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, 16:00 UTC
main
[Model Runner V2] Change bookkeeping logic in preparation for spec decoding (#29194)
Woosuk Kwon committed 12 months ago
[Bugfix] Use lazy string reference for DeepseekV3Config in config registry (#28958)
Luke committed 12 months ago
[ROCm][CI] Fix "Cannot re-initialize CUDA in forked subprocess" in test_pynccl.py (#29119)
Micah Williamson committed 12 months ago
[CI/Build][AMD] Add check for flash_att_varlen_func to test_tree_attention.py (#29252)
rasmith committed 12 months ago
rasmith committed 12 months ago
[Model Runner V2] Minor fix for cudagraph_utils (#29256)
Woosuk Kwon committed 12 months ago
[Doc] Update more docs with respect to V1 (#29188)
Cyrus Leung committed 12 months ago
docs: fixes distributed executor backend config for multi-node vllm (#29173)
Michael Act committed 12 months ago
[Fix] Add SM check to flashinfer MOE backend (#29144)
jiahanc committed 12 months ago
Patch DeepEP when building docker image with CUDA 13 (#29154)
Qidong Su committed 12 months ago
[Chore] Update batch invariant code owner (#29246)
Wentao Ye committed 12 months ago
Refactor: Move CUDA graph dispatch logic earlier (#27382)
Yizhou committed 12 months ago
[BugFix] Fix chunked prompt logprobs + preemption (#29071)
Nick Hill committed 12 months ago
[CI/build] Removes source compilation from runtime image (#26966)
Benjamin Bartels committed 12 months ago
[Bugfix] Use HF config fields as fallback when loading Mistral config (#29239)
Cyrus Leung committed 12 months ago
[perf][cpu] Accelerate paged attention GEMMs (QK, PV) on Arm CPUs with NEON (#29193)
Fadi Arafeh committed 12 months ago
chore: add RTX_PRO_6000 GLM4.6-FP8 kernel tuning (#29240)
Federico committed 12 months ago
Fix EVS crash when using `video_embeds` inputs in Qwen2.5-VL (#29232)
ZiTian Zhao committed 12 months ago
[Misc] Fix pre-commit (#29238)
Cyrus Leung committed 12 months ago
[BugFix] Fix returned logprobs with spec decode + prefill chunking (#29216)
Nick Hill committed 12 months ago