张量

# 张量类似于numpy的ndarray，但还可以在GPU上是用来加速
from __future__ import print_function
import torch

# 创建一个没有初始化的矩阵
x = torch.empty(5,3)
print(x)
"""
tensor([[8.9082e-39, 8.9082e-39, 8.9082e-39],
        [1.0194e-38, 9.1837e-39, 8.4490e-39],
        [9.6429e-39, 8.4490e-39, 9.6429e-39],
        [9.2755e-39, 1.0286e-38, 9.0919e-39],
        [8.9082e-39, 9.2755e-39, 8.4490e-39]])
"""

# 创建一个随机初始化矩阵
x = torch.rand(5,3)
print(x)
"""
tensor([[0.6363, 0.4911, 0.4214],
        [0.9973, 0.4196, 0.3368],
        [0.0134, 0.2469, 0.3792],
        [0.4146, 0.5446, 0.4023],
        [0.2710, 0.3975, 0.5112]])

"""

# 构造一个填满数据且数据类型为long的矩阵
x = torch.zeros(5,3,dtype = torch.long)
print(x)
"""
tensor([[0, 0, 0],
        [0, 0, 0],
        [0, 0, 0],
        [0, 0, 0],
        [0, 0, 0]])

"""

# 直接从数据构造张量
x = torch.tensor([5.5,3])
print(x)  # tensor([5.5000, 3.0000])


# 根据已有的tensor建立新的tensor，除非用户提供新的值，否则这些方法将重用输入张量的属性，例如dtype等
x = x.new_ones(5,3,dtype = torch.double)
print(x)

"""
tensor([[1., 1., 1.],
        [1., 1., 1.],
        [1., 1., 1.],
        [1., 1., 1.],
        [1., 1., 1.]], dtype=torch.float64)
"""

x = torch.randn_like(x,dtype = torch.float)
print(x)
"""
tensor([[ 0.4324,  1.2643,  0.3443],
        [ 0.5959,  1.3176,  0.0298],
        [-0.3860, -0.1308, -0.9114],
        [ 0.8100, -1.3739,  1.6596],
        [ 1.2363,  1.9997,  0.2527]])

"""


# 运算
# 形式一
y = torch.rand(5,3)
print(x + y)
"""
tensor([[ 1.4126e+00,  6.8308e-01,  4.7527e-01],
        [-1.2185e+00,  1.0213e+00,  2.3724e-01],
        [-3.1333e-01,  1.6759e+00,  2.0732e+00],
        [ 7.4825e-01, -1.9758e-03, -8.2048e-02],
        [ 1.2351e-01, -4.2529e-01,  9.7085e-01]])
"""

# 形式二：
print(torch.add(x,y))
"""
tensor([[-0.5781,  0.3660,  0.5735],
        [ 0.4229,  0.0500,  0.2018],
        [ 0.6387,  0.6206,  1.2428],
        [ 0.5858, -1.0471,  0.3354],
        [ 1.8888,  1.6495,  1.1323]])
"""

# 形式三：给出一个张量作为参数
result = torch.empty(5,3)
torch.add(x,y,out = result)
print(result)
"""
tensor([[ 0.3352,  0.2303,  0.0473],
        [ 0.2402, -0.0209,  0.5980],
        [ 0.8451,  0.1239,  1.8851],
        [ 0.1244,  2.1331, -0.1232],
        [ 1.1081,  0.5348, -1.9197]])

"""

# 形式四：原位/原地操作（in-place)
# add x to y
y.add_(x)
print(y)
"""
tensor([[ 1.5857, -1.4062, -0.5115],
        [ 0.6804,  1.5849,  0.1429],
        [ 1.7859,  1.9659,  1.1634],
        [-0.1099, -1.1706,  1.5166],
        [-0.0537, -0.0459,  0.0303]])
"""
# 注意：任何一个in-place改变张量的操作都是一个_。例如：x.copy_(y)

# 索引
print(x[:,1])  # 打印所有行的第一列
# tensor([-0.3033,  0.0998,  0.1000,  0.1890, -1.4164])

# 改变形状:如果想改变形状，可以使用torch.view
x = torch.randn(4,4)
y = x.view(16)
z = x.view(-1,8)  # -1根据后面的维度自动计算
print(x.size(),y.size(),z.size())  # torch.Size([4, 4]) torch.Size([16]) torch.Size([2, 8])

# 对于只有一个元素的tensor，可以使用item（）来获得对应的python数值
x = torch.randn(1)
print(x)  # tensor([0.8120])
print(x.item())  # 0.8119511008262634

# 将torch的Tensor转换为Numpy数组
# torch张量和Numpy数组将共享他们底层内存位置，因此当一个改变时，另外一个也会改变
a = torch.ones(5)
print(a)  # tensor([1., 1., 1., 1., 1.])

b = a.numpy()
print(b)  # [1. 1. 1. 1. 1.]

a.add_(1)  # 当a改变时，b也会随着改变
print(a)  # tensor([2., 2., 2., 2., 2.])
print(b)  # [2. 2. 2. 2. 2.]

# 将numpy数组转换为Torch张量
import numpy as np
a = np.ones(5)
b = torch.from_numpy(a)
np.add(a,1,out = a)
print(a)  # [2. 2. 2. 2. 2.]
print(b)  # tensor([2., 2., 2., 2., 2.], dtype=torch.float64)