pytorch基础

一、Tensor (张量)

import torch
from torch.autograd import Variable
import numpy as np

#tensor 张量
a = torch.Tensor([[2,3],[4,8],[7,9]])
print('a is:{}'.format(a))
print('a size is:{}'.format(a.size()))

a is:tensor([[2., 3.],
        [4., 8.],
        [7., 9.]])
a size is:torch.Size([3, 2])

#指定想要的类型
b = torch.LongTensor([[2,3],[4,8],[7,9]])
print('b is:{}'.format(b))

b is:tensor([[2, 3],
        [4, 8],
        [7, 9]])

#全空的tensor
c = torch.zeros((3,2))
print('zero tensor:{}'.format(c))

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

#正态分布作为随机初始值的tensor
d = torch.randn((3,2))
print('normal random is:{}'.format(d))

normal random is:tensor([[-1.1648,  1.0327],
        [ 0.9533,  0.3417],
        [ 0.4022, -0.7862]])

#用索引取值
a[0,1]=100
print('changed a is:{}'.format(a))

changed a is:tensor([[  2., 100.],
        [  4.,   8.],
        [  7.,   9.]])

#和numpy之间转换
np_b = b.numpy()
print('convert to numpy is \n {}'.format(np_b))

e = np.array([[2,3],[4,5]])
torch_e = torch.from_numpy(e)
print('from numpy to torch.Tensor is {}'.format(torch_e))

convert to numpy is 
 [[2 3]
 [4 8]
 [7 9]]
from numpy to torch.Tensor is tensor([[2, 3],
        [4, 5]])

二、Variable (变量)

#variable变量
#标量求导
x = Variable(torch.Tensor([1]),requires_grad=True)
w = Variable(torch.Tensor([2]),requires_grad=True)
b = Variable(torch.Tensor([3]),requires_grad=True)

y = w*x+b

#这一步即自动求导
y.backward()

print(x.grad)
print(w.grad)
print(b.grad)

tensor([2.])
tensor([1.])
tensor([1.])

#矩阵求导
x = torch.randn(3)
x = Variable(x,requires_grad = True)

y = x*2
print(y)

y.backward(torch.FloatTensor([1,0.1,0.01]))
print(x.grad)

tensor([ 3.3212, -3.4426, -0.1929], grad_fn=<MulBackward0>)
tensor([2.0000, 0.2000, 0.0200])