ponder-lab / Hybridize-Functions-Refactoring / #3918

90%

push

github

Recognize a Keras model reached through any of its Python bases (#1196)

The receiver check behind MAY_BUILD_LAYERS (#1172) asked the class
hierarchy whether the receiver's class is a subclass of tf.keras.Model.
The engine records a single superclass per class, so for
`class Net(Mixin, tf.keras.Model)`, in either order of bases, the check
followed the mixin and missed the model, and a model's `call` that
builds a sublayer was refused.

The check now uses the transitive base walk the Keras-callable test
already uses, which follows every declared Python base. The walk also
follows each class's superclass, for a class that records no bases, and
resolves a base in the loader of the class declaring it.


Claude-Session: https://claude.ai/code/session_01NTnpjPjJj2aZy1ZkUD1eAn

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

13 of 13 new or added lines in 2 files covered. (100.0%)

5974 of 6669 relevant lines covered (89.58%)

0.9 hits per line