pytorch创建tensor_torch 创建tenor

**

一、import from numpy

**

In [40]: a = np.array([2,3.3])

In [41]: a
Out[41]: array([2. , 3.3])

In [42]: b = torch.from_numpy(a)

In [43]: b
Out[43]: tensor([2.0000, 3.3000], dtype=torch.float64)

In [44]: a = np.ones([2,3])

In [46]: b = torch.from_numpy(a)

In [47]: b
Out[47]:
tensor([[1., 1., 1.],
        [1., 1., 1.]], dtype=torch.float64)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

**

二、import from list​

**
torch.tensor()#接收现有的数据，以List输入
torch.Tensor() torch.FloatTensor()#接收现有的数据或者数据的维度

In [48]: torch.tensor([2.,3.2])
Out[48]: tensor([2.0000, 3.2000])

In [50]: torch.FloatTensor([2.,3.2])
Out[50]: tensor([2.0000, 3.2000])

In [51]: torch.FloatTensor([[2.,3.2],[1.,22.3]])
Out[51]:
tensor([[ 2.0000,  3.2000],
        [ 1.0000, 22.3000]])
1
2
3
4
5
6
7
8
9
10

**

三、uninitialized未初始化数据

**
Torch.empty()
Torch.FloatTensor(d1,d2,d3)
Torch.IntTensor(d1,d2,d3)
未初始化的tensor一定要跟写入数据的后续步骤

In [1]: import torch

In [2]: torch.empty(1)
Out[2]: tensor([-6.0475e+26])

In [3]: torch.Tensor(2,3)
Out[3]:
tensor([[-4.4539e+19,  4.5912e-41, -1.4528e+26],
        [ 6.2218e-43,  0.0000e+00, -0.0000e+00]])

In [4]: torch.FloatTensor(2,3)
Out[4]:
tensor([[ 0.0000e+00,  0.0000e+00, -1.4539e+26],
        [ 6.2218e-43,  0.0000e+00, -0.0000e+00]])

In [5]: torch.IntTensor(2,3)
Out[5]:
tensor([[ -535132572,       32764,  -461452272],
        [        444,           0, -2147483648]], dtype=torch.int32)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

set default type

In [7]: torch.tensor([1.2,3]).type()
Out[7]: 'torch.FloatTensor'

In [8]: torch.set_default_tensor_type(torch.DoubleTensor)

In [9]: torch.tensor([1.2,3]).type()
Out[9]: 'torch.DoubleTensor'
1
2
3
4
5
6
7

**

四、随机初始化

torch.rand()#随机产生均匀分布，输入参数为shape
torch.rand_like(a)#将a.shape读出来后，送给rand函数
torch.randint(1,10,[3,3])#随机产生1-10之间的整数，其shape为(3,3)
torch.randn(3,3)#随机产生标准正态分布，输入参数为shape，均值为0，方差为1

**

In [10]: torch.rand(3,3)#随机产生0-1之间的数值，不包括1
Out[10]:
tensor([[0.5588, 0.6363, 0.8763],
        [0.3796, 0.5534, 0.7435],
        [0.3892, 0.7436, 0.8083]])

In [11]: a = torch.rand(3,3)

In [12]: torch.rand_like(a)#把a.shape读出来后，送给rand函数
Out[12]:
tensor([[0.6652, 0.6416, 0.6140],
        [0.0278, 0.4688, 0.4863],
        [0.7347, 0.9787, 0.1331]])
1
2
3
4
5
6
7
8
9
10
11
12
13

In [16]: a = torch.randint(1,10,[3,3])#随机产生1-10之间的整数，其shape为(3,3)

In [17]: a.shape
Out[17]: torch.Size([3, 3])

In [18]: a
Out[18]:
tensor([[4, 9, 3],
        [4, 1, 4],
        [1, 1, 8]])
1
2
3
4
5
6
7
8
9
10

In [19]: torch.randn(3,3)
Out[19]:
tensor([[ 1.1848,  0.0742,  0.9433],
        [-0.5318, -0.9366, -0.0956],
        [-0.3990,  0.0978,  1.6421]])

In [20]: torch.normal(mean=torch.full([10],0),std=torch.arange(1,0,-0.1))
#均值全为0，方差为[1,0.9,...,0.1]
#第1个数据从N(0,1)中采样，第2个数据从N(0,0.9)中采样...第10个数据从N(0,0.1)中采样
Out[20]:
tensor([-0.9845,  0.0048,  0.0205, -0.9091, -0.0573,  0.7027,  0.0539,  0.3360,
         0.2018,  0.0567])
1
2
3
4
5
6
7
8
9
10
11
12

**

五、full函数

**

In [21]: torch.full([2,3],7)#2行3列张量
Out[21]:
tensor([[7., 7., 7.],
        [7., 7., 7.]])

