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 6, 05:45 UTC
main
[ROCm][CI] Update GSM8K eval config to use fp8-and-mixed models list (MI355) (#37721)
Andreas Karatzas committed 8 months ago
[Responses API] Add kv_transfer_params for PD disaggregation (#37424)
Bongwoo Bak committed 8 months ago
[ROCm][CI] Mark gemma3 as large GPU test to avoid OOM on MI250 (#37610)
Andreas Karatzas committed 8 months ago
Revert "[compile] Initialize passes at VllmBackend init" (#37733)
Simon Mo committed 8 months ago
Andreas Karatzas committed 8 months ago
[MoE Refactor] Mxfp4 oracle rebased (#37128)
Yongye Zhu committed 8 months ago
[Frontend] Remove librosa from audio dependency (#37058)
Isotr0py committed 8 months ago
Add get_device_uuid for rocm (#37694)
tmm77 committed 8 months ago
elastic_ep: Fix issues with repeated scale up/down cycles (#37131)
Itay Alroy committed 8 months ago
[BugFix] Allow qk_nope_head_dim=192 in FlashInfer MLA backend checks (#37475)
Kaihang Jiang committed 8 months ago
[Refactor] Remove unused dead code (#36171)
Wentao Ye committed 8 months ago
[Model Runner V2] Support Streaming Inputs (#37028)
Santino Ramos committed 8 months ago
[ROCm][CI] Setting some mi325_4 tests back to optional (in parity with upstream) (#37711)
Andreas Karatzas committed 8 months ago
[Bugfix] Disable monolithic TRTLLM MoE for Renormalize routing (#37591) (#37605)
Vadim Gimpelson committed 8 months ago
[compile] Initialize passes at VllmBackend init (#35216)
Angela Yi committed 8 months ago
[Model] Update Kimi-K25 and Isaac processors to fit HF-style (#37693)
Cyrus Leung committed 8 months ago
[Bugfix] Disable --calculate-kv-scales for hybrid GDN/Mamba+Attention… (#37565)
Le Yang committed 8 months ago
refactor: abstract deepgemm support into platform (#37519)
SherryC41 committed 8 months ago
[Attention] Support distinguishing between short extends and decodes (#37303)
Lucas Wilkinson committed 8 months ago
fix CUDAGraph memory being counted twice (#37426)
Peter Pan committed 8 months ago