Modelpublic

lalitkarthik/scholar-3b-gguf

sourceHugging Faceapache-2.0updated 2mo agoView on Hugging Face

0likes10downloads

Model Card

scholar-3b — GGUF

Q4KM quantisation of lalitkarthik/scholar-3b-v2 merged into Qwen/Qwen2.5-3B-Instruct, for CPU inference with llama.cpp.

Judged win rate against the base model: 85.7% over 60 held-out prompts.

bash

llama-server --hf-repo lalitkarthik/scholar-3b-gguf --hf-file scholar-3b-Q4_K_M.gguf