Skip to main content

oapython

oapython is the Python distribution for OA, a cross-vendor Vulkan 1.4 GPU compute, machine-learning, vision, audio, and cryptography framework. Install oapython; import it as oa.

0.7 development preview. The bindings execute real OA C++ objects and Vulkan kernels, but the Python API and binary ABI are not frozen.

Install

python -m pip install oapython

The current wheel targets Linux x86-64, CPython 3.12, and glibc 2.39+. The host must provide a Vulkan loader and a working vendor ICD.

First computation

from oa import *

a = OaFnMatrix.Ones([2, 3])
b = OaFnMatrix.Full([2, 3], 2.0)
c = (a + b) * 0.5
print(OaFnMatrix.CopyToHost(c))

Import is host-only. The first device-backed request creates the native OA host; beginner code does not initialize an engine manually.

Public surface

Owner Surface
OaMatrix, OaFnMatrix matrices, shapes, dtypes, factories, operators, and numerical operations
OaModule, OaFnLoss, OaFnAutograd modules, autograd, losses, optimizers, training, and metrics
OaImage, OaFnImage, OaImageDecoder, OaImageEncoder image operations, still-image codecs, video, and capture
OaAudio, OaFnAudio, OaAudioDecoder, OaAudioEncoder semantic audio decode, processing, and encode
OaFnHash and Crypto values host cryptography and GPU public-data hashing when included in the wheel

Python uses the same compute graph and kernels as native C++; it is not a NumPy or CPU fallback implementation. Public names retain their C++ PascalCase spelling so a Python prototype translates directly to the C++ API. Lowercase domain modules remain compatibility aliases, not the canonical tutorial vocabulary.

Core, ML, Audio, Vision, end-to-end MNIST, and the complete 16-entry NLP suite live under Tutorial/Py. The current desktop matrix and preview-to-preview performance history are published in the NLP benchmark.

Links

OA is source-available under the Business Source License 1.1 and converts to Apache-2.0 on the Change Date stated in the license.

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distributions

No source distribution files available for this release.See tutorial on generating distribution archives.

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

oapython-0.7.7-cp312-cp312-manylinux_2_39_x86_64.whl (12.0 MB view details)

Uploaded CPython 3.12manylinux: glibc 2.39+ x86-64

File details

Details for the file oapython-0.7.7-cp312-cp312-manylinux_2_39_x86_64.whl.

File metadata

File hashes

Hashes for oapython-0.7.7-cp312-cp312-manylinux_2_39_x86_64.whl
Algorithm Hash digest
SHA256 af1abb06eb5f0be27111ef9c2d3a4fc8835033e0f571a8d09b1357ddf9bd8c88
MD5 99f828055302b1520c9e88750c98f1fd
BLAKE2b-256 3c10da4934cea0ee77c7fdc96abdfaf1cb607d30af775d56b26468da69edc9d9

See more details on using hashes here.

Provenance

The following attestation bundles were made for oapython-0.7.7-cp312-cp312-manylinux_2_39_x86_64.whl:

Publisher: ci.yml on realminc/oa

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Release history Release notifications | RSS feed

0.7.32

1 file

0.7.29

1 file

0.7.26

1 file

0.7.23

1 file

0.7.20

1 file

0.7.18

1 file

0.7.17

1 file

0.7.16

1 file

0.7.14

1 file

0.7.13

1 file

0.7.12

1 file

0.7.11

1 file

0.7.9

1 file

This release

0.7.7 This release

1 file

0.7.6

1 file

0.7.5

1 file

0.7.4

1 file

0.7.3

1 file

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page