aj9o9/Qwen3.8-27B-Escha-W2-GGUF

Upload README.md with huggingface_hub

Document MTP speculative decoding

Add MTP draft (F16 layer, Q8_0 embed, Q4_K head)

Full context under 20 GB; VRAM table for smaller cards

Full context fits under 20 GB, measured

Fix wrong information

Add model card

Upload Escha-Qwen3.8-27B-W2.gguf with huggingface_hub

Upload Escha-Qwen3.8-27B-W2-Q8E.gguf with huggingface_hub

initial commit