torch7
文章目录
- torch.max
- tensor元素读取
torch.max
格式：
torch.max(input, dim, keepdim=False, out=None) -> (Tensor, LongTensor)
th> a=torch.rand(2,3)
0.1055 0.5020 0.1127
0.4437 0.4049 0.6918
[torch.DoubleTensor of size 2x3]
th> torch.max(a) --a中最大值
0.69182460033335
th> torch.max(a,1) --返回每一列中最大值的那个元素，且返回索引
0.4437 0.5020 0.6918
[torch.DoubleTensor of size 1x3]
2 1 2
[torch.LongTensor of size 1x3]
th> torch.max(a,2)–返回每一行中最大值的那个元素，且返回索引
0.5020
0.6918
[torch.DoubleTensor of size 2x1]
2
3
[torch.LongTensor of size 2x1]
th> b=torch.rand(2,3,3)
(1,.,.) =
0.2968 0.2864 0.0240
0.0565 0.6799 0.5388
0.4088 0.4470 0.1924
(2,.,.) =
0.9951 0.4827 0.7110
0.1648 0.8386 0.2206
0.9690 0.0007 0.6369
[torch.DoubleTensor of size 2x3x3]
th> torch.max(b)
0.99509069975466
th> torch.max(b,1)
(1,.,.) =
0.9951 0.4827 0.7110
0.1648 0.8386 0.5388
0.9690 0.4470 0.6369
[torch.DoubleTensor of size 1x3x3]
(1,.,.) =
2 2 2
2 2 1
2 1 2
[torch.LongTensor of size 1x3x3]
th> torch.max(b,2)
(1,.,.) =
0.4088 0.6799 0.5388
(2,.,.) =
0.9951 0.8386 0.7110
[torch.DoubleTensor of size 2x1x3]
(1,.,.) =
3 2 2
(2,.,.) =
1 2 1
[torch.LongTensor of size 2x1x3]
tensor元素读取
th> a= torch.rand(2,3)
th>a
0.9646 0.2998 0.8082
0.4970 0.5252 0.2022
[torch.DoubleTensor of size 2x3]
th> a[1] --[]
0.9646
0.2998
0.8082
[torch.DoubleTensor of size 3]
th> a[{{1}}] --[{}]
0.9646 0.2998 0.8082
[torch.DoubleTensor of size 1x3]
th> a[{{1},{2}}]
0.2998
[torch.DoubleTensor of size 1x1]
th> b=torch.rand(2,3,4)
th> b
(1,.,.) =
0.5474 0.4723 0.0516 0.0136
0.9878 0.3986 0.0787 0.0547
0.8163 0.7833 0.3281 0.8037
(2,.,.) =
0.1795 0.4234 0.5883 0.4228
0.3473 0.3866 0.3203 0.3479
0.9887 0.4204 0.4081 0.0496
[torch.DoubleTensor of size 2x3x4]
th> b[{{1}}]
(1,.,.) =
0.5474 0.4723 0.0516 0.0136
0.9878 0.3986 0.0787 0.0547
0.8163 0.7833 0.3281 0.8037
[torch.DoubleTensor of size 1x3x4]
th> b[{{},{2}}]
(1,.,.) =
0.9878 0.3986 0.0787 0.0547
(2,.,.) =
0.3473 0.3866 0.3203 0.3479
[torch.DoubleTensor of size 2x1x4]
th> b[{{1},{2},{3}}]
(1,.,.) =
0.01 *
7.8706
[torch.DoubleTensor of size 1x1x1]