A lightweight AXIR playground to run numerical kernels across CPU and OpenCL
Project description
AXIOMOS – Minimal Public Showcase
This repository is a public showcase proving that AXIR – the universal AI runtime IR – works without exposing any proprietary core code.
It relies on the public PyPI package axiomos (CPU + OpenCL), already published.
✨ Goal: let anyone (YC, investors, reviewers) test in 2 minutes that the IR executes correctly and results are verifiable across backends.
⚡️ Quickstart
1) Environment setup
python -m venv .venv
# Windows
.\.venv\Scripts\activate
# macOS/Linux
# source .venv/bin/activate
2) Install the public runtime
CPU only:
python -m pip install axiomos
With OpenCL (if GPU/drivers are available):
python -m pip install "axiomos[opencl]"
3) Generate two example fixtures
python make_fixtures.py
This creates:
examples/
├─ vector_add_small.axir.json
└─ softmax2d_small.axir.json
✅ Demo 1 — Environment check
axiomos-doctor
If OpenCL is not available, you can still use --backend-a cpu --backend-b cpu.
✅ Demo 2 — “Hello AXIR” (vector add)
Compare CPU vs CPU on a simple vector:
axiomos-verify examples/vector_add_small.axir.json --buffer hC --backend-a cpu --backend-b cpu
Expected output: RESULT: PASS.
✅ Demo 3 — Softmax2D (8×8)
Without GPU (CPU vs CPU):
axiomos-verify examples/softmax2d_small.axir.json --buffer hY --backend-a cpu --backend-b cpu
With GPU (CPU vs OpenCL):
# Windows PowerShell
$env:AXIOMOS_OCL_REAL="1"
# macOS/Linux
# export AXIOMOS_OCL_REAL=1
axiomos-verify examples/softmax2d_small.axir.json --buffer hY --backend-a cpu --backend-b opencl
Expected output:
SHAPES : CPU(8, 8) vs OPENCL(8, 8)
max_abs_err: 0.0
ALLCLOSE : True (atol=1e-06, rtol=1e-05)
[SMOKE] RESULT: PASS
🔥 Demo 4 — Small benchmark (optional)
axiomos-smoke --axir examples/softmax2d_small.axir.json --buffer hY --backend-a opencl --backend-b cpu --inproc --time --warmup 3 --repeat 10
Produces an OpenCL vs CPU latency comparison.
If no GPU is present, use --backend-a cpu --backend-b cpu.
🧩 What this showcase proves
Portability – the same .axir.json runs on both CPU and OpenCL.
Verifiability – axiomos-verify checks numerical equality with strict tolerance (allclose).
Simplicity – 2 fixtures, 3 commands, no access to proprietary code needed.
The full runtime (optimizations, advanced kernels, Ed25519 signature, etc.) remains private.
This showcase demonstrates only the critical path: IR → execution → verification.
🪪 License
MIT (for this public showcase only). The full runtime remains proprietary.
'@ | Out-File -Encoding utf8 -FilePath .\README.md -NoNewline
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file axiomos-0.1.9.tar.gz.
File metadata
- Download URL: axiomos-0.1.9.tar.gz
- Upload date:
- Size: 6.8 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.13.6
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
78acb162946aaec582ca57617bc834f718f72f8e70f8a5d477b9011944ae569f
|
|
| MD5 |
1bdef76b256cc3289e1e980ea39c20e5
|
|
| BLAKE2b-256 |
0d72dbbf03981cddb2fc18aab6e266afa5635f35b3137eb6ae993f1705b90530
|
File details
Details for the file axiomos-0.1.9-py3-none-any.whl.
File metadata
- Download URL: axiomos-0.1.9-py3-none-any.whl
- Upload date:
- Size: 8.0 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.13.6
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
a70335e1d4f2134278bb718dd1446f04d4cc13f05bf89bdf26c674d3dce05276
|
|
| MD5 |
257df7ebc67a2625a420b4ceb4c0098f
|
|
| BLAKE2b-256 |
c92c946aa50caffc8a594e74d6d843dc027c5dc76f3896f89b96579574928c25
|