ピトーチテンソルインデックス

2022-02-21 03:25:59

i. pytorchは最も値の大きいインデックスを返す。

1 公式ドキュメント

1.1 torch.argmax()の紹介

最大値の添え字を返します。

Function.
     torch.argmax(input, dim, keepdim=False) → LongTensor

Return value.
     Returns the indices of the maximum values of a tensor across a dimension.

Parameters.
	input (Tensor) - the input tensor.
	If None, the argmax of the flattened input is returned.
	keepdim (bool) - whether the output tensor has dim retained or not. Ignored if dim=None.

1.2 torch.argmin()の紹介

最小の値の添え字を返します。

Function.
     torch.argmin(input, dim, keepdim=False) → LongTensor

Return value.
     Returns the indices of the mimimum values of a tensor across a dimension.

Parameters.
	input (Tensor) - the input tensor.
	If None, the argmax of the flattened input is returned.
	keepdim (bool) - whether the output tensor has dim retained or not. Ignored if dim=None.

2 コード例

2.1 torch.argmax()のコード例

>>> import torch
>>> Matrix = torch.randn(2,2,2)
>>> print(Matrix)
tensor([[ 0.3772, -0.1143],
         [[ 0.2217, -0.1897]],

        [[ 0.1488, -0.8758]],
         [ 1.7734, -0.5929]]])
>>> print(Matrix.argmax(dim=0))
tensor([[0, 0],
        [1, 0]])
>>> print(Matrix.argmax(dim=1))
tensor([[0, 0],
        [1, 1]])
>>> print(Matrix.argmax(dim=2))
tensor([[0, 0],
        [0, 0]])
>>> print(Matrix.argmax())
tensor(6)

2.2 torch.argmin()のコードからわかること

>>> import torch
>>> Matrix = torch.randn(2,2,2)
>>> print(Matrix)
tensor([[ 0.5821, 0.2889],
         [[ 0.4669, -0.3135]],

        [[-0.4567, 0.2975]],
         [-1.5084, 0.7320]]])
>>> print(Matrix.argmin(dim=0))
tensor([[1, 0],
        [1, 0]])
>>> print(Matrix.argmin(dim=1))
tensor([[1, 1],
        [1, 0]])
>>> print(Matrix.argmin(dim=2))
tensor([[1, 1],
        [0, 0]])
>>> print(Matrix.argmin())
tensor(6)

II. pytorchは任意の値のインデックスを返す。

tens = tensor([ 101, 146, 1176, 21806, 1116, 1105, 18621, 119, 102, 0,
             0, 0, 0],
        [ 101, 1192, 1132, 1136, 1184, 146, 1354, 1128, 1127, 117,
          1463, 119, 102],
        [ 101, 6816, 1905, 1132, 14918, 119, 102, 0, 0, 0, 0,
             0, 0, 0]])
idxs = torch.tensor([(i == 101).nonzero() for i in tens])

from torch import tensor                                                                       

tens = torch.tensor([[ 101, 146, 1176, 21806, 1116, 1105, 18621, 119, 102, 0, 
    ... : 0, 0, 0], 
    ... : [ 101, 1192, 1132, 1136, 1184, 146, 1354, 1128, 1127, 117, 
    ... : 1463, 119, 102], 
    ... : [ 101, 6816, 1905, 1132, 14918, 119, 102, 0, 0, 0, 
    ... : 0, 0, 0]])                                                                

(tens == 101).nonzero()[:, 1]                                                                  
tensor([0, 0, 0])

iii. pytorchはテンソルの最大値または最小値のみを保持し、他の位置はゼロに設定します。

以下のように、xは入力テンソル、dimは次元を指定し、maxはminで置き換えることができる。


import torch

if __name__ == '__main__':
    
    x = torch.randn([1, 3, 4, 4]).cuda()

    mask = (x == x.max(dim=1, keepdim=True)[0]).to(dtype=torch.int32)
    result = torch.mul(mask, x)

    print(x)
    print(mask)
    print(result)

効果を出力します。

tensor([[[[-0.8807, 0.1029, 0.0184, 1.2695],
          [-0.0934, 1.0650, -0.2927, 0.0049],
          [ 0.2338, -1.8663, 1.2763, 0.7248],
          [ -1.5138, 0.6834, 0.1463, 0.0650]],

         [[ 0.5020, 1.6078, -0.0104, 1.2042],
          [ 1.8859, -0.4682, -0.1177, 0.5197],
          [ 1.7649, 0.4585, 0.6002, 0.3350],
          [-1.1384, -0.0325, 0.8490, 0.6080]],

         [[-0.5618, 0.5388, -0.0572, -0.7240]],
          [-0.3458, 1.3494, -0.0603, -1.1562],
          [-0.3652, 1.1885, 1.6293, 0.4134],
          [ 1.3009, 1.2027, -0.8711, 1.3321]]]], device='cuda:0')
tensor([[[[0, 0, 1, 1],
          [0, 0, 0, 0],
          [0, 0, 0, 1],
          [0, 0, 0, 0, 0]],

         [[1, 1, 0, 0],
          [1, 0, 0, 1],
          [1, 0, 0, 0],
          [0, 0, 1, 0]],

         [[0, 0, 0, 0],
          [0, 1, 1, 0],
          [0, 1, 1, 0],
          [1, 1, 0, 1]]]], device='cuda:0', dtype=torch.int32)
tensor([[[[-0.0000, 0.0000, 0.0184, 1.2695],
          [-0.0000, 0.0000, -0.0000, 0.0000],
          [ 0.0000, -0.0000, 0.0000, 0.7248],
          [ -0.0000, 0.0000, 0.0000, 0.0000]],

         [[ 0.5020, 1.6078, -0.0000, 0.0000],
          [ 1.8859, -0.0000, -0.0000, 0.5197],
          [ 1.7649, 0.0000, 0.0000, 0.0000],
          [-0.0000, -0.0000, 0.8490, 0.0000]],

         [[-0.0000, 0.0000, -0.0000, -0.0000],
          [-0.0000, 1.3494, -0.0603, -0.0000],
          [-0.0000, 1.1885, 1.6293, 0.0000],
          [ 1.3009, 1.2027, -0.0000, 1.3321]]]], device='cuda:0')

Process finished with exit code 0