Skip to content

knowledge: daily intake batch 13 - #158

Open
tzhouam wants to merge 1 commit into
mainfrom
knowledge/intake-2026-09-13
Open

tzhouam wants to merge 1 commit into
mainfrom
knowledge/intake-2026-09-13

Conversation

@tzhouam

@tzhouam tzhouam commented Sep 13, 2026

Copy link
Copy Markdown
Collaborator

Daily knowledge-intake batch from omni-reviewbot: executable rules distilled from merged PRs and copilot bugfix runs, append-only per the knowledge contract. Merging this PR is the human promotion gate — review the rules like any other knowledge edit.

Proposed rules

  • knowledge/repos/vllm-omni/components/diffusion/rules-output-lifecycle.mdDIFF-1ai (sources: PR #6615)
  • knowledge/repos/vllm-omni/components/diffusion/rules-output-lifecycle.mdDIFF-1aj (sources: PR #7126)
  • knowledge/repos/vllm-omni/components/host-weight-runtime/rules.mdHWR-1f (sources: PR #7128)
  • knowledge/repos/vllm-omni/components/diffusion/rules-lora.mdDIFF-2ag (sources: PR #7349)
  • knowledge/repos/vllm-omni/components/serving/rules-metrics.mdSERV-2g (sources: PR #6561)
  • knowledge/repos/vllm-omni/ci/rules.mdOMNI-CI-1g (sources: PR #7371)
  • knowledge/repos/vllm-omni/models/minimax-h3/rules-media.mdMMH3-2r (sources: PR #7281)
  • knowledge/repos/vllm-omni/components/serving/rules-session-lifecycle.mdSERV-6j (sources: PR #7363)

Source events

  • merged_pr PR #6615: [Core][Diffusion] Add a typed pre-D2H video media contract
  • merged_pr PR #7126: [Bugfix] Escalate diffusion worker shutdown and retain survivors
  • merged_pr PR #7128: [Bugfix] Bound HWR domain initialization lock waits
  • merged_pr PR #6938: [CI] Isolate layerwise offload memory measurements
  • merged_pr PR #7145: [Misc] Add standalone safetensors retention diagnostic
  • merged_pr PR #6560: [Model] Add Cosmos3 mixed W8A8/W8A16 and W4A4/W4A16 denoising
  • merged_pr PR #7362: [Test] Use public render_jinja_template in MiniCPM-o native template test
  • merged_pr PR #7363: [Bugfix] Fix video prewarm cache retention and cancel-restart delay
  • merged_pr PR #6460: Cosmos3 action policy improvements
  • merged_pr PR #7371: [BugFix][CI] Restore diff-aware source filtering for post-merge L3
  • merged_pr PR #7349: [Bugfix] Fail when a diffusion LoRA adapter binds no layer
  • merged_pr PR #6561: [Bugfix] Fix host-memory leak on aborted /v1/images/generations (#6462)
  • merged_pr PR #6171: [Refactor] Declare model-local KV held outside the paged manager
  • merged_pr PR #7339: [Realtime] Emit current (non-beta) OpenAI audio/transcript event names
  • merged_pr PR #6956: [Bugfix][Core] Clean up failed HWR atomic metadata writes
  • merged_pr PR #7281: [Bugfix] Keep MiniMax-H3 reference audio budgets separate
  • merged_pr PR #6930: [Bugfix] Fix Helios USP: per-component split for correct sequence parallelism
  • merged_pr PR #7005: [Perf][Diffusion] Optimize HSDP startup via Rank-0 shared weight loading and accelerated LoRA delta computation
  • merged_pr PR #5559: [Example] Migrate HunyuanImage-3.0 to model_extras + shared task examples
  • merged_pr PR #7172: [Model] Avoid scalar synchronizations in GLM-Image preparation

