-
Notifications
You must be signed in to change notification settings - Fork 4.4k
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.
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
[14_0_X] Skip evaluation of TF model if one of the input tensors is empty #45145
Conversation
A new Pull Request was created by @valsdav for CMSSW_14_0_X. It involves the following packages:
@cmsbuild, @wpmccormack, @valsdav can you please review it and eventually sign? Thanks. cms-bot commands are listed here
|
cms-bot internal usage |
please test |
urgent
|
+1 Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-574d79/39722/summary.html Comparison SummarySummary:
|
@cms-sw/ml-l2 , any reason not to sign this ? |
+1 |
This pull request is fully signed and it will be integrated in one of the next CMSSW_14_0_X IBs (tests are also fine) and once validation in the development release cycle CMSSW_14_1_X is complete. This pull request will now be reviewed by the release team before it's merged. @sextonkennedy, @rappoccio, @antoniovilela (and backports should be raised in the release meeting by the corresponding L2) |
+1 |
Backport of #45139
PR description:
Solves: #44481
Evaluating TensorFlow models with empty tensors is problematic. We observed different behavior on different architectures. It crashes on machines with AVX512 registers available, whereas it works fine and returns empty tensors.
In some cases incompatible shapes for operators are reported by the TF engine.
Related issues: #45136, #44333
This PR introduces a simple check for empty inputs in the central TensorFlow interface to avoid evaluation of the model for empty tensors.
PR validation:
A new test has been added calling the TF inference with empty tensors.