NVIDIA DIGITS Installation Tutorial

 

DIGITS (Deep learning GPU Training System) is a web application for training deep learning models. Currently supported frameworks are: Caffe, Torch and Tensorflow.

You can download the source code here: https://gitee.com/NVIDIA_Developer_Community/DIGITS

1. Installation environment and background

DIGITS supports basic training frameworks including Caffe, TensorFlow, and Torch. This article uses Caffe as an example.

DIGITS can currently support Ubuntu 14.04 and Ubuntu 16.04. This article will not repeat the basic operation of system installation. If necessary, you can find a lot of tutorials online.

Environment configuration:

• Ubuntu 16.04
• MIRACLES 9.0
• CUDNN 7.0
• Caffe-0.15

Second, the detailed installation process

    1.cuda installation (if you have already installed it, you don't need to perform this step):      

# For Ubuntu 16.04
CUDA_REPO_PKG=http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/cuda-repo-ubuntu1604_8.0.61-1_amd64.deb
ML_REPO_PKG=http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1604/x86_64/nvidia-machine-learning-repo-ubuntu1604_1.0.0-1_amd64.deb

    2. Install some required libraries:  

sudo apt-get install --no-install-recommends git graphviz python-dev python-flask python-flaskext.wtf python-gevent python-h5py python-numpy python-pil python-pip python-scipy python-tk

     Note If you don't have CAFFE installed, you can find a detailed tutorial here:

     https://github.com/NVIDIA/DIGITS/blob/digits-6.0/docs/BuildCaffe.md

    3. Download the source code:

# example location - can be customized
DIGITS_ROOT=~/digits
git clone https://github.com/NVIDIA/DIGITS.git $DIGITS_ROOT

    4. Install some Python packages:

sudo pip install -r $DIGITS_ROOT/requirements.txt

     5. Finally, install DIGITS:

sudo pip install -e $DIGITS_ROOT

After the installation is complete, you can start DIGITS with the following command:

./digits-devserver

Then open a browser at random (you can use the browser to access this machine, or you can use the browser of any computer in the same network to access), and visit:http://Your Digits Server IP:5000/