szl-formulas
Software kernel for SZL formula composition. Not a model. No weights.
Control before action. Evidence after.
Product · Proof · GitHub organization
This GitHub tree is the source. The Hub package is the publish mirror: kernels/SZLHOLDINGS/szl-formulas. Card: SZLHOLDINGS/szl-formulas.
The only writer of both Hub repositories is hf-kernel-publish, configured in .github/hf-kernels.json. A published release builds build/torch-cpu and build/torch-universal from torch-ext/ at the release commit, pushes them fast-forward to the kernel repo's main and v1 branches with a SZL_KERNEL_SOURCE.json binding that names that commit, verifies the result through get_kernel in a fresh job, and only then turns the model repository into a pointer card to the kernel repo. A weekly job re-verifies what the Hub serves. Nothing under build/ is committed here; the tests run against torch-ext and both generated variants (tests/conftest.py).
What
Python package under torch-ext/szl_formulas/. Formula composer + canonical formula table. Apache-2.0.
What this is NOT
-
model.joblibis QUARANTINED executable serialization. It is absent from the Hub model repository andhf-kernel-publishkeeps it pruned. Do notjoblib.loadany copy. GitHub source is the approved path. -
Not trained weights, not a LoRA, not GGUF
-
Not a CUDA/Triton speedup claim (no MEASURED benches in this repo)
-
Not the TypeScript product
ouroborosand notlutar-lean
Load
Set SZL_FORMULAS_HF_REVISION to the immutable first-class Kernel Hub commit
from a verified publication of kernels/SZLHOLDINGS/szl-formulas. Use the kernels
client version qualified with that publication. The GitHub source commit,
model-repository commit, and Kernel Hub commit are separate identities.
An observed head, a branch name, or a successful import does not qualify a release.
A qualifying revision is a commit of the kernel repo's v1 branch written by a
hf-kernel-publish run whose verify job passed; SZL_KERNEL_SOURCE.json at that
commit names the GitHub source commit. version=1 (shown on the
kernel card) resolves the moving head of that branch, not a
qualified release.
trust_remote_code=True permits execution of the selected repository's Python.
Review that exact revision, its provenance and publication evidence before enabling it.
The format check below only rejects missing or mutable revision inputs; it does not
verify hashes, publisher authorization or compatibility. If that evidence is unavailable,
stop the Hub load and use separately reviewed local source for development.
import os
import re
hf_revision = os.environ.get("SZL_FORMULAS_HF_REVISION", "")
if re.fullmatch(r"[0-9a-f]{40}", hf_revision) is None:
raise ValueError("A verified immutable Kernel Hub revision is required")
from kernels import get_kernel
get_kernel("SZLHOLDINGS/szl-formulas", revision=hf_revision, trust_remote_code=True)
Source-only development
Take the exact 40-character GitHub source revision from the qualifying publication
receipt, then review https://github.com/szl-holdings/szl-formulas/tree/<SOURCE_SHA>/torch-ext/szl_formulas
at that immutable revision, separately from any Hub release.
With the source's dependencies already available, run from the reviewed checkout root:
import sys
from pathlib import Path
sys.path.insert(0, str(Path("torch-ext").resolve()))
import szl_formulas as local_kernel
This selects local Python source rather than calling the Hub loader. Importing local source also executes Python. This documentation check does not run that import, install dependencies, qualify a runtime or establish a Hub publication.
Honesty
| Claim | Label |
|---|---|
| Source on GitHub | REACHABLE |
| CUDA benches | UNAVAILABLE |
| Weights | not applicable |
| Λ | Conjecture 1 (advisory, never a theorem) |
| NVIDIA Nemotron 3 Nano | NVIDIA result, arXiv:2512.20848. Not an SZL formula. |
Revision-bound Lean reference
load_locked_formula_reference() exposes a deterministic reference to the
24 named Lean carrier declarations for the locked eight at exact lutar-lean
commit c4c42baf107f2cd5ec11ac3958fbd8fd06a875c8. The repository vendors the
exact workflow evidence and its external receipt, validates their byte digests,
source tree, toolchain, carrier set, and axiom sets, and ships the derived
reference with every kernel variant.
The observed level is build/elaboration plus Lean.collectAxioms at that pin.
lean4checker --fresh and comparator/external-checker validation are NOT_RUN.
The reference keeps F-number-to-registry mapping, Lean-to-Python refinement, and
informal claim semantics UNVERIFIED; it grants no execution, paper-promotion,
publication, deployment, or runtime authority. See Lean's official
proof-validation ladder
and axiom guidance.
Doctrine v11 LOCKED. Owner: Stephen Lutar / SZL Holdings. The locked formulas stay SZL's. Nemotron stays NVIDIA's published model.
License
Apache-2.0. Copyright 2026 SZL Holdings.
Metadata
Release files for szl-formulas 0.1.3
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| szl_formulas-0.1.3.tar.gz | 79.8 kB | With provenance | Details for szl_formulas-0.1.3.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| szl_formulas-0.1.3-py3-none-any.whl | Python 3 | none | any | Details for szl_formulas-0.1.3-py3-none-any.whl |
Total release size: 119.4 kB
Release files / szl_formulas-0.1.3.tar.gz
| Download URL | szl_formulas-0.1.3.tar.gz |
|---|---|
| Size | 79.8 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
005fabfb35fe830f4496ef160befb45b1037c6d331e2ba02dc41f1fcbfeae25b
|
|
BLAKE2b-256 checksum How to use checksums |
5ed0d620d7112dea3d66116ac779c7b985c17d7d05763cabc1cfe69ac70a2ccc
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
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 Nov 16, 2026.
Transparency logRelease files / szl_formulas-0.1.3-py3-none-any.whl
| Download URL | szl_formulas-0.1.3-py3-none-any.whl |
|---|---|
| Size | 39.6 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
c46da35bfb680416a403b03b79a3bb94db6ae3b9df0e188b9a5278b8f5dc8b6b
|
|
BLAKE2b-256 checksum How to use checksums |
6a93b1a0dccd8019b81febe1023c1b88a569709a2a6cdd53ee0ead65a26a230d
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
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 Nov 16, 2026.
Transparency logSecurity for szl-formulas 0.1.3
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.
Full provenance
Both files in this release have been uploaded with provenance metadata.
Attestation for szl_formulas-0.1.3.tar.gz
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 Nov 16, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.