1 import torch
2
3 # 1prepare dataset
4 # x,y是矩阵，3行1列 也就是说总共有3个数据，每个数据只有1个特征
5 x_data = torch.tensor([[1.0], [2.0], [3.0]])
6 y_data = torch.tensor([[2.0], [4.0], [6.0]])
7
8 # 2design model using class
9 """
10 our model class should be inherit from nn.Module, which is base class for all neural network modules.
11 member methods __init__() and forward() have to be implemented
12 class nn.linear contain two member Tensors: weight and bias
13 class nn.Linear has implemented the magic method __call__(),which enable the instance of the class can
14 be called just like a function.Normally the forward() will be called
15 """
16 class LinearModel(torch.nn.Module):
17 def __init__(self):
18 super(LinearModel, self).__init__()
19 # (1,1)是指输入x和输出y的特征维度，这里数据集中的x和y的特征都是1维的
20 # 该线性层需要学习的参数是w和b 获取w/b的方式分别是~linear.weight/linear.bias
21 self.linear = torch.nn.Linear(1, 1)
22
23 def forward(self, x):
24 y_pred = self.linear(x)
25 return y_pred
26
27 model = LinearModel()
28
29 # 3construct loss and optimizer
30 # criterion = torch.nn.MSELoss(size_average = False)
31 criterion = torch.nn.MSELoss(reduction='sum')
32 optimizer = torch.optim.SGD(model.parameters(), lr=0.01) # model.parameters()自动完成参数的初始化操作
33
34 # 4training cycle forward, backward, update
35 for epoch in range(100):
36 y_pred = model(x_data) # forward:predict
37 loss = criterion(y_pred, y_data) # forward: loss
38 print(epoch, loss.item())
39
40 optimizer.zero_grad() # the grad computer by .backward() will be accumulated. so before backward, remember set the grad to zero
41 loss.backward() # backward: autograd，自动计算梯度
42 optimizer.step() # update 参数，即更新w和b的值
43
44 print('w = ', model.linear.weight.item())
45 print('b = ', model.linear.bias.item())
46
47 x_test = torch.tensor([[4.0]])
48 y_test = model(x_test)
49 print('y_pred = ', y_test.data)
