tf.data.Dataset creation in TensorFlow - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of Dataset from_tensor_slices with nested lists

What is the output of the following code snippet when iterated over?

import tensorflow as tf

data = [[1, 2], [3, 4], [5, 6]]
dataset = tf.data.Dataset.from_tensor_slices(data)
output = [element.numpy().tolist() for element in dataset]

TensorFlow

import tensorflow as tf

data = [[1, 2], [3, 4], [5, 6]]
dataset = tf.data.Dataset.from_tensor_slices(data)
output = [element.numpy().tolist() for element in dataset]
print(output)

A [[1], [2], [3], [4], [5], [6]]

B [1, 2, 3, 4, 5, 6]

C [[1, 3, 5], [2, 4, 6]]

D [[1, 2], [3, 4], [5, 6]]

💡 Hint

Remember that from_tensor_slices slices the first dimension, so each element corresponds to one sublist.

The from_tensor_slices method slices the input tensor along the first dimension. Since the input is a list of lists, each element yielded is one sublist, resulting in [[1, 2], [3, 4], [5, 6]].

❓ data_output

intermediate

Number of elements in Dataset from_generator

Given the following generator and dataset creation code, how many elements does the dataset contain?

import tensorflow as tf

def gen():
    for i in range(5):
        yield i * 2

dataset = tf.data.Dataset.from_generator(gen, output_signature=tf.TensorSpec(shape=(), dtype=tf.int32))
count = sum(1 for _ in dataset)

TensorFlow

import tensorflow as tf

def gen():
    for i in range(5):
        yield i * 2

dataset = tf.data.Dataset.from_generator(gen, output_signature=tf.TensorSpec(shape=(), dtype=tf.int32))
count = sum(1 for _ in dataset)
print(count)

A 0

B 10

C 5

D 1

💡 Hint

Count how many times the generator yields values.

The generator yields 5 values (0, 2, 4, 6, 8), so the dataset contains 5 elements.

🔧 Debug

advanced

Identify the error in Dataset creation from dict

What error does the following code raise when executed?

import tensorflow as tf

data = {'a': [1, 2], 'b': [3, 4, 5]}
dataset = tf.data.Dataset.from_tensor_slices(data)
for element in dataset:
    print(element)

TensorFlow

import tensorflow as tf

data = {'a': [1, 2], 'b': [3, 4, 5]}
dataset = tf.data.Dataset.from_tensor_slices(data)
for element in dataset:
    print(element)

A TypeError: Expected list or tuple

B ValueError: All components must have the same size

C No error, prints elements

D AttributeError: 'dict' object has no attribute 'numpy'

💡 Hint

Check if all lists in the dictionary have the same length.

from_tensor_slices requires all components to have the same size along the first dimension. Here, 'a' has length 2 and 'b' has length 3, causing a ValueError.

❓ visualization

advanced

Visualize Dataset elements after map transformation

What is the output list after applying the map function to the dataset?

import tensorflow as tf

data = [1, 2, 3, 4]
dataset = tf.data.Dataset.from_tensor_slices(data)
dataset = dataset.map(lambda x: x * x)
output = [element.numpy() for element in dataset]

TensorFlow

import tensorflow as tf

data = [1, 2, 3, 4]
dataset = tf.data.Dataset.from_tensor_slices(data)
dataset = dataset.map(lambda x: x * x)
output = [element.numpy() for element in dataset]
print(output)

A [1, 4, 9, 16]

B [1, 2, 3, 4]

C [2, 4, 6, 8]

D [1, 8, 27, 64]

💡 Hint

The map function squares each element.

The map applies the lambda function x*x to each element, so the output is the squares of the original list.

🚀 Application

expert

Creating a Dataset from multiple numpy arrays with different shapes

You have two numpy arrays:

import numpy as np
import tensorflow as tf

arr1 = np.array([[1, 2], [3, 4], [5, 6]])
arr2 = np.array([10, 20, 30])

Which code snippet correctly creates a tf.data.Dataset that yields tuples of corresponding elements from arr1 and arr2?

TensorFlow

import numpy as np
import tensorflow as tf

arr1 = np.array([[1, 2], [3, 4], [5, 6]])
arr2 = np.array([10, 20, 30])

# Choose the correct dataset creation code

A dataset = tf.data.Dataset.from_tensor_slices((arr1, arr2))

B dataset = tf.data.Dataset.from_tensor_slices(arr1).concatenate(tf.data.Dataset.from_tensor_slices(arr2))

C dataset = tf.data.Dataset.from_tensor_slices(arr1).zip(tf.data.Dataset.from_tensor_slices(arr2))

D dataset = tf.data.Dataset.from_tensor_slices(arr1 + arr2)

💡 Hint

from_tensor_slices can take a tuple of arrays with matching first dimension.

Option A correctly creates a dataset yielding tuples (arr1[i], arr2[i]). Option A creates a zipped dataset but is more verbose. Option A tries to add arrays element-wise which is invalid for slicing. Option A concatenates datasets, not pairs elements.