Repository navigation

[None][chore] Mass merge release/1.3 branch back to main (batch 2) - #20843

Open

SimengLiu-nv wants to merge 23 commits into

NVIDIA:mainfrom

SimengLiu-nv:port/release-1.3-missing-to-main-batch2

Open

[None][chore] Mass merge release/1.3 branch back to main (batch 2)#20843
SimengLiu-nv wants to merge 23 commits into
NVIDIA:mainfrom
SimengLiu-nv:port/release-1.3-missing-to-main-batch2

Commits on Oct 26, 2026
- [https://nvbugs/6665906][fix] Cherry-pick #19788 from release/1.3: Restore Mistral Large 3 multimodal inference
  
  pranav-nvidia
  authored and
  SimengLiu-nv
  committed
  1c9f751
- [https://nvbugs/6562942][fix] Cherry-pick #19815 from release/1.3: Load Qwen3.5 packed INT4 checkpoints
  
  pranav-nvidia
  authored and
  SimengLiu-nv
  committed
  da07e21
- [https://nvbugs/6877539][fix] Cherry-pick #20029 from release/1.3: Fold GQA group into SDPA query axis in MiniMax-M3 decode attention
  
  authored and
  SimengLiu-nv
  committed
  f811713
- [None][test] Cherry-pick #20132 from release/1.3: Move cases to multinode test list
  
  crazydemo
  authored and
  SimengLiu-nv
  committed
  682c90e
- [https://nvbugs/6463987][fix] Cherry-pick #20136 from release/1.3: Pin GB300 GLM-5 ctx workers to their GPU's NUMA node in perf-sanity
  
  authored and
  SimengLiu-nv
  committed
  eaeba34
- [https://nvbugs/6859934][fix] Cherry-pick #19847 from release/1.3: Process-wide workspace caches keyed by `Mapping` with a per-mapping lock…
  
  authored and
  SimengLiu-nv
  committed
  ca66314
- [https://nvbugs/6792615][fix] Cherry-pick #20152 from release/1.3: Keep session-prefetch output off sys.stdout
  
  sunnyqgg
  authored and
  SimengLiu-nv
  committed
  e6ec4c5
- [https://nvbugs/6918015][doc] Cherry-pick #20193 from release/1.3: Fix spelling and capitalization in release/1.3 docs
  
  StanleySun639
  authored and
  SimengLiu-nv
  committed
  455c15f
- [https://nvbugs/6379406][fix] Cherry-pick #20156 from release/1.3: Use coherent managed memory for EPLB when GDRCopy is unavailable
  
  sunnyqgg
  authored and
  SimengLiu-nv
  committed
  50dbedc
- [https://nvbugs/6911393][fix] Cherry-pick #20195 from release/1.3: Run the launcher test's python3 stand-in with -S so start-up hooks cannot…
  
  authored and
  SimengLiu-nv
  committed
  5c078b5
- [https://nvbugs/6825983][perf] Cherry-pick #20264 from release/1.3: Share TRTLLM-Gen FMHA cubins across isolated MPI ranks
  
  chenfeiz0326
  authored and
  SimengLiu-nv
  committed
  cbecf90
- [https://nvbugs/6912133][fix] Cherry-pick #20248 from release/1.3: Detect dead MPI pool manager thread in MpiPoolSession
  
  asfiyab-nvidia
  authored and
  SimengLiu-nv
  committed
  b7cd5eb
- [https://nvbugs/6862662][fix] Cherry-pick #20415 from release/1.3: Bound FP8 MLA page-table loads for partial tiles
  
  yuxianq
  authored and
  SimengLiu-nv
  committed
  ab40fa8
- [None][fix] Cherry-pick #20381 from release/1.3: Follow up PrimTS refresh backport review tests and docs
  
  yuxianq
  authored and
  SimengLiu-nv
  committed
  ca51b44
- [https://nvbugs/6932241][fix] Cherry-pick #20588 from release/1.3: Scope NCCL window leases in helix attention tests
  
  yuxianq
  authored and
  SimengLiu-nv
  committed
  e16a156
- [https://nvbugs/6553508][fix] Cherry-pick #20344 from release/1.3: Reject conflicting LoRA adapter identities
  
  yibinl-nvidia
  authored and
  SimengLiu-nv
  committed
  57b691f
- [https://nvbugs/6857649][fix] Cherry-pick #20346 from release/1.3: Validate native transfer destinations before submission
  
  yibinl-nvidia
  authored and
  SimengLiu-nv
  committed
  a9b9c77
- [https://nvbugs/6384589][fix] Cherry-pick #20531 from release/1.3: Mount a 640G /dev/shm for online-EPLB perf cases
  
  xxi-nv
  authored and
  SimengLiu-nv
  committed
  1190531
- [https://nvbugs/6944374][fix] Cherry-pick #20657 from release/1.3: Fix rank-asymmetric NCCL window allocation hang during PARALLEL autotuning
  
  lori-ren
  authored and
  SimengLiu-nv
  committed
  5020a21
- [https://nvbugs/6885370][fix] Cherry-pick #20284 from release/1.3: Keep Nemotron-H SSM cache dtype unset
  
  dominicshanshan
  authored and
  SimengLiu-nv
  committed
  11f6764
- [https://nvbugs/6887489][fix] Cherry-pick #20651 from release/1.3: Restrict coordinator HTTP cluster storage to loopback
  
  yibinl-nvidia
  authored and
  SimengLiu-nv
  committed
  dc6104d
- [https://nvbugs/6964933][test] Cherry-pick #20794 from release/1.3: Use FlashInfer for Gemma4 NVFP4 perf on SM107
  
  ruodil
  authored and
  SimengLiu-nv
  committed
  ab6d70e
- [None][fix] Cherry-pick #20777 from release/1.3: Validate telemetry deployment IDs as UUIDs
  
  Mgluhovskoi
  authored and
  SimengLiu-nv
  committed
  034f5dd

Commits

Commits on Oct 26, 2026