Installing Tensorflow (with GPU support) on Windows
MM 0413/2018
The machine are suggested to meet requirements of (64-bit, x86 desktop or laptops) and (windows 7 or later).
The tensorFlow programs with GPU support typically run significantly faster on a GPU than on a CPU. If you have a NVIDIA GPU meeting the pre-requisites, you can install this `Tensorflow with GPU support' version.
Requirements to run TensorFlow with GPU support
Follwing NVIDIA software must be installed on systems:
CIDA Toolkit 9.0. Ensure that you append the relevant Cuda pathnames to teh %PATH% environment variable as desribed in the NVIDIA documentation.
The NVIDIA drivers associated wtih CUDA Toolkit 9.0.
cuDNN v7.0. Note that cuDNN is typically installed in differet location from the other CUDA DLLs. Ensure that you add the directory where you installed the cuDNN to your %PATH% environment variable.
GPU card with CUDA Compute Capability 3.0 or higher for building from source and 3.5 or higher for our binaries. There are a list of supported GPU cards in NVIDIA documents.
If you have a different version of one o fthe preceding packages, please change to teh specified versions. In particular, the cuDNN version must match exactly: TensorFlow will not load if it can not find cuDNN64_7.dll. To use a different version of cuDNN, you must build from source.
[0]