gpu-jupyter

Leverage the power of Jupyter and use your NVIDEA GPU and use Tensorflow and Pytorch in collaborative notebooks.

Contents

1. Requirements
2. Quickstart
3. Deployment
4. Configuration
5. Trouble-Shooting

Requirements

1. Install Docker version 1.10.0+

2. Install Docker Compose version 1.6.0+

3. Get access to use your GPU via the CUDA drivers, see this blog-post

4. Clone this repository

   git clone https://github.com/iot-salzburg/gpu-jupyter.git
   cd gpu-jupyter
   

Quickstart

As soon as you have access to your GPU locally (it can be tested via a Tensorflow or PyTorch), you can run these commands to start the jupyter notebook via docker-compose:

./start-local.sh

This will run jupyter on the default port localhost:8888. The general usage is:

./start-local.sh -p [port]  # port must be an integer with 4 or more digits.

In order to stop the local deployment, run:

./stop-local.sh

Deployment in the Docker Swarm