开源大模型推理框架发布 v0.30.0 版本更新，包含来自 315 位贡献者的 762 次提交。新版本支持了多款前沿大模型及架构优化，包括支持 MXFP8 格式 KV 缓存与 FlashMLA 的 DeepSeek-V4.1-Flash、支持 ROCm 与 LoRA 的 DeepSeek-V4-Flash-Vision-Exp、支持 EPLB 的 GLM-5.3-Flash，以及 K2-Horizon、Cohere Compass、Bailing V3 VL 等，并引入了基于 AVX512/AMX 的 CPU 后端支持。
vllm-proto 0.1.0
v0.29.0 Highlights This release features 594 commits from 277 contributors (91 new)! Model Runner V2 is now the default for all models ( #53183 ), completing the rollout that began with pooling models ( #48290 ). MRV2 also gained CUDA graph memory profiling for KV cache auto-sizing ( #53306 ), batch-sharded sampling that cuts per-step logits memory by 1/TP ( #50465 ), prompt embeds ( #42963 ), extract_hidden_states speculation ( #49811 ), padded FULL cudagraph dispatch for uniform decode under spec decode ( #53407 ), and DP-sync skipping before EAGLE/MTP draft prefill ( #53694 ). MRV1 remains in use for a few ROCm models and features MRV2 doe
[Bugfix][Core] Apply dense prefix cache default to hybrid models ( #55 …