tensorflowgpu--help.md

Tensorflow local GPU setup: Windows Native

Hardware Requirements

Note: TensorFlow binaries use AVX instructions which may not run on older CPUs.

• NVIDIA® GPU drivers version 450.80.02 or higher.
• CUDA® Toolkit 11.8.
• cuDNN SDK 8.6.0. (Nvidia login required).
• (Optional) TensorRT to improve latency and throughput for inference.

Software Requirements

• MultiPack Installer Microsoft Visual C++ .
• Install Miniconda or Anaconda

Create Environment

Open Anaconda Prompt.

conda create --name tf python=3.6

conda deactivate
conda activate tf

Install conda packages (Miniconda)

If you have installed Miniconda, follow the give commands

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

pip install --upgrade pip

pip install "tensorflow<2.11"

Installing conda packages (Anaconda)

• Open Anaconda and select tf environment and install the following packages:
• keras
• tensorflow
• tensorflow-gpu
• The dependency packages will be installed automatically.

Verify the installation

• Verify GPU setup:

python -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"

• Verify CPU setup:

python -c "import tensorflow as tf; print(tf.reduce_sum(tf.random.normal([1000, 1000])))"

Verify with Jupyter Notebook

Download verify-tf and import inside tf environment jupyter notebook