DEV Community

Super Kai (Kazuya Ito)
Super Kai (Kazuya Ito)

Posted on

1

pow in PyTorch

Buy Me a Coffee

*Memos:

pow() can get the 0D or more D tensor of zero or more powers from two of the 0D or more D tensors of zero or more elements or the 0D or more D tensor of zero or more elements and a scalar as shown below:

*Memos:

  • pow() can be used with torch or a tensor.
  • The 1st argument(input) with torch(Required-Type:tensor or scalar of int, float or complex) or using a tensor(Required-Type:tensor of int, float or complex). *torch must use a scalar without input=.
  • The 2nd argument with torch or the 1st argument with a tensor is exponent(Required-Type:tensor or scalar of int, float or complex). *A negative scalar cannot be used.
  • There is out argument with torch(Optional-Default:None-Type:tensor): *Memos:
    • out= must be used.
    • My post explains out argument.
  • The combination of a scalar(input or a tensor) and a scalar(exponent) cannot be used.
  • The combination of a tensor(input(bool) or a tensor(bool)) and a scalar(exponent(bool)) works.
import torch

tensor1 = torch.tensor(-3)
tensor2 = torch.tensor([-4, -3, -2, -1, 0, 1, 2, 3])

torch.pow(input=tensor1, exponent=tensor2)
tensor1.pow(exponent=tensor2)
# tensor([0, 0, 0, 0, 1, -3, 9, -27])

torch.pow(-3, exponent=tensor2)
# tensor([0, 0, 0, 0, 1, -3, 9, -27])

torch.pow(input=tensor1, exponent=3)
# tensor(-27)

tensor1 = torch.tensor([-3, 1, -2, 3, 5, -5, 0, -4])
tensor2 = torch.tensor([-4, -3, -2, -1, 0, 1, 2, 3])

torch.pow(input=tensor1, exponent=tensor2)
# tensor([0, 1, 0, 0, 1, -5, 0, -64])

torch.pow(-3, exponent=tensor2)
# tensor([0, 0, 0, 0, 1, -3, 9, -27])

torch.pow(input=tensor1, exponent=3)
# tensor([-27, 1, -8, 27, 125, -125, 0, -64])

tensor1 = torch.tensor([[-3, 1, -2, 3], [5, -5, 0, -4]])
tensor2 = torch.tensor([0, 1, 2, 3])

torch.pow(input=tensor1, exponent=tensor2)
# tensor([[1, 1, 4, 27], [1, -5, 0, -64]])

torch.pow(-3, exponent=tensor2)
# tensor([1, -3, 9, -27])

torch.pow(input=tensor1, exponent=3)
# tensor([[-27, 1, -8, 27], [125, -125, 0, -64]])

tensor1 = torch.tensor([[[-3, 1], [-2, 3]],
                        [[5, -5], [0, -4]]])
tensor2 = torch.tensor([2, 3])

torch.pow(input=tensor1, exponent=tensor2)
# tensor([[[9, 1], [4, 27]],
#         [[25, -125], [0, -64]]])

torch.pow(-3, exponent=tensor2)
# tensor([9, -27])

torch.pow(input=tensor1, exponent=3)
# tensor([[[-27, 1], [-8, 27]],
#         [[125, -125], [0, -64]]])

tensor1 = torch.tensor([[[-3., 1.], [-2., 3.]],
                        [[5., -5.], [0., -4.]]])
tensor2 = torch.tensor([2., 3.])

torch.pow(input=tensor1, exponent=tensor2)
# tensor([[[9., 1.], [4., 27.]],
#         [[25., -125.], [0., -64.]]])

torch.pow(-3., exponent=tensor2)
# tensor([9., -27.])

torch.pow(input=tensor1, exponent=3.)
# tensor([[[-27., 1.], [-8., 27.]],
#         [[125., -125.], [0., -64.]]])

tensor1 = torch.tensor([[[-3.+0.j, 1.+0.j], [-2.+0.j, 3.+0.j]],
                        [[5.+0.j, -5.+0.j], [0.+0.j, -4.+0.j]]])
tensor2 = torch.tensor([2.+0.j, 3.+0.j])

torch.pow(input=tensor1, exponent=tensor2)
# tensor([[[9.0000+1.5736e-06j, 1.0000+0.0000e+00j],
#          [4.0000+6.9938e-07j, 27.0000+0.0000e+00j]],
#         [[25.0000+0.0000e+00j, -125.0000-2.9812e-06j],
#          [0.0000-0.0000e+00j, -64.0000-1.5264e-06j]]])

torch.pow(-3.+0.j, exponent=tensor2)
# tensor([9.0000+1.5736e-06j, -27.0000-6.4394e-07j])

torch.pow(input=tensor1, exponent=3.+0.j)
# tensor([[[-27.+0.j, 1.+0.j],
#          [-8.+0.j, 27.+0.j]],
#         [[125.+0.j, -125.+0.j],
#          [0.+0.j, -64.+0.j]]])

my_tensor = torch.tensor([[[True, False], [True, False]],
                          [[False, True], [False, True]]])
torch.pow(input=my_tensor, exponent=True)
# tensor([[[True, False], [True, False]],
#         [[False, True], [False, True]]])
Enter fullscreen mode Exit fullscreen mode

Billboard image

The Next Generation Developer Platform

Coherence is the first Platform-as-a-Service you can control. Unlike "black-box" platforms that are opinionated about the infra you can deploy, Coherence is powered by CNC, the open-source IaC framework, which offers limitless customization.

Learn more

Top comments (0)

Image of Docusign

Bring your solution into Docusign. Reach over 1.6M customers.

Docusign is now extensible. Overcome challenges with disconnected products and inaccessible data by bringing your solutions into Docusign and publishing to 1.6M customers in the App Center.

Learn more