点乘/hadamard product

使用 torch.mul 。

  1. In [14]: import numpy as np
  3. In [15]: a = torch.from_numpy(np.arange(12).reshape(3, 4))
  5. In [16]: a
  6. Out[16]:
  7. tensor([[ 0,  1,  2,  3],
  8.         [ 4,  5,  6,  7],
  9.         [ 8,  9, 10, 11]], dtype=torch.int32)
  11. In [17]: b = torch.from_numpy(np.arange(12).reshape(3, 4))
  13. In [18]: c = torch.mul(a, b)
  15. In [19]: c
  16. Out[19]:
  17. tensor([[  0,   1,   4,   9],
  18.         [ 16,  25,  36,  49],
  19.         [ 64,  81, 100, 121]], dtype=torch.int32)

使用 torch.mm 进行矩阵乘法matrix multiply。

  1. In [21]: a
  2. Out[21]:
  3. tensor([[ 0,  1,  2,  3],
  4.         [ 4,  5,  6,  7],
  5.         [ 8,  9, 10, 11]], dtype=torch.int32)
  7. In [22]: b = torch.from_numpy(np.arange(12).reshape(4, 3))
  9. In [23]: d = torch.mm(a, b)
  11. In [24]: d
  12. Out[24]:
  13. tensor([[ 42,  48,  54],
  14.         [114, 136, 158],
  15.         [186, 224, 262]], dtype=torch.int32)