Keras2 - T5 Backbone Preset Tests fail

[sampathweb]

T5 backbone presets test are failing for Keras 2, but pass for Keras 3 (all backends).

Log: https://btx-internal.corp.google.com/invocations/5135a970-cae4-4635-a2be-c9e3e8a73215/targets/keras_nlp%2Fgithub%2Fubuntu%2Fgpu%2Fkeras2%2Fpresubmit/log

This test is now disabled for Keras 2 so the builds are green - #1362

_____________________ T5BackboneTest.test_smallest_preset ______________________

self = 

    @pytest.mark.large
    def test_smallest_preset(self):
>       self.run_preset_test(
            cls=T5Backbone,
            preset="t5_small_multi",
            input_data=self.input_data,
            expected_output_shape={
                "encoder_sequence_output": (2, 3, 512),
                "decoder_sequence_output": (2, 3, 512),
            },
            expected_partial_output={
                "encoder_sequence_output": ops.array(
                    [-0.0034, 0.0293, -0.0827, -0.1076]
                ),
                "decoder_sequence_output": ops.array(
                    [0.0097, 0.3576, -0.1508, 0.0150]
                ),
            },
        )

keras_nlp/models/t5/t5_backbone_test.py:59: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
keras_nlp/tests/test_case.py:400: in run_preset_test
    instance = cls.from_preset(preset, **init_kwargs)
keras_nlp/models/backbone.py:135: in from_preset
    return super(cls, calling_cls).from_preset(*args, **kwargs)
keras_nlp/models/backbone.py:123: in from_preset
    model.load_weights(weights)
/tmpfs/venv/lib/python3.9/site-packages/keras/src/utils/traceback_utils.py:61: in error_handler
    return fn(*args, **kwargs)
/tmpfs/venv/lib/python3.9/site-packages/keras/src/engine/training.py:3235: in load_weights
    return saving_api.load_weights(
/tmpfs/venv/lib/python3.9/site-packages/keras/src/saving/saving_api.py:301: in load_weights
    saving_lib.load_weights_only(
/tmpfs/venv/lib/python3.9/site-packages/keras/src/saving/saving_lib.py:333: in load_weights_only
    _load_state(
/tmpfs/venv/lib/python3.9/site-packages/keras/src/saving/saving_lib.py:457: in _load_state
    _load_state(
/tmpfs/venv/lib/python3.9/site-packages/keras/src/saving/saving_lib.py:435: in _load_state
    trackable.load_own_variables(weights_store.get(inner_path))
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

self = 
store = {}

    def load_own_variables(self, store):
        if not self.built:
            self.build()
>       self.embeddings.assign(store["0"])
E       KeyError: '0'

keras_nlp/layers/modeling/reversible_embedding.py:144: KeyError