cloudyu's picture
Qwen3.8-Flash-Next-E384-FP4-Local
129dc06
Raw History Blame Contribute Delete
452 Bytes
[build-system]
requires = ["setuptools>=61"]
build-backend = "setuptools.build_meta"
[project]
name = "q38-mtp-draft"
version = "0.1.0"
description = "Cheap lossless MTP drafter for Qwen3.8-Flash-Next (64k draft vocab + NVFP4 Marlin draft head / MTP linears) for vLLM 0.31 Model Runner V2"
requires-python = ">=3.10"
[project.entry-points."vllm.general_plugins"]
q38_mtp_draft = "q38_mtp_draft:main"
[tool.setuptools]
py-modules = ["q38_mtp_draft"]