DEV Community

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

Posted on • Edited on

1

cumsum and cumprod in PyTorch

Buy Me a Coffee

*Memos:

cumsum() can get the 0D or more D tensor of zero or more cumulative sum's elements from the 0D or more D tensor of zero or more elements as shown below:

*Memos:

  • cumsum() can be used with torch or a tensor.
  • The 1st argument(input) with torch or using a tensor(Required-Type:tensor of int, float, complex or bool).
  • The 2nd argument with torch or the 1st argument with a tensor is dim(Required-Type:int).
  • There is dtype argument with torch(Optional-Default:None-Type:dtype): *Memos:
    • If it's None, it's inferred from input or a tensor.
    • dtype= must be used.
    • My post explains dtype argument.
  • There is out argument with torch(Optional-Default:None-Type:tensor): *Memos:
    • out= must be used.
    • My post explains out argument.
import torch

my_tensor = torch.tensor([1, 2, 3, 4])

torch.cumsum(input=my_tensor, dim=0)
my_tensor.cumsum(dim=0)
torch.cumsum(input=my_tensor, dim=-1)
# tensor([1, 3, 6, 10])

my_tensor = torch.tensor([[1, 2, 3, 4],
                          [5, 6, 7, 8]])
torch.cumsum(input=my_tensor, dim=0)
torch.cumsum(input=my_tensor, dim=-2)
# tensor([[1, 2, 3, 4],
#         [6, 8, 10, 12]])

torch.cumsum(input=my_tensor, dim=1)
torch.cumsum(input=my_tensor, dim=-1)
# tensor([[1, 3, 6, 10],
#         [5, 11, 18, 26]])

my_tensor = torch.tensor([[[1, 2], [3, 4]],
                          [[5, 6], [7, 8]]])
torch.cumsum(input=my_tensor, dim=0)
torch.cumsum(input=my_tensor, dim=-3)
# tensor([[[1, 2], [3, 4]],
#         [[6, 8], [10, 12]]])

torch.cumsum(input=my_tensor, dim=1)
torch.cumsum(input=my_tensor, dim=-2)
# tensor([[[1, 2], [4, 6]],
#         [[5, 6], [12, 14]]])

torch.cumsum(input=my_tensor, dim=2)
torch.cumsum(input=my_tensor, dim=-1)
# tensor([[[1, 3], [3, 7]],
#         [[5, 11], [7, 15]]])

my_tensor = torch.tensor([[[1., 2.], [3., 4.]],
                          [[5., 6.], [7., 8.]]])
torch.cumsum(input=my_tensor, dim=0)
# tensor([[[1., 2.], [3., 4.]],
#         [[6., 8.], [10., 12.]]])

my_tensor = torch.tensor([[[1.+0.j, 2.+0.j], [3.+0.j, 4.+0.j]],
                          [[5.+0.j, 6.+0.j], [7.+0.j, 8.+0.j]]])
torch.cumsum(input=my_tensor, dim=0)
# tensor([[[1.+0.j, 2.+0.j], [3.+0.j, 4.+0.j]],
#         [[6.+0.j, 8.+0.j], [10.+0.j, 12.+0.j]]])

my_tensor = torch.tensor([[[True, False], [True, False]],
                          [[False, True], [False, True]]])
torch.cumsum(input=my_tensor, dim=0)
# tensor([[[1, 0], [1, 0]],
#         [[1, 1], [1, 1]]])
Enter fullscreen mode Exit fullscreen mode

cumprod() can get the 0D or more D tensor of zero or more cumulative product's elements from the 0D or more D tensor of zero or more elements as shown below:

*Memos:

  • cumprod() can be used with torch or a tensor.
  • The 1st argument(input) with torch or using a tensor(Required-Type:tensor of int, float, complex or bool).
  • The 2nd argument with torch or the 1st argument with a tensor is dim(Required-Type:int).
  • There is dtype argument with torch(Optional-Default:None-Type:dtype): *Memos:
    • If it's None, it's inferred from input or a tensor.
    • dtype= must be used.
    • My post explains dtype argument.
  • There is out argument with torch(Optional-Default:None-Type:tensor): *Memos:
    • out= must be used.
    • My post explains out argument.
import torch

my_tensor = torch.tensor([1, 2, 3, 4])

torch.cumprod(input=my_tensor, dim=0)
my_tensor.cumprod(dim=0)
torch.cumprod(input=my_tensor, dim=-1)
# tensor([1, 2, 6, 24])

my_tensor = torch.tensor([[1, 2, 3, 4],
                          [5, 6, 7, 8]])
torch.cumprod(input=my_tensor, dim=0)
torch.cumprod(input=my_tensor, dim=-2)
# tensor([[1, 2, 3, 4],
#         [5, 12, 21, 32]])

torch.cumprod(input=my_tensor, dim=1)
torch.cumprod(input=my_tensor, dim=-1)
# tensor([[1, 2, 6, 24],
#         [5, 30, 210, 1680]])

my_tensor = torch.tensor([[[1, 2], [3, 4]],
                          [[5, 6], [7, 8]]])
torch.cumprod(input=my_tensor, dim=0)
torch.cumprod(input=my_tensor, dim=-3)
# tensor([[[1, 2], [3, 4]],
#         [[5, 12], [21, 32]]])

torch.cumprod(input=my_tensor, dim=1)
torch.cumprod(input=my_tensor, dim=-2)
# tensor([[[1, 2], [3, 8]],
#         [[5, 6], [35, 48]]])

torch.cumprod(input=my_tensor, dim=2)
torch.cumprod(input=my_tensor, dim=-1)
# tensor([[[1, 2], [3, 12]],
#         [[5, 30], [7, 56]]])

my_tensor = torch.tensor([[[1., 2.], [3., 4.]],
                          [[5., 6.], [7., 8.]]])
torch.cumprod(input=my_tensor, dim=0)
# tensor([[[1., 2.], [3., 4.]],
#         [[5., 12.], [21., 32.]]])

my_tensor = torch.tensor([[[1.+0.j, 2.+0.j], [3.+0.j, 4.+0.j]],
                          [[5.+0.j, 6.+0.j], [7.+0.j, 8.+0.j]]])
torch.cumprod(input=my_tensor, dim=0)
# tensor([[[1.+0.j, 2.+0.j], [3.+0.j, 4.+0.j]],
#         [[ 5.+0.j, 12.+0.j], [21.+0.j, 32.+0.j]]])

my_tensor = torch.tensor([[[True, False], [True, False]],
                          [[False, True], [False, True]]])
torch.cumprod(input=my_tensor, dim=0)
# tensor([[[1, 0], [1, 0]],
#         [[0, 0], [0, 0]]])
Enter fullscreen mode Exit fullscreen mode

Heroku

Build apps, not infrastructure.

Dealing with servers, hardware, and infrastructure can take up your valuable time. Discover the benefits of Heroku, the PaaS of choice for developers since 2007.

Visit Site

Top comments (0)

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

👋 Kindness is contagious

Please leave a ❤️ or a friendly comment on this post if you found it helpful!

Okay