How to create a rank 3 tensor in Pytorch?

[pradeep_vasudev]

How to create a rank 3 tensor in Pytorch?
Please share the co

[J_Johnson]

import torch

x=torch.rand(8)

print(x.size())

x=x.view(2,2,2)
print(x.size())

You can also use .reshape() depending on the application.

[pradeep_vasudev]

@J_Johnson But when I checked the rank of the tensor that you provided using torch.linalg.matrix_rank() function it shows like tensor([2, 2]).
Here is the code

import torch
x=torch.rand(8)
print(x.size())
x=x.view(2,2,2)
print(x.size())
print(torch.linalg.matrix_rank(x))

Ideally, it should be 3 right if the rank is 3?

[J_Johnson]

@pradeep_vasudev

@J_Johnson But when I checked the rank of the tensor that you provided using torch.linalg.matrix_rank() function it shows like tensor([2, 2]).
Here is the code

import torch
x=torch.rand(8)
print(x.size())
x=x.view(2,2,2)
print(x.size())
print(torch.linalg.matrix_rank(x))

Ideally, it should be 3 right if the rank is 3?

That’s not the tensor rank.

https://pytorch.org/docs/stable/generated/torch.linalg.matrix_rank.html

See also:

Is the rank of a Tensor different from the rank of a Matrix?

[pradeep_vasudev]

@J_Johnson
Thank you.
If so what is the corresponding torch function or method to see the rank of a tensor?

[J_Johnson]

There is no built in method I’m aware of. But if you want the number of dims(i.e. tensor rank), you can just use len(x.size()).

[pradeep_vasudev]

@J_Johnson Thank you

[tom (Thomas V)]

@J_Johnson

len(x.size()).

x.dim()

Best regards

Thomas