Skip to content

Bump onnxruntime to 1.24.4 and add Python 3.14#177

Merged
cbourjau merged 5 commits into
conda-forge:mainfrom
jviehhauser:bump-1.24.4-py314
May 5, 2026
Merged

Bump onnxruntime to 1.24.4 and add Python 3.14#177
cbourjau merged 5 commits into
conda-forge:mainfrom
jviehhauser:bump-1.24.4-py314

Conversation

@jviehhauser
Copy link
Copy Markdown
Contributor

@jviehhauser jviehhauser commented Apr 1, 2026

Summary

  • Bumps onnxruntime from 1.24.2 to 1.24.4
  • Adds Python 3.14 support (drops Python 3.10 — upstream now requires >= 3.11)
  • Switches Linux runners from decommissioned cirun-openstack-cpu-4xlarge to Cirrus runners (ghcr.io/cirruslabs/ubuntu-runner-amd64:24.04-lg)
  • Removes deprecated self_hosted from conda-forge.yml (fixes linter warning)
  • Re-rendered with conda-smithy 3.59.0

Motivation

The Python 3.14 migration PR (#163) was blocked by a segfault in onnxruntime on Python 3.14 (upstream issue microsoft/onnxruntime#27392). This was fixed upstream in v1.24.3 via microsoft/onnxruntime#27413 (refcount bug in map input conversion). The reporter confirmed the fix on win-64, linux-64, and osx-arm64.

v1.24.4 is a patch release containing additional bug fixes. It also bumps python_requires to >= 3.11, so Python 3.10 is dropped from the build matrix.

All 4 existing patches were verified to still apply cleanly to v1.24.4.

CI runner changes

The Quansight Open GPU Server (OpenStack) was decommissioned on 2026-03-13 (conda-forge/.cirun#174), causing all Linux GHA jobs to time out after 24h waiting for runners that no longer exist. This PR switches Linux builds to Cirrus runners (access granted via conda-forge/admin-requests#1982). Windows builds remain on cirun-azure-windows-4xlarge.

Supersedes #163 and #176.

Test plan

  • osx-arm64 builds pass (all Python versions including 3.14)
  • Linux builds pass on Cirrus runners
  • Windows builds pass on cirun-azure
  • Python 3.10 builds are correctly skipped

@conda-forge-admin
Copy link
Copy Markdown
Contributor

conda-forge-admin commented Apr 1, 2026

Hi! This is the friendly automated conda-forge-linting service.

I wanted to let you know that I linted all conda-recipes in your PR (recipe/meta.yaml) and found some lint.

Here's what I've got...

For recipe/meta.yaml:

  • ❌ In conda-forge.yml: $.github_actions = {'self_hosted': True, 'timeout_minutes': 720, 'triggers': ['push', 'pull_request']}.

    {'self_hosted': True, 'timeout_minutes': 720, 'triggers': ['push', 'pull_request']} is not valid under any of the given schemas

    Schema
    {
      "anyOf": [
        {
          "$ref": "#/$defs/GithubActionsConfig"
        },
        {
          "type": "null"
        }
      ]
    }

For recipe/meta.yaml:

  • ℹ️ The recipe is not parsable by parser conda-souschef (grayskull). This parser is not currently used by conda-forge, but may be in the future. We are collecting information to see which recipes are compatible with grayskull.
  • ℹ️ The recipe is not parsable by parser conda-recipe-manager. The recipe can only be automatically migrated to the new v1 format if it is parseable by conda-recipe-manager.

This message was generated by GitHub Actions workflow run https://github.com/conda-forge/conda-forge-webservices/actions/runs/23869425105. Examine the logs at this URL for more detail.

@jviehhauser
Copy link
Copy Markdown
Contributor Author

@conda-forge-admin, please rerender

@conda-forge-admin
Copy link
Copy Markdown
Contributor

Hi! This is the friendly automated conda-forge-webservice.

I tried to rerender for you, but it looks like there was nothing to do.

This message was generated by GitHub Actions workflow run https://github.com/conda-forge/conda-forge-webservices/actions/runs/23870122707. Examine the logs at this URL for more detail.

@conda-forge-admin
Copy link
Copy Markdown
Contributor

Hi! This is the friendly automated conda-forge-webservice.

I tried to rerender for you, but it looks like there was nothing to do.

This message was generated by GitHub Actions workflow run https://github.com/conda-forge/conda-forge-webservices/actions/runs/23870136963. Examine the logs at this URL for more detail.

