PyTorch学习问题记录

Q1：def train() 中的model.train()的作用是什么？为什么要写？

A1：class torch.nn.Module中 train(mode=True)

　　Sets the module in training mode. This has any effect only on modules such as Dropout or BatchNorm.

　　参看 http://pytorch.org/docs/master/nn.html

Q2：torch.gather()函数的功能是什么？

 1 t = torch.Tensor([[1, 2], [3, 4]])
 2 print(t)
 3 a = torch.gather(t, 1, torch.LongTensor([[0,0], [1,0]]))
 4 print(a)
 5 '''
 6  1  2
 7  3  4
 8 [torch.FloatTensor of size 2x2]
 9 
10  1  1
11  4  3
12 [torch.FloatTensor of size 2x2]
13 '''

A2：

out[i][j][k] = input[index[i][j][k]][j][k] 　　# if dim == 0

out[i][j][k] = input[i][index[i][j][k]][k] 　　# if dim == 1

out[i][j][k] = input[i][j][index[i][j][k]] 　　# if dim == 2

out[i][j] = input[index[i][j]][j]

out[i][j] = input[i][index[i][j]]

out[0][0] = input[0][index[0][0]] = input[0][0] = 1

out[0][1] = input[0][index[0][1]] = input[0][0] = 1

out[1][0] = input[1][index[1][0]] = input[1][1] = 4

out[1][1] = input[1][index[1][1]] = input[1][0] = 3

Q3：torch.norm() 函数的功能是什么？

 1 a = torch.FloatTensor([[1, 2], [3, 4]])
 2 b = torch.norm(a)
 3 print(a)
 4 print(b)
 5 '''
 6  1  2
 7  3  4
 8 [torch.FloatTensor of size 2x2]
 9 
10 5.477225575051661
11 '''

A3：

norm() 函数是求范数，一般默认是2范数。平方和开根号。

参考博文：几种范数的简单介绍

normal() 函数是求正太分布。

Q4: topk()函数

• torch.Tensor.topk (Python method, in torch.Tensor) ||topk(k, dim=None, largest=True, sorted=True) -> (Tensor, LongTensor)
• torch.topk (Python function, in torch) ||torch.topk(input, k, dim=None, largest=True, sorted=True, out=None) -> (Tensor, LongTensor)

1 topi = torch.LongTensor([5])        # [torch.LongTensor of size 1]
2 topii = torch.LongTensor([[5]])     # [torch.LongTensor of size 1x1]
3 ni = topi[0]
4 nii = topii[0][0]
5 print(ni, nii)                           # 5 5

Q5:

1 loss = Variable(torch.FloatTensor([1]))
2 print(loss.data)                      # 1 [torch.FloatTensor of size 1]
3 print(loss.data[0])                   # 1.0