here is my code.

early_stopping = EarlyStopping(
    min_delta=0.001, # minimium amount of change to count as an improvement
    patience=5, # how many epochs to wait before stopping
    restore_best_weights=True,
)
learning_rate_reduce = k.callbacks.ReduceLROnPlateau(
    monitor='val_acc',   # Metric to monitor for changes (usually validation accuracy)
    patience=2,          # Number of epochs with no improvement after which learning rate will be reduced
    verbose=1,           # Verbosity mode (0: silent, 1: update messages)
    factor=0.5,          # Factor by which the learning rate will be reduced (e.g., 0.5 means halving)
    min_lr=0.00001       # Lower bound for the learning rate (it won't go below this value)
)
learning_rate_schedule = keras.optimizers.schedules.ExponentialDecay(
    initial_learning_rate=0.01,  # Initial learning rate for training
    decay_steps=1,            # Number of steps before decaying the learning rate
    decay_rate=0.5,              # Rate at which the learning rate decreases
)
lr_callback = LearningRateScheduler(learning_rate_schedule)
callback=[ lr_callback , learning_rate_reduce ,early_stopping ]


base_model = keras.applications.NASNetLarge(
    weights="imagenet", # Load weights pre-trained on ImageNet.
    input_shape=(IMG_SIZE, IMG_SIZE, 3),
    include_top=False,
)  # Do not include the ImageNet classifier at the top.

# Freeze the base_model
base_model.trainable = False

# Create new model on top
model3 = keras.Sequential([
        base_model,
        layers.GlobalAveragePooling2D(),
        layers.Dropout(0.5),
        layers.Dense(2, activation='softmax')
    ])
model3.summary()
model3.compile(
    optimizer=keras.optimizers.Adam(learning_rate = learning_rate_schedule),
    # optimizer="adam",
    loss='categorical_crossentropy',
    metrics=["accuracy"],
)
history = model3.fit(
    train_generator2,
    steps_per_epoch=train_generator2.samples // batch_size,
    epochs=10,
    validation_data=valid_generator2,
    validation_steps=valid_generator2.samples // batch_size,
    callbacks=[callback]
    )

I got error

“The output of the schedule function should be a float. Got: 0.009999999776482582″
but no errors when using google colab(keras 2.15.0)”

It works in google colab(keras 2.15.0) but it doesn’t work in jupyter notebook in vscode(keras 3.0.0).