Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

`linalg` routines fail with pathologically edge cases

See original GitHub issue

Description

This is discovered using the array-api-tests. I am not sure if the same behavior for handling 2D “scalar” matrices is applicable to all numpy.linalg routines, or just the two I posted here.

To Reproduce

>>> import numpy as np
>>> import cupy as cp
>>>
>>> # cholesky
>>> np.linalg.cholesky(np.empty((0, 0), dtype=np.float32))
array([], shape=(0, 0), dtype=float32)
>>> cp.linalg.cholesky(cp.empty((0, 0), dtype=cp.float32))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/leof/dev/cupy_cuda114/cupy/linalg/_decomposition.py", line 205, in cholesky
    potrf(
  File "cupy_backends/cuda/libs/cusolver.pyx", line 1205, in cupy_backends.cuda.libs.cusolver.spotrf
    cpdef spotrf(intptr_t handle, int uplo, int n, size_t A, int lda,
  File "cupy_backends/cuda/libs/cusolver.pyx", line 1212, in cupy_backends.cuda.libs.cusolver.spotrf
    check_status(status)
  File "cupy_backends/cuda/libs/cusolver.pyx", line 1062, in cupy_backends.cuda.libs.cusolver.check_status
    raise CUSOLVERError(status)
cupy_backends.cuda.libs.cusolver.CUSOLVERError: CUSOLVER_STATUS_INVALID_VALUE
>>>
>>> # matrix_power
>>> np.linalg.matrix_power(np.empty((0, 0), dtype=np.float32), n=-1)
array([], shape=(0, 0), dtype=float32)
>>> cp.linalg.matrix_power(cp.empty((0, 0), dtype=cp.float32), n=-1)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/leof/dev/cupy_cuda114/cupy/linalg/_product.py", line 374, in matrix_power
    M = _solve.inv(M)
  File "/home/leof/dev/cupy_cuda114/cupy/linalg/_solve.py", line 323, in inv
    cupyx.lapack.gesv(a, b)
  File "/home/leof/dev/cupy_cuda114/cupyx/lapack.py", line 68, in gesv
    getrf(handle, n, n, a.data.ptr, n, dwork.data.ptr, dipiv.data.ptr,
  File "cupy_backends/cuda/libs/cusolver.pyx", line 1402, in cupy_backends.cuda.libs.cusolver.sgetrf
    cpdef sgetrf(intptr_t handle, int m, int n, size_t A, int lda,
  File "cupy_backends/cuda/libs/cusolver.pyx", line 1409, in cupy_backends.cuda.libs.cusolver.sgetrf
    check_status(status)
  File "cupy_backends/cuda/libs/cusolver.pyx", line 1062, in cupy_backends.cuda.libs.cusolver.check_status
    raise CUSOLVERError(status)
cupy_backends.cuda.libs.cusolver.CUSOLVERError: CUSOLVER_STATUS_INVALID_VALUE

Installation

No response

Environment

No response

Additional Information

No response

Issue Analytics

State:
Created 2 years ago
Comments:5 (5 by maintainers)

Top GitHub Comments

1reaction

toslunarcommented, Dec 13, 2021

Fixed by #6164, #6167, #6210.

1reaction

toslunarcommented, Dec 6, 2021

I’ll fix eigh and eigvalsh.

Top Results From Across the Web

Linear algebra (numpy.linalg) — NumPy v1.24 Manual

Several of the linear algebra routines listed above are able to compute results for several matrices at once, if they are stacked into...

4 Linear Algebra - Numerical Methods

We first need to understand how Python's numpy library builds and stores vectors and matrices. The following exercises will give you some experience...

Numerical Linear Algebra - Introduction to Programming in Java

This textbook provides an interdisciplinary approach to the CS 1 curriculum. We teach the classic elements of programming, using an.

Linear algebra (scipy.linalg) — SciPy v1.9.3 Manual

Solves a standard or generalized eigenvalue problem for a complex Hermitian or real symmetric matrix. Solve real symmetric or complex Hermitian band matrix ......

Optimizing Matrix Multiply using PHiPAC: a Portable, High ...

on linear algebra kernels but this currently requires ex- ... matrix multiplication, these routines have a large design ... cludes error checking.