torch.max函数比较最大值

新建两个tensor列表，利用torch.max()进行两个列表的比较：

import torch


a=torch.tensor([1,2,3,4,5])
print("a: ", a)

b=torch.tensor([10,1,7,9,3])
print("b: ", b)

x=torch.max(a,b)

print("x: ", x)

比较结果如下：

# a:  tensor([1, 2, 3, 4, 5])
# b:  tensor([10,  1,  7,  9,  3])
# x:  tensor([10,  2,  7,  9,  5])

================================================
矩阵操作:

import numpy as np
import torch
import torch.nn as nn
import torch.tensor as tensor


scores = torch.randn(4, 3, 5).cuda()


a = torch.tensor([[1,2,3,1,4],
                  [65, 12, 45, 78, 99],
                  [7,8,9,6,12]])


b = torch.tensor([[12, 45, 23, 21, 78],
                  [4,5,6,7,8],
                  [23, 65, 24, 54, 12]])


c = torch.tensor([[90, 21, 12, 20, 21],
                  [123, 456, 12, 1, 45],
                  [45, 67, 87, 22, 56]])

d = torch.tensor([[11, 34, 21, 678, 12],
                  [21, 67, 23,110,1],
                  [67, 21, 23, 12, 23]])


scores[0]=a
scores[1]=b
scores[2]=c
scores[3]=d

fused_scores = scores.max(0)  
out = fused_scores.values

print("out: ", out)

结果如下:

out:  tensor([[ 90.,  45.,  23., 678.,  78.],
        [123., 456.,  45., 110.,  99.],
        [ 67.,  67.,  87.,  54.,  56.]], device='cuda:0')