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 4, 10:00 UTC
main
[ROCm][Deepseek] dsv3.2 further optimization (#41217)
Pleaplusone committed 6 months ago
[Perf] Warmup forward_native sampler kernel (#41375)
Artem Perevedentsev committed 6 months ago
[Kernel] Pack output and LSE in DCP A2A (#41160)
sungsoo ha committed 6 months ago
[kv_offload+HMA][13/N]: Enable HMA support (#41445)
Or Ozeri committed 6 months ago
[ROCm] Enable DBO (Dynamic Batch Optimization) on ROCm (#34726)
raviguptaamd committed 6 months ago
Refractor longcat loading to use AutoWeightsLoader (#41448)
Yuyi Ao committed 6 months ago
[Bugfix][Metrics] Fix RayPrometheusMetric.labels() returning shared labeled child (#40840)
Seiji Eicher committed 6 months ago
[ROCm][CI] Upgraded UCX and RIXL (#41210)
Andreas Karatzas committed 6 months ago
[Bugfix] Pass reasoning parser kwargs to structured output (#41199)
Bugen Zhao committed 6 months ago
[compile] Add FlashInfer FP8 async TP fusion and preserve allreduce fusion ordering #27893 (#39505)
baonudesifeizhai committed 6 months ago
Add nvfp4 kv cache support (#40177)
sychen52 committed 6 months ago
[kv_offload+HMA][12/N]: Scheduler-side support for sliding window groups (#41228)
Or Ozeri committed 6 months ago
[Kernel][MoE] Support GELU on TRT-LLM NvFP4 fused MoE for Gemma4 (#41050)
Juhi Mittal committed 6 months ago
[Bugfix] Disable FlashInfer CUTLASS MoE on SM110 (Jetson Thor AGX) (#40808)
Stefano Castagnetta committed 6 months ago
Ronen Schaffer committed 6 months ago
[Model] Add Moondream3 model support(only query and caption skills) (#32325)
Dong W committed 6 months ago
feat: Enable `prompt_embeds` Content Part Support in vLLM Chat Completions API (#40720)
Luis 🚀 committed 6 months ago
[ROCm][CI] Add ROCm score absolute tolerance floor (#41341)
Andreas Karatzas committed 6 months ago
(bugfix): block_size check for flex attn (#41363)
Soyaazz committed 6 months ago
[Core] Simplify handling of `scheduler_reserve_full_isl` option (#41064)
Nick Hill committed 6 months ago