Installing TensorFlow

DIGITS now supports TensorFlow as an optional alternative backend to Caffe or Torch.

Installation for Ubuntu

Installation for Mac

Requirements

DIGITS is current targeting tensorflow-gpu V1.2.1.

TensorFlow for DIGITS requires one or more NVIDIA GPUs with CUDA Compute Capbility of 3.0 or higher. See the official GPU support list to see if your GPU supports it.

Along with that requirement, the following should be installed

• One or more NVIDIA GPUs (details)
• An NVIDIA driver (details and installation instructions)
• A CUDA toolkit (details and installation instructions)
• cuDNN 5.1 (download page)

A Note About cuDNN and TensorFlow

Currently tensorflow v1.2.1 targets cuDNN 5.1. To have tensorflow v1.2.1 running in digits, you must have cuDNN 5.1 installed. To install it, use the following command in a terminal

sudo apt-get install libcudnn5

Installation

These instructions are based on [the official TensorFlow instructions] (https://www.tensorflow.org/versions/master/install/)

TensorFlow comes with pip, to install it, just simply use the command

pip install tensorflow-gpu==1.2.1

TensorFlow should then install effortlessly and pull in all its required dependices.

Getting Started With TensorFlow In DIGITS

Follow these instructions for information on getting started with TensorFlow in DIGITS