• Install Ubuntu 14.04 on Amazon AWS g2 with CUDA
  • Driver
  • CUDA
  • cuDNN
  • Install Anaconda
  • Try with keras
  • Remote Jupyter Notebook
    • Create profile
    • Generate password, copy the output from passwd

@(Cabinet)[cluster]

Install Ubuntu 14.04 on Amazon AWS g2 with CUDA

date: 2016-04-24

Based on http://tleyden.github.io/blog/2014/10/25/cuda-6-dot-5-on-aws-gpu-instance-running-ubuntu-14-dot-04/

Also got lots of help from my friend David.

sudo apt-get update 
sudo apt-get -y install build-essential linux-image-extra-virtual axel

sudo vim /etc/modprobe.d/blacklist-nouveau.conf

blacklist nouveau blacklist lbm-nouveau options nouveau modeset=0 alias nouveau off alias lbm-nouveau off

echo options nouveau modeset=0 | sudo tee -a /etc/modprobe.d/nouveau-kms.conf

sudo update-initramfs -u

sudo shutdown -r now 

sudo apt-get install -y linux-headers-$(uname -r)

Driver

axel -avn 30 http://us.download.nvidia.com/XFree86/Linux-x86_64/361.42/NVIDIA-Linux-x86_64-361.42.run

sudo bash ./NVIDIA-Linux-x86_64-361.42.run

CUDA

axel -avn 30 http://developer.download.nvidia.com/compute/cuda/7.5/Prod/local_installers/cuda_7.5.18_linux.run

sudo bash ./cuda_7.5.18_linux.run  
# Don't install the driver, install the toolkit only

vim ~/.bashrc

export PATH=/usr/local/cuda-7.5/bin:$PATH export LD_LIBRARY_PATH=/usr/local/cuda-7.5/lib64

cuDNN

wget https://www.dropbox.com/s/5yuh62lolheoyxk/cudnn-7.0-linux-x64-v4.0-prod.tgz

tar xvf cudnn-7.0-linux-x64-v4.0-prod.tgz

sudo cp cuda/lib64/* /usr/local/cuda/lib64/
sudo cp cuda/include/cudnn.h /usr/local/cuda/include/

Install Anaconda

wget https://3230d63b5fc54e62148e-c95ac804525aac4b6dba79b00b39d1d3.ssl.cf1.rackcdn.com/Anaconda2-4.0.0-Linux-x86_64.sh

bash Anaconda2-4.0.0-Linux-x86_64.sh
source ~/.bashrc

Try with keras

sudo apt-get -y install git htop

git clone https://github.com/fchollet/keras.git --depth=1

pip install ./keras

vim ~/.theanorc

[global]
floatX = float32
device = gpu0

cd keras/examples
python mnist_cnn.py

Remote Jupyter Notebook

Setup remote jupyter notebook server on port 9999 Running public notebook server, so can access remotely

http://jupyter-notebook.readthedocs.org/en/latest/public_server.html

sudo apt-get install -y libssl-dev

Create profile

jupyter notebook --generate-config

Generate password, copy the output from passwd

python
from notebook.auth import passwd
passwd()

vim ~/.jupyter/jupyter_notebook_config.py

c.NotebookApp.ip = '*'
c.NotebookApp.password = u'sha1:xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx'
c.NotebookApp.open_browser = False
c.NotebookApp.port = 1111