@conda-forge-admin
Copy link
Copy Markdown
Contributor

conda-forge-admin commented Apr 1, 2026

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe/meta.yaml) and found it was in an excellent condition.

I do have some suggestions for making it better though...

For recipe/meta.yaml:

  • ℹ️ The recipe is not parsable by parser conda-souschef (grayskull). This parser is not currently used by conda-forge, but may be in the future. We are collecting information to see which recipes are compatible with grayskull.
  • ℹ️ The recipe is not parsable by parser conda-recipe-manager. The recipe can only be automatically migrated to the new v1 format if it is parseable by conda-recipe-manager.

This message was generated by GitHub Actions workflow run https://github.com/conda-forge/conda-forge-webservices/actions/runs/25367859258. Examine the logs at this URL for more detail.

@jviehhauser jviehhauser marked this pull request as ready for review April 1, 2026 21:08
@hmaarrfk hmaarrfk added the automerge Merge the PR when CI passes label Apr 1, 2026
@cbourjau
Copy link
Copy Markdown
Contributor

cbourjau commented Apr 2, 2026

Do we expect the GitHub-hosted runners to be beefy enough to build the CUDA builds? Either way, the cirun ones are still referenced here. I don't think that this is ready, is it?

@conda-forge-admin
Copy link
Copy Markdown
Contributor

Hi! This is the friendly conda-forge automerge bot!

I considered the following status checks when analyzing this PR:

  • linter: passed
  • azure: passed
  • github-actions: failed

Thus the PR was not passing and not merged.

@cbourjau
Copy link
Copy Markdown
Contributor

cbourjau commented Apr 7, 2026

conda-forge/admin-requests#1982 provides an alternative for the Linux runners, but I'm not sure what the solution for Windows is, yet. Either way, this PR is not ready to be merged.

@cbourjau cbourjau removed the automerge Merge the PR when CI passes label Apr 7, 2026
@jviehhauser
Copy link
Copy Markdown
Contributor Author

jviehhauser commented Apr 7, 2026

Updated to switch Linux builds to Cirrus runners (see also conda-forge/.cirun#174). Windows builds remain on cirun-azure-windows-4xlarge for now.

@0xbe7a
Copy link
Copy Markdown
Member

0xbe7a commented Apr 9, 2026

Is this PR for now just blocked on maintainer approval for the workflow run, or are there other blockers? If it’s just that, can @janjagusch or @cbourjau approve it please?

@diegoferigo-rai
Copy link
Copy Markdown
Contributor

diegoferigo-rai commented Apr 9, 2026

@jviehhauser would it be possible to activate linux-aarch64 GPU entries in the build matrix? As I did in:

Relevant comment: #172 (comment).

@jviehhauser
Copy link
Copy Markdown
Contributor Author

@diegoferigo-rai Done — removed the skip directive for linux-aarch64 CUDA builds (as per your commit in #172).

@conda-forge-admin, please rerender

@cbourjau
Copy link
Copy Markdown
Contributor

@jviehhauser please reduce the build matrix to something minmal first (e.g. only python 3.14 and no -novec) so that we can try out the aarch64 builds.

@cbourjau
Copy link
Copy Markdown
Contributor

Is this PR for now just blocked on maintainer approval for the workflow run, or are there other blockers? If it’s just that, can @janjagusch or @cbourjau approve it please?

We don't have a way to build for Windows+CUDA, currently. We can't merge this as is. We either have to drop those builds or wait a little longer to see how this is resolved upstream (my preference).

@jviehhauser
Copy link
Copy Markdown
Contributor Author

@conda-forge-admin, please rerender

@diegoferigo-rai
Copy link
Copy Markdown
Contributor

We don't have a way to build for Windows+CUDA, currently. We can't merge this as is. We either have to drop those builds or wait a little longer to see how this is resolved upstream (my preference).

Would it make sense to drop the currently unavailable platforms for now and reintroduce them later? If resolving this upstream takes several weeks, blocking all other builds seems unnecessary. Updated CUDA linux-aarch64 binaries have already been missing for a while, it would be great to move forward in the meantime.

@jaimergp
Copy link
Copy Markdown
Member

Hey all, please check conda-forge/admin-requests#2008 for some news about larger runners. Either Namespace or Blacksmith should provide beefy machines for CPU builds across all platforms. Ping me for any questions!

