TensorFlow TensorFlow Deployment and TFLite 2

Question 1

What format does TFLite use to store converted models?

Accepted Answer

.tflite (FlatBuffer)

Answer

TFLite models are stored in FlatBuffer format with the .tflite extension, optimized for fast loading without parsing.

Question 2

What is TensorFlow.js used for?

Accepted Answer

Running and training TensorFlow models directly in a web browser or Node.js

Answer

TensorFlow.js allows training and inference directly in the browser using WebGL acceleration or in Node.js environments.

Question 3

Which method loads a TFLite model for inference using the Python interpreter?

Accepted Answer

tf.lite.Interpreter(model_path=...)

Answer

tf.lite.Interpreter loads a .tflite model and provides methods to allocate tensors and run inference.

Question 4

What is dynamic range quantization in TFLite?

Accepted Answer

Quantizes weights to int8 at conversion time but keeps activations as float at runtime

Answer

Dynamic range quantization reduces weight size by ~4x by storing them as int8, while activations are quantized dynamically during inference.

Question 5

Which TensorFlow tool helps deploy models to Edge TPU hardware?

Accepted Answer

Edge TPU Compiler and Coral runtime

Answer

The Edge TPU Compiler converts TFLite models for execution on Google Coral devices equipped with Edge TPU chips.

TensorFlow Practice Test