Installation and GPU setup in TensorFlow - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding GPU availability in TensorFlow

Which TensorFlow command correctly checks if a GPU is available for use?

A tf.check_gpu()

B tf.device('GPU')

C tf.gpu.is_available()

D tf.config.list_physical_devices('GPU')

💡 Hint

Look for the TensorFlow function that lists physical devices.

The function tf.config.list_physical_devices('GPU') returns a list of available GPU devices. If the list is empty, no GPU is available.

❓ Predict Output

intermediate

Output of TensorFlow GPU memory growth setting

What is the output of this code snippet when run on a system with one GPU?

TensorFlow

import tensorflow as tf

gpus = tf.config.list_physical_devices('GPU')
if gpus:
    try:
        for gpu in gpus:
            tf.config.experimental.set_memory_growth(gpu, True)
        print('Memory growth set')
    except RuntimeError as e:
        print(e)
else:
    print('No GPU found')

A Memory growth set

B No GPU found

C SyntaxError

D RuntimeError: Memory growth must be set before GPUs have been initialized

💡 Hint

If a GPU is present and memory growth is set before initialization, it prints confirmation.

The code sets memory growth on all GPUs before TensorFlow initializes them, so it prints 'Memory growth set'.

❓ Model Choice

advanced

Choosing the correct TensorFlow version for GPU support

Which TensorFlow version is required to support GPU acceleration on CUDA 11.2?

A TensorFlow 1.15

B TensorFlow 2.4

C TensorFlow 2.5

D TensorFlow 2.3

💡 Hint

Check TensorFlow release notes for CUDA compatibility.

TensorFlow 2.5 added support for CUDA 11.2, earlier versions do not support it.

❓ Hyperparameter

advanced

Setting GPU device in TensorFlow

Which code snippet correctly sets TensorFlow to use only the first GPU device?

A tf.config.set_visible_devices([tf.config.list_physical_devices('GPU')[0]], 'GPU')

B tf.config.set_visible_devices(tf.config.list_physical_devices('GPU')[0], 'GPU')

C tf.device('/GPU:0')

D tf.config.set_visible_devices('GPU:0')

💡 Hint

The function expects a list of devices, not a single device.

tf.config.set_visible_devices expects a list of physical devices to set visibility. Passing a single device without list causes error.

🔧 Debug

expert

Diagnosing TensorFlow GPU memory allocation error

Given this error message when running TensorFlow code on GPU: "ResourceExhaustedError: OOM when allocating tensor", what is the most likely cause?

A TensorFlow is not installed correctly

B The GPU ran out of memory due to large batch size or model size

C The CPU is overloaded and cannot handle the computation

D The GPU driver is outdated and incompatible

💡 Hint

OOM means out of memory on the GPU device.

ResourceExhaustedError with OOM indicates the GPU memory is insufficient for the requested tensor allocation, often due to large batch size or model.