Text Generation
PEFT
Safetensors
qwen2
sft
trl
hf_jobs
governed-ai
szl-holdings
doctrine-v11
identity
willay
conversational
Instructions to use SZLHOLDINGS/WILLAY with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use SZLHOLDINGS/WILLAY with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct") model = PeftModel.from_pretrained(base_model, "SZLHOLDINGS/WILLAY") - Notebooks
- Google Colab
- Kaggle
Commit History
17 days ago
-
eval: publish governed contract base comparison (#2)
- eval: publish governed contract base comparison (ee2b049612fe83ca34176aadb25fe7e7e61165c0)96a0ac5
19 days ago
-
card: publish the reviewed curated card (szl-forge willay/card/README.md, metadata/prose only) (#4)
- card: publish the reviewed curated card (szl-forge willay/card/README.md, metadata/prose only) (e9f59ac9cfdb294d08b6cb91c434e941545c6cb2)432646a
22 days ago
-
docs(card): estate audit 2026-09-25 — honest labels, dead links, metadata (#3)
- docs(card): estate audit 2026-09-25 — honest labels, dead links, metadata (f87c35c96bf49cf2d72d2fe8b3433e24888e37cf)e04bd72
-
Merged full-precision checkpoint (receipted) + config fixes
Verified signature375a007
-
Add per-repo CycloneDX BOM (canonical registry: datasets/SZLHOLDINGS/model-bom)
Verified signature7319d40 -
Add per-repo CycloneDX BOM (canonical registry: datasets/SZLHOLDINGS/model-bom)
Verified signature142ef2b
-
Materialize declared Apache-2.0 license
Verified signatureMaterializes the repository card's existing Apache-2.0 declaration. Ownership, lineage, consent, privacy, deployment, and production readiness remain unverified.2de2d9d
-
fix: correct metadata and quickstart — PEFT adapter, not a transformers checkpoint
Verified signatureThe card declared library_name: transformers with a quickstart calling AutoModelForCausalLM.from_pretrained on this repo id, but there is no config.json here, only adapter_config.json. Sets library_name: peft and base_model_relation: adapter, and rewrites the quickstart to load the declared base and apply the adapter. Also documents that the two adapters in this repo target DIFFERENT base models (root: Qwen/Qwen2.5-0.5B-Instruct r=16 q/v; adapter-unsloth: unsloth bnb-4bit r=8 across 7 modules) — pairing either with the wrong base loads and generates but is not the trained model. States plainly that no evaluation exists.95e3106 -
security(model): replace pickle training arguments with JSON
Verified signatureRemove executable serialized Trainer metadata; preserve only text-sourced, non-executable provenance. No pickle deserialization.eb3711d
-
chore(receipt): MEASURED unique Unsloth willay (eval none-this-run)
Verified signaturef068446 -
feat(adapter): unique Unsloth willay Doctrine mouth. rsLoRA rank-8 attn+mlp packing=false. Short ctx.
Verified signatureb375c3d -
feat(train): unique Unsloth GPU trainer (adapter-unsloth/, does not clobber R1 signed khipu)
Verified signature0b45aff -
weights(atelier): CPU NumPy silhouette seed=20260721 — not a GPU retrain
Verified signatureadbf083 -
atelier: one-of-one card — cut, silhouette, GitHub Python. YAML preserved.
Verified signature966a1a9 -
End of training
Verified signatureabdeb80
-
docs(card): INTI copy contract — ROADMAP. first sentence + roadmap tag
Verified signaturef39fe0e -
docs(card): Series A WILLAY signed-refusal card — cutting, no weights yet
Verified signature952fcf9 -
initial commit
Verified signature61185e4