switch to prebuilt wheel to beat build timeout

switch to prebuilt wheel to beat build timeout

fix: bump llama-cpp-python to 0.3.12+ for SmolLM3 GGUF support

revert: llama-cpp-python 0.3.20 build timed out on cpu-basic; revert to 0.3.9 with load_failed runtime status (documented)

fix: bump llama-cpp-python 0.3.9 -> 0.3.20 to support SmolLM3 GGUF architecture

touch: force rebuild 2026-04-17T20:50:23.682702Z

fix: repoint to unsloth/SmolLM3-3B-GGUF (ggml-org ref 404)

init: CPU LLM Space template

initial commit