release-create-tracker-issue

作者: pytorch

根據發布公告生成(並可選擇打開)PyTorch 發布追蹤器 / cherry-pick 追蹤問題,例如…

npx skills add https://github.com/pytorch/test-infra --skill release-create-tracker-issue

Create Release Tracker Issue

Generates the body for a PyTorch release tracker issue (the cherry-pick tracking issue cut alongside a release branch, like https://github.com/pytorch/pytorch/issues/180506) from a release announcement that contains the milestone (key dates) schedule, then opens the issue in pytorch/pytorch.

When to use this skill

Use when the user asks to:

  • Create a release tracker / release tracking issue for a PyTorch release
  • Create the cherry-pick tracking issue for a newly cut release branch
  • Turn a "PyTorch release X.Y key dates" announcement into a tracker issue

Instructions

Step 1: Determine parameters

Collect these parameters. Most can be derived; ask only for what is missing.

ParameterDescriptionExampleHow to obtain
Release announcementURL or pasted text of the "key dates" posthttps://dev-discuss.pytorch.org/t/pytorch-release-2-13-key-dates/3390From the user
Release versionFull version being released2.13.0From the user or announcement title
Branch versionMAJOR.MINOR of the release branch2.13Derived from release version (drop the patch)
Previous minorThe most recent prior minor release2.12Branch minor minus 1 (e.g. 2.13 → 2.12)
Release managersGitHub handles with cherry-pick dispensation@atalman, @malfetDefault @atalman, @malfet unless told otherwise

Step 2: Extract milestone dates from the announcement

If given a URL, fetch it (the dev-discuss forum is not a GitHub URL, so use WebFetch, not gh):

WebFetch(url=<announcement_url>, prompt="Extract the M3 (release branch cut), M4 (release branch finalized / feature classifications), M4.1 (tutorial drafts submission deadline), M5 (external-facing content finalized), and M6 (release day) dates verbatim, preserving any 'week of' prefix.")

If given pasted text, parse the dates directly from it.

Map the announcement milestones to the tracker fields. Keep the date format exactly as written in the announcement (e.g. DD/MM/YY, and keep a week of prefix if present):

Tracker fieldSource milestone
M3_DATEM3 — Release branch cut
M4_DATEM4 — Release branch finalized / feature classifications published
M4_1_DATEM4.1 — Tutorial drafts submission deadline
M5_DATEM5 — External-Facing Content Finalized
M6_DATEM6 — Release Day
PHASE_CUTOFFThe M4 date. If M4 is written as week of 22/6/26, use the bare date 22/6/26 for the phase cutoff.

If a milestone is missing from the announcement, ask the user rather than guessing.

Step 3: Generate the issue body

Fill the template below. Replace every {PLACEHOLDER}:

  • {VERSION} → full release version (e.g. 2.13.0)
  • {BRANCH} → branch version (e.g. 2.13)
  • {PREV_MINOR} → previous minor (e.g. 2.12)
  • {PHASE_CUTOFF}, {M3_DATE}, {M4_DATE}, {M4_1_DATE}, {M5_DATE}, {M6_DATE} → dates from Step 2
  • {RELEASE_MANAGERS} → e.g. @atalman, @malfet
We cut a [release branch](https://github.com/pytorch/pytorch/tree/release/{BRANCH}) for the {VERSION} release.

Our plan from this point is roughly:

* Phase 1 (until {PHASE_CUTOFF}): work on finalizing the release branch
* Phase 2 (after {PHASE_CUTOFF}): perform extended integration/stability/performance testing based on Release Candidate builds.

This issue is for tracking cherry-picks to the release branch.

## Release dates

* M3: Release branch cut ({M3_DATE})
* M4: Release branch finalized, Announce final launch date, Feature classifications published ({M4_DATE}) - Final RC is produced.
* M4.1: Tutorial drafts submission deadline ({M4_1_DATE})
* M5: External-Facing Content Finalized ({M5_DATE})
* M6: Release Day ({M6_DATE})

## Cherry-Pick Criteria

**Phase 1 (until {PHASE_CUTOFF}):**

Only low-risk changes may be cherry-picked from main:

