Skip to content

Test change for test_rocm_wheels enablement in rockrel #7383

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
jonatluu wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Test change for test_rocm_wheels enablement in rockrel #7383

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
30 changes: 29 additions & 1 deletion .github/workflows/multi_arch_release_linux.yml
View file
Open in desktop

ghost Aug 14, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

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

Always keep Linux and Windows workflows in sync: https://github.com/ROCm/TheRock/blob/main/.github/workflows/multi_arch_release_windows.yml

Original file line number Diff line number Diff line change
Expand Up @@ -250,6 +250,34 @@ jobs:
"ref": "${{ inputs.ref }}"
}

# Wheel tests are dispatched asynchronously (like native install tests) so
# publish is not blocked on GPU runner queues. See https://github.com/ROCm/TheRock/issues/5212
trigger_test_wheels_per_family:
needs: [build_python_packages]
name: Trigger Test Wheels (${{ matrix.amdgpu_family }}, py${{ matrix.python_version }})
if: ${{ toJSON(fromJSON(inputs.build_config).test_python_packages_matrix) != '[]' }}
runs-on: ubuntu-24.04
permissions:
actions: write
strategy:
fail-fast: false
matrix:
include: ${{ fromJSON(inputs.build_config).test_python_packages_matrix }}
steps:
- name: Trigger test_rocm_wheels
uses: benc-uk/workflow-dispatch@31e2b3319479a63f0ab15bf800eff9e913504e26 # v1.3.2
with:
workflow: test_rocm_wheels.yml
inputs: |
{ "amdgpu_family": "${{ matrix.amdgpu_family }}",
"test_runs_on": "${{ matrix.test_runs_on }}",
"package_find_links_url": "${{ needs.build_python_packages.outputs.kpack_split == 'true' && needs.build_python_packages.outputs.package_find_links_url || format('{0}/{1}/index.html', needs.build_python_packages.outputs.package_find_links_url, matrix.amdgpu_family) }}",
"python_version": "${{ matrix.python_version }}",
"rocm_version": "${{ inputs.rocm_package_version }}",
"container_image_name": "${{ matrix.container_image_name }}",
"container_image_url": "${{ matrix.container_image_url }}",
"kpack_split": "${{ needs.build_python_packages.outputs.kpack_split }}"
}
Comment on lines +270 to +280

ghost Aug 14, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

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

Ah this isn't great - the matrix is going to dispatch lots of jobs here.

I would rather have releases dispatch closer to what test_artifacts does: one workflow run per amdgpu_family, not per [amdgpu_family, container, python_version].

Need to look at the workflow architecture a bit to see how we might simplify there.

ghost Aug 17, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

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

I haven't had much time to take a more detailed look here yet.

One idea would be to add a new workflow that takes fromJSON(inputs.build_config).test_python_packages_matrix as an input and have that workflow expand the matrix itself.

CI would run that new workflow here:

TheRock/.github/workflows/multi_arch_ci_linux.yml

Lines 285 to 313 in 4444ac2

test_python_packages_per_family:
needs: [build_python_packages]
name: Test Python
if: ${{ !failure() && !cancelled() && toJSON(fromJSON(inputs.build_config).test_python_packages_matrix) != '[]' }}
strategy:
fail-fast: false
matrix:
include: ${{ fromJSON(inputs.build_config).test_python_packages_matrix }}
uses: ./.github/workflows/test_rocm_wheels.yml
secrets: inherit
with:
amdgpu_family: ${{ matrix.amdgpu_family }}
test_runs_on: ${{ matrix.test_runs_on }}
# TODO: Simplify to just `needs.build_python_packages.outputs.package_find_links_url`
# once kpack split is always enabled.
package_find_links_url: >-
${{
needs.build_python_packages.outputs.kpack_split == 'true'
&& needs.build_python_packages.outputs.package_find_links_url
|| format('{0}/{1}/index.html',
needs.build_python_packages.outputs.package_find_links_url,
matrix.amdgpu_family)
}}
python_version: ${{ matrix.python_version }}
rocm_version: ${{ inputs.rocm_package_version }}
container_image_name: ${{ matrix.container_image_name }}
container_image_url: ${{ matrix.container_image_url }}
repository: ${{ inputs.repository }}
ref: ${{ inputs.ref }}

We've gone through a few iterations on the workflow matrix shape, see also


# TODO: Consider switch pulling ROCm packages from the CDN instead of the artifact
# bucket if the job order is not altered.
Expand Down Expand Up @@ -300,7 +328,7 @@ jobs:

notify_quartz_completed:
if: ${{ always() && vars.NOTIFY_QUARTZ_ENABLED == 'true' }}
needs: [build_artifacts, build_tarballs, build_python_packages, build_native_deb_packages, build_native_rpm_packages, publish_to_release_buckets, trigger_test_artifacts_per_family, trigger_test_native_install, trigger_release_pytorch_wheels, trigger_release_jax_wheels]
needs: [build_artifacts, build_tarballs, build_python_packages, build_native_deb_packages, build_native_rpm_packages, publish_to_release_buckets, trigger_test_artifacts_per_family, trigger_test_native_install, trigger_test_wheels_per_family, trigger_release_pytorch_wheels, trigger_release_jax_wheels]
name: "Quartz - completed - release Linux"
runs-on: ubuntu-24.04
timeout-minutes: 10
Expand Down