All pull requests

Pull requests

• [Fix] Correct GLM-5.3-Flash assistant boundary labels

  #2198·Sun-Jiatuhao opened Oct 16, 2026·

  ·

• [WIP] support GLM5.3-flash RL

  #2197·PengchengShi00 opened Oct 16, 2026Collaborator·

  ·

• [Fix] Discard dirty GLM-5.3-Flash multimodal samples instead of halting training

  #2196·ShilohYu opened Oct 16, 2026Collaborator·

  ·

• [Fix] Tolerate plain-string content parts in multimodal sample collection

  #2195·ShilohYu opened Oct 16, 2026Collaborator·

  ·

• [Fix] Render GLM-5.3-Flash assistant media as placeholder markers

  #2193·ShilohYu opened Oct 16, 2026Collaborator·

  ·

• [Fix] Render structured GLM-5.3-Flash tool outputs via visible_text

  #2192·ShilohYu opened Oct 16, 2026Collaborator·

  ·

• [Fix] Render GLM-5.3-Flash tool result blocks in tool_calls order

  #2191·ShilohYu opened Oct 16, 2026Collaborator·

  ·

• [Fix] Append GLM-5.3-Flash training EOS only after the final assistant

  #2190·ShilohYu opened Oct 16, 2026Collaborator·

  ·

• [WIP] Support value head and critic loss

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

  ·
  2

• [WIP] Support PPO with shared actor and critic packing

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

  ·
  15

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

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

  ·
  1

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

  #2184·gulixin0922 opened Oct 16, 2026·

  ·

• [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·

  ·

• [Fix] Preserve tools in SessionServer raw-token requests

  #2176·matrix72c opened Oct 15, 2026Contributor·

  ·

• [Fix] Restore SwapAdamW state and parameter aliases after checkpoint loading

  #2175·matrix72c opened Oct 15, 2026Contributor·

  ·

• [Fix] Load rollout KV cache before resuming checkpoint production

  #2174·matrix72c opened Oct 15, 2026Contributor·

  ·

• [Fix] Distinguish scored task deadlines from sandbox infrastructure failures

  #2173·matrix72c opened Oct 15, 2026Contributor·

  ·

• [Test] Add OPD e2e autotest cases and launch_script support for RL ETE

  #2171·YanhuiDua opened Oct 14, 2026Collaborator·

  ·

• [WIP] Restore env vars after colocate/disagg weight-update tests

  #2165·PengchengShi00 opened Oct 13, 2026Collaborator·

  ·

• [Fix] Remesh FP32 ignored params onto world-wide replicas

  #2162·YanhuiDua opened Oct 12, 2026Collaborator·

  ·
  9

• [Fix] Unify Muon matrix layout for NS and lr_ratio

  #2161·YanhuiDua opened Oct 12, 2026Collaborator·

  ·
  8

• test: add trainer-free AutoModel to existing HF/XT parity test

  #2152·xs1997zju opened Oct 10, 2026Collaborator·

  ·

• [Enhance] extend sandbox creation args

  #2142·braisedpork1964 opened Oct 9, 2026Collaborator·

  ·

• [Refactor] Separate sandbox environment ownership from Runner execution

  #2135·matrix72c opened Oct 8, 2026Contributor·

  ·