Loading and inference in TensorFlow - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of model prediction after loading

Given the following TensorFlow code that loads a saved model and runs inference on a sample input, what will be the printed output?

TensorFlow

import tensorflow as tf
import numpy as np

model = tf.keras.models.load_model('my_model')
sample_input = np.array([[1.0, 2.0, 3.0]])
prediction = model.predict(sample_input)
print(prediction)

A [[0.3 0.4 0.3]]

B Raises a FileNotFoundError

C [[0.1 0.7 0.2]]

D A numpy array with shape (1, 3) containing probabilities

💡 Hint

The output shape and type depend on the model's last layer and input shape.

The model.predict returns a numpy array with shape matching the output layer. Since the model is loaded successfully, it returns a numpy array of probabilities or values, not a fixed hardcoded array.

❓ Model Choice

intermediate

Choosing the correct method to load a TensorFlow model

Which TensorFlow method correctly loads a saved Keras model for inference?

A tf.keras.load('path_to_model')

B tf.saved_model.load('path_to_model')

C tf.keras.models.load_model('path_to_model')

D tf.load_model('path_to_model')

💡 Hint

Keras models have a specific loading function.

The correct method to load a saved Keras model is tf.keras.models.load_model. The other options are either invalid or used for different purposes.

❓ Hyperparameter

advanced

Effect of batch size on inference speed

When performing inference with a loaded TensorFlow model, how does increasing the batch size generally affect inference speed and memory usage?

A Inference speed per sample increases and memory usage increases

B Inference speed per sample decreases and memory usage increases

C Inference speed per sample decreases and memory usage decreases

D Inference speed per sample increases and memory usage decreases

💡 Hint

Think about processing multiple inputs at once.

Increasing batch size allows the model to process more samples simultaneously, often improving throughput (speed per sample), but it also requires more memory to hold the batch data.

❓ Metrics

advanced

Interpreting inference output probabilities

A loaded TensorFlow classification model outputs the following prediction for a single input: [[0.05, 0.9, 0.05]]. What does this output represent?

A The model predicts class 2 with 90% confidence

B The model predicts class 1 with 90% confidence

C The model predicts class 0 with 90% confidence

D The model output is invalid because probabilities do not sum to 1

💡 Hint

Look at the highest probability and its index.

The output is a probability distribution over classes. The highest value 0.9 is at index 1, so the model predicts class 1 with 90% confidence.

🔧 Debug

expert

Identifying the cause of an error during model loading

You run the following code to load a TensorFlow model saved in 'model_dir', but get the error: "OSError: SavedModel file does not exist at: model_dir". What is the most likely cause?

TensorFlow

import tensorflow as tf
model = tf.keras.models.load_model('model_dir')

A The path 'model_dir' is incorrect or the directory does not contain a saved model

B The model was saved using tf.saved_model.save instead of tf.keras.models.save_model

C The model file is corrupted and cannot be loaded

D TensorFlow version is incompatible with the saved model format

💡 Hint

Check the file path and directory contents.

The error indicates the specified path does not contain the expected saved model files. This usually means the path is wrong or the model was not saved there.