TensorFlow.js conversion - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is TensorFlow.js conversion?

TensorFlow.js conversion is the process of changing a machine learning model created in TensorFlow or Keras into a format that can run in web browsers using TensorFlow.js.

beginner

Which command-line tool is used to convert TensorFlow SavedModel or Keras models to TensorFlow.js format?

The tool is called tensorflowjs_converter. It converts models into a format that TensorFlow.js can load and run in the browser.

intermediate

What are the common input formats for TensorFlow.js conversion?

Common input formats include TensorFlow SavedModel, Keras HDF5 (.h5) files, and TensorFlow Hub modules.

beginner

Why convert models to TensorFlow.js format instead of running them on a server?

Converting models to TensorFlow.js format allows them to run directly in the browser, which means faster responses, no server needed, and better privacy since data stays on the user’s device.

beginner

What is the output of the TensorFlow.js conversion process?

The output is a set of JSON and binary weight files that TensorFlow.js can load to recreate the model in the browser.

Which file format is commonly converted to TensorFlow.js format?

A TXT

B CSV

C XML

D Keras HDF5 (.h5)

Keras HDF5 (.h5) files are commonly converted to TensorFlow.js format using the tensorflowjs_converter tool.

What does the tensorflowjs_converter tool produce?

A JSON and binary weight files

B Python scripts

C CSV data files

D HTML pages

The converter produces JSON files describing the model architecture and binary files containing the model weights.

Why might you want to run a model in the browser with TensorFlow.js?

A To increase server load

B To avoid sending data to a server

C To slow down predictions

D To require internet connection

Running models in the browser keeps data on the user's device, improving privacy and reducing server load.

Which of these is NOT a valid input format for TensorFlow.js conversion?

A JPEG image

B Keras HDF5 (.h5)

C TensorFlow Hub module

D TensorFlow SavedModel

JPEG images are data, not model formats, so they cannot be converted by tensorflowjs_converter.

What command would you use to convert a Keras model to TensorFlow.js?

A tensorflow convert model.h5

B python convert_model.py

C tensorflowjs_converter --input_format keras model.h5 tfjs_model

D keras_to_tfjs model.h5

The correct command uses tensorflowjs_converter with the --input_format keras flag.

Explain the steps to convert a Keras model to TensorFlow.js format and why this is useful.

Think about the command and the advantages of browser-based models.

Describe the types of model files that can be converted to TensorFlow.js and what the output looks like.

Focus on input formats and output files.