Repository navigation

feat(slurm): integrate Slurm execution for 1.0.0 release - #1040

Draft

nabinchha wants to merge 75 commits into

mainfrom

feat/slurm-execution

Draft

feat(slurm): integrate Slurm execution for 1.0.0 release#1040
nabinchha wants to merge 75 commits into
mainfrom
feat/slurm-execution

Commits on Oct 27, 2026
- feat: add optional Slurm package (#856)
  
  andreatnvidia
  authored and
  nabinchha
  committed
  707c4fd
- feat: define public generation invocation contract (#854)
  
  andreatnvidia
  authored and
  nabinchha
  committed
  57a779a
- feat: add lazy CLI extensions (#863)
  
  andreatnvidia
  and
  nabinchha
  committed
  b0ffab1
- feat: define Slurm runtime and state record contracts (#878)
  nabinchha
  committed
  553187a
- feat: define Slurm config and plan contracts (#879)
  
  andreatnvidia
  and
  nabinchha
  committed
  fbff83f
- feat: integrate Slurm plan and runtime state (#889)
  nabinchha
  committed
  021cc2d
- test: add deterministic Slurm fakes (#888)
  
  andreatnvidia
  authored and
  nabinchha
  committed
  92dc167
- feat: add Slurm image registry foundation (#891)
  nabinchha
  committed
  049ca8e
- feat: add Slurm command client and renderer (#892)
  nabinchha
  committed
  acd270f
- feat: resolve Slurm serving deployments (#890)
  nabinchha
  committed
  c00aeb2
- feat: compile deterministic Slurm run plans (#893)
  
  andreatnvidia
  authored and
  nabinchha
  committed
  f8a83c3
- feat: add Slurm service foundation (#894)
  
  andreatnvidia
  authored and
  nabinchha
  committed
  500f531
- feat: add Slurm image lifecycle jobs (#896)
  nabinchha
  committed
  d239c7d
- fix: harden Slurm image publication (#907)
  nabinchha
  committed
  baae0a9
- feat: add Slurm state writer (#908)
  nabinchha
  committed
  ac72753
- feat: add one-node Slurm runtime (#909)
  nabinchha
  committed
  59a09a9
- feat: add allocation-local Slurm client worker (#911)
  
  andreatnvidia
  authored and
  nabinchha
  committed
  9c79e27
- feat: finalize Slurm shard winners (#910)
  nabinchha
  committed
  fc5e96c
- feat(slurm): wire M2 services and CLI (#926)
  
  andreatnvidia
  authored and
  nabinchha
  committed
  aa63100
- fix(slurm): close one-node M2 integration (#929)
  
  andreatnvidia
  authored and
  nabinchha
  committed
  5e5b0dd
- feat: reconcile persisted Slurm state (#913)
  nabinchha
  committed
  edbcbab
- fix(slurm): restore plugin activation boundary (#930)
  nabinchha
  committed
  306916e
- fix: harden Slurm release artifacts (#916)
  nabinchha
  committed
  588e6da
- feat(slurm): harden distributed allocation runtime (#914)
  nabinchha
  committed
  d9a6433
- feat(slurm): add profile setup and image lifecycle wiring (#933)
  
  andreatnvidia
  authored and
  nabinchha
  committed
  fe86eec
- feat(slurm): wire retry and merge commands (#931)
  
  andreatnvidia
  authored and
  nabinchha
  committed
  a2f2c0d
- feat(slurm): add deterministic benchmark workflows (#932)
  
  andreatnvidia
  authored and
  nabinchha
  committed
  d2bfc2e
- fix(slurm): use allocation-local scratch for runtime (#938)
  
  andreatnvidia
  authored and
  nabinchha
  committed
  d1920a1
- docs(slurm): add public workflow and release checks (#939)
  
  andreatnvidia
  authored and
  nabinchha
  committed
  2859588
- fix(slurm): run collection with native pinned runtime (#965)
  nabinchha
  committed
  5523a29
- feat(slurm): harden inference proxy
  nabinchha
  committed
  2c5a970
- feat(slurm): run client natively with serving image (#970)
  nabinchha
  committed
  bc86924
- test(slurm): repair runtime test fixtures (#980)
  nabinchha
  committed
  4adb41f
- fix(slurm): show CLI submission progress (#979)
  nabinchha
  committed
  6c6cbe6
- fix(slurm): support Enroot 3.5 OCI imports (#988)
  nabinchha
  committed
  08b0dac
- feat(slurm): improve CLI output and image logs (#992)
  nabinchha
  committed
  9d51b03
- fix(slurm): stabilize job and GPU replica startup (#993)
  nabinchha
  committed
  966f304
- feat(slurm): support multiple GPU partitions (#998)
  nabinchha
  committed
  2e6c125
- fix(slurm): create profile parent directories (#999)
  nabinchha
  committed
  0e43897
- fix(slurm): expose vLLM logs and honor JSON output (#1000)
  nabinchha
  committed
  1aa46f4
- fix(slurm): address retry, startup, and proxy failure paths (#1005)
  
  andreatnvidia
  authored and
  nabinchha
  committed
  e0b6320
- refactor(slurm): remove unused config flags and set client CPU default (#1012)
  nabinchha
  committed
  9074b1d
- fix(slurm): leave host patterns empty by default (#1014)
  nabinchha
  committed
  33d52db
- feat(slurm): improve image add progress and removal (#1016)
  nabinchha
  committed
  71c48cb
- fix(slurm): simplify image details output (#1018)
  nabinchha
  committed
  d4685f3
- fix(slurm): improve preparation and allocation diagnostics (#1019)
  nabinchha
  committed
  8fda05d
- fix(slurm): choose Parquet partitions at merge (#1021)
  nabinchha
  committed
  7ef0b6d
- fix(slurm): reuse final batch files for candidates (#1023)
  nabinchha
  committed
  bbd7a58
- fix(slurm): isolate Harmony tokenizer caches (#1024)
  nabinchha
  committed
  7b28145
- feat(slurm): merge runs by ID with default output (#1025)
  nabinchha
  committed
  d59f497
- feat(slurm): accept nonempty partial shard outputs (#1026)
  nabinchha
  committed
  37c1642
- fix(slurm): merge reordered nested shard schemas (#1028)
  nabinchha
  committed
  730e483
- fix(slurm): benchmark one array task per case (#1037)
  nabinchha
  committed
  f8499d5
- feat(slurm): colocate distinct models with GPU quotas (#1034)
  
  andreatnvidia
  authored and
  nabinchha
  committed
  cf1217a
- fix: recognize locked dependency licenses
  nabinchha
  committed
  8262492
- fix(slurm): deduplicate runtime inventory paths
  
  andreatnvidia
  authored and
  nabinchha
  committed
  d5094b2
- fix(slurm): sync checked examples with documentation
  
  andreatnvidia
  authored and
  nabinchha
  committed
  f568775
- fix(slurm): harden run lifecycle and diagnostics (#1044)
  nabinchha
  committed
  f892066
- refactor(slurm): simplify preflight and run preparation (#1047)
  nabinchha
  committed
  d09b675
- fix(slurm): tag submitted jobs for telemetry (#1060)
  nabinchha
  committed
  9e355b5
- fix(slurm): speed up allocation artifact checks (#1071)
  
  andreatnvidia
  authored and
  nabinchha
  committed
  3127dfe
- fix(slurm): remove vLLM release allowlist (#1086)
  nabinchha
  committed
  b75b9c7
- feat(slurm): support hosted models through native DD provider settings (#1080)
  
  dhruvnathawani
  and
  nabinchha
  committed
  e6d3277
- fix(slurm): align rebased branch with main
  nabinchha
  committed
  672a4c1
- test(slurm): align JSON storage result fixture
  nabinchha
  committed
  bb54b50
Commits on Oct 28, 2026
- fix(slurm): avoid ephemeral port collisions (#1098)
  andreatnvidia
  authored
  28442e5
- test(slurm): stabilize CLI truncation fixture (#1101)
  andreatnvidia
  authored
  0a5b0e7
Commits on Oct 29, 2026
- feat(slurm): support cross-node tensor and expert parallelism (#1112)
  nabinchha
  authored
  0ffd8cd
Commits on Oct 30, 2026
- fix(slurm): complete proxy lifecycle and overload validation (#1104)
  nabinchha
  authored
  14df42e
Commits on Nov 9, 2026
- Merge branch 'main' into feat/slurm-execution
  andreatnvidia
  committed
  607836c
- fix(slurm): default pipeline parallelism to one (#1116)
  
  nabinchha
  authored and
  andreatnvidia
  committed
  3c3ac9d
- fix: deliver package licenses and required notices (#1133)
  
  nabinchha
  authored and
  andreatnvidia
  committed
  3e0a627
- chore: add provenance and candidates to phone sampling data (#1134)
  andreatnvidia
  committed
  4d93b7d
- chore: remove NANPA screening from phone sampling data (#1139)
  andreatnvidia
  committed
  2c49446
- fix: add license headers to Wikidata SPARQL queries (#1145)
  andreatnvidia
  authored
  1455008

Commits

Commits on Oct 27, 2026

Commits on Oct 28, 2026

Commits on Oct 29, 2026

Commits on Oct 30, 2026

Commits on Nov 9, 2026