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 7, 22:45 UTC
main
[Model Runner V2] Enable CUDA graph for Eagle3 (#35040)
Woosuk Kwon committed 9 months ago
Fix apply_top_k_top_p_triton called by non-cuda logits Tensor (#35030)
Xiao Li committed 9 months ago
[Benchmark] Use `sns.relplot` for plotting (#35027)
Cyrus Leung committed 9 months ago
[New Model] Add ColModernVBERT (#34558)
Athrael Soju committed 9 months ago
[CI] Bump mteb version to `mteb[bm25s]>=2, <3` for pooling model unit tests (#34961)
Wentao Ye committed 9 months ago
[CI] Stabilizing ROCm amd-ci signal and minor name fix in upstream (#35008)
Andreas Karatzas committed 9 months ago
[Model Runner V2] Support attention group (#35036)
Woosuk Kwon committed 9 months ago
[Model Bash][DSR1] Add selective dynamic shape marking for CustomOp (#34900)
Vadim Gimpelson committed 9 months ago
[Model Runner V2] Support Eagle3 (no CUDA graph) (#35029)
Woosuk Kwon committed 9 months ago
[CI/Build] Fix gRPC version mismatch (#35013)
Cyrus Leung committed 9 months ago
[Frontend] Add automatic language detection for Whisper transcription (#34342)
Roman committed 9 months ago
[Bugfix] Gate 256-bit instructions to CUDA 12.9+ (#34791)
Huy Do committed 9 months ago
[Benchmark] Improve benchmarks (#35012)
Cyrus Leung committed 9 months ago
[Doc] Fix example of eagle3 (#34960)
petrpechman committed 9 months ago
[Core] Minor structured-output related scheduler optimization (#34765)
Nick Hill committed 9 months ago
[PD] Change kv_load_failure_policy Default from "recompute" to "fail" (#34896)
Nicolò Lucchesi committed 9 months ago
[ROCm] Enable bitsandbytes quantization support on ROCm (#34688)
BADAOUI Abdennacer committed 9 months ago
[ROCM] Optimize ROCM_AITER_FA spec decode eagle performance (#34541)
jennyyyyzhen committed 9 months ago
[ROCm][CI] Fix spec decode logprobs flakiness and parametrize tree attention backends (#34599)
Andreas Karatzas committed 9 months ago
[ROCm][AITER] Fix aiter paged_attention_v1 decode for sliding window and head_size < 64 (#34570)
Andreas Karatzas committed 9 months ago