Ingot-IV-27B-Q2 三值量化模型发布
原标题:dendriteholdings/Ingot-IV-27B-Q2
AI 摘要
Hugging Face 上发布了 Ingot-IV-27B-Q2,这是 Bittensor 子网 Ingot 竞赛中获胜矿工提交的三值量化模型,基于 Prism ML 的 Ternary Bonsai 2 27B 微调,重新调整了分块缩放与归一化权重以更接近 Qwen3.8-27B 教师模型。该模型提供 1.75 bpw(5.95 GB)和 2.125 bpw(7.21 GB)两个 GGUF 文件,在 SWE-bench Multilingual、GPQA Diamond、τ³-bench Banking 等基准上优于 Bonsai 2,可在 16 GB MacBook 上运行,但需使用 PrismML 或 DendriteLabs 的 llama.cpp 分支。
正文节选
--- license: apache-2.0 base_model: prism-ml/Ternary-Bonsai-2-27B-gguf base_model_relation: finetune library_name: llama.cpp pipeline_tag: text-generation tags: - gguf - ternary - ingot - bittensor --- # Ingot-IV-27B-Q2 A public mirror of a model that won on Ingot, a Bittensor subnet where miners quantize a fixed teacher model and the validator crowns the copy that stays closest to it. It is Prism ML's [Ternary Bonsai 2 27B](https://huggingface.co/prism-ml/Ternary-Bonsai-2-27B-gguf), refined: