One . torch.sort() Function analysis

1. The website links

torch.sort(), As shown in the figure below ：

2. torch.sort() Function analysis

torch.sort(input, dim=- 1, descending=False, stable=False, *, out=None)

Input input, stay dim Sort dimensions , The default is dim=-1 Sort the last dimension ,descending Indicates whether to arrange in descending order , The default is False, Output the sorted value and the corresponding value in the original input imput Subscript in

3. The code for

3.1 dim = -1 Indicates that the elements in each row are sorted in ascending order ,descending=False Indicates ascending sort

x = torch.randn(3, 4)
sorted, indices = torch.sort(x)
x,sorted,indices

 The output is as follows ：
(tensor([[-1.3864,  0.5811, -0.1056, -0.3237],
         [-0.2136, -1.4806,  0.4986,  0.9382],
         [-0.2820,  0.1171, -0.3983, -0.8061]]),
 tensor([[-1.3864, -0.3237, -0.1056,  0.5811],
         [-1.4806, -0.2136,  0.4986,  0.9382],
         [-0.8061, -0.3983, -0.2820,  0.1171]]),
 tensor([[0, 3, 2, 1],
         [1, 0, 2, 3],
         [3, 2, 0, 1]]))

3.2 dim = 0 Indicates that the elements in each column are sorted in ascending order ,descending=False Indicates ascending sort

x = torch.randn(3, 4)
sorted, indices = torch.sort(x,dim=0)
x,sorted,indices

 The output is as follows ：
(tensor([[ 0.7081,  1.0502,  2.0434, -0.2592],
         [ 1.2052,  0.8809,  0.5771,  1.2978],
         [-1.5873, -0.4808, -2.1774, -0.2503]]),
 tensor([[-1.5873, -0.4808, -2.1774, -0.2592],
         [ 0.7081,  0.8809,  0.5771, -0.2503],
         [ 1.2052,  1.0502,  2.0434,  1.2978]]),
 tensor([[2, 2, 2, 0],
         [0, 1, 1, 2],
         [1, 0, 0, 1]]))

3.3 dim = 0 Means to sort the elements in each column in descending order ,descending=True Represents a descending sort

x = torch.randn(3, 4)
sorted, indices = torch.sort(x,dim=0,descending=True)
x,sorted,indices

 The output is as follows ：
(tensor([[ 0.9142, -0.2178,  0.5602,  2.3951],
         [-0.6977,  0.4915,  0.3988,  0.6406],
         [ 0.4880,  1.1646, -0.3466,  0.5801]]),
 tensor([[ 0.9142,  1.1646,  0.5602,  2.3951],
         [ 0.4880,  0.4915,  0.3988,  0.6406],
         [-0.6977, -0.2178, -0.3466,  0.5801]]),
 tensor([[0, 2, 0, 0],
         [2, 1, 1, 1],
         [1, 0, 2, 2]]))

3.4 dim = 1 Means to sort the elements in each row in descending order ,descending=True Represents a descending sort

x = torch.randn(3, 4)
sorted, indices = torch.sort(x,dim=1,descending=True)
x,sorted,indices

 The output is as follows ：
(tensor([[-0.3048, -1.9915, -0.0888,  0.3881],
         [ 1.0677, -1.3520,  0.2944, -0.0772],
         [-0.9409, -0.9630, -0.7946,  1.4400]]),
 tensor([[ 0.3881, -0.0888, -0.3048, -1.9915],
         [ 1.0677,  0.2944, -0.0772, -1.3520],
         [ 1.4400, -0.7946, -0.9409, -0.9630]]),
 tensor([[3, 2, 0, 1],
         [0, 2, 3, 1],
         [3, 2, 0, 1]]))

Two .torch.argsort() Function analysis

1. The website links

torch.argsort(), As shown in the figure below ：

2. torch.argsort() Function analysis

Usage follow the above torch.sort() The function is the same , The difference is torch.argsort() Return only the original input corresponding to the sorted value input The subscript , namely torch.sort() Back to indices

3. The code for

dim = 1 Means to sort the elements in each row in descending order ,descending=True Represents a descending sort , The output result is the original input corresponding to the returned sorted value input The subscript indices

x = torch.randn(3, 4)
indices = torch.argsort(x,dim=1,descending=True)
x,indices

 The output is as follows ：
(tensor([[-0.6069, -0.9252, -0.9177,  0.6997],
         [ 0.3245, -0.0665,  0.4600,  0.0722],
         [-1.0662,  2.2669, -0.1171, -0.9208]]),
 tensor([[3, 0, 2, 1],
         [2, 0, 3, 1],
         [1, 2, 3, 0]]))