LiteRT Model Analyzer

Licensed under the Apache License, Version 2.0 (the "License");

Run in Google Colab View source on GitHub Download notebook

LiteRT Model Analyzer API helps you analyze models in LiteRT format by listing a model's structure.

Model Analyzer API

The following API is available for the LiteRT Model Analyzer.

tf.lite.experimental.Analyzer.analyze(model_path=None,
                                      model_content=None,
                                      gpu_compatibility=False)

You can find the API details from tf.lite.experimental.Analyzer or run help(tf.lite.experimental.Analyzer.analyze) from a Python terminal.

Basic usage with simple Keras model

The following code shows basic usage of Model Analyzer. It shows contents of the converted Keras model in TFLite model content, formatted as a flatbuffer object.

import tensorflow as tf

model = tf.keras.models.Sequential([
  tf.keras.layers.Flatten(input_shape=(128, 128)),
  tf.keras.layers.Dense(256, activation='relu'),
  tf.keras.layers.Dropout(0.2),
  tf.keras.layers.Dense(10)
])

fb_model = tf.lite.TFLiteConverter.from_keras_model(model).convert()

tf.lite.experimental.Analyzer.analyze(model_content=fb_model)

Basic usage with MobileNetV3Large Keras model

This API works with large models such as MobileNetV3Large. Since the output is large, you might want to browse it with your favorite text editor.

model = tf.keras.applications.MobileNetV3Large()
fb_model = tf.lite.TFLiteConverter.from_keras_model(model).convert()

tf.lite.experimental.Analyzer.analyze(model_content=fb_model)

Check GPU delegate compatibility

The ModelAnalyzer API provides a way to check the GPU delegate compatibility of the given model by providing gpu_compatibility=True option.

Case 1: When model is incompatible

The following code shows a way to use gpu_compatibility=True option for simple tf.function which uses tf.slice with a 2D tensor and tf.cosh which are not compatible with GPU delegate.

You will see GPU COMPATIBILITY WARNING per every node which has compatibility issue(s).

import tensorflow as tf

@tf.function(input_signature=[
    tf.TensorSpec(shape=[4, 4], dtype=tf.float32)
])
def func(x):
  return tf.cosh(x) + tf.slice(x, [1, 1], [1, 1])

converter = tf.lite.TFLiteConverter.from_concrete_functions(
    [func.get_concrete_function()], func)
converter.target_spec.supported_ops = [
    tf.lite.OpsSet.TFLITE_BUILTINS,
    tf.lite.OpsSet.SELECT_TF_OPS,
]
fb_model = converter.convert()

tf.lite.experimental.Analyzer.analyze(model_content=fb_model, gpu_compatibility=True)

Case 2: When model is compatible

In this example, the given model is compatible with GPU delegate.

model = tf.keras.models.Sequential([
  tf.keras.layers.Flatten(input_shape=(128, 128)),
  tf.keras.layers.Dense(256, activation='relu'),
  tf.keras.layers.Dropout(0.2),
  tf.keras.layers.Dense(10)
])

fb_model = tf.lite.TFLiteConverter.from_keras_model(model).convert()

tf.lite.experimental.Analyzer.analyze(model_content=fb_model, gpu_compatibility=True)