This release is a pre-release and may not be stable for production use.
mlx-sparse
mlx-sparse is an attempt at an MLX-native sparse array package. The public API is Python,
while performance-critical operations are implemented as MLX primitives in C++
with CPU backends and Metal kernels for fixed-shape sparse operations.
The supported format surface is COO, CSR, and CSC for 2D sparse arrays. Current
functionality includes construction, validation, format conversions,
canonicalization, sparse-dense products, batched dense RHS products,
sparse-sparse products, transpose, and Hermitian transpose. mlx-sparse also
supports sparse linalg solvers (cg, bicgstab, gmres, minres),
reusable direct solves (factorized), sparse spectral routines (eigsh,
eigs, svds), sparse
Cholesky/LU factors, sparse triangular solves, sparse dot/vdot, and autodiff
through sparse values and dense RHS operands, including complex64.
Supported value dtypes are float32, float16, bfloat16, and complex64.
Supported index dtypes are int32 and int64 on CPU and GPU.
Quick Start
Install from PyPI:
python -m pip install mlx-sparse
Published macOS wheels are built with optional Apple Accelerate sparse direct solver support enabled. Published Linux wheels are CPU-only. For editable/source macOS builds, enable the same Accelerate path with:
CMAKE_ARGS="-DMLX_SPARSE_ENABLE_ACCELERATE=ON" python -m pip install -e .
See the full installation guide for platform requirements, feature gates, and verification commands.
import mlx.core as mx
import numpy as np
import mlx_sparse as ms
ms.use_cpu()
data = [2.0, -1.0, 4.0]
row = [0, 0, 1]
col = [0, 2, 1]
a = ms.coo_array((data, (row, col)), shape=(2, 3)).tocsr(canonical=True)
x = mx.array(np.array([3.0, 10.0, 7.0], dtype=np.float32))
y = a @ x
dense = a.todense()
at = a.T
b = mx.array(np.array([1.0, 2.0], dtype=np.float32))
spd = ms.csr_array(
(
[4.0, 1.0, 1.0, 3.0],
[0, 1, 0, 1],
[0, 2, 4],
),
shape=(2, 2),
canonical=True,
)
solution, info = ms.linalg.cg(spd, b)
factor = ms.linalg.sparse_cholesky(spd)
The package build compiles src/sparse/*.metal into
mlx_sparse/mlx_sparse.metallib when the macOS Metal toolchain is available,
and macOS wheels ship that metallib beside the Python package. Linux wheels do
not build or ship Metal assets.
Development
For contributors, use an editable install from the repository root. This builds the native extension and installs the development tooling.
python -m pip install -e ".[dev]"
License
This package is licensed under the Apache License 2.0.
1 maintainer
Metadata
1 maintainer
Release files for mlx-sparse 0.0.6b0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| mlx_sparse-0.0.6b0.tar.gz | 1.1 MB | Details |
Built distributions (wheels)
| File | Reset | |||
|---|---|---|---|---|
| mlx_sparse-0.0.6b0-cp313-cp313-manylinux_2_35_x86_64.whl | CPython 3.13 | CPython 3.13 | Linux glibc 2.35+ x86-64 | Details |
| mlx_sparse-0.0.6b0-cp313-cp313-macosx_14_0_arm64.whl | CPython 3.13 | CPython 3.13 | macOS 14.0+ ARM64 | Details |
| mlx_sparse-0.0.6b0-cp312-cp312-manylinux_2_35_x86_64.whl | CPython 3.12 | CPython 3.12 | Linux glibc 2.35+ x86-64 | Details |
| mlx_sparse-0.0.6b0-cp312-cp312-macosx_14_0_arm64.whl | CPython 3.12 | CPython 3.12 | macOS 14.0+ ARM64 | Details |
| mlx_sparse-0.0.6b0-cp311-cp311-manylinux_2_35_x86_64.whl | CPython 3.11 | CPython 3.11 | Linux glibc 2.35+ x86-64 | Details |
| mlx_sparse-0.0.6b0-cp311-cp311-macosx_14_0_arm64.whl | CPython 3.11 | CPython 3.11 | macOS 14.0+ ARM64 | Details |
| mlx_sparse-0.0.6b0-cp310-cp310-manylinux_2_35_x86_64.whl | CPython 3.10 | CPython 3.10 | Linux glibc 2.35+ x86-64 | Details |
| mlx_sparse-0.0.6b0-cp310-cp310-macosx_14_0_arm64.whl | CPython 3.10 | CPython 3.10 | macOS 14.0+ ARM64 | Details |
Total release size: 43.8 MB
Release files / mlx_sparse-0.0.6b0.tar.gz
| Download URL | mlx_sparse-0.0.6b0.tar.gz |
|---|---|
| Size | 1.1 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
17addfd20db5834ff07d5469e9db70b5acaa77a9a6971d85e178406cb7a74d5c
|
|
BLAKE2b-256 checksum How to use checksums |
9d581ed8e5d1a57867fc0f25b0b12a380b1f0417bbb03229dd13adbccaa3fbbc
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jul 22, 2026.
