All pull requests

Pull requests

• [None][fix] Merge back #19593 from release/1.3: Make MoE communication timeouts follow the warmup phase

  #20246·xxi-nv opened Oct 14, 2026Collaborator·1 task done·

  ·
  2

  

• [None][perf] Cut per-response overhead in the LLM API streaming result path

  #20244·jthomson04 opened Oct 13, 2026Collaborator·1 task done·

  ·
  6

  

• [https://nvbugs/6800150][test] Unwaive the KV budget split tests and fix their stale mocks

  #20243·moraxu opened Oct 13, 2026Collaborator·1 task done·

  ·
  4

  

• [None][fix] Honor separate V token stride in context FMHA

  #20242·achartier opened Oct 13, 2026Collaborator·2 tasks done·

  ·
  

• [https://nvbugs/6227108][perf] extend AllReduce tuning to the resolved token budget

  #20241·dhansen-nvidia opened Oct 13, 2026Collaborator·1 task done·

  ·
  1

  

• [TRTLLMINF-486][infra] Get sonar-scanner-cli from internal Artifactory and verify its SHA-256

  #20240·sharabeshj-nvda opened Oct 13, 2026Collaborator·

  ·
  4

  

• [None][infra] Reuse rank processes across visual_gen multi-GPU tests

  VisualGen
  #20239·ishovkun opened Oct 13, 2026Contributor·1 task done·

  ·
  

• [TRTLLM-20237][feat] Enable PPLX decider support

  Community want to contribute
  #20238·anurags25 opened Oct 13, 2026Contributor·

  ·
  1

  3

  

• [https://nvbugs/6919428][fix] Keep DeepSeek-V4 TP decode all-reduces off NCCL

  #20233·chenfeiz0326 opened Oct 13, 2026Collaborator·1 task done·

  ·
  1

  

• [None][chore] Mass merge release/1.13 branch back to main

  VisualGen
  #20232·SimengLiu-nv opened Oct 13, 2026Collaborator·1 task·

  ·
  15

  

• [https://nvbugs/6613027][fix] Responses API: pass image input to multimodal models

  #20231·JunyiXu-nv opened Oct 13, 2026Collaborator·1 task done·

  ·
  13

  

• [https://nvbugs/6682814][fix] Use node IPs for perf-sanity disagg servers and report unresolvable bind hosts

  #20230·JunyiXu-nv opened Oct 13, 2026Collaborator·1 task done·

  ·
  14

  

• [None][feat] Locality domain support for the C++ KV cache manager v2 and two-stream attention

  api-compatible
  #20229·farazkh80 opened Oct 13, 2026Collaborator·1 task done·

  ·
  

• [None][fix] Honor deterministic mode in fused mHC dispatch

  #20228·shikicloud opened Oct 13, 2026Collaborator·1 task done·

  ·
  1

  

• [None][fix] Propagate deterministic flags to MoE finalize selection

  #20227·shikicloud opened Oct 13, 2026Collaborator·1 task done·

  ·
  1

  

• [None][fix] Keep GDN decode recurrence stable in deterministic mixed batches

  #20226·shikicloud opened Oct 13, 2026Collaborator·1 task done·

  ·
  3

  

• [None][chore] Run test_disaggregated_videomme[qwen3vl_2b_instruct] in isolation

  #20225·HuiGao-NV opened Oct 13, 2026Collaborator·1 task done·

  ·
  4

  

• [None][feat] Support PrimsTS SageAttention for VisualGen VSA and SOL

  VisualGen
  #20224·heyuhhh opened Oct 13, 2026Collaborator·1 task done·

  ·
  2

  

• [None][fix] Size the Blackwell spec-dec tree mask for the SwapsMmaAb kernel layout

  #20223·lori-ren opened Oct 13, 2026Contributor·1 task done·

  ·
  3

  

• [TRTLLM-17005][refactor] Component-local C++ Tests (Part 2)

  #20222·Funatiq opened Oct 13, 2026Collaborator·1 task·

  ·
  3

  

• [https://nvbugs/6911105][fix] Remove sparse metadata preparation waiver

  #20220·ziyixiong-nv opened Oct 13, 2026Collaborator·1 task·

  ·
  4

  

• [TRTLLM-15863][feat] Export the Attention module-layer classes from _torch.attention lazily

  #20218·YihuiLu512 opened Oct 13, 2026Collaborator·1 task done·

  ·
  11

  

• [None][feat] Add the blob store driver boundary and an in-memory test driver

  #20217·nv-xtf opened Oct 13, 2026Collaborator·1 task done·

  ·
  3

  

• [None][infra] add Jenkins instance name to Slurm identifiers

  #20216·lyxxn0414 opened Oct 13, 2026Collaborator·

  ·
  3

  

• [TRTLLM-15864][chore] Add deprecated forwarding shims for the old VisualGen attention paths

  ci: full pre-merge approvedVisualGen
  #20215·YihuiLu512 opened Oct 13, 2026Collaborator·1 task done·

  ·
  9