Keras 3 compatibility, Model patch, shape helpers, graph fixes

[KMChris]

Summary

Modernizes visualkeras for Keras 3/TF >=2.16, removes deprecated APIs, adds robust shape helpers, applies a local patch for nested-model tuple issues, and refines graph_view layout. All tests pass.

Key changes

• Introduced .gitignore to exclude common Python artifacts and build files.
• Updated graph_view and layered_view functions to improve compatibility with Keras 2 and 3.
• Added helper functions to retrieve layer input and output shapes robustly.
• Applied a patch to ensure single-item outputs are unwrapped correctly in Keras models.

Testing

• Full suite passes (46 passed, ~20 warnings)
• Minor, intentional visual diffs in graph_view dimensions

[Soontosh]

Thanks so much for the PR! I'll run the tests on my machine first to see if they pass and then merge.

[Soontosh]

Thanks again for the PR, but I observed one issue.

from .layer_utils import *
ensure_singleton_sequence_unwrap_patched()

Importing visualkeras now runs ensure_singleton_sequence_unwrap_patches which changes how all keras.Model objects behave (see above code snippet from graph.py). That helper rewrites Model.__call__:

# visualkeras/layer_utils.py
def ensure_singleton_sequence_unwrap_patched():
    import keras
    _orig_call = keras.models.Model.__call__
    def _vk_model_call(self, *args, **kwargs):
        out = _orig_call(self, *args, **kwargs)
        if isinstance(out, (list, tuple)) and len(out) == 1:
            return out[0]
        return out
    keras.models.Model.__call__ = _vk_model_call

After this reassignment, model.__call__ no longer carries certain attributes, so operations like tf.saved_model.save(model) or any code requesting model.__call__.get_concrete_function(...) will fail with AttributeError

Since this happens at import time, anybody who just does import visualkeras would suddenly lose the usual list/tuple output from their own models.

This should be fixed. Do you have any ideas? Maybe something like functools.wraps?