vllm-project/vllm-metal v0.3.0.dev20260819070634

on GitHub

What's Changed

• Support Hugging Face config sources for local GGUF weights by @LxYuan0420 in #624
• Add M5 NAX prefill kernel for paged GQA attention by @WindChimeRan in #623

Full Changelog: v0.3.0.dev20260818075955...v0.3.0.dev20260819070634