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 4, 22:30 UTC
main
[BUGFIX] Fix Pixtral consolidated format vision weight loading (#39916)
Julien Denize committed 7 months ago
[XPU] [torch.compile] Skipping CUDA graph memory estimation to avoid startup errors. (#39977)
Chaojun Zhang committed 7 months ago
[Qwen][Bugfix] Fixes sigmoid activation in torch impl of RMSNormGated. (#40245)
Tao He committed 7 months ago
fix: Do not make function calls when request has no tools for /v1/responses (#40314)
Yuan Tang committed 7 months ago
[Bugfix][Responses API] Fix streaming tool calls on /v1/responses (#39892)
Hoang Nguyen committed 7 months ago
[Bugfix] Properly initialize `PerTensorScaleParameter` for fused-on-disk checkpoints (#39765)
Shinichi Hemmi committed 7 months ago
[Doc] Fix typos in token_embed pooling documentation (#40266)
Lxx committed 7 months ago
[BugFix][XPU] fix lora ops bgmv_expand size not match (#39989)
Liangliang Ma committed 7 months ago
[ci] Make ecr authenticate non blocking (#40305)
Kevin H. Luu committed 7 months ago
[ROCm] Fix cu_seqlens_q off-by-one in AITER FA speculative decode path (#39120)
Andrew Barnes committed 7 months ago
Fix MoE backend selection for LoRA (unquantized MoE) (#40273)
danisereb committed 7 months ago
Optimize nemotron VL image/video preprocessing (#40283)
Netanel Haber committed 7 months ago
[FEAT] [Perf] [Gemma4] Fused Gemma4 Routing Function Triton (#39083)
TJian committed 7 months ago
Flora Feng committed 7 months ago
[KV Offload] Pass request context (#39185)
omerpaz95 committed 7 months ago
[KV Connector] Allow metrics of multiple connectors of same types in multi connector. (#40010)
omerpaz95 committed 7 months ago
[Frontend] Preserve structured output special tokens in offline LLM.chat (#39352)
Luciano Martins committed 7 months ago
[Bugfix] Guard mxfp4_experts_quant bindings on ENABLE_NVFP4_SM100 (#40191)
mysterious hhhh committed 7 months ago
[Attention] TurboQuant: remove redundant random signs, add prior art attribution (#40194)
Dan Alistarh committed 7 months ago
Added general ND x ND matmul and unit test for it (#39909)
Yusuf Mohammad committed 7 months ago