These are quantizations of the model Jackrong / Qwopus3.6-27B-v1-preview
I've added the MTP layer on it.
My personal speed improvement on my 7900XTX with the vulkan backend has been from ~40 tps to around ~60 tps.
An imatrix has been calulated for coding tasks, as such it is specialized for coding.
Quick Start
- Download the latest release of llama.cpp.
- Download your preferred model variant from below.