ENVIRONMENT DETAILS:

  1. Operating System: Windows 11 Pro
  2. Graphics Card: NVIDIA GPU RTX-3050 Ti

RESTRICTIONS (https://www.tensorflow.org/extras/cert/Setting_Up_TF_Developer_Certificate_Exam.pdf):

  1. Python version has to be 3.9
  2. The TensorFlow version has to be 2.13

CONFIGURATION STEPS:

  1. Install the driver corresponding to my video card:

    1. What video card do I have?

      Untitled

    2. What driver do I need to install?

      This is the driver I got: 545.84-notebook-win10-win11-64bit-international-dch-whql

      Untitled

    c. Check if the driver was properly installed with nvidia-smi command

  2. Install CUDA

    1. According to the tensorflow.org/install/source documentation, TensorFlow 2.13 is compatible with CUDA 11.8

    Untitled

b. This is where I took the CUDA kit from:

Untitled

  1. Download and copy CuDNN files

    1. Download the CuDNN 8.6, according to the dependency list above.

    2. Unzip the archive.

    3. Copy the content of the following subfolders contained in the CuDNN archive:

      1. cudnn-windows-x86_64-8.6.0.163_cuda11-archive\bin → C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\bin
      2. cudnn-windows-x86_64-8.6.0.163_cuda11-archive\include → C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\include
      3. cudnn-windows-x86_64-8.6.0.163_cuda11-archive\lib\x64 → C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\lib\x64

    4. Set the environment variables:

      Add C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\bin, Toolkit\CUDA\v11.8\extras\CUPTI\lib64, C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\include and C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8 to the PATH variable

  2. Install Anaconda

    Untitled

  3. Create a virtual environment using Anaconda and install everything needed for running on GPU there

    conda create --name VIRTUALENVIRONMENTNAME python=3.9conda create --name

    conda activate VIRTUALENVIRONMENTNAME

    conda install -c conda-forge cudatoolkit=11.2 cudnn=8.1.0

    pip install --upgrade pip

    pip install tensorflow==2.13.*

    python -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))" #IT DOES NOT WORK 😞

    conda deactivate