Pytorch---Pytorch的张量基础--2

一、张量的数据类型转换

1.指定类型的转换

代码

import torch
import numpy as np

if __name__ == '__main__':
    t1 = torch.rand(3, 3)
    print(t1.dtype)
    # 进行数据类型转换
    t1 = t1.type(torch.float16)
    print(t1.dtype)

结果

2.快速转换为常用类型

代码

import torch
import numpy as np

if __name__ == '__main__':
    t1 = torch.rand(3, 3)
    print(t1.dtype)
    # 进行数据类型转换
    t1 = t1.long()
    print(t1.dtype)
    # 进行数据类型转换
    t1 = t1.float()
    print(t1.dtype)

结果

二、张量的运算

1.广播原则

代码

import torch
import numpy as np

if __name__ == '__main__':
    t1 = torch.rand(3, 3)
    print(t1)
    t1 = t1 + 3
    print(t1)

结果

代码

import torch
import numpy as np

if __name__ == '__main__':
    t1 = torch.rand(3, 3)
    t2 = torch.ones(3, 3)
    print(t1)
    t1 = t1 + t2
    print(t1)

结果

代码

import torch
import numpy as np

if __name__ == '__main__':
    t1 = torch.rand(3, 3)
    t2 = torch.ones(3, 3)
    print(t1)
    t1 = t1.add(t2)
    print(t1)

结果

2.直接改变原值

在张量的操作方法后加 _ 将会直接改变原值

代码

import torch
import numpy as np

if __name__ == '__main__':
    t1 = torch.rand(3, 3)
    t2 = torch.ones(3, 3)
    print(t1)
    t1.add_(t2)
    print(t1)

结果

3.矩阵乘法

代码

import torch
import numpy as np

if __name__ == '__main__':
    t1 = torch.rand(3, 4)
    t2 = torch.randn(4, 3)

    t = t1.matmul(t2)
    print(t.size())

    tt = t1 @ t2
    print(tt.shape)

结果

4.将维度为0的Tensor转换为Python的数据类型

代码

import torch
import numpy as np

if __name__ == '__main__':
    t1 = torch.rand(3, 4)
    t = t1.sum()
    print(t.item())

结果

5.将Tensor与numpy数据类型的相互转换

代码

import torch
import numpy as np

if __name__ == '__main__':
    t = torch.from_numpy(np.random.randn(3, 4))
    print(t)

    print('-------------------------------------')

    t = t.numpy()
    print(t)

结果