Pytorch系列(一):基础操作篇

版权提示: 参考 PyTorch中文文档

1、 构建5*3的矩阵,不进行初始化

import torch
x = torch.empty(5, 3)
print(x)

输出

tensor([[1.0682e-11, 4.5757e-41, 1.6970e-36],
        [0.0000e+00, 0.0000e+00, 0.0000e+00],
        [0.0000e+00, 0.0000e+00, 0.0000e+00],
        [0.0000e+00, 0.0000e+00, 0.0000e+00],
        [0.0000e+00, 0.0000e+00, 0.0000e+00]])

2、构建5*3的随机初始化的矩阵

x = torch.rand(5, 3)
print(x)

输出

tensor([[0.8665, 0.2635, 0.6159],
        [0.6626, 0.2808, 0.7473],
        [0.1187, 0.0978, 0.3235],
        [0.4837, 0.7860, 0.4497],
        [0.0208, 0.6852, 0.1181]])

3、构造一个矩阵全为0, 而且数据类型是long

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]])

4、 构造一个张量,直接使用数据

x = torch.tensor([5.5, 3])
print(x)

tensor([5.5000, 3.0000])

5、 创建一个tensor,基于已存在的tensor

x = x.new_ones(5, 3, dtype=torch.double)
print(x)

x = torch.randn_like(x, dtype=torch.float)
print(x)

tensor([[1., 1., 1.],
        [1., 1., 1.],
        [1., 1., 1.],
        [1., 1., 1.],
        [1., 1., 1.]], dtype=torch.float64)
tensor([[-0.0781,  1.2222,  0.4896],
        [ 0.9935, -0.2359,  2.1339],
        [-1.5652, -2.9325, -1.4721],
        [-0.6683,  1.3144,  0.2216],
        [-1.3604, -0.7052,  1.0775]])

6、 获取维度信息

torch.size 是一个元祖,所以支持左右的元祖操作

print(x.size())
print(x.size()[0])

torch.Size([5, 3])
5

7、 加法操作 + 和 add

y = torch.rand(5, 3)
print(x+y)

print(torch.add(x, y))

完全一样

tensor([[ 0.3728,  1.6224,  1.3346],
        [ 1.4414,  0.3598,  2.5921],
        [-1.4853, -1.9702, -1.0021],
        [ 0.1606,  1.3632,  0.7417],
        [-1.2048,  0.0495,  1.3192]])
tensor([[ 0.3728,  1.6224,  1.3346],
        [ 1.4414,  0.3598,  2.5921],
        [-1.4853, -1.9702, -1.0021],
        [ 0.1606,  1.3632,  0.7417],
        [-1.2048,  0.0495,  1.3192]])

8、 提供一个输出Tensor作为参数

result = torch.empty(5, 3)
torch.add(x, y, out=result)
print(result)

tensor([[ 0.3728,  1.6224,  1.3346],
        [ 1.4414,  0.3598,  2.5921],
        [-1.4853, -1.9702, -1.0021],
        [ 0.1606,  1.3632,  0.7417],
        [-1.2048,  0.0495,  1.3192]])

9、 加法 in-place y.add(x) 和 y.add_(x)

y.add(x) 只减了一次, y.add_(x) 则一直在减

for i in range(4):
    y.add(x)
    print(y)
for i in range(4):
    y.add_(x)
    print(y)

tensor([[0.4509, 0.4002, 0.8449],
        [0.4478, 0.5957, 0.4582],
        [0.0800, 0.9623, 0.4701],
        [0.8289, 0.0489, 0.5201],
        [0.1557, 0.7547, 0.2417]])
tensor([[0.4509, 0.4002, 0.8449],
        [0.4478, 0.5957, 0.4582],
        [0.0800, 0.9623, 0.4701],
        [0.8289, 0.0489, 0.5201],
        [0.1557, 0.7547, 0.2417]])
tensor([[0.4509, 0.4002, 0.8449],
        [0.4478, 0.5957, 0.4582],
        [0.0800, 0.9623, 0.4701],
        [0.8289, 0.0489, 0.5201],
        [0.1557, 0.7547, 0.2417]])
tensor([[0.4509, 0.4002, 0.8449],
        [0.4478, 0.5957, 0.4582],
        [0.0800, 0.9623, 0.4701],
        [0.8289, 0.0489, 0.5201],
        [0.1557, 0.7547, 0.2417]])
tensor([[ 0.3728,  1.6224,  1.3346],
        [ 1.4414,  0.3598,  2.5921],
        [-1.4853, -1.9702, -1.0021],
        [ 0.1606,  1.3632,  0.7417],
        [-1.2048,  0.0495,  1.3192]])
tensor([[ 0.2947,  2.8446,  1.8242],
        [ 2.4349,  0.1239,  4.7260],
        [-3.0505, -4.9027, -2.4742],
        [-0.5076,  2.6776,  0.9633],
        [-2.5652, -0.6556,  2.3966]])
tensor([[ 0.2166,  4.0667,  2.3138],
        [ 3.4285, -0.1120,  6.8600],
        [-4.6157, -7.8352, -3.9463],
        [-1.1759,  3.9919,  1.1849],
        [-3.9257, -1.3608,  3.4741]])
tensor([[  0.1385,   5.2889,   2.8035],
        [  4.4220,  -0.3479,   8.9939],
        [ -6.1810, -10.7677,  -5.4184],
        [ -1.8442,   5.3063,   1.4066],
        [ -5.2861,  -2.0660,   4.5516]])

10、 使用numpy类似的索引操作

print(x[:, 1])

tensor([ 1.2222, -0.2359, -2.9325,  1.3144, -0.7052])

11、改变Tensor大小的形状

如果你想改变一个tensor的大小或者形状, 可以使用torch.view

x = torch.randn(4, 4)
y = x.view(16)
z = x.view(-1, 8)
print(x.size(), y.size(), z.size())

torch.Size([4, 4]) torch.Size([16]) torch.Size([2, 8])

12、如果一个tensor只有一个元素,获取元素值

可以用 item()来获取这个value.注意只能有一个元素,否则会报错

x = torch.randn(1)
print(x)
print(x.item())

x = torch.tensor([3.14])
print(x.item())

tensor([-0.6945])
-0.6944621801376343
3.140000104904175