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 8, 07:30 UTC
main
[Feature] Pipeline Parallel Async send/recv, 2.9% E2E throughput improvement (#33368)
Wentao Ye committed 9 months ago
[Core] Move pause and resume functions into engine (#34125)
Aaron Hao committed 9 months ago
[KVConnector] Clean up redundant code in KV connectors (#34147)
Martin Hickey committed 9 months ago
[Perf] fused_moe: add int4_w4a16 benchmark support and tuning config (#34130)
Matthias Gehre committed 9 months ago
[Bugfix] fix the import path in moe test utils.py (#34245)
Marek Michalowski committed 9 months ago
[Bug Fix] Fix MambaManager.cache_blocks() crash on null blocks in align mode (#34418)
haosdent committed 9 months ago
[BugFix] Fix and optimize max_num_blocks_per_req calculation for MambaSpec (#34440)
Harry Huang committed 9 months ago
[New Model] support new model ovis2.6 (#34426)
myselvess committed 9 months ago
[Refactor] Call renderer for online IO processor request (#34490)
Cyrus Leung committed 9 months ago
[Bugfix] Fix mamba state dtype setting for Qwen3-Next and Qwen3.5 (#34489)
Roger Wang committed 9 months ago
[Refactor] Pass full VllmConfig to Renderer (#34485)
Cyrus Leung committed 9 months ago
[CI/Build] Fix CUDA re-initialization error in distributed model tests (#34491)
Cyrus Leung committed 9 months ago
[Bugfix] Fix encoder cache underestimation for GLM-4V/GLM-OCR single image (#34483)
haosdent committed 9 months ago
[Bugfix] Standardize getting number of image patches/tokens (#34358)
Cyrus Leung committed 9 months ago
[ROCm][CI] Fix serving tokens test failures (#34047)
Andreas Karatzas committed 9 months ago
Add new sections to CODEOWNERS (#34309)
Cyrus Leung committed 9 months ago
[Hybrid] Fix and optimize block-aligned splitting in mamba cache align mode (#33706)
Harry Huang committed 9 months ago
[Bugfix] Fix Random Dataset Prefix Length Inaccuracy (#33907)
Frank Wang committed 9 months ago
[Kernel] [Helion] [5/N] Add Helion Autotuning infrastructure (#34025)
Yanan Cao committed 9 months ago
[Bugfix] Delete unused redundant code in Kimi-K2.5 (#34427)
LoganJane committed 9 months ago