All pull requests

Pull requests

• [worker] fix seq loss averaging across micro-batches and ranks

  #11016·zsxm1998 opened Oct 4, 2026·

  ·

• Take the last standalone digit as the android_gui answer

  #11015·shaurya416 opened Sep 24, 2026·

  ·
  1

• fix: prevent dynamic_batching micro-batch deadlock under uneven visual tokens

  #11012·houzl3416 opened Sep 5, 2026·3 of 4 tasks·

  ·
  1

  1

• Bring EasyR1 onto newer Transformers (>5) + vLLM (≥0.20) and add Qwen…

  #11001·erow opened Jul 14, 2026·

  ·
  10

• docs: add CONTRIBUTING.md guide

  #10999·dashitongzhi opened May 19, 2026·

  ·
  6

• code to patch for qwen3.5 grpo training

  #10957·leah1985 opened Apr 6, 2026·

  ·
  8

• Fix #648: ValueError: max_num_batched_tokens should be greater than p...

  #3425·JiwaniZakir opened Apr 4, 2026·

  ·
  3

• Use weights_only=True in DataProto deserialization

  #656·RinZ27 opened Mar 2, 2026·

  ·
  7

• [FEAT] Support GLM4.1V series

  #534·ZiyiTsang opened Oct 15, 2025·

  ·
  21