1. Fixes to regressions against the most recent minor release (e.g. {PREV_MINOR}.x for this release; see [module: regression issue list](https://github.com/pytorch/pytorch/issues?q=is%3Aissue+is%3Aopen+label%3A%22module%3A+regression%22+))
2. Critical fixes for: [silent correctness](https://github.com/pytorch/pytorch/issues?q=is%3Aissue+is%3Aopen+label%3A%22topic%3A+correctness+%28silent%29%22), [backwards compatibility](https://github.com/pytorch/pytorch/issues?q=is%3Aissue+is%3Aopen+label%3A%22topic%3A+bc-breaking%22+), [crashes](https://github.com/pytorch/pytorch/issues?q=is%3Aissue+is%3Aopen+label%3A%22topic%3A+crash%22+), [deadlocks](https://github.com/pytorch/pytorch/issues?q=is%3Aissue+is%3Aopen+label%3A%22topic%3A+deadlock%22+), (large) [memory leaks](https://github.com/pytorch/pytorch/issues?q=is%3Aissue+is%3Aopen+label%3A%22topic%3A+memory+usage%22+)
3. Critical fixes to new features introduced in the most recent minor release (e.g. {PREV_MINOR}.x for this release)
4. Test/CI fixes
5. Documentation improvements
6. Compilation fixes or ifdefs required for different versions of the compilers or third-party libraries
7. Release branch specific changes (e.g. change version identifiers)

Any other change requires special dispensation from the release managers (currently {RELEASE_MANAGERS} ). If this applies to your change please write "Special Dispensation" in the "Criteria Category:" template below and explain.

**Phase 2 (after {PHASE_CUTOFF}):**

Note that changes here require us to rebuild a Release Candidate and restart extended testing (likely delaying the release). Therefore, the only accepted changes are **Release-blocking** critical fixes for: [silent correctness](https://github.com/pytorch/pytorch/issues?q=is%3Aissue+is%3Aopen+label%3A%22topic%3A+correctness+%28silent%29%22), [backwards compatibility](https://github.com/pytorch/pytorch/issues?q=is%3Aissue+is%3Aopen+label%3A%22topic%3A+bc-breaking%22+), [crashes](https://github.com/pytorch/pytorch/issues?q=is%3Aissue+is%3Aopen+label%3A%22topic%3A+crash%22+), [deadlocks](https://github.com/pytorch/pytorch/issues?q=is%3Aissue+is%3Aopen+label%3A%22topic%3A+deadlock%22+), (large) [memory leaks](https://github.com/pytorch/pytorch/issues?q=is%3Aissue+is%3Aopen+label%3A%22topic%3A+memory+usage%22+)

Changes will likely require a discussion with the larger release team over VC or Slack.

## Cherry-Pick Process

1. Ensure your PR has landed in master. This does not apply for release-branch specific changes (see Phase 1 criteria).
2. Create (but do not land) a PR against the [release branch](https://github.com/pytorch/pytorch/tree/release/{BRANCH}).
   <details>

    ```bash
    # Find the hash of the commit you want to cherry pick
    # (for example, abcdef12345)
    git log

    git fetch origin release/{BRANCH}
    git checkout release/{BRANCH}
    git cherry-pick -x abcdef12345

    # Submit a PR based against 'release/{BRANCH}' either:
    # via the GitHub UI
    git push my-fork

    # via the GitHub CLI
    gh pr create --base release/{BRANCH}
    ```

    You can also use the `@pytorchbot cherry-pick` command to cherry-pick your PR. To do this, just add a comment in your merged PR. For example:

    ```
    @pytorchbot cherry-pick --onto release/{BRANCH} -c docs
    ```
    (`-c docs` - is the category of your changes - adjust accordingly):

    For more information, see [pytorchbot cherry-pick docs](https://github.com/pytorch/pytorch/wiki/Bot-commands#cherry-pick).

