TensorFlow.js conversion - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of TensorFlow.js model prediction after conversion

Given a TensorFlow model converted to TensorFlow.js format and loaded in the browser, what will be the output shape of the prediction for an input tensor of shape [1, 28, 28, 1]?

TensorFlow

const input = tf.zeros([1, 28, 28, 1]);
const model = await tf.loadLayersModel('model.json');
const prediction = model.predict(input);
prediction.shape;

A [10]

B [28, 28, 1]

C [1, 784]

D [1, 10]

💡 Hint

Think about the output layer size of a typical classification model for 10 classes.

The model outputs predictions for 10 classes, so the output shape for a single input is [1, 10].

❓ Model Choice

intermediate

Choosing the correct TensorFlow.js model format for conversion

Which TensorFlow saved model format is compatible for direct conversion to TensorFlow.js using the tensorflowjs_converter tool?

A TensorFlow Lite flatbuffer file

B TensorFlow SavedModel directory with saved_model.pb and variables/

C TensorFlow checkpoint files only

D Keras HDF5 (.h5) file only

💡 Hint

The converter expects a full SavedModel directory structure.

The tensorflowjs_converter tool accepts the TensorFlow SavedModel directory format for conversion.

❓ Hyperparameter

advanced

Effect of quantization during TensorFlow.js model conversion

What is the main effect of applying 8-bit quantization during TensorFlow.js model conversion?

A Converts the model to use float64 precision for better stability

B Removes dropout layers from the model

C Reduces model size and speeds up inference with slight accuracy loss

D Increases model size but improves accuracy

💡 Hint

Quantization trades precision for smaller size and faster speed.

8-bit quantization reduces model size and speeds up inference but may slightly reduce accuracy.

🔧 Debug

advanced

Identifying error when loading a TensorFlow.js model

You try to load a TensorFlow.js model with tf.loadLayersModel('model.json') but get a 404 error. What is the most likely cause?

A The model.json file path is incorrect or the file is missing on the server

B The model.json file is corrupted and cannot be parsed

C The input tensor shape does not match the model input shape

D The browser does not support WebGL

💡 Hint

A 404 error means the file was not found at the specified URL.

A 404 error indicates the model.json file path is wrong or the file is missing on the server.

🧠 Conceptual

expert

Understanding the difference between TensorFlow.js Layers and Graph models

Which statement correctly describes the difference between TensorFlow.js Layers API models and Graph models?

A Layers API models are high-level and easier to use; Graph models are low-level and support arbitrary TensorFlow graphs

B Graph models are deprecated and replaced by Layers API models

C Layers API models only support CPU execution; Graph models support GPU

D Graph models are only for training; Layers API models are only for inference

💡 Hint

Think about abstraction levels and flexibility.

Layers API provides a user-friendly way to build models; Graph models allow loading any TensorFlow graph for inference.