Skip to content

[ci] Add weekly ete test workflow for docker build and ETE runs - #2038

Merged
zhulinJulia24 merged 4 commits into
mainfrom
optimize_ete_case
Aug 21, 2026
Merged

[ci] Add weekly ete test workflow for docker build and ETE runs#2038
zhulinJulia24 merged 4 commits into
mainfrom
optimize_ete_case

Conversation

@zhulinJulia24

Copy link
Copy Markdown
Collaborator

No description provided.

JuliaLin and others added 3 commits August 21, 2026 15:16
Consolidate scheduled GPU ETE into a manual weekly workflow that builds a tagged image and runs tests against CI_ETE_TRAIN_IMAGE, and install lmdeploy from InternLM bump-version.

Co-authored-by: Cursor <cursoragent@cursor.com>
Prefer PyPI lmdeploy==0.16.0 in image builds and retain LMDEPLOY_URL for source installs when the version is cleared.

Co-authored-by: Cursor <cursoragent@cursor.com>
Run weekly docker build and ETE at 17:00 UTC+8 (09:00 UTC), and restore NPU ETE schedule.

Co-authored-by: Cursor <cursoragent@cursor.com>
@zhulinJulia24
zhulinJulia24 marked this pull request as ready for review August 21, 2026 08:01
@zhulinJulia24
zhulinJulia24 merged commit b934f46 into main Aug 21, 2026
5 checks passed
@zhulinJulia24
zhulinJulia24 deleted the optimize_ete_case branch August 21, 2026 08:02

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Pull request overview

Adds a new GitHub Actions workflow intended to run end-to-end (ETE) tests using a freshly built Docker image, while updating the Docker image build inputs and enabling an ETE train-image override in the autotest harness.

Changes:

  • Introduces a new weekly_ete_test workflow to build/push a Docker image and run cluster ETE cases with configurable case selection and parallelism.
  • Pins lmdeploy installation via LMDEPLOY_VERSION in the image build script and passes it through as a Docker build arg/label.
  • Allows ETE runs to override the training image via CI_ETE_TRAIN_IMAGE, and removes cron schedules from existing GPU/NPU ETE workflows.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 2 comments.

Show a summary per file
File Description
image_build.sh Pins lmdeploy version and forwards it to Docker build as an arg/label.
autotest/utils/common_utils.py Adds support for overriding the training step image via CI_ETE_TRAIN_IMAGE.
.github/workflows/weekly_ete_test.yaml New workflow to build/push an image and run ETE tests, with report artifact/upload and report publishing.
.github/workflows/ete_test_npu.yaml Removes the scheduled trigger (cron).
.github/workflows/ete_test_gpu.yaml Removes the scheduled trigger (cron).
Suppressed comments (3)

.github/workflows/weekly_ete_test.yaml:9

  • This workflow is titled "weekly" and the existing GPU/NPU ETE workflows removed their schedule triggers, but this new workflow has no on.schedule entry. As a result, there may be no scheduled ETE coverage at all unless something external dispatches it.
on:
  workflow_dispatch:
    inputs:

.github/workflows/weekly_ete_test.yaml:14

  • The repo_org input is described as an "organization name", but it is used as the repository: value for actions/checkout and defaults to InternLM/xtuner (owner/repo). This can confuse users about what to enter.
      repo_org:
        required: false
        description: 'Tested repository organization name. Default is InternLM/xtuner'
        type: string
        default: 'InternLM/xtuner'

.github/workflows/weekly_ete_test.yaml:96

  • The build job checkout ignores the repo_org input (it only sets ref). If someone dispatches this workflow with a different repo_org, the image will still be built from the workflow repository while the ETE job clones repo_org, creating a code/image mismatch.
      - name: Checkout repository
        uses: actions/checkout@v6
        with:

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

Comment on lines +4 to +6
permissions:
contents: read

Comment thread image_build.sh
Comment on lines +15 to +16
export LMDEPLOY_VERSION="${LMDEPLOY_VERSION:-0.16.0}"
export LMDEPLOY_URL="${LMDEPLOY_URL:-https://github.com/InternLM/lmdeploy@bump-version}"
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