    </details>

3. Make a request below with the following format:

```
Link to landed trunk PR (if applicable):
*

Link to release branch PR:
*

Criteria Category:
*
```

1. Someone from the release team will reply with approved / denied or ask for more information.
2. If approved, someone from the release team will merge your PR once the tests pass. **Do not land the release branch PR yourself.**

**NOTE: Our normal tools (ghstack / ghimport, etc.) do not work on the release branch.**

Please note HUD Link with branch CI status and link to the HUD to be provided here.
[HUD](https://hud.pytorch.org/hud/pytorch/pytorch/release%2F{BRANCH})

cc @seemethere @malfet @pytorch/pytorch-dev-infra

### Versions

{VERSION}

Step 4: Review, then create the issue

  1. Show the rendered body to the user for confirmation. Creating a GitHub issue is an outward-facing action — do not create it until the user approves the content (or has clearly asked you to open it directly).
  2. Create the issue in pytorch/pytorch with the exact title and labels:
    • Title: [v.{VERSION}] Release Tracker (note the v. prefix — e.g. [v.2.13.0] Release Tracker)
    • Labels: release tracker, triaged
gh issue create \
  --repo pytorch/pytorch \
  --title "[v.{VERSION}] Release Tracker" \
  --label "release tracker" \
  --label "triaged" \
  --body-file release_tracker_body.md

If gh is unavailable, POST to https://api.github.com/repos/pytorch/pytorch/issues with {"title": ..., "labels": ["release tracker", "triaged"], "body": ...} using an authenticated token.

  1. Report the new issue URL back to the user.

Example usage

Example — Create the 2.13.0 tracker from the key-dates announcement:

Create a release tracker issue for 2.13.0 from https://dev-discuss.pytorch.org/t/pytorch-release-2-13-key-dates/3390

For that announcement the milestones resolve to:

FieldValue
M3 (branch cut)week of 8/6/26
M4 (finalized)week of 22/6/26
M4.1 (tutorial drafts)30/6/26
M5 (content finalized)1/7/26
M6 (release day)8/7/26
Phase cutoff22/6/26
Previous minor2.12

Producing title [v.2.13.0] Release Tracker with labels release tracker, triaged.

Notes

  • Phase cutoff = the M4 date. Both the "Phase 1 (until …)" / "Phase 2 (after …)" lines and the "Phase 1/2" cherry-pick criteria headers use the bare M4 date (strip a week of prefix for the cutoff, even though the M4 line in the Release dates section keeps it).
  • Preserve the announcement's date format verbatim (DD/MM/YY). Do not reformat or convert dates.
  • The release tracker label must already exist in the repo (it does in pytorch/pytorch). If creating in a repo without it, create the label first or drop it.
  • Keep the trailing cc @seemethere @malfet @pytorch/pytorch-dev-infra line unless the user specifies different reviewers.
  • The release managers handles default to @atalman, @malfet; update only if the user names different managers.

來自 pytorch 的更多技能

zephyr
pytorch
為嵌入式開發板建置並配置 ExecuTorch 作為 Zephyr RTOS 模組。用於設定包含 ET 的 Zephyr 工作區、新增開發板支援(覆蓋層、…)
aoti-debug
pytorch
調試 AOTInductor (AOTI) 錯誤與崩潰。用於遇到 AOTI 段錯誤、設備不匹配錯誤、常量加載失敗或運行時錯誤時…
skill-writer
pytorch
為 Claude Code 建立結構化 Agent Skills 的指南,包含最佳實踐與驗證方法。涵蓋完整的 Skill 生命週期:範圍界定、檔案結構、YAML 前置資料驗證、內容組織與測試流程。強制執行嚴格的命名規則(小寫、連字號、最多 64 個字元)與描述要求(特定觸發條件、檔案類型、「什麼」與「何時」子句)。提供常見模式的範本,包括唯讀 Skills、基於腳本的 Skills,以及多檔案 Skills 搭配...
triaging-issues
pytorch
根據路由將GitHub問題分派給值班團隊、套用標籤,並關閉提問。適用於處理新的PyTorch問題,或當被要求對某個問題進行分類時…
wheel-size-analyzer
pytorch
使用 GitHub Actions artifacts API 分析 PyTorch 夜間版 wheel 在日期範圍內的大小。用於追蹤二進位檔案大小變化、識別 wheel 大小…
release-go-live-binary-build-matrix
pytorch
當 PyTorch 版本正式發佈時,更新 tools/scripts/generate_binary_build_matrix.py。將 CURRENT_STABLE_VERSION 推進至新的穩定版本,並提升…
pr-review
pytorch
審查 PyTorch 的拉取請求,針對程式碼品質、測試覆蓋率、安全性及向後相容性。適用於審查 PR 時、被要求審查程式碼變更時…
qualcomm
pytorch
建置、測試或開發 QNN(Qualcomm AI Engine Direct)後端。在處理 backends/qualcomm/、建置 QNN(使用…