Comment thread recipe/conda_build_config.yaml Outdated
@jviehhauser
Copy link
Copy Markdown
Contributor Author

@conda-forge-admin, please rerender

1 similar comment
@jviehhauser
Copy link
Copy Markdown
Contributor Author

@conda-forge-admin, please rerender

@jviehhauser jviehhauser requested a review from jaimergp April 12, 2026 21:15
Comment thread recipe/meta.yaml Outdated
@cbourjau
Copy link
Copy Markdown
Contributor

Upstream runners seem to be sorted out. Let's see if the Windows builds as expected. If so, I'll fan-out to to the full build matrix.

@cbourjau
Copy link
Copy Markdown
Contributor

@conda-forge-admin, please rerender

@conda-forge-admin
Copy link
Copy Markdown
Contributor

Hi! This is the friendly automated conda-forge-webservice.

I tried to rerender for you but ran into some issues. Please check the output logs of the GitHub Actions workflow below for more details. You can also ping conda-forge/core (using the @ notation) for further assistance or you can try rerendering locally.

The following suggestions might help debug any issues:

  • Is the recipe/{meta.yaml,recipe.yaml} file valid?
  • If there is a recipe/conda-build-config.yaml file in the feedstock make sure that it is compatible with the current global pinnings.
  • Is the fork used for this PR on an organization or user GitHub account? Automated rerendering via the webservices admin bot only works for user GitHub accounts.

This message was generated by GitHub Actions workflow run https://github.com/conda-forge/conda-forge-webservices/actions/runs/24410596181. Examine the logs at this URL for more detail.

@cbourjau
Copy link
Copy Markdown
Contributor

Current status: There still seems to be some issue with the Windows builds. I had to manually edit some generated files to get the build going, which isn't viable, but at least allows us to see if they work otherwise. Cross-referencing the conda-simthy issue: conda-forge/conda-smithy#2527

@jviehhauser
Copy link
Copy Markdown
Contributor Author

Update: Whenever conda-forge/conda-smithy#2529 lands (just merged) in a new release, we should be good to go with this.

@cbourjau
Copy link
Copy Markdown
Contributor

Update: Whenever conda-forge/conda-smithy#2529 lands (just merged) in a new release, we should be good to go with this.

Actually, it turns out that the issue in question only affects Blacksmith runners, not Namespace which is what is used since the latest commit.

@diegoferigo-rai
Copy link
Copy Markdown
Contributor

Hi, what's the current status of this PR? Sorry to raise this again, but if the issues are limited to a single platform, could we temporarily disable that platform and proceed with the others? It seems unreasonable to keep releases blocked for months over one platform's outstanding issues.

@cbourjau
Copy link
Copy Markdown
Contributor

cbourjau commented May 4, 2026

#171 ist almost done and keeps Python 3.10. If I can't fix the remaining issue there tomorrow I'll trigger the build.

@cbourjau cbourjau force-pushed the bump-1.24.4-py314 branch from b60300c to ca79009 Compare May 5, 2026 09:13
Copy link
Copy Markdown
Contributor

@cbourjau cbourjau left a comment

Choose a reason for hiding this comment

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

I squashed the commits a bit. This ran through much better than I expected 🚀

@cbourjau cbourjau merged commit 0b5773b into conda-forge:main May 5, 2026
59 checks passed
@jviehhauser
Copy link
Copy Markdown
Contributor Author

Thanks, @cbourjau

@larsoner
Copy link
Copy Markdown

larsoner commented May 5, 2026

The commit on main seems to be failing

https://github.com/conda-forge/onnxruntime-feedstock/actions/runs/25391311054/job/74466252157

Once it finishes we can ping the conda-forge folks to restart it, but it suggests some limits might need to be adjusted

@larsoner
Copy link
Copy Markdown

larsoner commented May 5, 2026

@jviehhauser can you see if you have permissions to restart the job? If not I can spam the conda help-python folks, but maybe now that it's running on GHA (rather than Azure) they have permissions set up to allow it... 🤞

@diegoferigo-rai
Copy link
Copy Markdown
Contributor

Thank you @cbourjau, appreciate all your work! 🚀

@larsoner
Copy link
Copy Markdown

larsoner commented May 6, 2026

Yes thanks @cbourjau (and thanks @jtilly for restarting the failed build twice)!

@jviehhauser jviehhauser deleted the bump-1.24.4-py314 branch May 6, 2026 16:24
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.

8 participants