pytorch基础入门（一）

pytorch基础入门（一）

一、tensor属性

import torch

a = torch.tensor([1,2,3], dtype = int)

print(a)

tensor([1, 2, 3])

a.type

<function Tensor.type>

b = torch.tensor([4,5,6], dtype = float)

print(b)

tensor([4., 5., 6.], dtype=torch.float64)

tensor = torch.tensor([[1,2,3],
                       [4,5,6]])

# 查看数据的维度
tensor.ndim

2

# 查看数据的形状
tensor.shape  # 属性

torch.Size([2, 3])

tensor.size()  # 方法

torch.Size([2, 3])

二、数据生成

import torch

# 生成权重为1的两行三列的矩阵
torch.ones(2,3)

tensor([[1., 1., 1.],
        [1., 1., 1.]])

# 生成权重为0的三行三列的矩阵
torch.zeros(3,3)

tensor([[0., 0., 0.],
        [0., 0., 0.],
        [0., 0., 0.]])

# 生成三行四列的随机数矩阵
torch.rand(3,4)

tensor([[0.2095, 0.2981, 0.9930, 0.2970],
        [0.5243, 0.6877, 0.2672, 0.1739],
        [0.3329, 0.4882, 0.5752, 0.0886]])

# 生成0-10之间两行三列的随机整数矩阵
torch.randint(0,10,(2,3))

tensor([[1, 6, 6],
        [9, 4, 7]])

# 生成三行四列的符合正态分布矩阵
torch.randn(3,4)

tensor([[ 1.6634,  0.8821,  0.3271,  0.7346],
        [-0.7800, -0.8522,  0.3166, -1.0605],
        [-0.2146, -0.4478,  1.0225,  0.7094]])

a = torch.tensor([[1,2],
                  [3,4],
                  [5,6]])

a

tensor([[1, 2],
        [3, 4],
        [5, 6]])

# 根据a的形状生成b
b = torch.rand_like(a, dtype = float)

b

tensor([[0.0610, 0.8480],
        [0.9012, 0.8032],
        [0.8593, 0.0249]], dtype=torch.float64)

# 改变数据形状
d = b.view(2,3)

d

tensor([[0.0610, 0.8480, 0.9012],
        [0.8032, 0.8593, 0.0249]], dtype=torch.float64)

# 改变数据形状
d = d.reshape(6)

d

tensor([0.0610, 0.8480, 0.9012, 0.8032, 0.8593, 0.0249], dtype=torch.float64)

d[1]

tensor(0.8480, dtype=torch.float64)

# 转化为正常数据，非tensor,只能改变一个
d[1].item()

0.8479788102877875

import numpy as np

# 将tenser转numpy array
array = np.array(d)

array

array([0.06101681, 0.84797881, 0.90117927, 0.80323335, 0.85928866,
       0.02492348])

tensor = torch.tensor(array)

tensor

tensor([0.0610, 0.8480, 0.9012, 0.8032, 0.8593, 0.0249], dtype=torch.float64)

三、数据索引

import torch

tensor = torch.arange(2,14)
tensor

tensor([ 2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13])

print(tensor[2])

tensor(4)

tensor[1:4]

tensor([3, 4, 5])

tensor[2:-1]

tensor([ 4,  5,  6,  7,  8,  9, 10, 11, 12])

tensor[:5]

tensor([2, 3, 4, 5, 6])

tensor[::2]

tensor([ 2,  4,  6,  8, 10, 12])

tensor[-3:]

tensor([11, 12, 13])

index = [1,4,5,5,7]
tensor[index]

tensor([3, 6, 7, 7, 9])

四、torch的基本运算操作

import torch

a = torch.randint(1,5,(2,3))
b = torch.randint(1,5,(2,3))
print(a)
print(b)

tensor([[4, 2, 2],
        [2, 2, 3]])
tensor([[4, 1, 3],
        [3, 1, 2]])

a+b

tensor([[8, 3, 5],
        [5, 3, 5]])

torch.add(a,b)

tensor([[8, 3, 5],
        [5, 3, 5]])

result = torch.zeros(2,3)
torch.add(a,b,out= result)

tensor([[8., 3., 5.],
        [5., 3., 5.]])

# a + b
# 注意 任何使张量tensor会发生变化的操作都有一个前缀'_'，例如：a.add_()
a.add_(b)

tensor([[8, 3, 5],
        [5, 3, 5]])

tensor = torch.ones(3,5)
tensor

tensor([[1., 1., 1., 1., 1.],
        [1., 1., 1., 1., 1.],
        [1., 1., 1., 1., 1.]])

a = a.float()

# 矩阵乘法
torch.matmul(a,tensor)

tensor([[16., 16., 16., 16., 16.],
        [13., 13., 13., 13., 13.]])

a = torch.rand(3,2)
a

tensor([[0.3879, 0.3897],
        [0.6362, 0.4619],
        [0.9510, 0.0311]])

torch.sum(a)

tensor(2.8577)

torch.min(a)

tensor(0.0311)

# 求最小值所在位置（索引）
torch.argmin(a)

tensor(5)

# 平均值
torch.mean(a)

tensor(0.4763)

# 中位数
torch.median(a)

tensor(0.3897)

torch.sqrt(a)

tensor([[0.6228, 0.6243],
        [0.7976, 0.6796],
        [0.9752, 0.1763]])

五、自动求导

import torch

x = torch.ones(2,2,requires_grad=True)
x

tensor([[1., 1.],
        [1., 1.]], requires_grad=True)

y = x + 2
y

tensor([[3., 3.],
        [3., 3.]], grad_fn=<AddBackward0>)

z = y * y * 3
z

tensor([[27., 27.],
        [27., 27.]], grad_fn=<MulBackward0>)

out = z.mean()
out

tensor(27., grad_fn=<MeanBackward0>)

# 求导
out.backward()

print(x.grad)

tensor([[4.5000, 4.5000],
        [4.5000, 4.5000]])

参考：31 项目实战 猫狗分类模型预测 - YouTube