Windows 10
Supported system: Python 3.5-3.8, Windows 7 or later (with C++ redistributable) , Ubuntu 16.04 or later.
*Step 1-3 can be skipped if python3 has been installed.
1) Install the Microsoft Visual C++ Redistributable for Visual Studio 2015, 2017, and 2019.
2) Enable longpath for windows.
4) Create virtual environment:
python -m venv --system-site-packages .\venv #Create a new virtual environment by choosing a Python interpreter and making a .\venv directory to hold it
> .\venv\Scripts\activate # Activate the virtual environment
> pip install --upgrade pip # upgrade pip in virtual environment
> pip list # show packages installed within the virtual environment
> deactivate # Deactivate virtual environment. don't exit until you're done using TensorFlow
5) Install Tensorflow via pip:
> pip install --upgrade tensorflow
> python -c "import tensorflow as tf;print(tf.reduce_sum(tf.random.normal([1000, 1000])))" # verifying installation
6) GPU setup:
Install NVIDIA GPU driver: CUDA 11 requires > 450.x (* can skip this as CUDA toolkit includes GPU driver)
Install CUDA Toolkit: Tensorflow 2.4.0 supports CUDA 11.0 (Tensorflow 2.3 supports 10.2)
Download cuDNN that matches with the CUDA version, and move those files (lib, bin, include) into the CUDA folder.
Add the CUDA®, CUPTI, and cuDNN installation directories to the %PATH% environmental variable.
SET PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\bin;%PATH%
SET PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\extras\CUPTI\lib64;%PATH%
SET PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\include;%PATH%
** Note, use import tensorflow.compat.v1 as tf. instead of import tensorflow as tf. for tensorflow 1.x code.