Skip to page content
SAVRN
Search Plan a project

Open-weight model

szl-formulas

by SZL Holdings SZLHOLDINGS/szl-formulas

szl-formulas is an open-weight model from SZL Holdings, released under Apache License 2.0. Its published files total 236.9 KB.

GitHub source: szl-holdings/szl-formulas @ 977f344ccf248b55299a623d91db162f266eda29. Artifacts: BENCH.laptop-blackwell.json, OPERATIONAL.json. Software kernel for SZL formula composition. Not a model. No weights.

Parameters—
Context—
Weights236.9 KB
Licenseapache-2.0
AccessOpen weights
Monthly Downloads—

Model Card

By SZL Holdings, published under apache-2.0, revision 8bfdf0f62dd9.

GitHub source: szl-holdings/szl-formulas @ 977f344ccf248b55299a623d91db162f266eda29. Artifacts: BENCH.laptop-blackwell.json, OPERATIONAL.json. Software kernel for SZL formula composition. Not a model. No weights. IS: a pure-Python, stdlib-only governance kernel — offline replay of the 21 canonical formulas plus the governed-loop composer. Public surface includes registrycount(), lambdaaggregate, LOCKEDPROVENFORMULAIDS, rungovernedloop, PROOFSTATUS (mirrored verbatim). IS NOT: trained weights, LoRA, GGUF,.pt, or a classifier. Hub model.joblib is QUARANTINED executable serialization (and is not on this tree) — do not joblib.load it. Not a CUDA/Triton speedup. Not lutar-lean and not the…

Read SZL Holdings's full model card

Operational (MEASURED laptop-Blackwell)

STATUS: tests PASS. get_kernel import-LIVE. Unsloth/LoRA is the wrong tool. Receipted kernels, not silent CUDA.

Thing Label Method / N / date / what-NOT
tests (PYTHONPATH=torch-ext) PASS MEASURED 2026-08-29T15:52:52Z host betterwithage Windows-10-10.0.26200-SP0. torch 2.10.0+cu128. GPU NVIDIA GeForce RTX 5050 Laptop GPU arch Blackwell. pytest 19 passed in 0.04s. Failed nodes: none. What-NOT: not a leaderboard. torch.compile fullgraph failures on Windows Blackwell (cl is not found) are MEASURED, not hidden.
Kernel Hub get_kernel import-LIVE kernels 0.16.1. Default: get_kernel("SZLHOLDINGS/szl-formulas", revision="main", trust_remote_code=True) → True. backend="cpu" → True. trust_remote_code=False → ValueError (SZLHOLDINGS is not a trusted publisher). repo_type=kernel required (kernels 0.16). What-NOT: not a weight load; do not pickle/joblib.load.
formula-tax ADVISORY locked-8 F1 F4 F7 F11 F12 F18 F19 F22. registry_count=21. Λ geomean 1.0. uniqueness Conjecture 1 (never a theorem).
I1–I8 catalog I1 receipt-chain-continuity; I2 ledger-failure-shape; I3 served-run-has-model; I4 signed-columns-atomic; I5 loop-steps-positive; I6 receipt-ed25519-verify; I7 receipt-columns-consistent; I8 flywheel-lineage. Executed by SZLHOLDINGS/szl-invariants. Statuses never coerced. Λ untouched.
CUDA speedup / tokens/s / joules UNAVAILABLE Not claimed. Receipted kernels, not silent CUDA.

GitHub source: szl-holdings/szl-formulas @ 977f344ccf248b55299a623d91db162f266eda29. Artifacts: BENCH.laptop-blackwell.json, OPERATIONAL.json.

from kernels import get_kernel
k = get_kernel("SZLHOLDINGS/szl-formulas", revision="main", trust_remote_code=True)

szl-formulas

The cut, in one line: twenty-one formulas walk in; exactly eight walk out proven.

Software kernel for SZL formula composition. Not a model. No weights.

IS: a pure-Python, stdlib-only governance kernel — offline replay of the 21 canonical formulas plus the governed-loop composer. Public surface includes registry_count(), lambda_aggregate, LOCKED_PROVEN_FORMULA_IDS, run_governed_loop, PROOF_STATUS (mirrored verbatim).

IS NOT: trained weights, LoRA, GGUF, .pt, or a classifier. Hub model.joblib is QUARANTINED executable serialization (and is not on this tree) — do not joblib.load it. Not a CUDA/Triton speedup. Not lutar-lean and not the TypeScript ouroboros product.

Canonical GitHub source: https://github.com/szl-holdings/szl-formulas
Hub package: https://huggingface.co/kernels/SZLHOLDINGS/szl-formulas
This model-type repo is the publish / card mirror. Apache-2.0.

The cut

Papers have appendices. We have a kernel of formulas whose sorry-count is public (Doctrine v11: 749 / 14 / 163).

A formula registry that cannot drift from the Lean tree.

