All pull requests

Pull requests

• [Refactor] Reuse single-media pack tensors without torch.cat copies

  #2239·ShilohYu opened Oct 29, 2026Collaborator·

  ·

• [Perf] Add opt-in Muon and Ulysses layout fast paths

  #2238·Denny991 opened Oct 29, 2026Collaborator·

  ·

• [Refactor] Unify FP32 precision declarations in PrecisionPolicy

  #2237·jayhenry opened Oct 28, 2026Collaborator·

  ·

• [Fix] Make flash_mla_cudnn SparseMLA backward deterministic under XTUNER_DETERMINISTIC

  #2235·ShilohYu opened Oct 28, 2026Collaborator·

  ·

• [ ci ] add rl case about opd

  #2233·kkscilife opened Oct 28, 2026Collaborator·

  ·

• [Docs] Harden align-automodel-xtuner-text skill from field runs and cross-model reviews

  #2232·Denny991 opened Oct 28, 2026Collaborator·2 of 3 tasks·

  ·

• [Fix] Connect GLM53 multimodal inputs to training-side OPD

  #2231·xs1997zju opened Oct 28, 2026Collaborator·

  ·
  4

• [Fix] Preserve literal added tokens in GLM-5.3 multimodal SFT

  #2230·JT-Ushio opened Oct 27, 2026Collaborator·

  ·
  10

• [Fix] Align flash-linear-attention pin with LMDeploy

  #2229·matrix72c opened Oct 27, 2026Contributor·

  ·

• [Fix] Load SwapAdamW state in place to keep pinned host residency

  #2225·PengchengShi00 opened Oct 27, 2026Collaborator·

  ·

• [Feature] Add opt-in SP leader-only full-batch loading

  #2223·xs1997zju opened Oct 27, 2026Collaborator·

  ·

• [Fix] Preserve FP32 target subtraction in NPU Liger CE backward

  #2222·PussyCat0700 opened Oct 27, 2026·

  ·

• [Fix] Normalize inline system messages and tool argument order in SessionServer traces

  #2221·matrix72c opened Oct 26, 2026Contributor·

  ·

• [Docs] Fix LLaVA dataset guide links in zh-CN config READMEs

  #2212·pratikgx opened Oct 21, 2026·

  ·

• [Feature] Share GLM/Qwen lazy images, SP decode and async prefetch

  #2210·xs1997zju opened Oct 20, 2026Collaborator·

  ·

• [Fix] Bound GLM53 original-image residency with per-image preprocessing

  #2202·xs1997zju opened Oct 17, 2026Collaborator·

  ·

• [Feature] Add opt-in GLM53 lazy media and SP-group decode

  #2201·xs1997zju opened Oct 17, 2026Collaborator·

  ·

• [WIP] Diagnose SP media duplication and prototype pack sharing

  #2200·xs1997zju opened Oct 17, 2026Collaborator·

  ·

• [feat] support GLM5.3-flash RL

  #2197·PengchengShi00 opened Oct 16, 2026Collaborator·

  ·
  10

• [feat] Support value head and critic loss

  #2188·YanhuiDua opened Oct 16, 2026Collaborator·3/4·

  ·
  16

• [Feature] Support actor–critic training with GAE

  #2186·YanhuiDua opened Oct 16, 2026Collaborator·4/4·

  ·
  27

• [Refactor] Add index-based data packing and worker-side materialization

  #2185·YanhuiDua opened Oct 16, 2026Collaborator·2/4·

  ·
  7

• [Fix] Align GLM-5.3-Flash chat rendering and SFT loss masks

  #2184·gulixin0922 opened Oct 16, 2026Contributor·

  ·

• [Feature] Add opt-in training progress diagnostics

  #2183·xs1997zju opened Oct 16, 2026Collaborator·

  ·

• feat(tools): add model normalization tooling with reference-aligned output

  #2182·caixianzhang opened Oct 16, 2026Collaborator·

  ·