Framework semantic default

Keras 3 migration guide SavedModel load_model model.save export TFSMLayer official

Saving or loading a TensorFlow SavedModel-style directory fails under standalone Keras.

Keras 3Machine learningFramework semantic defaultkeras3

Agent Quick Fix

Repair against the current Keras 3 contract, then keep the change narrow and source-backed.

Product: Keras 3
Current-contract area: Keras 3 no longer accepts SavedModel path in model.save/load_model workflow
Likely root cause: Saving or loading a TensorFlow SavedModel-style directory fails under standalone Keras.
Repair direction: Verifier checks behavior across multiple requests/build modes, not merely compilation. Check model outputs and save/load behavior, not only construction.

Symptom

Saving or loading a TensorFlow SavedModel-style directory fails under standalone Keras.

Why This Happens

Local code often type-checks under both old and new framework semantics; the failure depends on the installed framework version and current runtime defaults.

Common Wrong Fixes

Changing local code without checking the current external contract.
Retrying the same install, build, or API call with no version/source change.
Applying a broad unsafe bypass when a narrow compatibility fix is available.

Codex Search Keywords

These are the search terms observed in a neutral Codex validation run for this failure shape.

Keras 3 migration guide SavedModel load_model model.save export TFSMLayer official
https://keras.io/guides/migrating_to_keras_3/