Upload llm.mnn.json with huggingface_hub

Upload README.md with huggingface_hub

Add runtime_config.json with optimal spec decode settings

Update README with honest 3x averaged benchmark numbers

Add config_opencl.json for OpenCL draft backend support

v3: KL-distilled 0.6B from Qwen3-8B (10K samples, KL=0.339, +41% uplift on SM8850)

docs: update README — abliterated draft with benchmarks

feat: switch to abliterated draft (Huihui-Qwen3-0.6B-abliterated-v2) for better acceptance

feat: switch to abliterated draft (Huihui-Qwen3-0.6B-abliterated-v2) for better acceptance

feat: switch to abliterated draft (Huihui-Qwen3-0.6B-abliterated-v2) for better acceptance

feat: switch to abliterated draft (Huihui-Qwen3-0.6B-abliterated-v2) for better acceptance

feat: switch to abliterated draft (Huihui-Qwen3-0.6B-abliterated-v2) for better acceptance

Upload folder using huggingface_hub

initial commit