[pull] main from NVIDIA:main#1130

Merged

pull[bot] merged 7 commits intoxinhe-nv:mainfrom

Mar 13, 2026

Commits on Mar 13, 2026

[None][feat] Add flashinfer api for TRTLLMGenFusedMoE (#10453)
rosong11
authored
[None][Test] Add multinode e2e and accuracy cases on DGX-Spark (#12110)

JennyLiu-nv
and
Jenny Liu
authored
[TRTLLM-11207][requirements] Update numpy version to 2 (#11280)
Funatiq
authored
[None][chore] Fix KVCacheManagerV2 shrink for last level and improve init_ratio (#12112)
lowsfer
authored
[TRTLLM-10319][feat] Dynamic draft length on spec decode one-model path (#10860)
zheyuf
authored
[TRTLLM-11288][feat] Configurable warmup shapes for VisualGen (#12107)
luyiyun1021
authored
[None][feat] add trtllm-gen kernels for glm4.7 and support groupsTokensHeadsQ + e2m1 output (#11643)
PerkzZheng
authored