=True)
)
FlattenLayer()
Linear(in_features=784, out_features=10, bias=True)

四、._modules

print(net._modules)

print(net._modules.items())

for name,layer in net._modules.items():
    print(name)
    print(layer)

输出：

OrderedDict([('flatten', FlattenLayer()), 
('linear', Linear(in_features=784, out_features=10, bias=True))])

odict_items([('flatten', FlattenLayer()), 
('linear', Linear(in_features=784, out_features=10, bias=True))])

flatten
FlattenLayer()

linear
Linear(in_features=784, out_features=10, bias=True)