《深度学习框架PyTorch入门与实践》——Tensor基本操作(2)

一、Tensor操作

**************************************************************************************************************

（1）归并操作

函数	功能
mean/sum/median/mju	均值/和/中位数/众数
nom/dist	范数/距离
std/var	标准差/方差
cumsum/cumprod	累加/累乘

import torch as t
b=t.ones(2,3)
print(b.sum(dim=0,keepdim=True))
print(b.sum(dim=0,keepdim=False))
print(b.sum(dim=1))
a=t.arange(0,6).view(2,3)
print(a)
a=a.cumsum(dim=1)#沿着行累加
print(a)

                       ----------------------在pycharm中的运行结果-----------------------

**************************************************************************************************************

 （2）比较

函数	功能
gt/lt/ge/le/eq/ne	大于/小于/大于等于/小于等于/等于/不等
topk	最大的K的个数
sort	排序
max/min	比较两个tensor的最大最小值

import torch as t
a=t.linspace(0,15,6).view(2,3)
print(a)
b=t.linspace(15,0,6).view(2,3)
print(b)
print(a>b)
print(a[a>b])
print(t.max(a))
print(t.max(b,dim=1))
print(t.max(a,b))
print(t.clamp(a,min=10))

-------------------------------------在pycharm中的运行结果-------------------------------------------

(3)线性代数

函数	功能
trace	对角线元素之和
diag	对角线元素
triu/tril	矩阵上三角/下三角
mm/bmm	矩阵乘法/batch的矩阵乘法
addmm/addbmm/addmv	矩阵运算
t	转置
dot/cross	内积/外积
inverse	求逆矩阵
svd	奇异值分解

import torch as t
a=t.linspace(0,15,6).view(2,3)
b=a.t()
print(b.is_contiguous())
print(b.contiguous())

-------------------------------------在pycharm中的运行结果-------------------------------------------

（4）tensor和numpy

import torch as t
import numpy as np
a=np.ones([2,3],dtype=np.float32)
print(a)
b=t.from_numpy(a)
print(b)
b=t.Tensor(a)
print(b)
a[0,1]=100
print(b)
c=b.numpy
print(c)

-------------------------------------在pycharm中的运行结果-------------------------------------------

 (5)

import torch as t
import numpy as np
a=t.ones(3,2)
b=t.zeros(2,3,1)
print(a+b)
print(a.view(1,3,2).expand(2,3,2)+b.expand(2,3,2))
print(a.unsqueeze(0).expand(2,3,2)+b.expand(2,3,2))
e=a.unsqueeze(0).expand(10000000000000,3,2)
print(e)

(6)

a[1]=100
print(b)
c=a[2:]
print(c.storage())
print(c.data_prt())
print(a.data_ptr())
c[0]=-100
print(a)
d=t.Tensor(c.storage())
d[0]=6666
print(b)
print(id(a.storage())==id(b.storage())==id(c.storage())==id(d.storage()))
print(a.storage_offset())
print(c.storage_offset())
print(d.storage_offset())
e=b[::2,::2]
print(id(a.storage())==id(a.storage()))
print(b.stride())
print(e.stride())
print(e.is_contiguous())

(7)

import torch as t
import timeit
a=t.arange(0,6)
print(a.storage())
b=a.view(2,3)
print(b.storage())
print(id(b.storage())==id(a.storage()))
def for_loop_add(x,y):
    result = []
    for i, j in zip(x, y):
        result.append(i + j)
    return t.Tensor(result)

a = t.zeros(100)
b = t.ones(100)
t2 = timeit.Timer('a+b')
print(t2)
t1=timeit.Timer('for_loop_add(a,b)')
print(t1)

(8)

import torch as t
a=t.arange(0,2000000)
print(a[-1],a[-2])
b=t.LongTensor()
t.arange(0,20000000,out=b)
print(b[-1])
print(b[-2])
a=t.randn(2,3)
print(a)
t.set_printoptions(precision=10)
print(a)

二、线性回归项目

import torch as t
from matplotlib import pyplot as plt
from IPython import display
#设置随机数种子，保证在不同计算机上运行时下面的输出一致
import os
os.environ["KMP_DUPLICATE_LIB_OK"]="TRUE"
t.manual_seed(1000)
def get_fake_data(batch_size=8):
    '''产生随机数据：y=x*2+3,加上了一些噪音'''
    x = t.rand(batch_size, 1)*20   #生成size为(batch_size, 1)的二维数组，并元素乘20
    y = x * 2 + (1+t.rand(batch_size, 1))*3   #用于生成噪音
    return x,y
#来看看产生的x-y分布，输出如图所示
x, y = get_fake_data()
plt.scatter(x.squeeze().numpy(), y.squeeze().numpy())
plt.show()

 

import torch as t
from matplotlib import pyplot as plt
from IPython import display
import os
os.environ["KMP_DUPLICATE_LIB_OK"]="TRUE"


def get_fake_data(batch_size=8):
    x = t.rand(batch_size, 1) * 20  
    y = x * 2 + (1 + t.rand(batch_size, 1)) * 3  
    return x, y
w = t.rand(1, 1)
b = t.zeros(1, 1)
lr = 0.001  
for ii in range(2000):  
    x, y = get_fake_data()  
    y_pred = x.mm(w) + b.expand_as(y)
    loss = 0.5 * (y_pred - y) ** 2  
    loss = loss.sum()
    dloss = 1
    dy_pred = dloss * (y_pred - y)
    dw = x.t().mm(dy_pred)
    db = dy_pred.sum()
    w.sub_(lr * dw)
    b.sub_(lr * db)
    if ii % 1000 == 0: 
        display.clear_output(wait=True)
        x = t.arange(0, 20).view(-1, 1).float()
        y = x.mm(w) + b.expand_as(x)
        plt.plot(x.numpy(), y.numpy())
        x2, y2 = get_fake_data(batch_size=20)
        plt.scatter(x2.numpy(), y2.numpy())  # true data
        plt.xlim(0, 20)
        plt.ylim(0, 41)
        plt.show()
        plt.pause(0.5)
print(w.squeeze()[0], b.squeeze()[0])