DEV Community

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

Posted on • Edited on

Set `out` argument in PyTorch

Buy Me a Coffee

*Memos:

You can set out as shown below:

*Memos:

  • I selected some popular out argument functions such as arange(), rand() add(), mean(), median(), min(), max(), all(), any() and matmul().
  • Basically, out(Optional-Default-None-Type:tensor) can have a returned tensor. *Sometimes, out(Optional-Default-None-Type:tuple(tensor, tensor) or list(tensor, tensor)).
  • Basically, out can be used with torch but not with a tensor.
  • Basically, out= must be used.
  • Sometimes, out needs to be used with dim.
  • I recommend not to use out argument because it is useless at all.

arange(). *My post explains arange():

import torch

torch.arange(start=5, end=15, step=4)
# tensor([5, 9, 13])

my_tensor = torch.tensor([0, 1, 2])

torch.arange(start=5, end=15, step=4, out=my_tensor)
# tensor([5, 9, 13])

tensor1 = torch.tensor([0, 1, 2])

tensor2 = torch.arange(start=5, end=15, step=4, out=tensor1)

tensor1, tensor2
# (tensor([5, 9, 13]), tensor([5, 9, 13]))
Enter fullscreen mode Exit fullscreen mode

rand(). *My post explains rand():

import torch

tensor1 = torch.tensor([0., 1., 2.])

tensor2 = torch.rand(size=(3,), out=tensor1)

tensor1, tensor2
# (tensor([0.3379, 0.9394, 0.5509]), tensor([0.3379, 0.9394, 0.5509]))
Enter fullscreen mode Exit fullscreen mode

add(). *My post explains add():

import torch

tensor1 = torch.tensor([1, 2, 3])
tensor2 = torch.tensor([4, 5, 6])
tensor3 = torch.tensor([7, 8, 9])

tensor4 = torch.add(input=tensor1, other=tensor2, out=tensor3)

tensor1, tensor2, tensor3, tensor4
# (tensor([1, 2, 3]), tensor([4, 5, 6]), tensor([5, 7, 9]), tensor([5, 7, 9]))
Enter fullscreen mode Exit fullscreen mode

mean(). *My post explains mean():

import torch

tensor1 = torch.tensor([5., 4., 7., 7.])
tensor2 = torch.tensor(9.)

tensor3 = torch.mean(input=tensor1, dim=0, out=tensor2)

tensor1, tensor2, tensor3
# (tensor([5., 4., 7., 7.]), tensor(5.7500), tensor(5.7500))
Enter fullscreen mode Exit fullscreen mode

median(). *My post explains median():

import torch

tensor1 = torch.tensor([5., 4., 7., 7.])
tensor2 = torch.tensor(9.)
tensor3 = torch.tensor(6)

tensor4 = torch.median(input=tensor1, dim=0, out=(tensor2, tensor3))

tensor1, tensor2, tensor3, tensor4
# (tensor([5., 4., 7., 7.]),
#  tensor(5.),
#  tensor(0),
#  torch.return_types.median_out(
#  values=tensor(5.),
#  indices=tensor(0)))
Enter fullscreen mode Exit fullscreen mode

min(). *My post explains min():

import torch

tensor1 = torch.tensor([5, 4, 7, 7])
tensor2 = torch.tensor(9)
tensor3 = torch.tensor(6)

tensor4 = torch.min(input=tensor1, dim=0, out=(tensor2, tensor3))

tensor1, tensor2, tensor3, tensor4
# (tensor([5, 4, 7, 7]),
#  tensor(4),
#  tensor(1),
#  torch.return_types.min_out(
#  values=tensor(4),
#  indices=tensor(1)))
Enter fullscreen mode Exit fullscreen mode

max(). *My post explains max():

import torch

tensor1 = torch.tensor([5, 4, 7, 7])
tensor2 = torch.tensor(9)
tensor3 = torch.tensor(6)

tensor4 = torch.max(input=tensor1, dim=0, out=(tensor2, tensor3))

tensor1, tensor2, tensor3, tensor4
# (tensor([5, 4, 7, 7]),
#  tensor(7),
#  tensor(2),
#  torch.return_types.max_out(
#  values=tensor(7),
#  indices=tensor(2)))
Enter fullscreen mode Exit fullscreen mode

all(). *My post explains all():

import torch

tensor1 = torch.tensor([True, False, True, False])
tensor2 = torch.tensor(True)

tensor3 = torch.all(input=tensor1, out=tensor2)
tensor3 = torch.all(input=tensor1, dim=0, out=tensor2)

tensor1, tensor2, tensor3
# (tensor([True, False, True, False]), tensor(False), tensor(False))
Enter fullscreen mode Exit fullscreen mode

any(). *My post explains any():

import torch

tensor1 = torch.tensor([True, False, True, False])
tensor2 = torch.tensor(True)

tensor3 = torch.any(input=tensor1, out=tensor2)
tensor3 = torch.any(input=tensor1, dim=0, out=tensor2)

tensor1, tensor2, tensor3
# (tensor([True, False, True, False]), tensor(True), tensor(True))
Enter fullscreen mode Exit fullscreen mode

matmul(). *My post explains matmul():

import torch

tensor1 = torch.tensor([2, -5, 4])
tensor2 = torch.tensor([3, 6, -1])
tensor3 = torch.tensor(7)

tensor4 = torch.matmul(input=tensor1, other=tensor2, out=tensor3)

tensor1, tensor2, tensor3, tensor4
# (tensor([2, -5, 4]), tensor([3, 6, -1]), tensor(-28), tensor(-28))
Enter fullscreen mode Exit fullscreen mode

Sentry image

Hands-on debugging session: instrument, monitor, and fix

Join Lazar for a hands-on session where you’ll build it, break it, debug it, and fix it. You’ll set up Sentry, track errors, use Session Replay and Tracing, and leverage some good ol’ AI to find and fix issues fast.

RSVP here →

Top comments (0)

Image of Timescale

Timescale – the developer's data platform for modern apps, built on PostgreSQL

Timescale Cloud is PostgreSQL optimized for speed, scale, and performance. Over 3 million IoT, AI, crypto, and dev tool apps are powered by Timescale. Try it free today! No credit card required.

Try free

👋 Kindness is contagious

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

Okay