Pytorch入门

张量

新建张量

• 1.生成一个空张量torch.empty

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

• 2.rand(*size, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) -> Tensor返回一个张量，该张量由间隔上均匀分布的随机数填充：math:[0，1）

import torch
x = torch.rand(2, 3)
print(x)

• 3.randn(*size, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) -> Tensor返回一个张量，该张量由均值为“0”、方差为“1”的正态分布中的随机数填充（也称为标准正态分布）

import torch
x = torch.randn(2, 3)
print(x)

• 4.normal(mean, std, *, generator=None, out=None) -> Tensor给定平均值和标准差；返回对应**正太分布**数值

import torch
x = torch.normal(2, 3, size=(1, 4))
print(x)

• linspace(…)
  linspace(start, end, steps=100, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False) -> Tensor在start和end之间等距的返回数据

import torch
x = torch.linspace(3, 10, steps=5)
print(x)

tensor([ 3.0000,  4.7500,  6.5000,  8.2500, 10.0000])

• **torch.zeros(2, 3)**生成一个全零的张量

import torch
x = torch.zeros(2, 3)
print(x)
tensor([[0., 0., 0.],
        [0., 0., 0.]])

• **torch.tensor([[0.1, 1.2], [2.2, 3.1], [4.9, 5.2]])**已知矩阵转换为张量

x = torch.tensor([[0.1, 1.2], [2.2, 3.1], [4.9, 5.2]])
print(x)
tensor([[0.1000, 1.2000],
        [2.2000, 3.1000],
        [4.9000, 5.2000]])

• 生成一个同样维度的矩阵

import torch
import numpy as np
x = torch.ones(5,5, 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.],
        [1., 1., 1., 1., 1.],
        [1., 1., 1., 1., 1.]], dtype=torch.float64)
tensor([[ 0.2423,  1.7955, -0.5171,  1.1406, -0.5407],
        [-1.1997,  0.1341, -1.2628, -1.4078, -0.3429],
        [ 1.1838,  0.4697,  0.2860, -0.0708,  0.2232],
        [-0.0636, -0.1558, -0.3082, -0.2701, -0.5039],
        [ 0.9021, -1.5196, -0.8247, -1.4582,  0.8148]])

• 获取张量的形状print(x.size())

print(x.size())
torch.Size([5, 5])

张量运算

• 加法

y = torch.rand(5,5)
# 加法1
print(x + y)
#加法2
print(torch.add(x,y))
# 加法 3
y.add_(x)
print(y)

tensor([[-0.5748,  0.5915, -0.1207,  0.9979, -0.0416],
        [ 1.1396,  0.4685,  0.9498,  1.1800, -2.0215],
        [ 0.3359, -1.4478, -1.2943,  0.4292,  0.1471],
        [-0.3918,  0.5140, -1.2918,  0.4770, -0.1626],
        [ 0.4617,  2.1801,  0.1918,  0.0635,  2.1010]])
tensor([[-0.5748,  0.5915, -0.1207,  0.9979, -0.0416],
        [ 1.1396,  0.4685,  0.9498,  1.1800, -2.0215],
        [ 0.3359, -1.4478, -1.2943,  0.4292,  0.1471],
        [-0.3918,  0.5140, -1.2918,  0.4770, -0.1626],
        [ 0.4617,  2.1801,  0.1918,  0.0635,  2.1010]])
        tensor([[-0.5748,  0.5915, -0.1207,  0.9979, -0.0416],
        [ 1.1396,  0.4685,  0.9498,  1.1800, -2.0215],
        [ 0.3359, -1.4478, -1.2943,  0.4292,  0.1471],
        [-0.3918,  0.5140, -1.2918,  0.4770, -0.1626],
        [ 0.4617,  2.1801,  0.1918,  0.0635,  2.1010]])

