How to Check if Tensorflow is Using GPU?

GPU is abbreviated as Graphics Processing Unit. It is a specialized processor designed to handle the complex and repetitive calculations required for video encoding or decoding, graphics rendering and other computational intensive tasks.

It is mainly suited to perform large-scale parallel computations, which makes it ideal for machine learning and other data-based applications.

GPUs in machine learning have become more popular as they reduce the time required to train complex neural networks. TensorFlow, PyTorch, and Keras are built-in frameworks of machine learning which support GPU acceleration.

The following are the steps to check if TensorFlow is using GPU ?

Installing TensorFlow

First we have to install TensorFlow in the Python environment by using the below command ?

pip install tensorflow

If you see the following output, then TensorFlow is installed successfully ?

Collecting tensorflow
  Downloading tensorflow-2.12.0-cp310-cp310-win_amd64.whl (1.9 kB)
Collecting tensorflow-intel==2.12.0
  Downloading tensorflow_intel-2.12.0-cp310-cp310-win_amd64.whl (272.8 MB)
     ---------------------------------------- 272.8/272.8 MB 948.3 kB/s eta 
Installing collected packages: tensorflow
Successfully installed tensorflow-2.12.0

Method 1: Using list_physical_devices()

The most reliable way to check if TensorFlow can access GPU is using list_physical_devices() ?

import tensorflow as tf

# Check all available devices
print("All devices:", tf.config.list_physical_devices())

# Check specifically for GPU devices
gpu_devices = tf.config.list_physical_devices('GPU')
print("GPU devices:", gpu_devices)

# Check if GPU is available
if gpu_devices:
    print("GPU is available")
    print(f"Number of GPUs: {len(gpu_devices)}")
else:
    print("GPU is not available")

All devices: [PhysicalDevice(name='/physical_device:CPU:0', device_type='CPU')]
GPU devices: []
GPU is not available

Method 2: Using built_with_cuda()

Check if TensorFlow was built with CUDA support ?

import tensorflow as tf

# Check if TensorFlow was built with CUDA support
print("Built with CUDA:", tf.test.is_built_with_cuda())

# Check TensorFlow version
print("TensorFlow version:", tf.__version__)

Built with CUDA: False
TensorFlow version: 2.12.0

Method 3: Checking Device Placement

You can also check which device TensorFlow uses for operations by creating a simple computation ?

import tensorflow as tf

# Create a simple operation
with tf.device('/CPU:0'):
    a = tf.constant([1, 2, 3])
    b = tf.constant([4, 5, 6])
    c = tf.add(a, b)

print("Operation result:", c.numpy())
print("Device used:", c.device)

Operation result: [5 7 9]
Device used: /job:localhost/replica:0/task:0/device:CPU:0

Comparison of Methods

Conclusion

Use tf.config.list_physical_devices('GPU') as the primary method to check GPU availability in TensorFlow. This is the most reliable and modern approach recommended by the TensorFlow team.