Add Q4_K_M GGUF (merged Qwen2.5-0.5B + LoRA)

Fix: handle tar.gz release format for Ubuntu llama.cpp binaries

Add F16 GGUF (merged Qwen2.5-0.5B + LoRA)

Use pre-built llama.cpp binaries instead of cmake build

Install cmake via apt-get before building llama-quantize

Remove redundant pip install (gguf/numpy already in PEP 723 header)

Fix: install gguf package directly instead of llama.cpp requirements.txt

Add GGUF conversion script

initial commit