Commits

Breadcrumbs

History for

onmain

User selector

All users

All time

Commit history

Commits on Feb 23, 2026

• [rollout] fix: Include routed_experts in ToolAgentLoop return value to support R3 router replay (#5368)

  Show description for c6255ae
  mirrorboat

  authored
  c6255aeView commit details

  Copy full SHA for c6255ae

  View code at this pointBrowse repository at this point

Commits on Feb 14, 2026

• [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 13, 2026

• [worker, tool] fix: stabilize agent loop extra fields schema (#5301)

  Show description for 9682330
  

  3 peopleauthored
  9682330View commit details

  Copy full SHA for 9682330

  View code at this pointBrowse repository at this point

Commits on Feb 12, 2026

• [reward] fix: reward model args and reward_kwargs bug (#5289)

  Show description for 92c454a
  yyDing1

  authored
  92c454aView commit details

  Copy full SHA for 92c454a

  View code at this pointBrowse repository at this point

Commits on Feb 11, 2026

• [reward] fix: backward compatibility with old reward config (#5287)

  Show description for 018f8dc
  yyDing1

  authored
  018f8dcView commit details

  Copy full SHA for 018f8dc

  View code at this pointBrowse repository at this point

• [fsdp, megatron] refactor: Refactor Fully Async Implementation via Engine Workers (#5269)

  Show description for aa4ca31
  ZLiao097

  authored
  aa4ca31View commit details

  Copy full SHA for aa4ca31

  View code at this pointBrowse repository at this point

• [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

• [reward] fix: preserve input non_tensor_batch in AgentLoopManager when reward_loop_worker_handles is None (#5195)

  Show description for 28688e7
  none0663

  authored
  28688e7View commit details

  Copy full SHA for 28688e7

  View code at this pointBrowse repository at this point

Commits on Feb 6, 2026

• [fsdp] fix: Support trust_remote_code during FSDP HugginFace checkpoint save (#5200)

  Show description for f245e2d
  

  4 peopleauthored
  f245e2dView commit details

  Copy full SHA for f245e2d

  View code at this pointBrowse repository at this point

• [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

• [recipe] refactor: refactor ray trainer for separate recipe use. (fully async / one step off) (#5184)

  Show description for 32f2d3d
  ArronHZG

  authored
  32f2d3dView commit details

  Copy full SHA for 32f2d3d

  View code at this pointBrowse repository at this point

• [reward] fix: Add assert to prevent reward NaN caused by overlong_cfg.len=0 (#5216)

  Show description for 21f4e49
  ZLiao097

  authored
  21f4e49View commit details

  Copy full SHA for 21f4e49

  View code at this pointBrowse repository at this point

Commits on Feb 4, 2026

• [megatron] Add Megatron-Bridge support in fully async policy (#5196)

  Show description for 3205067
  eternally-z

  authored
  3205067View commit details

  Copy full SHA for 3205067

  View code at this pointBrowse repository at this point

• [vllm] fix: handle multimodal inputs correctly in full async mode (#5160)

  Show description for cadaf86
  Silas-11

  authored
  cadaf86View commit details

  Copy full SHA for cadaf86

  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 3, 2026

• [data] fix: keyword video_metadata (#5177)

  Show description for aa16e55
  sophiayyya

  authored
  aa16e55View commit details

  Copy full SHA for aa16e55

  View code at this pointBrowse repository at this point

• [ci] fix: fully async ci break (#5166)

  Show description for b97ebfd
  yyDing1

  authored
  b97ebfdView commit details

  Copy full SHA for b97ebfd

  View code at this pointBrowse repository at this point

Commits on Feb 2, 2026

• [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

• [training_utils] fix: Resolved bugs and conflicts in the fully async caused by multiple PRs (#5100)

  Show description for 72c109a
  ZLiao097

  authored
  72c109aView commit details

  Copy full SHA for 72c109a

  View code at this pointBrowse repository at this point

• [model, algo] feat: implement SAC algorithm and support Pi0.5 model (#5118)

  Show description for 6e17b90
  Miical

  authored
  6e17b90View commit details

  Copy full SHA for 6e17b90

  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

• [trainer] fix: resolve dataset config in agent loop (#5034)

  Show description for b63fe7f
  yyDing1

  authored
  b63fe7fView commit details

  Copy full SHA for b63fe7f

  View code at this pointBrowse repository at this point

Commits on Jan 27, 2026

• [fsdp, megatron] feat: Support fully-async training on Ascend NPU (#5043)

  Show description for 064b455
  acat-rw

  authored
  064b455View commit details

  Copy full SHA for 064b455

  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

• [trtllm] fix: Fixes for TRTLLM rollout (#5032)

  Show description for e46f700
  

  hchings

  and

  davidmlw

  authored
  e46f700View commit details

  Copy full SHA for e46f700

  View code at this pointBrowse repository at this point

Commits on Jan 23, 2026

• [BREAKING][worker, rollout, vllm] feat: implement vLLM colocated training-inference rollout with process separation (#4280)

  Show description for e9405d7
  

  jianjunzhong

  and

  wuxibin89

  authored
  e9405d7View commit details

  Copy full SHA for e9405d7

  View code at this pointBrowse repository at this point

Commits on Jan 22, 2026

• [reward] fix: conditionally include reward_extra_keys in meta_info based on rm_scores presence (#5005)

  Show description for cf16188
  none0663

  authored
  cf16188View commit details

  Copy full SHA for cf16188

  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 19, 2026

• [rollout,vllm] fix: num_preempted metrics fix and typo correction in vllm async server (#4976)

  Show description for 3b1c139
  RobotGF

  authored
  3b1c139View commit details

  Copy full SHA for 3b1c139

  View code at this pointBrowse repository at this point

• [megatron] fix: fix megatron sync_weights oom on user_trainer_do_validate mode (#4944)

  Show description for fbf1190
  chenjiaoAngel

  authored
  fbf1190View commit details

  Copy full SHA for fbf1190

  View code at this pointBrowse repository at this point

• [megatron] feat: Using MTP in RL Training and Inference (#4936)

  Show description for df340d7
  ArronHZG

  authored
  df340d7View commit details

  Copy full SHA for df340d7

  View code at this pointBrowse repository at this point

• [sglang, rollout] feat: support sglang as rollout engine in fully async policy (#4191)

  Show description for f104dfa
  

  3 peopleauthored
  f104dfaView commit details

  Copy full SHA for f104dfa

  View code at this pointBrowse repository at this point