In [22]: torch.full([],7)#标量
Out[22]: tensor(7.)

In [23]: torch.full([1],7)#1维矢量
Out[23]: tensor([7.])
1
2
3
4
5
6
7
8
9
10

**

六、arange/range函数

**

In [24]: torch.arange(0,10)#步长为1，从0到10，含0不含10
Out[24]: tensor([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])

In [25]: torch.arange(0,10,2)#步长为2，从0到10，含0不含10
Out[25]: tensor([0, 2, 4, 6, 8])

In [26]: torch.range(0,10)#即将丢弃不再使用，建议不使用
D:\Anaconda3\Scripts\ipython:1: UserWarning: torch.range is deprecated in favor of torch.arange and will be removed in 0.5. Note that arange generates values in [start; end), not [start; end].
Out[26]: tensor([ 0.,  1.,  2.,  3.,  4.,  5.,  6.,  7.,  8.,  9., 10.])
1
2
3
4
5
6
7
8
9

**

七、linspace/logspace函数

**

In [27]: torch.linspace(0,10,steps=4)#0-10，含0且含10，等分4份
Out[27]: tensor([ 0.0000,  3.3333,  6.6667, 10.0000])

In [28]: torch.linspace(0,10,steps=10)#0-10，含0且含10，等分10份
Out[28]:
tensor([ 0.0000,  1.1111,  2.2222,  3.3333,  4.4444,  5.5556,  6.6667,  7.7778,
         8.8889, 10.0000])

In [29]: torch.linspace(0,11,steps=11)#0-11，含0且含11，等分11份
Out[29]:
tensor([ 0.0000,  1.1000,  2.2000,  3.3000,  4.4000,  5.5000,  6.6000,  7.7000,
         8.8000,  9.9000, 11.0000])

In [30]: torch.linspace(0,10,steps=11)#0-10，含0且含10，等分11份
Out[30]: tensor([ 0.,  1.,  2.,  3.,  4.,  5.,  6.,  7.,  8.,  9., 10.])

In [31]: torch.logspace(0,-1,steps=10)#基底默认为10,10的0次方到10的-1次方，等分10份，可修改基底
Out[31]:
tensor([1.0000, 0.7743, 0.5995, 0.4642, 0.3594, 0.2783, 0.2154, 0.1668, 0.1292,
        0.1000])

In [32]: torch.logspace(0,1,steps=10)#基底默认为10,10的0次方到10的1次方，等分10份,可修改基底
Out[32]:
tensor([ 1.0000,  1.2915,  1.6681,  2.1544,  2.7826,  3.5938,  4.6416,  5.9948,
         7.7426, 10.0000])
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25

**

八、ones/zeros/eye函数

**

In [33]: torch.eye(3,3)
Out[33]:
tensor([[1., 0., 0.],
        [0., 1., 0.],
        [0., 0., 1.]])

In [34]: torch.zeros(3,3)
Out[34]:
tensor([[0., 0., 0.],
        [0., 0., 0.],
        [0., 0., 0.]])

In [35]: a = torch.ones(3,3)

In [36]: torch.ones_like(a)
Out[36]:
tensor([[1., 1., 1.],
        [1., 1., 1.],
        [1., 1., 1.]])
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

**

九、randperm函数

**
randperm随机打散索引

In [11]: torch.randperm(10)
Out[11]: tensor([4, 8, 9, 3, 6, 1, 0, 7, 5, 2])
1
2

In [2]: a = torch.rand(2,3)

In [3]: b = torch.rand(2,2)

In [4]: idx = torch.randperm(2)

In [5]: idx
Out[5]: tensor([1, 0])

In [7]: a[idx]
Out[7]:
tensor([[0.1356, 0.8077, 0.3847],
        [0.4126, 0.8919, 0.8108]])

In [8]: a
Out[8]:
tensor([[0.4126, 0.8919, 0.8108],
        [0.1356, 0.8077, 0.3847]])

In [9]: b[idx]
Out[9]:
tensor([[0.3102, 0.5754],
        [0.6128, 0.0977]])

In [10]: b
Out[10]:
tensor([[0.6128, 0.0977],
        [0.3102, 0.5754]])
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28