Commits

Breadcrumbs

History for

onmain

User selector

All users

All time

Commit history

Commits on Mar 31, 2025

• [fix] Add param to resolve custom model loading failure (#845)

  BearBiscuit05

  authored
  1f78e8bView commit details

  Copy full SHA for 1f78e8b

  View code at this pointBrowse repository at this point

• [BREAKING config] fix: move val_before_train to config yaml. Using trainer.val_before_train instead of +trainer.val_before_train going forward (#820)

  tongyx361

  authored
  64bddb6View commit details

  Copy full SHA for 64bddb6

  View code at this pointBrowse repository at this point

Commits on Mar 30, 2025

• [BREAKING config] feat: add mlflow val generation log and uri config (#822)

  Show description for 7fbf609
  ccclyu

  authored
  7fbf609View commit details

  Copy full SHA for 7fbf609

  View code at this pointBrowse repository at this point

• [sglang] fix: add memory saver support to sglang rollout to avoid OOMs (#756)

  Show description for 5138a22
  

  SwordFaith

  and

  ocss884

  authored
  5138a22View commit details

  Copy full SHA for 5138a22

  View code at this pointBrowse repository at this point

• Megatron checkpoint default not save hf_models, and provide model merge tool. (#780)

  Show description for ccab836
  ETOgaosion

  authored
  ccab836View commit details

  Copy full SHA for ccab836

  View code at this pointBrowse repository at this point

Commits on Mar 29, 2025

• [misc] add deepseek v3 flops compute func (#814)

  BearBiscuit05

  authored
  0cf4ca4View commit details

  Copy full SHA for 0cf4ca4

  View code at this pointBrowse repository at this point

• [megatron] fix: remove redundant return value for hf_config (#722)

  caaatch22

  authored
  f3913d0View commit details

  Copy full SHA for f3913d0

  View code at this pointBrowse repository at this point

• docs: update checkpoint doc (#800)

  Show description for 50cba4a
  ETOgaosion

  authored
  50cba4aView commit details

  Copy full SHA for 50cba4a

  View code at this pointBrowse repository at this point

Commits on Mar 28, 2025

• [trainer] fix: skip the update step when encountering gradient overflow (#789)

  Show description for 093e959
  zhao9797

  authored
  093e959View commit details

  Copy full SHA for 093e959

  View code at this pointBrowse repository at this point

• Fix padding length for sglang rollout in veRL (#773)

  Show description for 52e80fc
  

  GeLee-Q

  and

  GeLee-Q

  authored
  52e80fcView commit details

  Copy full SHA for 52e80fc

  View code at this pointBrowse repository at this point

• Fix: Undefined variable name in the print_rank_0 statement (#788)

  Show description for 44ac51f
  jinqinn

  authored
  44ac51fView commit details

  Copy full SHA for 44ac51f

  View code at this pointBrowse repository at this point

Commits on Mar 27, 2025

• refactor: remove redundant response_mask computation (#792)

  tongyx361

  authored
  f09e365View commit details

  Copy full SHA for f09e365

  View code at this pointBrowse repository at this point

• doc: upgrade to vllm 0.8.2 (#769)

  hiyouga

  authored
  1b8b0e8View commit details

  Copy full SHA for 1b8b0e8

  View code at this pointBrowse repository at this point

• prevent division by zero in SFT (#786)

  cpchenpi

  authored
  076acdfView commit details

  Copy full SHA for 076acdf

  View code at this pointBrowse repository at this point

• fix: prevent NaN when all items are intentionally masked by adding epsilon to masked_mean (#782)

  The-Hierophant

  authored
  e9bcd2bView commit details

  Copy full SHA for e9bcd2b

  View code at this pointBrowse repository at this point

• fix: set torch dtype to auto (#749)

  Show description for a1dd922
  ccclyu

  authored
  a1dd922View commit details

  Copy full SHA for a1dd922

  View code at this pointBrowse repository at this point

Commits on Mar 26, 2025

• fix: prompt_token_ids should be list[int] instead of np.array (#772)

  Show description for fc27caf
  

  0x404

  and

  hiyouga

  authored
  fc27cafView commit details

  Copy full SHA for fc27caf

  View code at this pointBrowse repository at this point

• [Bug fix] change max_model_len to be configurable in vllm_rollout_spmd (#677)

