Metadata-Version: 2.4
Name: axiomos
Version: 0.1.14
Summary: A lightweight AXIR playground to run numerical kernels across CPU and OpenCL
Author: Axiomos
License: MIT
Project-URL: Homepage, https://github.com/Aidenkuro10/axiomos-vitrine
Project-URL: Documentation, https://github.com/Aidenkuro10/axiomos-vitrine#readme
Project-URL: Repository, https://github.com/Aidenkuro10/axiomos-vitrine
Keywords: opencl,kernels,gemm,axir,numpy,runtime,AI
Classifier: Development Status :: 3 - Alpha
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: numpy>=1.23
Requires-Dist: jsonschema>=4.21
Requires-Dist: tqdm>=4.0
Provides-Extra: opencl
Requires-Dist: pyopencl>=2022.3; extra == "opencl"
Dynamic: license-file

# AXIOMOS — Public Showcase (Minimal)

[PyPI](https://pypi.org/project/axiomos/) • [GitHub](https://github.com/Aidenkuro10/axiomos-vitrine)  
![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)

> 🔒 Public showcase build — proves the critical path: **AXIR JSON → CPU/OpenCL execution → numeric verification**.  
> The private core (optimizers, advanced kernels, schedulers, trust layer) remains **under NDA**.

---

## TL;DR (works anywhere, outside the repo)

### Windows (PowerShell)
```powershell:
pip install axiomos
axiomos-doctor
axiomos-demo

macOS / Linux (bash):
pip install axiomos
axiomos-doctor
axiomos-demo

That’s it.

axiomos-doctor prints versions + device info and runs a tiny softmax sanity check.
axiomos-demo auto-creates a small AXIR fixture and verifies CPU ↔ OPENCL.
If no OpenCL runtime is present, it gracefully falls back to OPENCL(cpu-fallback) and still verifies parity.

That’s it.

axiomos-doctor prints versions + device info and runs a tiny softmax sanity check.

axiomos-demo auto-creates a small AXIR fixture and verifies CPU ↔ OPENCL.
If no OpenCL runtime is present, it gracefully falls back to OPENCL(cpu-fallback) and still verifies parity.

Why AXIOMOS (Vision):

Modern AI stacks are fragmented. AXIR (Axiomos IR) is a universal, hardware-agnostic IR aimed at:

Portability — compile once, run on CPU, GPU, accelerators.

Determinism & Reproducibility — numerically verifiable parity across backends.

Trust — measurable parity today; cryptographic provenance/signatures in the private build.

Longevity — a stable IR beyond today’s frameworks and vendor APIs.

This public build is intentionally minimal: it shows AXIR JSON → multi-backend execution → verification without exposing the private core.

Optional tools

List OpenCL devices:
axiomos-devices
If PyOpenCL + vendor drivers are available, this lists platforms/devices. Otherwise it prints a clear hint.

Tiny latency “smoke” (indicative only):
axiomos-smoke --size 512 --warmup 3 --repeat 30 --seed 0


Shows p50/p95/mean/best and an indicative GFLOP/s (CPU numpy).
Not a performance claim; the public OpenCL path is minimal (not optimized).

(Optional) PyTorch → AXIR mini-export:
pip install torch
axiomos-export-torch
python -m axiomos.verify examples/pytorch_softmax.axir.json --buffer hY --backend-a cpu --backend-b opencl --seed 0

Generates a tiny pytorch_softmax.axir.json in examples/ and verifies parity like the demo.

If you cloned the repo (extra fixtures):
python make_fixtures.py


Creates:

examples/
 ├─ vector_add_small.axir.json
 └─ softmax2d_small.axir.json


Verify examples:

# CPU ↔ CPU (Hello AXIR)
axiomos-verify examples/vector_add_small.axir.json --buffer hC --backend-a cpu --backend-b cpu --seed 0

# CPU ↔ OpenCL (Softmax 8×8)
python -m axiomos.verify examples/softmax2d_small.axir.json --buffer hY --backend-a cpu --backend-b opencl --seed 0


If PyOpenCL + drivers are present → OPENCL(…) uses the device path.

Otherwise → OPENCL(cpu-fallback).

Output includes: SHAPES, max_abs_err, ALLCLOSE (atol=1e-6, rtol=1e-5), CPU time, OPENCL time.
Note: the public OpenCL path is minimal (not optimized).

What this proves (today):

Portability — the same .axir.json runs on CPU and OpenCL (if present).

Verifiability — strict numeric parity (allclose, max_abs_err) with shapes & timings.

Reproducibility — seedable runs and versioned fixtures.

Operable UX — install, run, verify in minutes; no private code exposure.


What’s intentionally not here:

Private optimizer passes, schedulers, and advanced kernels.

Full operator coverage and vendor-tuned implementations.

Cryptographic provenance & signatures (available in the private build).

➡️ For the full demo (under NDA), please contact us.

License
MIT (for this public showcase only). The full runtime remains proprietary