Public SDK curation result

{
  "batch_id": "sha256:5301f35935eee5b05c0e986b5aab54a472ed7cfc6bec19acb2078aee2203c86d",
  "success": true,
  "attempted": 8,
  "applied": 8,
  "accepted_indexes": [
    0,
    1,
    2,
    3,
    4,
    5,
    6,
    7
  ],
  "rejected_indexes": [],
  "updated_document_ids": [
    "knowledge/repos/vllm-omni/ci/rules.md",
    "knowledge/repos/vllm-omni/components/diffusion/rules-lora.md",
    "knowledge/repos/vllm-omni/components/diffusion/rules-output-lifecycle.md",
    "knowledge/repos/vllm-omni/components/host-weight-runtime/rules.md",
    "knowledge/repos/vllm-omni/components/serving/rules-metrics.md",
    "knowledge/repos/vllm-omni/components/serving/rules-session-lifecycle.md",
    "knowledge/repos/vllm-omni/models/minimax-h3/rules-media.md"
  ],
  "updated_on": "2026-09-13",
  "validators": [
    {
      "validator_id": "knowledge/tools/check_knowledge_tree.py",
      "passed": true,
      "status": "passed",
      "returncode": 0,
      "output": "提醒:文件接近拆分线:general/review/guides/review-execution-contract.md (183 个非空行,21548 bytes)\n提醒:文件接近拆分线:repos/vllm-omni/ci/rules.md (256 个非空行,28918 bytes)\n提醒:文件接近拆分线:repos/vllm-omni/components/configuration/rules.md (174 个非空行,30403 bytes)\n提醒:目录有 15 个普通页面,应该考虑分类:repos/vllm-omni/components/configuration\n提醒:文件接近拆分线:repos/vllm-omni/components/diffusion/rules-attention.md (190 个非空行,25118 bytes)\n提醒:文件接近拆分线:repos/vllm-omni/components/diffusion/rules-checkpoint-loading.md (211 个非空行,26702 bytes)\n提醒:文件接近拆分线:repos/vllm-omni/components/diffusion/rules-output-lifecycle.md (177 个非空行,23831 bytes)\n提醒:文件接近拆分线:repos/vllm-omni/components/diffusion/rules-system-runtime.md (189 个非空行,27138 bytes)\n提醒:文件接近拆分线:repos/vllm-omni/components/diffusion/rules.md (172 个非空行,27612 bytes)\n提醒:目录有 22 个普通页面,应该考虑分类:repos/vllm-omni/components/diffusion\n提醒:文件接近拆分线:repos/vllm-omni/components/model-executor/rules-bridge-batch.md (134 个非空行,18619 bytes)\n提醒:文件接近拆分线:repos/vllm-omni/components/model-executor/rules.md (193 个非空行,31439 bytes)\n提醒:文件接近拆分线:repos/vllm-omni/components/scheduler/rules.md (299 个非空行,32496 bytes)\n提醒:文件接近拆分线:repos/vllm-omni/components/serving/rules-engine-lifecycle.md (246 个非空行,32271 bytes)\n提醒:文件接近拆分线:repos/vllm-omni/components/serving/rules.md (196 个非空行,31197 bytes)\n提醒:目录有 12 个普通页面,应该考虑分类:repos/vllm-omni/components/serving\n提醒:文件接近拆分线:repos/vllm-omni/models/hunyuan-image3/rules.md (137 个非空行,25387 bytes)\n提醒:文件接近拆分线:repos/vllm-omni/models/minicpm-o-4-5/rules.md (258 个非空行,30441 bytes)\n提醒:文件接近拆分线:repos/vllm-omni/models/minimax-h3/rules-cache-task.md (204 个非空行,28481 bytes)\n提醒:文件接近拆分线:repos/vllm-omni/models/minimax-h3/rules-deployment.md (236 个非空行,27707 bytes)\n提醒:文件接近拆分线:repos/vllm-omni/models/minimax-h3/rules.md (183 个非空行,32421 bytes)\n提醒:文件接近拆分线:repos/vllm-omni/models/qwen3-tts/rules.md (190 个非空行,20981 bytes)\n知识目录检查通过:0 个错误,22 个提醒"
    },
    {
      "validator_id": "knowledge/tools/check_wiki_lint.py",
      "passed": true,
      "status": "passed",
      "returncode": 0,
      "output": "wiki lint 通过:0 个错误,0 个提醒(共 387 页)"
    }
  ],
  "rolled_back": false
}

The vllm-omni release audit runs in CI on this PR as usual.

@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Sep 13, 2026

Copy link
Copy Markdown

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review Completed 2026-09-13T07:30:05.445433Z 24c15d4 PR opened
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 24c15d456d

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment on lines +148 to +152
- 强制:视频声轨与 standalone audio 各自遵守本模态的 2–15 秒(及既有编码后单段长度)
预算,并在 encoding 前分别校验。拼接 conditioning rows 之后,不得再对
`sum(audio_lengths)` 施加单一 15 秒/600 latent-frame 总预算。各 `ref_blocks` 种类
(如 `video_audio` 与 `audio`)与长度列表必须原样保留。
- 禁止:把“各 ≤15 s”误读成“embedded+standalone 合计 ≤15 s”;在合法分账输入上因合计

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Reconcile the superseded aggregate audio budget

This new rule requires embedded and standalone audio to have separate budgets, but MMH3-2b in this same file still mandates embedded+standalone audio conditioning 总时长≤15 s at lines 72–73. A future change following that earlier hard rule will reject the newly documented valid 2–3 s video-audio plus 15 s standalone case, so update the existing conclusion rather than leaving mutually exclusive requirements.

AGENTS.md reference: AGENTS.md:L7-L11

Useful? React with 👍 / 👎.

- 禁止:仅因 `BUILDKITE_BRANCH=main` 就禁用 source 过滤;把 post-merge L3 跑成全量 E2E
却声称仍是 diff-aware;把 scheduled full upload 与 post-merge L3 混为同一策略。
- 验收:单测断言 `main` 且 `force_all=False/e2e_only=False` 返回 diff;`force_all`/`e2e`
返回 `None`;bootstrap 在非 WEEKLY 的 main merge 上传不带 `--all`。^[PR #7371]

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Record the batch PRs in frontmatter sources

All seven modified pages retain their previous sources: arrays even though this batch adds paragraph citations for PRs #7371, #7349, #6615, #7126, #7128, #6561, #7363, and #7281. doc/knowledge/SCHEMA.md lines 15–18 and 56–59 require PR-learning provenance in both page-level sources: and paragraph markers, so provenance consumers inspecting frontmatter cannot discover any of this batch's evidence; add each cited PR to its corresponding page metadata.

AGENTS.md reference: AGENTS.md:L8-L11

Useful? React with 👍 / 👎.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants