Commits

Breadcrumbs

History for

onmain

User selector

All users

All time

Commit history

Commits on Feb 19, 2026

• [reward] fix: empty class_dict for standalone reward model resource pool (#5348)

  Show description for eec88a0
  yyDing1

  authored
  eec88a0View commit details

  Copy full SHA for eec88a0

  View code at this pointBrowse repository at this point

Commits on Feb 14, 2026

• [rollout] fix: make skip rollout compatible with async mode (#5320)

  Show description for cf97337
  ChengQianqian

  authored
  cf97337View commit details

  Copy full SHA for cf97337

  View code at this pointBrowse repository at this point

• [misc] refactor: delete remaining batch-mode code in single controller (#5319)

  Show description for 395938b
  ji-huazhong

  authored
  395938bView commit details

  Copy full SHA for 395938b

  View code at this pointBrowse repository at this point

Commits on Feb 11, 2026

• [BREAKING][reward] refactor: the full reward configuration (#5255)

  Show description for 67dc84b
  yyDing1

  authored
  67dc84bView commit details

  Copy full SHA for 67dc84b

  View code at this pointBrowse repository at this point

Commits on Feb 9, 2026

• [BREAKING][reward] refactor: deprecate batch reward manager (#5237)

  Show description for 3df21a6
  yyDing1

  authored
  3df21a6View commit details

  Copy full SHA for 3df21a6

  View code at this pointBrowse repository at this point

Commits on Feb 6, 2026

• [BREAKING][reward] refactor: remove reward model worker code and invocation (#5194)

  Show description for 4ef45d0
  yyDing1

  authored
  4ef45d0View commit details

  Copy full SHA for 4ef45d0

  View code at this pointBrowse repository at this point

Commits on Feb 4, 2026

• [perf] fix: infer server profiler args fix (#5121)

  Show description for e4915bd
  mengchengTang

  authored
  e4915bdView commit details

  Copy full SHA for e4915bd

  View code at this pointBrowse repository at this point

• [reward] refactor: migrate all reward managers to the new asynchronous reward manager (#5189)

  Show description for 2cd9283
  yyDing1

  authored
  2cd9283View commit details

  Copy full SHA for 2cd9283

  View code at this pointBrowse repository at this point

Commits on Feb 2, 2026

• [trainer] feat: move save_ckpt before update_weights and validate (#5137)

  Show description for 7b24498
  RobotGF

  authored
  7b24498View commit details

  Copy full SHA for 7b24498

  View code at this pointBrowse repository at this point

• [reward] feat: split reward loop manager and agent loop manager (#5134)

  Show description for 84d9d8f
  yyDing1

  authored
  84d9d8fView commit details

  Copy full SHA for 84d9d8f

  View code at this pointBrowse repository at this point

Commits on Jan 30, 2026

• Revert "[rollout] feat: automatically resume generation on abort" (#5127)

  PeterSH6

  authored
  ce1b78bView commit details

  Copy full SHA for ce1b78b

  View code at this pointBrowse repository at this point

Commits on Jan 29, 2026

• [rollout] feat: automatically resume generation on abort (#5071)

  Show description for 3f0e085
  wuxibin89

  authored
  3f0e085View commit details

  Copy full SHA for 3f0e085

  View code at this pointBrowse repository at this point

Commits on Jan 28, 2026

• Revert "[reward] fix: fix reward computation in _validate when use_reward_loop=True and reward_model.enable=True" (#5091)

  Show description for f904162
  wuxibin89

  authored
  f904162View commit details

  Copy full SHA for f904162

  View code at this pointBrowse repository at this point

• [reward] fix: fix reward computation in _validate when use_reward_loop=True and reward_model.enable=True (#5054)

  Show description for 7102a9a
  none0663

  authored
  7102a9aView commit details

  Copy full SHA for 7102a9a

  View code at this pointBrowse repository at this point

• [ckpt,rollout] fix: sleep_replicas before save_ckpt to avoid OOM (#5079)

  Show description for 59a41f2
  RobotGF

  authored
  59a41f2View commit details

  Copy full SHA for 59a41f2

  View code at this pointBrowse repository at this point

Commits on Jan 27, 2026

• [doc, trainer] fix: shoudn't use rollout routing replay data for R2 (#4973)

  Show description for 7edf8f0
  HollowMan6

  authored
  7edf8f0View commit details

  Copy full SHA for 7edf8f0

  View code at this pointBrowse repository at this point

• [rollout, perf, cfg] fix: Add global step info and support more profile control params for rollout profiling (sglang backend) (#5025)

  Show description for 11f14a7
  bithighrr

  authored
  11f14a7View commit details

  Copy full SHA for 11f14a7

  View code at this pointBrowse repository at this point

• [ckpt] feat: add CheckpointEngineManager (#5031)

