Use triton from pytorch in trustyai image #1614
Conversation
JIRA: https://issues.redhat.com/browse/RHOAIENG-34086 Triton here is a dependency of torch, and is just being explicitly re-declared so that the index to pull it from can be customized. Although torch was being pulled from the PyTorch index, all of its dependencies are still defaulting to coming from PyPI (or whatever default index is used for a build). In the triton case, PyPI only has x86_64 wheels for this version, but the PyTorch index has both x86_64 and aarch64 wheels, matching the arches that are available for torch. Since it's a dependency of torch, I've also added same platform exclusion marker to omit for ppc64le.
JIRA: https://issues.redhat.com/browse/RHOAIENG-34086 This was removed recently because it was thought that it wasn't currently possible to build it for Arm. It is, it just needed a dependency fix. That fix is proposed here: red-hat-data-services/notebooks#1614
|
/build-konflux odh-workbench-jupyter-trustyai-cpu-py312 |
@atheo89 I think that may not have an Arm build until red-hat-data-services/konflux-central#761 (I'm actually surprised that there's even a PR job for the release branch, I thought those were only for main). It's a good signal that the change isn't breaking anything else though. Here's an example of a build of the PR on main (#1613) where it did have do an Arm build, and that task succeeded: https://konflux-ui.apps.stone-prod-p02.hjvn.p1.openshiftapps.com/ns/rhoai-tenant/applications/automation/pipelineruns/odh-workbench-jupyter-trustyai-cpu-py312-on-pull-request-g4rbg/ |
|
Thank you for incorporating these changes in! Build looks fine |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: atheo89 The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
atheo89
merged commit 044602e
into
red-hat-data-services:rhoai-2.25
2 participants
JIRA: https://issues.redhat.com/browse/RHOAIENG-34086
Triton here is a dependency of torch, and is just being explicitly re-declared so that the index to pull it from can be customized. Although torch was being pulled from the PyTorch index, all of its dependencies are still defaulting to coming from PyPI (or whatever default index is used for a build). In the triton case, PyPI only has x86_64 wheels for this version, but the PyTorch index has both x86_64 and aarch64 wheels, matching the arches that are available for torch.
Since it's a dependency of torch, I've also added same platform exclusion marker to omit for ppc64le.