This repository contains ggml-model-tq2_0-qvac-v1.gguf, a qvac-compatible GGUF conversion of Microsoft's bitnet-b1.58-2B-4T checkpoint for Lumen.

It was converted with the pinned qvac-fabric-llm.cpp revision and must be used with the qvac BitNet compatibility patch in the Lumen application. The conversion preserves the model's offline AutoBitLinear scales and uses the checkpoint's ReLU-squared FFN activation.

SHA-256:

This is a derived distribution of the upstream Microsoft model. Refer to the upstream model card and license for terms, provenance, and limitations.