OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-06 12:20:52 +01:00
Code Issues Actions 38 Packages Projects Releases Wiki Activity

Fix for ambiguity in linalg.norm()'s ord argument of +2 & -2 (#155148)

Browse Source 
Fixes #136453

### Description
---
Fixed the ambiguity by referencing a hyperlink to wikipedia's SVD/Singular Values section as per past discussion (by other contributors) on the above thread.

In the ord argument, for values `+2` and `-2`, the `singular value` now points to [this section of singular values on the wiki SVD page](https://en.wikipedia.org/wiki/Singular_value_decomposition#Singular_values,_singular_vectors,_and_their_relation_to_the_SVD).

### Why not mention SVD
---
For conciseness (expanding 'largest singular value' -> 'largest singular value of a SVD' is too much, i think, wrt rest of the table)

---

I hope this is satisfactory. Please let me know if I have missed anything essential; cheers.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/155148
Approved by: https://github.com/Skylion007, https://github.com/lezcano
This commit is contained in:
ANotFox 2025-06-04 21:15:20 +00:00 committed by PyTorch MergeBot
parent b084e1b81c
commit 65a5eb8d27
1 changed files with 18 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

33
torch/linalg/__init__.py
View File

@ -1369,21 +1369,21 @@ Whether this function computes a vector or matrix norm is determined as follows:
:attr:`ord` defines the norm that is computed. The following norms are supported:
====================== ========================= ========================================================
:attr:`ord` norm for matrices norm for vectors
====================== ========================= ========================================================
`None` (default) Frobenius norm `2`-norm (see below)
`'fro'` Frobenius norm -- not supported --
`'nuc'` nuclear norm -- not supported --
`inf` `max(sum(abs(x), dim=1))` `max(abs(x))`
`-inf` `min(sum(abs(x), dim=1))` `min(abs(x))`
`0` -- not supported -- `sum(x != 0)`
`1` `max(sum(abs(x), dim=0))` as below
`-1` `min(sum(abs(x), dim=0))` as below
`2` largest singular value as below
`-2` smallest singular value as below
other `int` or `float` -- not supported -- `sum(abs(x)^{ord})^{(1 / ord)}`
====================== ========================= ========================================================
====================== ========================== ======================================================
:attr:`ord` norm for matrices norm for vectors
====================== ========================== ======================================================
`None` (default) Frobenius norm `2`-norm (see below)
`'fro'` Frobenius norm -- not supported --
`'nuc'` nuclear norm -- not supported --
`inf` `max(sum(abs(x), dim=1))` `max(abs(x))`
`-inf` `min(sum(abs(x), dim=1))` `min(abs(x))`
`0` -- not supported -- `sum(x != 0)`
`1` `max(sum(abs(x), dim=0))` as below
`-1` `min(sum(abs(x), dim=0))` as below
`2` largest `singular value`_ as below
`-2` smallest `singular value`_ as below
other `int` or `float` -- not supported -- `sum(abs(x)^{ord})^{(1 / ord)}`
====================== ========================== ======================================================
where `inf` refers to `float('inf')`, NumPy's `inf` object, or any equivalent object.
@ -1483,6 +1483,9 @@ Using the :attr:`dim` argument to compute matrix norms::
tensor([ 3.7417, 11.2250])
>>> LA.norm(A[0, :, :]), LA.norm(A[1, :, :])
(tensor(3.7417), tensor(11.2250))
.. _singular value:
https://en.wikipedia.org/wiki/Singular_value_decomposition#Singular_values,_singular_vectors,_and_their_relation_to_the_SVD
""",
)