EchoLabs33/zamba2-2.7b-instruct-hxq

feat: add HXQ_AFFINE_6 GGUF (6.27 bpw, RTX 3090 benchmarked)

docs: add GGUF runtime benchmark tables (RTX 3090, 4-way comparison)

Update model card: LoRA fine-tuning now supported via HelixLinearSTE

fix: model card metadata (language, results, branding, ratios)

rebrand: -helix → -hxq in model card

Recompressed with skip-tensor fix (147 modules, 162 SSM params restored, conversion_receipt PASS)

Upload README.md with huggingface_hub

Upload config.json with huggingface_hub

Upload folder using huggingface_hub

initial commit