爽歪歪666
不以物喜,不以己悲,努力才是永恒的主题。
 1 """
 2 torch.float64对应torch.DoubleTensor
 3 torch.float32对应torch.FloatTensor
 4 将真实函数的数据点能够拟合成一个多项式
 5 eg:y = 0.9 +0.5×x + 3×x*x + 2.4 ×x*x*x
 6 """
 7 import torch
 8 
 9 from torch import nn
10 
11 def make_features(x):
12     x = x.unsqueeze(1)#在原来的基础上扩充了一维
13     return torch.cat([x ** i for i in range(1,4)], 1)
14 
15 
16 def get_batch(batch_size=32):
17 
18     random = torch.randn(batch_size)
19     # print('random')
20     # print(random) #32个数
21 
22     x = make_features(random)#进行维度扩充,扩充后32*1,又进行1,2,3次幂运算,拼接后32*3
23 
24     '''Compute the actual results'''
25     y = f(x) # 32*3 *3*1
26     if torch.cuda.is_available():
27         return torch.autograd.Variable(x).cuda(), torch.autograd.Variable(y).cuda()
28     else:
29         return torch.autograd.Variable(x), torch.autograd.Variable(y)
30 
31 
32 w_target = torch.FloatTensor([0.5,3,2.4]).unsqueeze(1)#三行一列
33 b_target = torch.FloatTensor([0.9])
34 
35 
36 def f(x):
37     return x.mm(w_target)+b_target[0]
38 
39 class poly_model(nn.Module):
40     def __init__(self):
41         super(poly_model, self).__init__()
42         self.poly = nn.Linear(3, 1)# 输入是3维,输出是1维
43 
44     def forward(self, x):
45         out = self.poly(x)
46         return out
47 
48 if torch.cuda.is_available():
49  model = poly_model().cuda()
50 else:
51  model = poly_model()
52 
53 criterion = nn.MSELoss()
54 optimizer = torch.optim.SGD(model.parameters(), lr=1e-3)
55 
56 epoch = 0
57 for epoch in range(20):
58     batch_x,batch_y = get_batch()#batch_x 和get_batch里面的x是一样的
59     output = model(batch_x)
60     loss = criterion(output,batch_y)
61     print_loss = loss
62     print(loss.item()) # 0.4版本之后使用loss.item()从标量中获得Python number
63     optimizer.zero_grad()
64     loss.backward()
65     optimizer.step()
66 print('finished')

 

posted on 2019-10-21 09:35  爽歪歪666  阅读(206)  评论(0编辑  收藏  举报