Siddhartha Menon

@penpornk @keerthanakadiri Please let me know if you'd like to see any further changes or if you're happy to merge

@penpornk Thank you for the approval. I see the import/copybara job is failing. Is this something I need to address on my end?

@keerthanakadiri @penpornk During some end-to-end testing of tensorflow against the latest versions it seems there is some unexpected behaviour on some BF16 models. I'd advise **against** merging this just now....

@penpornk The aformentioned issue has been [fixed in oneDNN main](https://github.com/uxlfoundation/oneDNN/commit/aeaa73fb4fd7361d30e85aaac939624bbf43cff5) and [included here as a patchfile](https://github.com/tensorflow/tensorflow/pull/84975/files#diff-f8a2591ced2fdc6ba068b516fb851c2e670fef2625ecae78e2096f58371d38a2). Please let me know if everything seems to be in order.

I think since the time this PR was first opened, the `third_party/` paths have shifted around a bit. To match what seems to be the current state of `master`, I...

>Thank you for the updates. Let's try. Looks like it's passed the internal test suite? Would you like any further changes or are you happy with the patch as-is?

@mihaimaruseac @penpornk Thanks for the approval. >It seems there's some XLA failure right now: https://github.com/openxla/xla/actions/runs/14225255197/job/39863144414 Looking at the failing log it seems like it is failing because it is trying...

@MichaelHudgins Thanks for having a look. Is there any further required from my end? Or can this be resolved and merged by the TensorFlow/XLA dev teams?

Requesting a re-approval (@penpornk) after manual merge-conflict resolution

@penpornk @mihaimaruseac Could you advise on how we can proceed with the PR? Unfortunately the CI failure caused a delay during which a merge-conflict arose. I have fixed it but...