  Show description for 7f10e57
  

  Tian Wang

  authored
  7f10e57View commit details

  Copy full SHA for 7f10e57

  View code at this pointBrowse repository at this point

• [Feat] add max_ckpt_to_keep for old ckpt removal (#724)

  Show description for afb9f9f
  guanning03

  authored
  afb9f9fView commit details

  Copy full SHA for afb9f9f

  View code at this pointBrowse repository at this point

Commits on Mar 25, 2025

• fix: prime refactor ignores extra_info (#717)

  ganler

  authored
  8594627View commit details

  Copy full SHA for 8594627

  View code at this pointBrowse repository at this point

• refactor: unify ulysses flash attention patch to avoid single model patches (#735)

  Show description for ab7b941
  wuxibin89

  authored
  ab7b941View commit details

  Copy full SHA for ab7b941

  View code at this pointBrowse repository at this point

• Add tqdm progress bar to RayPPOTrainer to visualize training progress (#615)

  Show description for 36c10bf
  

  BeSkyer

  and

  hiyouga

  authored
  36c10bfView commit details

  Copy full SHA for 36c10bf

  View code at this pointBrowse repository at this point

Commits on Mar 23, 2025

• fix: slicing returns DataProto not DataProtoItem (#718)

  ganler

  authored
  44a65f9View commit details

  Copy full SHA for 44a65f9

  View code at this pointBrowse repository at this point

• [feat] Megatron checkpoint support for current Llama and Qwen models (#687)

  Show description for 5d0a7ea
  

  4 peopleauthored
  5d0a7eaView commit details

  Copy full SHA for 5d0a7ea

  View code at this pointBrowse repository at this point

• skip special tokens (#715)

  Show description for a339f6f
  gameofdimension

  authored
  a339f6fView commit details

  Copy full SHA for a339f6f

  View code at this pointBrowse repository at this point

• Fix checkpoint loading in fsdp_checkpoint_manager.py and ray_trainer.py (#712)

  haoy-zzz

  authored
  c523a31View commit details

  Copy full SHA for c523a31

  View code at this pointBrowse repository at this point

Commits on Mar 22, 2025

• fix: support transformers==4.50.0 (#704)

  Show description for 3f6d45d
  dirtyDan0

  authored
  3f6d45dView commit details

  Copy full SHA for 3f6d45d

  View code at this pointBrowse repository at this point

Commits on Mar 21, 2025

• [Bug Fix] Fix SGLang rollout error under multi node (#652)

  ocss884

  authored
  612823aView commit details

  Copy full SHA for 612823a

  View code at this pointBrowse repository at this point

• [misc] Add Ulysses parallel config precheck (#674)

  Show description for e67dea6
  uygnef

  authored
  e67dea6View commit details

  Copy full SHA for e67dea6

  View code at this pointBrowse repository at this point

Commits on Mar 20, 2025

• [tracking] swanlab add `verl` config (#663)

  Show description for 9478885
  Zeyi-Lin

  authored
  9478885View commit details

  Copy full SHA for 9478885

  View code at this pointBrowse repository at this point

• Make Math-Verify Optional (#683)

  Show description for 529a4fe
  yyDing1

  authored
  529a4feView commit details

  Copy full SHA for 529a4fe

  View code at this pointBrowse repository at this point

Commits on Mar 18, 2025

• misc: change main_task to TaskRunner actor (#648)

  Show description for c6dc8b7
  wuxibin89

  authored
  c6dc8b7View commit details

  Copy full SHA for c6dc8b7

  View code at this pointBrowse repository at this point

• [Bug Fix] Revert the RLHFDataset truncation config (#645)

  Show description for ff13794
  ETOgaosion

  authored
  ff13794View commit details

  Copy full SHA for ff13794

  View code at this pointBrowse repository at this point

Commits on Mar 17, 2025

• [rollout] feat: add SGLang as rollout engine to verl (#490)

  Show description for 333e6d6
  

  ocss884

  and

  zhaochenyang20

  authored
  333e6d6View commit details

  Copy full SHA for 333e6d6

  View code at this pointBrowse repository at this point

Commits on Mar 15, 2025

• [misc] fix: validation batch repeat before feed into rollout (#614)

  PeterSH6

  authored
  cb943beView commit details

  Copy full SHA for cb943be

  View code at this pointBrowse repository at this point