Transparency logRelease files / mlx_sparse-0.0.6b0-cp313-cp313-manylinux_2_35_x86_64.whl
| Download URL | mlx_sparse-0.0.6b0-cp313-cp313-manylinux_2_35_x86_64.whl |
|---|---|
| Size | 5.1 MB |
| Tags | CPython 3.13 Linux glibc 2.35+ x86-64 |
|
SHA-256 checksum How to use checksums |
4ddae60649cb599e7197f8f0c91c431e432a99c7990925091955951413ec360c
|
|
BLAKE2b-256 checksum How to use checksums |
afd5d03b84fddafa54d5365b193d2c1000ed6409c1aa6830fb6897d45d153fe7
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jul 22, 2026.
Transparency logRelease files / mlx_sparse-0.0.6b0-cp313-cp313-macosx_14_0_arm64.whl
| Download URL | mlx_sparse-0.0.6b0-cp313-cp313-macosx_14_0_arm64.whl |
|---|---|
| Size | 5.5 MB |
| Tags | CPython 3.13 macOS 14.0+ ARM64 |
|
SHA-256 checksum How to use checksums |
b0619f8583578dc9b9181b04a7a419b697b8ec063778a3d382800977ec44893f
|
|
BLAKE2b-256 checksum How to use checksums |
bfd01d46b003cbdbd45666c9d45d8da19cc137927d04e327169fb02fd5e23d72
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jul 22, 2026.
Transparency logRelease files / mlx_sparse-0.0.6b0-cp312-cp312-manylinux_2_35_x86_64.whl
| Download URL | mlx_sparse-0.0.6b0-cp312-cp312-manylinux_2_35_x86_64.whl |
|---|---|
| Size | 5.1 MB |
| Tags | CPython 3.12 Linux glibc 2.35+ x86-64 |
|
SHA-256 checksum How to use checksums |
087229f69b872b4e6feb40b408e4fa8ccdc96306def46de0f1ff797c733c6d02
|
|
BLAKE2b-256 checksum How to use checksums |
85d9820768af7b2a6a426a4b8ec446388c0546851f4e1d6b7db02a2ea99847b5
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jul 22, 2026.
Transparency logRelease files / mlx_sparse-0.0.6b0-cp312-cp312-macosx_14_0_arm64.whl
| Download URL | mlx_sparse-0.0.6b0-cp312-cp312-macosx_14_0_arm64.whl |
|---|---|
| Size | 5.5 MB |
| Tags | CPython 3.12 macOS 14.0+ ARM64 |
|
SHA-256 checksum How to use checksums |
c32332dcf1af7c856e5486969cbc0252ce2ebdf358ca46b8a24a33f7164cb1b6
|
|
BLAKE2b-256 checksum How to use checksums |
32633571713faf1d9bff700bd3847abca5ba37ec4e8aa7ee34f6ee60bf14e5aa
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jul 22, 2026.
Transparency logRelease files / mlx_sparse-0.0.6b0-cp311-cp311-manylinux_2_35_x86_64.whl
| Download URL | mlx_sparse-0.0.6b0-cp311-cp311-manylinux_2_35_x86_64.whl |
|---|---|
| Size | 5.1 MB |
| Tags | CPython 3.11 Linux glibc 2.35+ x86-64 |
|
SHA-256 checksum How to use checksums |
60ac590399ed597ebcde9295a5cb027ba02f575259fea958a675d8749312ce01
|
|
BLAKE2b-256 checksum How to use checksums |
02cab83882817b3df511f66018b2068a4f5009fa6914276b6528041ef48d76d1
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jul 22, 2026.
Transparency logRelease files / mlx_sparse-0.0.6b0-cp311-cp311-macosx_14_0_arm64.whl
| Download URL | mlx_sparse-0.0.6b0-cp311-cp311-macosx_14_0_arm64.whl |
|---|---|
| Size | 5.5 MB |
| Tags | CPython 3.11 macOS 14.0+ ARM64 |
|
SHA-256 checksum How to use checksums |
c8ff3f99238e5c6bb387d7e8292ac6ab89e545945cc704d8ee965e7746e0cbd3
|
|
BLAKE2b-256 checksum How to use checksums |
b4fcfdd8f9739ec91964b7fbea48bdee89f262ff0e61aa2afefffdca37b72f0b
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jul 22, 2026.
Transparency logRelease files / mlx_sparse-0.0.6b0-cp310-cp310-manylinux_2_35_x86_64.whl
| Download URL | mlx_sparse-0.0.6b0-cp310-cp310-manylinux_2_35_x86_64.whl |
|---|---|
| Size | 5.1 MB |
| Tags | CPython 3.10 Linux glibc 2.35+ x86-64 |
|
SHA-256 checksum How to use checksums |
d47d05a79a8b5932bdbd05ff542c4e6b5f29b6721028a18c989d9a5f6f84859c
|
|
BLAKE2b-256 checksum How to use checksums |
5a8abe882b3c2db455bd984cb6908e26c153b0ced8f1c02e5692f5de5f2a0092
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jul 22, 2026.
Transparency logRelease files / mlx_sparse-0.0.6b0-cp310-cp310-macosx_14_0_arm64.whl
| Download URL | mlx_sparse-0.0.6b0-cp310-cp310-macosx_14_0_arm64.whl |
|---|---|
| Size | 5.5 MB |
| Tags | CPython 3.10 macOS 14.0+ ARM64 |
|
SHA-256 checksum How to use checksums |
993697dec1e880d614aa4b9316ee9288aebb68aa521d30a7a7c098e68e56af54
|
|
BLAKE2b-256 checksum How to use checksums |
c99e2f68d4ba27ca138fcccb040db30eb3a954a5c93d46044791923bde3d4fa3
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jul 22, 2026.
Transparency log
