[oneDNN] Fixing eager-related nightly test failures #52204
Conversation
google-ml-butler
bot
added
size:S
|
@saxenasaurabh It sounds like related to the eager-op-as-function project, Could you take a look at this? |
| @@ -727,6 +727,16 @@ Status PopulateRetMap(FunctionDef* fdef, const AbstractOpAttrs* op_attrs, | |||
| return Status::OK(); | |||
| } | |||
|
|
|||
| #ifdef INTEL_MKL | |||
| void GetMKLNodeDef(NodeDef* ndef) { | |||
There was a problem hiding this comment.
Can we make this an inline function?
|
We found a Contributor License Agreement for you (the sender of this pull request), but were unable to find agreements for all the commit author(s) or Co-authors. If you authored these, maybe you used a different email address in the git commits than was used to sign the CLA (login here to double check)? If these were authored by someone else, then they will need to sign a CLA as well, and confirm that they're okay with these being contributed to Google. ℹ️ Googlers: Go here for more info. |
sachinmuradi
force-pushed
the
fixing-eager-related-tests
branch
from
5b693ac
to
36e3320
Compare
|
All (the pull request submitter and all commit authors) CLAs are signed, but one or more commits were authored or co-authored by someone other than the pull request submitter. We need to confirm that all authors are ok with their commits being contributed to this project. Please have them confirm that by leaving a comment that contains only Note to project maintainer: There may be cases where the author cannot leave a comment, or the comment is not properly detected as consent. In those cases, you can manually confirm consent of the commit author(s), and set the ℹ️ Googlers: Go here for more info. |
|
@googlebot I consent. |
This PR fixes : //tensorflow/python/kernel_tests:matmul_op_test and run_eager_op_as_function_test,
These tests which are run with eager op as function enabled, fail with error , "Could not find device for node:"
With a recent commit on public tensorflow : 1224326
Eager ops are wrapped, inside function WrapInCallOp, after the device has been inferred for op.
For MKL : we expect WrapInCallOp to be called before we infer device, which marks the '_kernel' attribute as NameChangeOp label for nodeDef of funcDef.
Since with new commit, WrapInCallOp is called AFTER the device inference, no device is found for the MKL Op as no kernel match is found without namechange label.
In this PR: we mark nodeDef with '_kernel' attribute as NameChange label at both places : a) before inferring the device and b) inside WrapInCallOp (already present in current code base)