tensorflowcudaversioncompatibilitycudnn

Which TensorFlow and CUDA version combinations are compatible?

I have noticed that some newer TensorFlow versions are incompatible with older CUDA and cuDNN versions. Does an overview of the compatible versions or even a list of officially tested combinations exist? I can't find it in the TensorFlow documentation.

Solution

  • TL;DR) See this table: https://www.tensorflow.org/install/source#gpu

    Generally:

    Check the CUDA version:

    cat /usr/local/cuda/version.txt

    and cuDNN version:

    grep CUDNN_MAJOR -A 2 /usr/local/cuda/include/cudnn.h

    and install a combination as given below in the images or here.

    The following images and the link provide an overview of the officially supported/tested combinations of CUDA and TensorFlow on Linux, macOS and Windows:

    Minor configurations:

    Since the given specifications below in some cases might be too broad, here is one specific configuration that works:

    The corresponding cudnn can be downloaded here.

    Tested build configurations

    Please refer to https://www.tensorflow.org/install/source#gpu for a up-to-date compatibility chart (for official TF wheels).

    (figures updated May 20, 2020)

    Linux GPU

    enter image description here

    Linux CPU

    enter image description here

    macOS GPU

    enter image description here

    macOS CPU

    enter image description here

    Windows GPU

    enter image description here

    Windows CPU

    enter image description here

    Updated as of Dec 5 2020: For the updated information please refer Link for Linux and Link for Windows.