TensorFlow Lite conversion - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to create a TensorFlow Lite converter from a saved model.

TensorFlow

import tensorflow as tf
converter = tf.lite.TFLiteConverter.from_[1]('saved_model_dir')

Drag options to blanks, or click blank then click option'

Akeras_model

Bsaved_model

Cconcrete_function

Dgraph_def

Attempts:

3 left

💡 Hint

Common Mistakes

Using from_keras_model() when the model is saved as a saved model directory.

Confusing from_concrete_function() with from_saved_model().

The method from_saved_model() creates a converter from a saved TensorFlow model directory.

2fill in blank

medium

Complete the code to convert the TensorFlow model to TensorFlow Lite format.

TensorFlow

tflite_model = converter.[1]()

Drag options to blanks, or click blank then click option'

Asave

Bexport

Cconvert

Dcompile

Attempts:

3 left

💡 Hint

Common Mistakes

Using save() which does not exist on the converter object.

Using compile() which is unrelated to conversion.

The convert() method runs the conversion and returns the TensorFlow Lite model as bytes.

3fill in blank

hard

Fix the error in saving the TensorFlow Lite model to a file.

TensorFlow

with open('model.tflite', 'wb') as f:
    f.[1](tflite_model)

Drag options to blanks, or click blank then click option'

Awrite

Bwritelines

Csave

Dappend

Attempts:

3 left

💡 Hint

Common Mistakes

Using writelines() which causes a TypeError with bytes.

Using save() which is not a file method.

The write() method writes bytes to a file. writelines() expects an iterable of strings.

4fill in blank

hard

Fill both blanks to enable post-training quantization for the converter.

TensorFlow

converter.optimizations = [[1].DEFAULT]
converter.target_spec.supported_types = [[2].int8]

Drag options to blanks, or click blank then click option'

Atf.lite.Optimize

Btf.lite.Quantize

Ctf

Dtf.lite.Converter

Attempts:

3 left

💡 Hint

Common Mistakes

Using tf.lite.Quantize which does not exist.

Confusing Converter class with optimization enums.

tf.lite.Optimize.DEFAULT enables default optimizations including quantization. tf.int8 specifies INT8 data type support.

5fill in blank

hard

Fill all three blanks to convert a Keras model to TensorFlow Lite and save it.

TensorFlow

import tensorflow as tf
converter = tf.lite.TFLiteConverter.from_[1](keras_model)
tflite_model = converter.[2]()
with open('model.tflite', '[3]') as f:
    f.write(tflite_model)

Drag options to blanks, or click blank then click option'

Akeras_model

Bsaved_model

Cconvert

Dwb

Attempts:

3 left

💡 Hint

Common Mistakes

Using from_saved_model() for Keras model.

Opening file in text mode instead of binary mode.

Using save() instead of convert().

Use from_keras_model() to create converter from Keras model, convert() to convert, and 'wb' mode to write binary file.