ponder-lab / Hybridize-Functions-Refactoring / #3346

88%

main: 88%

push

github

Follow a summary's trampoline to the Keras endpoint it dispatches to (#1038)

A Keras training endpoint such as `Model.fit` that the engine summarizes
resolves to a trampoline, `$tensorflow.keras.models.Model.fit.trampoline…`,
whose class is not a TensorFlow class by name. The eager-only check took
that as a non-TensorFlow target and stopped blocking the call. Each target
that is neither program code nor TensorFlow is now replaced by what it
dispatches to, so the trampoline counts as the endpoint behind it.


Claude-Session: https://claude.ai/code/session_01NTnpjPjJj2aZy1ZkUD1eAn
(cherry picked from commit 513289e17)

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

8 of 8 new or added lines in 1 file covered. (100.0%)

4798 of 5456 relevant lines covered (87.94%)

0.88 hits per line