Use Amazon cloud GPU

[gngdb]

Train a model with the Amazon cloud services and you get this achievement!

[matt-graham]

Instructions written by Kez on getting an Amazon EC2 GPU instance up and running for a pylearn2 model. Some of libraries and environment variables dependencies will be a bit different here

---

1. use the AMI: ubuntu-trusty-14.04-amd64-server-20141125 (ami-898dd9b9) - it should say (HVM)

2. follow this setup bash script to configure the environment

   chmod 400 first.pem
   ssh -v -i first.pem ubuntu@public_dsn  # copypaste public_dsn from AWS EC2 console
   
   sudo apt-get update
   sudo apt-get -y dist-upgrade
   sudo apt-get install -y gcc g++ gfortran build-essential git wget linux-image-generic libopenblas-dev python-dev python-pip python-nose python-numpy python-scipy python-matplotlib
   sudo pip install --upgrade --no-deps git+git://github.com/Theano/Theano.git
   sudo wget http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1404/x86_64/cuda-repo-ubuntu1404_6.5-14_amd64.deb
   sudo dpkg -i cuda-repo-ubuntu1404_6.5-14_amd64.deb
   sudo apt-get update
   sudo apt-get install -y cuda
   echo -e "\nexport PATH=/usr/local/cuda-6.5/bin:$PATH\n\nexport LD_LIBRARY_PATH=/usr/local/cuda-6.5/lib64" >> ~/.bashrc
   sudo ldconfig /usr/local/cuda-6.5/lib64 #more reliable: last answer of http://askubuntu.com/questions/210884/setting-ld-library-path-for-cuda
   sudo reboot
   
   chmod 400 first.pem
   ssh -v -i first.pem ubuntu@publicdsn
   
   cd /usr/local/cuda-6.5/samples/1_Utilities/deviceQuery
   sudo make
   ./deviceQuery
   echo -e "\n[global]\nfloatX=float32\ndevice=gpu\nmode=FAST_RUN\n\n[nvcc]\nfastmath=True\n\n[blas]ldflags=\n\n[cuda]\nroot=/usr/local/cuda-6.5" >> ~/.theanorc
   
   cd ~
   git clone git://github.com/lisa-lab/pylearn2.git
   cd pylearn2
   sudo python setup.py install
   cd ~
   
   sudo pip install pandas
   
   echo "\nexport DATA_DIR=~/data\nexport EXPERIMENTS_DIR=~/experiments" >> ~/.bashrc
   source ~/.bashrc

3. add any extra environment vars as required e.g. DATA_DIR, PYLEARN2_DATA_PATH
4. if installing theano, pylearn2 with sudo, you need to sudo python. However, we lose the environment vars in sudo default so the call becomes sudo -E python etc.py ...
5. convert the datasets offline and scp them over
6. adjust the test.py script to handle this data without the pipeline cruft
7. source .bashrc updates the environment in the current shell after editing .bashrc file