Setting up TensorFlow 0.9 with Python 3.5 on AWS GPU-instance

Description

setup-aws-tensorflow.bash installs the following things on the ec2 g2.2xlarge instance running Ubuntu 14.04:

• Required linux packages
• CUDA 7.5
• cuDNN v4
• Anaconda with Python 3.5
• TensorFlow 0.9
• Keras
• GPU usage tool gpustat

Also configures the Jupyter Notebook for interactive use

Forked from: https://github.com/Avsecz/aws-tensorflow-setup

Based on the blog post: http://max-likelihood.com/2016/06/18/aws-tensorflow-setup/.

Preliminaries -- broad strokes steps

• Install the Amazon Web Services CLI or use the AWS Dashboard online
• Setup your AWS credentials
• Create a security group with the following ports open to TCP/IP traffic: 22 (SSH), 80 (HTTP), 9999 (Jupyter Notebook)
• Launch a g2.2xlarge instance running Ubuntu server (ami-fce3c696)
• Log in to your instance via SSH

For more details, see the beginning of this blog post: http://max-likelihood.com/2016/06/18/aws-tensorflow-setup/. Though make sure to open the additional ports if you wish to use the Jupyter Notebook.

Setting up TensorFlow + Keras with GPU support

• Update apt-get and packages

sudo apt-get update && sudo apt-get -y upgrade

• Install git

sudo apt-get install -y git

• Clone this repo

git clone https://github.com/jwittenbach/aws-tensorflow-setup

• Run the installation script

./aws-tensorflow-setup/setup-aws-tensorflow.bash

• Source your bash profile to set environment variables

source ~/.bash_profile

• (Optional) Launch a Jupyter Notebook server

jupyter Notebook

and access it by pointing your browser to [publicIP]:9999

Shutting down

You have two options for shutting down the instance:

1. If you stop the instance, then you can later start the instance back up again, and you will not need to go through the setup again. However, the disk image will be stored on AWS S3, which does cost a small fee.
2. If you terminate the instance, then you will need to launch a branch new instance the next time, which will entail redoing the setup process.