李沐视频04代码动手敲

import torch
x = torch.arange(12,0,-1)
print(x)
# shape返回torch.Size (行为类似一个元组 tuple)
print(x.shape)

# numel()返回int
print(x.numel())

x = x.reshape(3,4)

print(x)
# shape返回torch.Size (行为类似一个元组 tuple)
print(x.shape)

# numel()返回int
print(x.numel())

print(torch.zeros(2,3,4))
print(torch.ones(2,3,4))

y=torch.tensor([[2,1,4,3],[1,2,3,4],[6,7,8,9]])
print(y)
print(y.shape) 

print(torch.exp(y))

xx = torch.arange(12,dtype=torch.float32).reshape((3,4))
yy= torch.tensor([[2.0,1,4,3],[1,2,6,8],[2,4,6,8]])
print(torch.cat((xx,yy),dim=0))
print(torch.cat((xx,yy),dim=1))

print(xx==yy)
# 属性不加括号，函数加括号
print(xx.sum())

a = torch.arange(3).reshape((3,1))
b = torch.arange(2).reshape((1,2))
print(a)
print(b)
# 广播机制
print(a+b)


# 原地操作
z = torch.zeros_like(x)
before = id(z)
z[:]=x+y
print(id(z)==before)

# 也可以使用x[:]=x+y或x+=y减少内存

# numpy类型和tensor类型的互相转化
print(x)
aa = x.numpy()
bb = torch.tensor(aa)
print(aa)
print(type(aa))
print(bb)
print(type(bb))

a = torch.tensor([3.5])
print(a)
print(type(a))
print(a.item())
print(type(a.item()))
print(int(a))
print(type(int(a)))

输出

tensor([12, 11, 10,  9,  8,  7,  6,  5,  4,  3,  2,  1])
torch.Size([12])
12
tensor([[12, 11, 10,  9],
        [ 8,  7,  6,  5],
        [ 4,  3,  2,  1]])
torch.Size([3, 4])
12
tensor([[[0., 0., 0., 0.],
         [0., 0., 0., 0.],
         [0., 0., 0., 0.]],

        [[0., 0., 0., 0.],
         [0., 0., 0., 0.],
         [0., 0., 0., 0.]]])
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.]]])
tensor([[2, 1, 4, 3],
        [1, 2, 3, 4],
        [6, 7, 8, 9]])
torch.Size([3, 4])
tensor([[7.3891e+00, 2.7183e+00, 5.4598e+01, 2.0086e+01],
        [2.7183e+00, 7.3891e+00, 2.0086e+01, 5.4598e+01],
        [4.0343e+02, 1.0966e+03, 2.9810e+03, 8.1031e+03]])
tensor([[ 0.,  1.,  2.,  3.],
        [ 4.,  5.,  6.,  7.],
        [ 8.,  9., 10., 11.],
        [ 2.,  1.,  4.,  3.],
        [ 1.,  2.,  6.,  8.],
        [ 2.,  4.,  6.,  8.]])
tensor([[ 0.,  1.,  2.,  3.,  2.,  1.,  4.,  3.],
        [ 4.,  5.,  6.,  7.,  1.,  2.,  6.,  8.],
        [ 8.,  9., 10., 11.,  2.,  4.,  6.,  8.]])
tensor([[False,  True, False,  True],
        [False, False,  True, False],
        [False, False, False, False]])
tensor(66.)
tensor([[0],
        [1],
        [2]])
tensor([[0, 1]])
tensor([[0, 1],
        [1, 2],
        [2, 3]])
True
tensor([[12, 11, 10,  9],
        [ 8,  7,  6,  5],
        [ 4,  3,  2,  1]])
[[12 11 10  9]
 [ 8  7  6  5]
 [ 4  3  2  1]]
<class 'numpy.ndarray'>
tensor([[12, 11, 10,  9],
        [ 8,  7,  6,  5],
        [ 4,  3,  2,  1]])
<class 'torch.Tensor'>
tensor([3.5000])
<class 'torch.Tensor'>
3.5
<class 'float'>
3
<class 'int'>