• 指定输出到某一个参数

result = torch.empty(5,3)
torch.add(x,y,out = result)
print(result)
print(x)
print(x[:, 1])
tensor([[-0.9804,  0.1685,  0.8569,  2.0596,  0.2225],
        [ 1.0814,  1.4838,  0.5020,  1.5157,  0.8952],
        [-0.3903,  1.9247,  0.8741,  1.9728,  0.2694],
        [ 1.1786, -0.7924, -0.8760,  0.3188,  1.7267],
        [ 1.6438,  1.4482,  0.7289,  0.9638,  0.8927]])
        tensor([[ 0.2748, -0.0774, -1.2880, -0.0365,  0.8686],
        [ 0.1119,  0.8125,  1.2119,  0.5379, -0.2646],
        [-0.1414,  0.8118, -1.0719, -0.5585,  0.0922],
        [ 0.1249,  0.5992, -0.7017,  0.0810, -0.2659],
        [-0.6018, -0.3716,  2.8113,  0.7393, -0.6450]])
tensor([-0.0774,  0.8125,  0.8118,  0.5992, -0.3716])

• 改变张量的维度

x = torch.randn(4, 4)
y = x.view(16)
z = x.view(-1, 4)  # the size -1 is inferred from other dimensions
print(x.size(), y.size(), z.size())
torch.Size([4, 4]) torch.Size([16]) torch.Size([4, 4])

张量的转换

在用CPU的时候numpy和pytorch共用物理内存

• 张量和numpy之间的转换

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)
print(a)
print(b)
"""
tensor([2., 2., 2., 2., 2.])
[2. 2. 2. 2. 2.]
"""

• numpy转pytorch

import numpy as np
a = np.ones(5)
b = torch.from_numpy(a)
np.add(a, 1, out=a)
print(a)
print(b)
"""
[2. 2. 2. 2. 2.]
tensor([2., 2., 2., 2., 2.], dtype=torch.float64)
"""

• GPU下数据转换

# let us run this cell only if CUDA is available
# We will use ``torch.device`` objects to move tensors in and out of GPU


if torch.cuda.is_available():
    device = torch.device("cuda")          # a CUDA device object
    y = torch.ones_like(x, device=device)  # directly create a tensor on GPU
    x = x.to(device)                       # or just use strings ``.to("cuda")``
    z = x + y
    print(z)
    print(z.to("cpu", torch.double))       # ``.to`` can also change dtype together!
"""
tensor([-0.4743], device='cuda:0')
tensor([-0.4743], dtype=torch.float64)
"""

Transfrom的初步使用

from PIL import Image
from torch.utils.tensorboard import SummaryWriter
from torchvision import transforms

#图片转换为PIL格式
writer = SummaryWriter("logs")
img = Image.open("image/648.jpg")
print(img)

#用tensorboard显示 totensor
trans_totensor = transforms.ToTensor()
img_temsor = trans_totensor(img)
writer.add_image("Totensor", img_temsor)

#Normalize
#output[channel] = (input[channel] - mean[channel]) / std[channel]
print(img_temsor[0][0][1])#0层0行0列
trans_norm = transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])
img_norm = trans_norm(img_temsor)
print(img_norm[0][0][1])
writer.add_image("Normalize", img_norm)


#resize
print(img.size)
#img PIL->resize->img_resize PIL
trans_resize = transforms.Resize((512, 512))
#img resize PIL-> totensor->img_resize tensor
img_resize = trans_resize(img)
#由于返回数据是PIL图像，需要转换为tensor才能进行tensorboard显示
img_resize = trans_totensor(img_resize)

writer.add_image("Resize", img_resize, 0)


#Compose-resize-2
trans_resize_2 = transforms.Resize(512)
#PTL->PIL->tensor
trans_compose = transforms.Compose([trans_resize_2, trans_totensor])
img_resize_2 = trans_compose(img)
writer.add_image("Resize", img_resize_2, 1)

#RandomCrop
trans_random = transforms.RandomCrop((255, 100))#随机裁剪
trans_cmpose_2 = transforms.Compose([trans_random, trans_totensor])
for i in range(10):
    img_crop = trans_cmpose_2(img)
    writer.add_image("RandomCrop", img_crop, i)

writer.close()