Freaksterz/Qwen3.8-27B-SmoothQuant-W8A8-INT8

v3: add MTP head (model-mtp.safetensors, BF16, `re:.*mtp.*` ignored) + index/config; card: group naming fix, MTP AL

v3: SmoothQuant + activation-aware GPTQ, mixed W8A8/W8A16 (288/112), no rotation — KLD 0.00556, DFlash2-compatible (AL 4.34)

Remove v1-era quantization scripts superseded by the rotation pipeline

Rotation + SmoothQuant + GPTQ W8A8 (KLD 0.01414 -> 0.01098); fix missing processor/tokenizer files (discussion #1)

Upload folder using huggingface_hub

initial commit