How to convert python model to TensorFlow Lite

Debabrata Roy

26 October 2021

TensorFlow is a powerful tool that help us to integrate python in Mobile app. I already wrote a blog about the TensorFlow implementation in iOS. Click here to Visit blog.

In this blog, I will show how to convert a model to .tflite. We assume that you have already created a model in Python.You have the following two options for using the converter:

Python API (recommended): This makes it easier to convert models as part of the model development pipeline, apply optimisations, add metadata and has many more features.
Command line: This only supports basic model conversion.

Let start:

1. Using Python: Suppose we have model in directory fruits.

import tensorflow as tf

# Convert the model
converter = tf.lite.TFLiteConverter.from_saved_model(fruits_dir) 
tflite_model = converter.convert()

# Save the model.
with open('fruits_model.tflite', 'wb') as f:
  f.write(tflite_model)

import tensorflow as tf

# Convert the model

converter = tf.lite.TFLiteConverter.from_saved_model(fruits_dir)

tflite_model = converter.convert()

# Save the model.

with open('fruits_model.tflite', 'wb') as f:

f.write(tflite_model)

Note: you can execute this code from local or colab.

2. Using Command: To execute the cmd you have to install TensorFlow with pip in your system. Here is the step to install TenserFlow.

Now run below command.

tflite_convert \
  --saved_model_dir=/tmp/fruits_model \
  --output_file=/tmp/fruits_model.tflite

tflite_convert \

--saved_model_dir=/tmp/fruits_model \

--output_file=/tmp/fruits_model.tflite

Visit git reference to know about command.
Note: TensorFlow not recommended this because it only supports basic model conversion.

Conclusion:

Using this code and command we can easily create MobileNet mode.

I hope this Blog will help you to create .tflite file. If you feel any doubt or query please comment below.

Visit my other blogs click here.

Thank you.