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, 20:30 UTC
main
[Bugfix] Fix minimax_m2 tool parser when stream interval > 1 (#35895)
Flora Feng committed 8 months ago
[XPU][Doc] Remove manual OneAPI install step, now handled by torch-xpu (#36831)
Kunshang Ji committed 8 months ago
[Frontend] Exclude anthropic billing header to avoid prefix cache miss (#36829)
Nick Hill committed 8 months ago
[Perf] Optimize compute maxsim using batched version, 3.2% E2E throughput improvement (#36710)
Wentao Ye committed 8 months ago
[ROCm][CI/Build] Add gfx1152/gfx1153 (Krackan) to HIP supported architectures (#36499)
Matthias Gehre committed 8 months ago
[BUG] Fix async rlhf tests (#35811)
Aaron Hao committed 8 months ago
[Kernel] [Helion] [15/N] Split config files into per-platform files (#36698)
Yanan Cao committed 8 months ago
Yanan Cao committed 8 months ago
[Kernel] [Helion] [14/N] Set autotune_ignore_errors=True during autotuning (#36683)
Yanan Cao committed 8 months ago
[Model Runner V2] Add WhisperModelState [6/N] (#35790)
Woosuk Kwon committed 8 months ago
[Model Runner V2] Add probabilistic rejection sampling for spec decoding (#35461)
Giancarlo Delfin committed 8 months ago
[LMCache] Pass TP size in lookup for MLA multi-reader locking (#36129)
maobaolong committed 8 months ago
[BugFix][kv_offload] Fix offloading decodes with async scheduling (#33881)
Or Ozeri committed 8 months ago
[Bugfix][ROCm] Strip block_size before attention backend validation (#36274)
jennyyyyzhen committed 8 months ago
[BugFix]: add bagel to MM_PREFIX_LM_MODELS (#36316)
汪志鹏 committed 8 months ago
[Refactor] Remove dead code in KV connector (#36424)
Wentao Ye committed 8 months ago
[ROCm] add tuned moe_wna16_triton kernel configs for CDNA4 (#35093)
Amanzhol Salykov committed 8 months ago
Make Gemma and Gemma 2 accept `inputs_embeds` like Gemma 3 (#36787)
Harry Mellor committed 8 months ago
[torch.compile] Add support for non-contiguous fused RMSNorm + group quant (#36551)
Luka Govedič committed 8 months ago
[KVConnector] Support worker -> scheduler metadata (#31964)
Or Ozeri committed 8 months ago