qwen3guard for Ollaya

Ollaya package of Qwen/Qwen3Guard-Gen-0.6B by the Qwen team, Alibaba Cloud. Ollaya runs open decision models locally, the way Ollama runs LLMs: typed questions in, calibrated answers out, behind a TypeSafe-compatible API.

ollaya run qwen3guard

What is in this repository

This repository holds only the files Ollaya derives, with no weights. Each graph is an ONNX export of the original model whose weights reference the authors' own weight files by byte offset, so ollaya pull downloads the weights from the upstream repositories, unmodified and pinned to a commit, and verifies their sha256.

Tag	Upstream	Files
qwen3guard:0.6b	Qwen/Qwen3Guard-Gen-0.6B@fada3b2	0.6b/model-fp32.onnx, 0.6b/decision.json, 0.6b/calibration.json, 0.6b/questions.json

Each tag has an fp32 graph, used on CPU and GPU. Each tag also has decision.json (sequence layout, special tokens) and calibration.json (temperatures). questions.json holds the built-in questions: the model answers those and no others, so requests leave questions out.

Parity

Ollaya's Rust runtime matches the transformers fp32 reference exactly on 74 texts (148 rows, 296 built-in questions). The token ids are identical, and so is the decision on every question. Probabilities are within 1.4e-5, on CPU and CUDA.

License

Same as the upstream model (Apache-2.0). Ollaya itself is Apache-2.0.