Fix: validate trainable field type in Layer.__init__

[Jah-yee]

Good day

Issue

deserialize_keras_object accepts non-boolean values for the trainable field without validation (#22699).

Reproduction

from keras.saving import deserialize_keras_object
config = {
    "class_name": "Dense",
    "module": "keras.layers",
    "config": {
        "units": 32,
        "trainable": "yes"  # string instead of bool
    }
}
layer = deserialize_keras_object(config)
print(type(layer.trainable), layer.trainable)  # str, "yes"

Fix

Added an explicit type check in Layer.__init__ that raises a clear TypeError when trainable is not a bool:

if not isinstance(trainable, bool):
    raise TypeError(
        f"Expected `trainable` to be a bool, but got "
        f"{type(trainable).__name__}: {trainable!r}"
    )

This ensures validation happens at layer construction time (during deserialization), catching the bad type early with a descriptive error.

Testing

Verified that valid bool values (True/False) work correctly, while invalid values like "yes", 1, 0, etc. now raise TypeError as expected.

Thank you for your attention. If there are any issues or suggestions, please leave a comment and I will address them promptly.

Warmly,
RoomWithOutRoof

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[gemini-code-assist]

Code Review

This pull request introduces input validation for the trainable parameter in layers and the config dictionary during serialization to ensure they are of the correct types. The review feedback correctly points out that ValueError should be used instead of TypeError for these user input validations, adhering to the Keras Repository Style Guide.

[gemini-code-assist]

The Keras Repository Style Guide (line 178) specifies that ValueError should be used for user input validation instead of TypeError. Additionally, consider that isinstance(trainable, bool) is quite strict and will reject 0 or 1, which are often used as boolean substitutes in Python. However, if strict boolean enforcement is intended to prevent common configuration errors (like the string "False" being truthy), please at least update the exception type to ValueError.

Suggested change

	if not isinstance(trainable, bool):
	raise TypeError(
	f"Expected `trainable` to be a bool, but got "
	f"{type(trainable).__name__}: {trainable!r}"
	)
	if not isinstance(trainable, bool):
	raise ValueError(
	f"Expected `trainable` to be a bool, but got "
	f"{type(trainable).__name__}: {trainable!r}"
	)

References

1. In general, to validate user input, always use ValueError and avoid assert. ^(link)

[gemini-code-assist]

Please use ValueError instead of TypeError to adhere to the Keras Repository Style Guide (line 178).

Suggested change

	if (
	inner_config is not None
	and not isinstance(inner_config, dict)
	and class_name != "function"
	):
	raise TypeError(
	f"Expected `config` to be a dict, but got {type(inner_config).__name__}: "
	f"{inner_config}"
	)
	if (
	inner_config is not None
	and not isinstance(inner_config, dict)
	and class_name != "function"
	):
	raise ValueError(
	f"Expected `config` to be a dict, but got {type(inner_config).__name__}: "
	f"{inner_config}"
	)

References

1. In general, to validate user input, always use ValueError and avoid assert. ^(link)

[codecov-commenter]

Codecov Report

❌ Patch coverage is 33.33333% with 4 lines in your changes missing coverage. Please review.
✅ Project coverage is 82.95%. Comparing base (e94cb07) to head (25a0477).

Files with missing lines	Patch %	Lines
keras/src/layers/layer.py	0.00%	1 Missing and 1 partial ⚠️
keras/src/saving/serialization_lib.py	50.00%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #22722      +/-   ##
==========================================
- Coverage   82.95%   82.95%   -0.01%     
==========================================
  Files         596      596              
  Lines       69252    69257       +5     
  Branches    10814    10816       +2     
==========================================
+ Hits        57451    57452       +1     
- Misses       8973     8975       +2     
- Partials     2828     2830       +2     

Flag	Coverage Δ
keras	82.77% <33.33%> (-0.01%)	⬇️
keras-jax	58.71% <33.33%> (-0.01%)	⬇️
keras-numpy	54.56% <33.33%> (-0.01%)	⬇️
keras-openvino	59.42% <33.33%> (-0.01%)	⬇️
keras-tensorflow	60.28% <33.33%> (-0.01%)	⬇️
keras-torch	59.05% <33.33%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[keerthanakadiri]

Hi @Jah-yee, Can you please sign the CLA? Thank you!

[hertschuh]

This issue was addressed by #22740