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
eval: publish governed contract base comparison
Bounded local evidence only. Eight deterministic fixtures compare WILLAY at 7319d4041577b8da0d968a03f38ea3c392e9c0e1 with exact Qwen base 7ae557604adf67be50417f59c2c2f167def9a775. WILLAY underperformed the base on this fixture; this is not a safety certification or general benchmark.
Preserving the historical 2026-09-08 eight-case contract comparison at its recorded revisions, not a fresh held-out evaluation or a qualification of current weights. Evidence class: REPORTED until independent replay. The adapter did not beat the base in this comparison. The newer failed evaluation and release blockers in the current model card remain authoritative. This PR adds only the two historical evidence files; it must not modify weights, configuration, tokenizer, or the current README.