ValueError: prune_low_magnitude can only prune an object of the following types: keras.models.Sequential, keras functional model, keras.layers.Layer, list of keras.layers.Layer. You passed an object of type: Sequential.

[RifatUllah102]

I am trying to prune a CNN model given below in the python script.

Describe the bug
I am trying to execute code, which gives me the error every time. Is it a version mismatch or something else?

System information

TensorFlow version (installed from source or binary): 2.17.1

TensorFlow Model Optimization version (installed from source or binary): 0.8.0

Python version: 3.10.12 [GCC 11.4.0]

Code to reproduce the issue

import tensorflow as tf
import tensorflow_model_optimization as tfmot
from tensorflow.keras import layers, models
import numpy as np

# Define the pruning parameters
pruning_schedule = tfmot.sparsity.keras.PolynomialDecay(
    initial_sparsity=0.0, 
    final_sparsity=0.5,
    begin_step=2000, 
    end_step=4000
)

# Create the model using keras.models.Sequential
base_model = models.Sequential([
    layers.Conv2D(32, (3, 3), activation='relu', input_shape=(32, 32, 3)),
    layers.MaxPooling2D((2, 2)),
    layers.Conv2D(64, (3, 3), activation='relu'),
    layers.MaxPooling2D((2, 2)),
    layers.Conv2D(64, (3, 3), activation='relu'),
    layers.Flatten(),
    layers.Dense(64, activation='relu'),
    layers.Dense(10)
])

# Build the model
base_model.build(input_shape=(None, 32, 32, 3))

# Apply pruning wrapper
model_for_pruning = tfmot.sparsity.keras.prune_low_magnitude(
    base_model,
    pruning_config={
        'pruning_schedule': pruning_schedule
    }
)

# Callbacks
callbacks = [
    tf.keras.callbacks.ModelCheckpoint(
        "pruned_model_checkpoint.h5",
        save_best_only=True,
        monitor="val_accuracy"
    ),
    tf.keras.callbacks.EarlyStopping(
        monitor="val_accuracy",
        patience=3
    ),
    tfmot.sparsity.keras.UpdatePruningStep()
]

# Compile the model
model_for_pruning.compile(
    optimizer='adam',
    loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
    metrics=['accuracy']
)

# Train the model with pruning
history = model_for_pruning.fit(
    train_images, train_labels,
    batch_size=32,
    epochs=10,
    validation_data=(test_images, test_labels),
    callbacks=callbacks
)

# Evaluate the pruned model
test_loss, test_acc = model_for_pruning.evaluate(test_images, test_labels, verbose=2)
print(f"Pruned Test Accuracy: {(test_acc * 100):.2f}%")

# Strip pruning wrappers for final model
final_model = tfmot.sparsity.keras.strip_pruning(model_for_pruning)

# Save the pruned and stripped model
final_model.save("cifar10_pruned_cnn_model.h5")
print("Pruned model saved as 'cifar10_pruned_cnn_model.h5'")

Screenshots
N/A

Additional context
Please help me to solve the issue.

[ShardulNalegave]

Hey! Did you find a solution to this? I am facing the same issue, I want to prune the MobileNetV2 model which is a Keras functional model. Even though they are supported, this fails.

If I instead prune each layer one by one and create a new Sequential model, that too doesn't seem to work as I get an error when passing Conv2D and Dense layers.