Downloads · 30 days
5
100% of all-time downloads
SZLHOLDINGS/szl-nemo
szl-nemo is a machine learning model from SZLHOLDINGS. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
<p<a href="https://huggingface.co/spaces/SZLHOLDINGS/szl-command-lab"<img src="https://raw.githubusercontent.com/szl-holdings/.github/main/profile/assets/szl/logos/szlmarkholographic.svg" alt="SZL Holdings" width="112…
Downloads · 30 days
5
100% of all-time downloads
All-time downloads
5
Public
Repo size
30.1 KB
Likes
0
Public
Click a slice to open those files.
.json31.3 KB · 40%
From the Hugging Face model README
Inspect historical recipe-conformance scripts and receipt records.
Artifact: Historical software and receipt mirror · Stage: Scorer quarantined
Explore in Command Lab · Build · Evidence
Historical recipe-conformance records; the scorer and its generator remain quarantined.
This model-style Hub listing is a historical scripts-and-receipt mirror. It
contains no approved loadable checkpoint, Nemotron weights, generative model,
or qualified kernel loading path. No joblib.load, retraining, or future scorer
publication is recommended by this card.
56c429886bf5ea98826bf99b27313626d70f01cc.szl-holdings/szl-nemo@3860c3cb92cf5dee7f7f15635e878713330b219c.nemo/card/README.md. A card-only change does not ship the complete newer canonical package.The historical TF-IDF/logistic-regression scorer and generator are quarantined.
Historical scripts, rule-check context, Modelfile prompt text, and receipt
records remain inspectable. An upstream Ollama-tag observation in
BASE_MODEL_MANIFEST.json is not a weight artifact or an immutable base binding.
Current canonical development describes deterministic R1–R5 doctrine checks and E1–E10 envelope witnesses. Those newer APIs, package files, and test claims must remain source-only until an exact source-bound Hub publication is verified. This older mirror does not establish complete current-package parity.
The unsigned TRAINING_RECEIPT.json reports the following historical scorer experiment:
| Field | Receipt-reported value and scope |
|---|---|
| Training time | 2026-07-21T02:52:42Z |
| Host/software | Replit 2-vCPU; scikit-learn 1.9.0 |
| Seed | 20260721 |
| Checker-labelled rows | 5620: 2638 conforming and 2982 violating; 80/20 stratified split |
| Fidelity against rule checker | 1.0; reported recipe-conformance experiment |
| Unseen paraphrases | 0.8333, N=12; small historical sample |
| Named scorer digest | model.joblib: d3f0cd7bebbb73fedbc9a0f098148f46f5834bf9184b43cd29b07f286a77ff5b |
| Scorer artifact in reviewed mirror | ABSENT / UNAVAILABLE; the named digest does not establish a published load path |
These are preserved receipt-reported results, not newly measured model quality, an LLM evaluation, a Nemotron benchmark, an independent benchmark, or a current latency result. The scorer cannot be replayed from the reviewed Hub snapshot because the named artifact is absent. Its historical digest must not be interpreted as permission to recover, retrain, or load that quarantined scorer.
The PROMOTION_READINESS_AUDIT.json is retained at the same reviewed revision. This documentation correction makes no new promotion, runtime qualification, release approval, or deployment claim.
Inspect historical recipe-conformance evidence and separately reviewed canonical software. Keep a deterministic policy decision authoritative outside any historical learned scorer.
from_pretrained path is established here.1.0 and paraphrase result 0.8333 on N=12 remain historical context.Apache-2.0 is declared for the repository's SZL files. A LICENSE is present in the reviewed mirror; this review does not establish upstream or downstream license coverage. No NVIDIA weights are republished by this card.
Lambda uniqueness remains Conjecture 1, open and advisory. Historical evidence and failed outcomes remain intact.
<!-- SZL-CARD-TECHNICAL:v1:END --> </details>