torch创建

pytorch , , 评论

@TOC

创建tensor

1. 指定tensor

  • 如果tensor都是整形，默认创建的都是torch.int64
  • 如果tensor有一个浮点型，创建的就是torch.float32
  • 也可以用tensor.type()来改变tensor类型
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
29
tensor = torch.tensor([[1, 2, 3], [4, 5, 6], [7, 8, 9]])
print(tensor, '\n', tensor.dtype)
'''
tensor([[1, 2, 3],
        [4, 5, 6],
        [7, 8, 9]]) 
 torch.int64
'''


tensor = torch.tensor([[1.1, 2.2, 3.3], [4.4, 5.5, 6.6], [7.7, 8.8, 9.9]])
print(tensor, '\n', tensor.dtype)
'''
tensor([[1.1000, 2.2000, 3.3000],
        [4.4000, 5.5000, 6.6000],
        [7.7000, 8.8000, 9.9000]]) 
 torch.float32
'''


tensor = torch.tensor([[1.1, 2.2, 3.3], [4.4, 5.5, 6.6], [7.7, 8.8, 9.9]])
tensor = tensor.type(torch.float64)
print(tensor, '\n', tensor.dtype)
'''
tensor([[1.1000, 2.2000, 3.3000],
        [4.4000, 5.5000, 6.6000],
        [7.7000, 8.8000, 9.9000]], dtype=torch.float64) 
 torch.float64
'''

2. 创建特殊类型的tensor

（1）空tensor(empty)

torch.empty(size)

返回一个未初始化的tensor（初始化比较随意）

Returns a tensor filled with uninitialized data. The shape of the tensor is defined by the variable argument size.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
tensor = torch.empty((2, 3))
print(tensor)
'''
tensor([[1.2332e+34, 4.5827e-41, 0.0000e+00],
        [0.0000e+00, 0.0000e+00, 0.0000e+00]])
'''


tensor = torch.tensor([[0, 0, 0],[0, 0, 0.]])
print(tensor)
'''
tensor([[0., 0., 0.],
        [0., 0., 0.]])
'''

（2）全为一(ones)

torch.ones(size)

返回一个全为1的tensor

Returns a tensor filled with the scalar value 1, with the shape defined by the variable argument size.

1
2
3
4
5
6
tensor = torch.ones((2, 3))
print(tensor)
'''
tensor([[1., 1., 1.],
        [1., 1., 1.]])
'''

（3）全为零(zeros)

torch.ones(size)

返回一个全为0的tensor

Returns a tensor filled with the scalar value 0, with the shape defined by the variable argument size.

1
2
3
4
5
6
tensor = torch.zeros((2, 3))
print(tensor)
'''
tensor([[0., 0., 0.],
        [0., 0., 0.]])
'''

（4）均匀分布(rand)

torch.rand(size)

返回符合的均匀分布：torch.float32

Returns a tensor filled with random numbers from a uniform distribution on the interval [0, 1)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
tensor = torch.rand(4)
print(tensor)
'''
tensor([0.1069, 0.4704, 0.1034, 0.8400])
'''


tensor = torch.rand((2, 3))
print(tensor)
print(tensor.type(), tensor.dtype)
'''
tensor([[0.8729, 0.0053, 0.2537],
        [0.5580, 0.3945, 0.5647]])
torch.FloatTensor torch.float32
'''

（5）正态分布(randn)

torch.randn(size)

返回的正态分布：torch.float32

Returns a tensor filled with random numbers from a normal distribution with mean 0 and variance 1 (also called the standard normal distribution).

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
tensor = torch.randn(4)
print(tensor)
'''
tensor([ 2.4390, -1.1983,  0.1438,  1.4247])
'''


tensor = torch.randn((2, 3))
print(tensor)
print(tensor.type(), tensor.dtype)
'''
tensor([[-0.2153,  1.9469, -1.0524],
        [-1.2857,  0.3539, -0.2951]])
torch.FloatTensor torch.float32
'''

（6）整数范围(randint)

torch.randint(low=0, high, size)

返回随机tensor，每一项的取值范围为：torch.int64

Returns a tensor filled with random integers generated uniformly between low (inclusive) and high (exclusive).

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
tensor = torch.randint(3, 5, (3,))
print(tensor)
'''
tensor([3, 4, 4])
'''


tensor = torch.randint(10, (2, 2))
print(tensor)
'''
tensor([[0, 9],
        [6, 2]])
'''


tensor = torch.randint(3, 10, (2, 2))
print(tensor)
print(tensor.type(), tensor.dtype)
'''
tensor([[3, 3],
        [7, 3]])
torch.LongTensor torch.int64
'''

（7）N范围内的随机序列(randperm)

torch.randperm(n)

返回之间的随机tensor序列：torch.int64

Returns a random permutation of integers from 0 to n - 1.

可以用torch.randperm()来将tensor打乱顺序，如下所示：x输出按照[2, 3, 0, 1]排列

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
tensor = torch.randperm(4)
print(tensor)
print(tensor.type(), tensor.dtype)
'''
tensor([2, 3, 0, 1])
torch.LongTensor torch.int64
'''


x = torch.randn(4, 2)
print(x)
print(x[tensor])
'''
tensor([[ 1.0033, -0.3932],
        [-0.1972, -0.0260],
        [-1.7497, -1.8479],
        [ 1.3488, -1.1916]])
tensor([[-1.7497, -1.8479],
        [ 1.3488, -1.1916],
        [ 1.0033, -0.3932],
        [-0.1972, -0.0260]])
'''

（8）全是value的tensor(full)

torch.full(size, full_value)

创建一个tensor，他的值全是value

Creates a tensor of size size filled with fill_value. The tensor’s dtype is inferred from fill_value.

1
2
3
4
5
6
tensor = torch.full((2, 3), 2.649)
print(tensor)
'''
tensor([[2.6490, 2.6490, 2.6490],
        [2.6490, 2.6490, 2.6490]])
'''

3. 创建于当前tensor相同大小的tensor

假如有一个tensor1，我们想创建一个和他一样大小的tensor2，PyTorch也给我们提供了*_like函数，其中的星号可以为上述讲过的ones，zeros等

例如我们的tensor1大小为(2, 3)，我们想创建一个大小为(2 ,3)并且是全1的tensor，就可以使用ones_like()函数

1
2
3
4
5
6
7
tensor1 = torch.tensor([[1, 2, 3],[4, 5, 6]])
tensor2 = torch.ones_like(tensor1)
print(tensor2)
'''
tensor([[1, 1, 1],
        [1, 1, 1]])
'''

与上述相同的还有torch.empty_like()、torch.ones_like()、torch.zeros_like()、torch.rand_like()、torch.randn_like()、torch.randint_like()，使用方法原函数相同，不同的地方在于size换成要与之相匹配的tensor即可

harryteaustc

computer vision