使用torch-gpu处理数据错误 01

今天运行下面的代码时候，出现了这个问题，百度半天没有找到合适的解决办法，思考半天，终于用两行代码解决了，记一下！！

#!/usr/bin/env python
# -*- coding: utf-8 -*-
# @Time : 2019/12/28 20:52
# @Author : LZQ
# @Software: PyCharm


import torch

x = torch.randn(5,4)
y=x.new_ones(5,4,dtype=float)
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
print(device)
x=x.to(device).cuda()
y=y.to(device).cuda()
z=x+y
print(z)

报错原因如下：

E:\lzqData\3d\anaconda3\Anaconda3\python.exe G:/pychram/workspace/python/study_pytorch/pytorch01/pytorch002.py
cuda:0
Traceback (most recent call last):
  File "G:/pychram/workspace/python/study_pytorch/pytorch01/pytorch002.py", line 19, in 
    z=x+y
RuntimeError: expected device cuda:0 and dtype Double but got device cuda:0 and dtype Float

Process finished with exit code 1

原因是，我的x,y的类型是float不对，但是gpu想要的是double，把float改为double并不行，解决办法如下：

#!/usr/bin/env python
# -*- coding: utf-8 -*-
# @Time : 2019/12/28 20:52
# @Author : LZQ
# @Software: PyCharm

import torch



x = torch.randn(5,4)
y=x.new_ones(5,4,dtype=float)
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
print(device)
x=x.double()  #为x添加double()函数
y = y.double()#为y添加double()函数
x=x.to(device).cuda()
y=y.to(device).cuda()
z=x+y
print(z)

结果：

E:\lzqData\3d\anaconda3\Anaconda3\python.exe G:/pychram/workspace/python/study_pytorch/pytorch01/pytorch002.py
cuda:0
tensor([[ 1.1086,  2.0426,  2.0449, -0.2139],
        [ 0.1211,  0.6647,  1.4036,  3.2326],
        [ 1.5524,  1.1472,  1.5518,  2.5580],
        [ 2.8068,  2.0145,  2.6142,  1.5480],
        [ 1.1827,  1.7224,  2.8095, -0.1689]], device='cuda:0',
       dtype=torch.float64)

Process finished with exit code 0

问题分析：因为类型不匹配，我们要把x和y转换成我们想要的double，但是直接在dtype中改，是不支持的，所以我们强制转换。就搞定啦。爽歪歪！！！