Setting up TensorFlow with GPU support can significantly accelerate your machine learning workflows. Follow this guide to configure TensorFlow for GPU usage on your system.

Prerequisites 📦

Before proceeding, ensure you have:

  • A compatible NVIDIA GPU (e.g., GTX 10-series, Titan, or RTX 30-series)
  • CUDA Toolkit installed (version matches TensorFlow requirements)
  • cuDNN library installed (aligned with CUDA version)
  • NVIDIA Driver updated to the latest version

Step-by-Step Installation 🔧

  1. Install NVIDIA Drivers
    Use your OS's package manager or download from NVIDIA官网.

    tensorflow_gpu_setup

  2. Install CUDA Toolkit
    Download and install the latest CUDA version from NVIDIA CUDA Toolkit.

    • For Ubuntu: sudo apt-get install cuda-toolkit
    • For Windows: Use the installer from the official site

  3. Install cuDNN
    Download cuDNN from NVIDIA cuDNN and extract the files.

  4. Set Environment Variables
    Add these paths to your .bashrc or system environment: 

    export PATH=/usr/local/cuda-11.x/bin:$PATH  
export LD_LIBRARY_PATH=/usr/local/cuda-11.x/lib64:$LD_LIBRARY_PATH

    Replace 11.x with your CUDA version.

  5. Install TensorFlow with GPU Support
    Use pip: 

    pip install tensorflow-gpu

    Or via conda: 

    conda install -c conda-forge tensorflow-gpu

Verify GPU Compatibility 🧪

Run this code to check if TensorFlow detects your GPU: 

import tensorflow as tf  
print("GPU Available:", tf.config.list_physical_devices('GPU'))

If it returns GPU Available: [GPU...], your setup is successful!

Further Reading 📚

For detailed steps on TensorFlow GPU configuration, visit our guide section.

🔧 Tip: Always match CUDA/cuDNN versions with your TensorFlow release.
💡 Note: If you encounter errors, ensure all dependencies are correctly installed.

Need help with specific steps? Explore our TensorFlow documentation for more insights.