Commits

Branch selector

 main

User selector

0x404

All time

Commit History

Commits on Jul 30, 2025

• [algo] feat: add GSPO-token policy loss computation function (#2775)

  Show description for b75b1f0
  

  0x404

  and

  BounharAbdelaziz

  authored
  b75b1f0

  Copy full SHA for b75b1f0

Commits on Jun 30, 2025

• [ckpt, doc] chore: add backward compatibility for model merger and sync docs (#2251)

  Show description for 024a8b8
  0x404

  authored
  024a8b8

  Copy full SHA for 024a8b8

• [algo] fix: correctly aggregate kl metrics in PPO actor (#2259)

  Show description for 6d9ac2f
  0x404

  authored
  6d9ac2f

  Copy full SHA for 6d9ac2f

Commits on Jun 23, 2025

• [fsdp] feat: support fsdp2 save hugging face model (#2138)

  Show description for 2ac410f
  0x404

  authored
  2ac410f

  Copy full SHA for 2ac410f

Commits on Jun 21, 2025

• [fsdp] refactor: set actor's strategy as default for critic and ref (#2130)

  Show description for 76f63cf
  0x404

  authored
  76f63cf

  Copy full SHA for 76f63cf

Commits on Jun 12, 2025

• [ckpt] refactor: enhance FSDP checkpoint manager flexibility (#1350)

  Show description for a1a152e
  

  0x404

  and

  ETOgaosion

  authored
  a1a152e

  Copy full SHA for a1a152e

Commits on May 23, 2025

• [megatron] feat: save hf model config in megatron checkpoint manager (#1562)

  Show description for 54a5e6e
  0x404

  authored
  54a5e6e

  Copy full SHA for 54a5e6e

Commits on May 21, 2025

• [minor] fix: use init_empty_weights instead of torch.device("meta") (#1587)

  0x404

  authored
  d475654

  Copy full SHA for d475654

• [constant scheduler] fix: model won't be updated on first training step (#1463)

  0x404

  authored
  80af51b

  Copy full SHA for 80af51b

Commits on May 18, 2025

• [merger] fix: avoid setting torch's global device to meta (#1564)

  Show description for 530154e
  0x404

  authored
  530154e

  Copy full SHA for 530154e

Commits on May 16, 2025

• [model merger] refactor model merger for better usage and maintainability (#1468)

  Show description for 3f4647f
  0x404

  authored
  3f4647f

  Copy full SHA for 3f4647f

Commits on May 15, 2025

• [lr_schedular] fix: implement proper min_lr_ratio support in cosine scheduler (#1400)

  Show description for e12edc7
  0x404

  authored
  e12edc7

  Copy full SHA for e12edc7

Commits on May 10, 2025

• [vllm rollout] minor fix: make vllm version determination stronger (#1401)

  0x404

  authored
  17f283b

  Copy full SHA for 17f283b

Commits on May 8, 2025

• [rollout] fix: missing trust_remote_code option in rollout initialization (#1423)

  Show description for 4ae9a0f
  0x404

  authored
  4ae9a0f

  Copy full SHA for 4ae9a0f

Commits on May 7, 2025

• docs: update config documentation with validation parameters (#1355)

  Show description for 8d36311
  0x404

  authored
  8d36311

  Copy full SHA for 8d36311

• [FSDPCheckpointManager] feat: save huggingface model when 'hf_model' in checkpoint_contents (#1288)

  Show description for ba6a2e0
  0x404

  authored
  ba6a2e0

  Copy full SHA for ba6a2e0

Commits on Apr 28, 2025

• [merger] fix: merged generation config is inconsistent with hf pre-trained model (#1277)

  Show description for ea4cd31
  0x404

  authored
  ea4cd31

  Copy full SHA for ea4cd31

Commits on Apr 25, 2025

• [rollout] feat: support rollout.n > 1 in hf_rollout (#1199)

  Show description for 281ed3a
  0x404

  authored
  281ed3a

  Copy full SHA for 281ed3a

Commits on Apr 17, 2025

• refactor: main generation should also use pad/unpad from verl.protocol (#1103)

  Show description for 833e7d7
  0x404

  authored
  833e7d7

  Copy full SHA for 833e7d7

Commits on Apr 11, 2025

• fix: use packaging to compre versions instead of str comparing (#1027)

  Show description for 6cbfa48
  0x404

  authored
  6cbfa48

  Copy full SHA for 6cbfa48

• [Breaking] dataset: support customized datasets for RayPPOTrainer (#924)

  Show description for 3256142
  0x404

  authored
  3256142

  Copy full SHA for 3256142

Commits on Apr 7, 2025

• [merger] fix: move megatron import into megatron related branch (#958)

  Show description for 8400beb
  0x404

  authored
  8400beb

  Copy full SHA for 8400beb

Commits on Apr 4, 2025

• [dataset] refactor: remove unused filter_prompts parameter from RLHFDataset (#889)

  Show description for 0407cad
  0x404

  authored
  0407cad

  Copy full SHA for 0407cad

• [dataset] refactor: use hf Dataset instead of pandas DataFrame in RLHFDataset for speedup (#890)

  Show description for 6974bba
  

  0x404

  and

  hiyouga

  authored
  6974bba

  Copy full SHA for 6974bba

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
  fc27caf

  Copy full SHA for fc27caf