v0.15.0

What's Changed

• Optimize Fused MoE LoRA Kernel PerformanceOptimize fused MoE LoRA kernel performance by @cwazai in #203
• Support vllm 0.15.0 by @ILikeIneine in #209

Full Changelog: v0.14.0...v0.15.0

v0.14.0

What's Changed

• Support vllm 0.14.0 by @ILikeIneine in #197
• support ray cluster test by @ILikeIneine in #191
• Optimize fused MoE LoRA kernel performance by @cwazai in #214

Full Changelog: v0.13.0...v0.14.0

v0.13.0

Release Note

This release is a routine update vLLM upstream v0.13.0, includes some upstream synchronization, feature enhancements, bug fixes, and build-system improvements.

What's Changed

• support vllm v0.13.0 by @ILikeIneine in #190
• Delete docker/reference.Dockerfile by @ILikeIneine in #193

Full Changelog: v0.12.0...v0.13.0

v0.12.0

Release Note

This release is a routine update vLLM upstream v0.12.0, includes some upstream synchronization, feature enhancements, bug fixes, and build-system improvements.

What's Changed

• [fix] fix assert failed with max_context_chunk when use dcp by @ILikeIneine in #184
• add test framework by @ILikeIneine in #181
• feat: support vllm v0.12.0 by @ILikeIneine in #185

Full Changelog: v0.11.2...v0.12.0

v0.11.2

Release Note

This release is a routine update focused on catching up with the vLLM upstream v0.11.2 release. It includes upstream synchronization, feature enhancements, bug fixes, and build-system improvements.

What's Changed

• chore: lazy import deep_gemm by @ILikeIneine in #110
• dockerfile: make some arrangement by @ILikeIneine in #111
• update: Add new models and update existing entries by @wing016 in #122
• vllm_metax.Dockerfile: install cleanup by @dtrifiro in #116
• feat!: support v0.11.1 by @ILikeIneine in #112
• [BugFix] fix moe quant error by @lianjiezh in #141
• Support vllm v0.11.2 by @ILikeIneine in #157
• update: support flash_attn mtp+cudagraph by @SherryC41 in #172
• update: remove redundant call in fused_moe by @SherryC41 in #175
• commits for release v0.11.2 by @ILikeIneine in #195

New Contributors

• @zkf-qwj made their first contribution in #91
• @wing016 made their first contribution in #122
• @dtrifiro made their first contribution in #116

Full Changelog: https://github.com/MetaX-MACA/vLLM-metax/commits/v0.11.2

v0.10.2

🚀 vllm-metax 0.10.2 Release Notes

This release delivers a regular iteration on vllm-metax, bringing improved compatibility, performance boosts, and important bug fixes.

✨ What's New

• 🔄 Upgraded compatibility with vLLM v0.10.2, ensuring smooth integration and feature alignment.
• ⚡ Performance optimizations for:
  • LoRA fine-tuning workflows
  • Qwen3-Next-80B
  • GLM4.5
• 🐞 Bug fixes addressing issues in:
  • deepseek_mtp
  • eagle speculative decoding