本文将介绍如何用PyTorch构建模型

torch.nn.Module和torch.nn.Parameter

除了Parameter之外，本视频中讨论的所有类都是torch.nn.Module的子类。这是PyTorch基类，用于封装PyTorch模型及其组件的特定行为。

torch.nn.Module的一个重要行为是注册参数。如果特定的Module子类具有学习权值，则这些权值表示为torch.nn.Parameter的实例。Parameter类是torch的子类。张量，具有特殊的行为，当它们被分配为模块的属性时，它们被添加到该模块的参数列表中。这些参数可以通过Module类的parameters（）方法访问。

作为一个简单的例子，这里有一个非常简单的模型，有两个线性层和一个激活函数。我们将创建它的一个实例，并要求它报告其参数：

import torch

class TinyModel(torch.nn.Module):

    def __init__(self):
        super(TinyModel, self).__init__()

        self.linear1 = torch.nn.Linear(100, 200)
        self.activation = torch.nn.ReLU()
        self.linear2 = torch.nn.Linear(200, 10)
        self.softmax = torch.nn.Softmax()

    def forward(self, x):
        x = self.linear1(x)
        x = self.activation(x)
        x = self.linear2(x)
        x = self.softmax(x)
        return x

tinymodel = TinyModel()

print('The model:')
print(tinymodel)

print('\n\nJust one layer:')
print(tinymodel.linear2)

print('\n\nModel params:')
for param in tinymodel.parameters():
    print(param)

print('\n\nLayer params:')
for param in tinymodel.linear2.parameters():
    print(param)

输出为：

The model:
TinyModel(
  (linear1): Linear(in_features=100, out_features=200, bias=True)
  (activation): ReLU()
  (linear2): Linear(in_features=200, out_features=10, bias=True)
  (softmax): Softmax(dim=None)
)


Just one layer:
Linear(in_features=200, out_features=10, bias=True)


Model params:
Parameter containing:
tensor([[-0.0451,  0.0361,  0.0902,  ..., -0.0564, -0.0323,  0.0335],
        [ 0.0668,  0.0843,  0.0506,  ...,  0.0162,  0.0668, -0.0089],
        [-0.0505, -0.0148,  0.0485,  ...,  0.0714, -0.0399,  0.0798],
        ...,
        [-0.0639,  0.0345, -0.0766,  ...,  0.0711, -0.0354, -0.0719],
        [ 0.0827, -0.0614,  0.0078,  ...,  0.0531, -0.0672,  0.0158],
        [-0.0577, -0.0733, -0.0662,  ..., -0.0263, -0.0143, -0.0904]],
       requires_grad=True)
Parameter containing:
tensor([-1.8241e-02, -8.1554e-02,  3.1390e-02, -9.7299e-02, -3.9416e-02,
        -3.4526e-02,  6.9457e-02,  9.3126e-02,  8.3945e-02,  2.5128e-02,
        -1.9594e-02,  1.4253e-02,  7.5062e-02, -2.5254e-02,  2.5275e-02,
         3.6509e-02, -5.4355e-02,  5.2070e-02, -1.1055e-02,  6.3872e-02,
        -4.2867e-02, -6.9062e-02, -9.6398e-02,  6.0366e-02,  8.6856e-02,
        -4.3543e-02,  7.1326e-02,  3.6623e-03,  5.4014e-02, -1.3758e-02,
         4.6091e-02,  4.6796e-03, -2.9959e-02, -5.0925e-02,  1.9598e-02,
         5.6875e-03, -2.5505e-02,  9.8728e-02,  4.3602e-02,  3