Silhouette → leave → SZL

Leader Take, then tweak
Anthropic Written principles, but ours compile.
NVIDIA Recipe as code.
Unsloth No.

No equivalent public combination was found among the leaders surveyed above (a snapshot, not an ecosystem-wide novelty claim).

Intended use

Import formulas. Check proof-status before citing.

Limitations

  • Λ uniqueness is Conjecture 1 — NOT a theorem.

Canonical GitHub: szl-holdings/szl-formulas

from kernels import get_kernel
fx = get_kernel("SZLHOLDINGS/szl-formulas", revision="main", trust_remote_code=True)
print(fx.registry_count())                   # 21
print(sorted(fx.LOCKED_PROVEN_FORMULA_IDS))  # exactly 8 F-ids
chain = fx.run_governed_loop([
    {"formula_name": "lambda_bounded", "args": [[0.9, 0.8, 0.95]]},
    {"formula_name": "reed_solomon_singleton", "args": [255, 223]},
])
print(chain["replay_ok"], chain["lambda_label"])  # True, ADVISORY (Conjecture 1)

Proof honesty: per-formula PROOF_STATUS is PROVEN / AXIOM / SORRY / CONJECTURE copied verbatim — that is not membership in the locked-proven set. Locked-proven is exactly 8: {F1, F4, F7, F11, F12, F18, F19, F22}. The F-number mapping onto these 21 registry names is UNKNOWN — never fabricated.

Claim Label
Source on GitHub REACHABLE
CUDA benches UNAVAILABLE
Weights not applicable
Λ Conjecture 1 (advisory, never a theorem)

Doctrine v11. Copyright 2026 SZL Holdings · Stephen P. Lutar · ORCID 0009-0001-0110-4173.

Identity and Version

Repository
SZLHOLDINGS/szl-formulas
Publisher
SZL Holdings
Task
Not stated by the source
Modality
Other
Library
kernels
Parameters
Not stated by the source
Languages
Not stated by the source
Revision
8bfdf0f62dd9415d95231769ddbfee9a7b427a03
First published
2026-07-12
Last updated
2026-09-30

Files and Weights

33 files, 236.9 KB in total.

Configuration27 files · 222.9 KB
Documentation3 files · 8.3 KB
Other2 files · 4.2 KB
Repository1 file · 1.5 KB
Every file
FileTypeSizeSHA-256
BENCH.laptop-blackwell.jsonConfiguration7.0 KB —
MODEL_PROVENANCE.jsonConfiguration12.0 KB —
OPERATIONAL.jsonConfiguration1.9 KB —
PROMOTION_READINESS_AUDIT.jsonConfiguration13.3 KB —
SZL_ESTATE_MANAGED.jsonConfiguration678 B —
SZL_SOURCE_BINDING.jsonConfiguration3.7 KB —
TRAINING_RECEIPT.jsonConfiguration5.1 KB —
bom/model-bom.cdx.jsonConfiguration6.8 KB —
build/torch-cpu/metadata.jsonConfiguration289 B —
build/torch-cpu/szl_formulas/__init__.pyConfiguration9.6 KB —
build/torch-cpu/szl_formulas/_aggregators.pyConfiguration18.3 KB —
build/torch-cpu/szl_formulas/_composer.pyConfiguration6.6 KB —
build/torch-cpu/szl_formulas/_formulas.pyConfiguration18.1 KB —
build/torch-cpu/szl_formulas/atlas.pyConfiguration837 B —
build/torch-cpu/szl_formulas/formula_atlas.v1.jsonConfiguration20.8 KB —
build/torch-cpu/szl_formulas/metadata.jsonConfiguration1.0 KB —
build/torch-universal/metadata.jsonConfiguration294 B —
build/torch-universal/szl_formulas/__init__.pyConfiguration9.6 KB —
build/torch-universal/szl_formulas/_aggregators.pyConfiguration18.3 KB —
build/torch-universal/szl_formulas/_composer.pyConfiguration6.6 KB —
build/torch-universal/szl_formulas/_formulas.pyConfiguration18.1 KB —
build/torch-universal/szl_formulas/atlas.pyConfiguration837 B —
build/torch-universal/szl_formulas/formula_atlas.v1.jsonConfiguration20.8 KB —
build/torch-universal/szl_formulas/metadata.jsonConfiguration1.0 KB —
scripts/eval.pyConfiguration1.6 KB —
scripts/forge.pyConfiguration13.1 KB —
tests/test_formulas.pyConfiguration6.6 KB —
LICENSEDocumentation776 B —
README.mdDocumentation5.7 KB —
SECURITY.mdDocumentation1.8 KB —
build.tomlOther58 B —
card/holo-banner.svgOther4.1 KB —
.gitattributesRepository1.5 KB —

License and Download

License
apache-2.0
Access
Open weights, no gate
Download from SZL Holdings

Released by SZL Holdings through its official repository on Hugging Face. Read the license.

Questions About szl-formulas

Can I use szl-formulas commercially?

Yes. szl-formulas is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.