models/canary/README.md

Canary packed resource for OpenCC

This is the existing accepted OpenCC Windows CPU packed artifact for Canary 1B v2. It is derived from NVIDIA Canary 1B v2 and the GGUF conversion by handy-computer. This publication copies the already used artifact byte for byte. It does not train a model, run a new conversion, or change the accepted Q8_0/F32 inference recipe.

The packed format is specific to the compatible OpenCC native runtime. It is not interchangeable with an arbitrary GGUF, CUDA profile or Metal model file. artifact.json records the output identity and the exact input GGUF identity.

Attribution: NVIDIA (original Canary model), handy-computer (GGUF distribution), OpenCC (runtime-specific packing/layout). The source GGUF is handy-computer/canary-1b-v2-gguf at e2d8e6d7. The original model is nvidia/canary-1b-v2. Applicable model terms: CC BY 4.0. The layout/container transformation is the modification described here; model authorship and original licensing remain with the respective upstream authors.

Publication is in progress. The complete application resource catalog and its HF/ModelScope integration are not yet declared complete. This card makes no new model quality or performance claims.