【Pytorch问题】RuntimeError: “max_pool2d“ not implemented for ‘Long‘

问题描述

将如下input作为torch.MaxPool2d的输入

input = torch.tensor([[1,2,0,3,1],
                      [0,1,2,3,1],
                      [1,2,1,0,0],
                      [5,2,3,1,1],
                      [2,1,0,1,1]])
input = torch.reshape(input, (-1, 1, 5, 5))

报错：
​RuntimeError: "max_pool2d" not implemented for 'Long'

解决办法

pytorch中的很多操作不支持Long类型的张量, 只需要把输入的张量改成浮点类型即可

input = torch.tensor([[1,2,0,3,1],
                      [0,1,2,3,1],
                      [1,2,1,0,0],
                      [5,2,3,1,1],
                      [2,1,0,1,1]], dtype = torch.float32)
input = torch.reshape(input, (-1, 1, 5, 5))

​