Pytorch学习之旅day02

1.	num_samples = 64 # N
2.	dim_in, dim_hid, dim_out = 1000, 100, 10  # IN H OUT
3.	x = torch.randn(num_samples, dim_in)  #  N * IN
4.	y = torch.randn(num_samples, dim_out) #  N * OUT
5.	
6.	# 提前定义模型
7.	model = torch.nn.Sequential(
8.	    torch.nn.Linear(dim_in, dim_hid, bias = False), #model[0]
9.	    torch.nn.ReLU(),
10.	    torch.nn.Linear(dim_hid, dim_out, bias = False),#model[2]
11.	)
12.	
13.	# torch.nn.init.normal_(model[0].weight) #修改一
14.	# torch.nn.init.normal_(model[2].weight)
15.	
16.	#提前定义loss函数和优化函数
17.	loss_fun = torch.nn.MSELoss(reduction='sum')
18.	eta = 1e-4 #修改二
19.	optimizer = torch.optim.Adam(model.parameters(), lr=eta)
20.	
21.	
22.	for i in range(1000):
23.	  #Forward pass
24.	  y_pred = model(x)
25.	  
26.	  #Loss
27.	  loss = loss_fun(y_pred, y)
28.	  print(it, loss.item())
29.	
30.	  optimizer.zero_grad()
31.	  # Backward pass
32.	  loss.backward()
33.	
34.	  # update model parameters
35.	  optimizer.step()