A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 12, 14:45 UTC
main
[ROCm][MLA] enable fp8 MLA decode on ROCm (#28032)
gbyu-amd committed 12 months ago
Pleaplusone committed 12 months ago
[Perf] Disable DeepGEMM MoE by default when TP=8 is used (#29346)
Michael Goin committed 12 months ago
Scheduled removal of `ParallelConfig`'s direct child EPLB fields (#29324)
Harry Mellor committed 12 months ago
[BugFix] Fix duplicate id tool-call race condition (#29355)
Nick Hill committed 12 months ago
[XPU] upgrade torch & ipex 2.9 on XPU platform (#29307)
Kunshang Ji committed 12 months ago
[Spec Decode] Add support for EAGLE3 heads that do not use_aux_hidden_states (#27688)
Hanjie Qiu committed 12 months ago
[UX] Raise error for attn backend of batch invariant (#29348)
Wentao Ye committed 12 months ago
[Bugfix] Only use triton_kernels for MXFP4 on SM90 and SM100 (#29339)
Michael Goin committed 12 months ago
[Hybrid Allocator] Better layer padding strategy for gpt-oss eagle (#29303)
Chen Zhang committed 12 months ago
Remove VLLM_SKIP_WARMUP tip (#29331)
Tyler Michael Smith committed 12 months ago
[Tests] Verify gpt_oss package is installed in harmony tests (#29336)
Nick Hill committed 12 months ago
[Model Runner V2] Simplify Eagle bookkeeping with num_rejected (#29347)
Woosuk Kwon committed 12 months ago
[XPU]fix Kimi-VL-A3B-thinking on xpu (#29309)
Yan Ma committed 12 months ago
[CI/Build] Moves to cuda-base runtime image while retaining minimal JIT dependencies (#29270)
Benjamin Bartels committed 12 months ago
[Model Runner V2] Add minor clarification comments for Eagle (#29332)
Woosuk Kwon committed 12 months ago
[Multimodal][Qwen3 Omni] Make Qwen3 Omni work with audio-in-video inputs in V1 engine. (#27721)
Chenheli Hua committed 12 months ago
[MoE][Refactor] Make select_experts a non-static method (#29067)
bnellnm committed 12 months ago
[Model Runner V2] Change Numba AoT to JIT (#29328)
Woosuk Kwon committed 12 months ago
[Model Runner V2] Implement Single-step Eagle 1 (#29300)
Woosuk Kwon committed 12 months ago