pytorch学习（七）

torch.max用法：

#coding=utf-8
import torch
from torch.autograd import Variable

#返回输入tensor中所有元素的最大值
a = torch.randn(1, 3)
print a
print torch.max(a)
print '--' * 10

#按维度dim 返回最大值  torch.max)(a,0) 返回每一列中最大值的那个元素，且返回索引（返回最大元素在这一列的行索引）
b = torch.randn(3,3)
b = Variable(b)
print b
print torch.max(b,0)
print '--' * 10

#torch.max(a,1) 返回每一行中最大值的那个元素，且返回其索引（返回最大元素在这一行的列索引）
print torch.max(b,1)

print '----' * 10
print torch.max(b,1)[0]   #只返回最大值

print '----' * 10
print torch.max(b,1)[1]   #只返回最大值所在的索引

结果：

0.8280 -0.3046  0.4422
[torch.FloatTensor of size 1x3]

0.828029453754
--------------------
Variable containing:
-0.2242  0.2954  1.3393
 0.2511  0.7484  1.4388
-0.1107  0.0453  2.2737
[torch.FloatTensor of size 3x3]

(Variable containing:
 0.2511  0.7484  2.2737
[torch.FloatTensor of size 1x3]
, Variable containing:
 1  1  2
[torch.LongTensor of size 1x3]
)
--------------------
(Variable containing:
 1.3393
 1.4388
 2.2737
[torch.FloatTensor of size 3x1]
, Variable containing:
 2
 2
 2
[torch.LongTensor of size 3x1]
)
----------------------------------------
Variable containing:
 1.3393
 1.4388
 2.2737
[torch.FloatTensor of size 3x1]

----------------------------------------
Variable containing:
 2
 2
 2
[torch.LongTensor of size 3x1]