  Show description for ef6eaa0
  wuxibin89

  authored
  ef6eaa0View commit details

  Copy full SHA for ef6eaa0

  View code at this pointBrowse repository at this point

Commits on Jan 21, 2026

• [perf] feat: verl profiler system support Agent Loop scenario and integrate torch.profiler (#4320)

  Show description for cd1926c
  

  mengchengTang

  and

  tardis-key

  authored
  cd1926cView commit details

  Copy full SHA for cd1926c

  View code at this pointBrowse repository at this point

Commits on Jan 20, 2026

• [ray,rollout,trtllm] feat: Adding tensorrt_llm as new rollout engine (#4665)

  Show description for f65fd72
  

  joyang-nv

  and

  Superjomn

  authored
  f65fd72View commit details

  Copy full SHA for f65fd72

  View code at this pointBrowse repository at this point

Commits on Jan 18, 2026

• [trainer] fix: pass scores device type to `group_mean_std` call (#4962)

  Show description for 5689fd7
  HollowMan6

  authored
  5689fd7View commit details

  Copy full SHA for 5689fd7

  View code at this pointBrowse repository at this point

Commits on Jan 15, 2026

• [data] feat: TransferQueue - Unify the return of reward (#4902)

  Show description for e204cd8
  

  walterchenchn

  and

  Copilot

  authored
  e204cd8View commit details

  Copy full SHA for e204cd8

  View code at this pointBrowse repository at this point

Commits on Jan 14, 2026

• [algo, doc] feat: trust region sequence masking - (1) k3 KL avg and (2) veto for max criterion (#4544)

  Show description for 97eac90
  

  3 peopleauthored
  97eac90View commit details

  Copy full SHA for 97eac90

  View code at this pointBrowse repository at this point

Commits on Jan 11, 2026

• [worker] feat: New engine share actor and ref for LoRA (#4867)

  Show description for 145a28b
  HollowMan6

  authored
  145a28bView commit details

  Copy full SHA for 145a28b

  View code at this pointBrowse repository at this point

Commits on Jan 9, 2026

• [megatron] feat: Share actor and ref in LoRA (#4673)

  Show description for e69998c
  HollowMan6

  authored
  e69998cView commit details

  Copy full SHA for e69998c

  View code at this pointBrowse repository at this point

Commits on Jan 8, 2026

• [fsdp] feat: add validate process on trainer node when use_trainer_do_validate=True (#4683)

  Show description for bb64f42
  

  chenjiaoAngel

  and

  Shangwei-Li

  authored
  bb64f42View commit details

  Copy full SHA for bb64f42

  View code at this pointBrowse repository at this point

• [training_utils] fix: json encode error in filelogger (#4811)

  Show description for df36930
  zhuangqh

  authored
  df36930View commit details

  Copy full SHA for df36930

  View code at this pointBrowse repository at this point

Commits on Jan 7, 2026

• [algo] feat: add optimal token baseline and variance proxy (#4678)

  Show description for 67a127b
  

  3 peopleauthored
  67a127bView commit details

  Copy full SHA for 67a127b

  View code at this pointBrowse repository at this point

• [perf] feat: Add MFU for Qwen3-VL dense (#4753)

  Show description for 11764c6
  zhihaofang1017

  authored
  11764c6View commit details

  Copy full SHA for 11764c6

  View code at this pointBrowse repository at this point

Commits on Jan 5, 2026

• [fsdp] feat: integrate PrefixGrouper for GRPO training acceleration (#4368)

  Show description for d6bc629
  kevssim

  authored
  d6bc629View commit details

  Copy full SHA for d6bc629

  View code at this pointBrowse repository at this point

Commits on Jan 3, 2026

• [data] feat: major refactor RLHFDataset for multi-modal data (#4759)

  Show description for 51d2737
  wuxibin89

  authored
  51d2737View commit details

  Copy full SHA for 51d2737

  View code at this pointBrowse repository at this point

Commits on Jan 2, 2026

• Revert "[algo] fix: Add seq mean mask denominator option" (#4769)

  Show description for 78014a2
  wuxibin89

  authored
  78014a2View commit details

  Copy full SHA for 78014a2

  View code at this pointBrowse repository at this point

Commits on Dec 31, 2025

• [trainer] fix: use dp_size instead of world_size in _balance_batch (#4697)

  yurekami

  authored
  b16e048View commit details

  Copy full SHA for b16e048

  View code at this pointBrowse repository at this point

Commits on Dec 28, 2025

• [misc] fix: deprecate rollout.mode config option (#4690)

  yurekami

  authored
  eb6991aView commit details

  Copy full SHA for eb6991a

  View code at this pointBrowse repository at this point

Commits on Dec 26, 2025

• [algo] SAPO algo by Qwen (#4345)

  BounharAbdelaziz

  authored
  c14652eView commit details

  Copy full SHA for c14652e

  View code at this pointBrowse repository at this point