Dataset from tensors in TensorFlow - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of creating a dataset from tensors

What is the output of this code snippet that creates a TensorFlow dataset from tensors and iterates over it?

TensorFlow

import tensorflow as tf

features = tf.constant([[1, 2], [3, 4], [5, 6]])
labels = tf.constant([0, 1, 0])
dataset = tf.data.Dataset.from_tensor_slices((features, labels))
output = []
for feature, label in dataset:
    output.append((feature.numpy().tolist(), label.numpy().item()))
print(output)

A [([1, 2], 1), ([3, 4], 0), ([5, 6], 1)]

B [([1, 2], 0), ([3, 4], 0), ([5, 6], 0)]

C [([1, 2], 0), ([3, 4], 1), ([5, 6], 0)]

D SyntaxError

💡 Hint

Remember that from_tensor_slices pairs elements by index from each tensor.

The dataset pairs each row of features with the corresponding label by index. So the first feature [1, 2] pairs with label 0, the second with 1, and the third with 0.

❓ data_output

intermediate

Number of elements in a dataset from tensors

Given this TensorFlow dataset created from tensors, how many elements does it contain?

TensorFlow

import tensorflow as tf

features = tf.constant([[10, 20], [30, 40], [50, 60], [70, 80]])
labels = tf.constant([1, 0, 1, 0])
dataset = tf.data.Dataset.from_tensor_slices((features, labels))
count = 0
for _ in dataset:
    count += 1
print(count)

A 4

B 2

C 3

D 0

💡 Hint

Count how many pairs are created from the tensors.

The dataset creates one element per row in the tensors. Since there are 4 rows, the dataset has 4 elements.

🔧 Debug

advanced

Identify the error when creating a dataset from tensors

What error will this code raise when trying to create a dataset from tensors of different shapes?

TensorFlow

import tensorflow as tf

features = tf.constant([[1, 2], [3, 4]])
labels = tf.constant([0, 1, 0])
dataset = tf.data.Dataset.from_tensor_slices((features, labels))

A ValueError: All input tensors must have the same first dimension size

B TypeError: Cannot convert list to tensor

C SyntaxError: invalid syntax

D No error, dataset created successfully

💡 Hint

Check the shapes of the tensors passed to from_tensor_slices.

The tensors must have the same size in the first dimension to pair elements by index. Here, features has shape (2, 2) and labels has shape (3,), causing a ValueError.

🚀 Application

advanced

Creating a batched dataset from tensors

Which option correctly creates a TensorFlow dataset from tensors and batches it with batch size 2?

TensorFlow

import tensorflow as tf

features = tf.constant([[1, 2], [3, 4], [5, 6], [7, 8]])
labels = tf.constant([0, 1, 0, 1])

A dataset = tf.data.Dataset.from_tensor_slices(features, labels).batch(2)

B dataset = tf.data.Dataset.from_tensor_slices((features, labels)).batch(2)

C dataset = tf.data.Dataset.from_tensors((features, labels)).batch(2)

D dataset = tf.data.Dataset.from_tensor_slices([features, labels]).batch(2)

💡 Hint

Check the correct method signature for from_tensor_slices and how to batch datasets.

Option B correctly uses from_tensor_slices with a tuple of tensors and then batches the dataset by 2 elements per batch.

🧠 Conceptual

expert

Understanding memory usage of datasets from tensors

When creating a TensorFlow dataset from large tensors using from_tensor_slices, what is the main memory behavior?

A The entire tensor data is copied into the dataset memory immediately

B The dataset compresses the tensors to save memory

C The dataset stores only the first slice of the tensors in memory

D The dataset holds references to the original tensors without copying data

💡 Hint

Think about how TensorFlow datasets handle input tensors internally.

TensorFlow datasets created from tensors hold references to the original tensors and slice them on demand, avoiding copying all data upfront.