Google Cloud GPU Instance

Instructions for setting up a GPU instance on Google Cloud

Install Google Cloud SDK on your laptop:
https://cloud.google.com/sdk/
Install file by executing the following in the terminal:
```
 ./google-cloud-sdk/install.sh  
```
Initialize gcloud in terminal (https://cloud.google.com/compute/docs/gcloud-compute/?hl=de):
```
 gcloud init  
```
Request limit increase for instances
Put in number of GPUs & CPUs, Size:
https://console.cloud.google.com/compute/quotas?hl=de&_ga=1.69181649.1124953923.1496188820
Create GPU instance:

GPU are only available in certain areas:

Choose Boot disk -> Ubuntu 16.04 LTS
Choose 30 GB HDD

Select zone, number of GPUs & CPUs and memory
Allow HTTP & HTTPS traffic
Instance is set up
Connect to instance:
```
gcloud compute config-ssh  
```

Or:

gcloud compute ssh instance-2 --zone us-west1-b
ssh –i ~/.ssh/google_compute_engine USERNAME@IP_ADDRESS

Install GPU drivers:

curl –O http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/cuda-repo-ubuntu1604_8.0.61-1_amd64.deb  
sudo dpkg -i cuda-repo-ubuntu1604_8.0.61-1_amd64.deb  
sudo apt-get update  
sudo apt-get install cuda -y

Insert lines in bashrc:

echo 'export CUDA_HOME=/usr/local/cuda' >> ~/.bashrc  
echo 'export PATH=$PATH:$CUDA_HOME/bin' >> ~/.bashrc  
echo 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CUDA_HOME/lib64' >> ~/.bashrc  
source ~/.bashrc

Download cuDNN from Nvidia Developer site -> Sign Up required

Open terminal on laptop and copy file to instance:

gcloud compute scp cudnn-8.0-linux-x64-v5.1.tgz *INSTANCE_NAME*:~

Go to terminal on instance and extract & install:

tar -xzvf cudnn-8.0-linux-x64-v5.1.tgz  
sudo cp cuda/lib64/* /usr/local/cuda/lib64/  
sudo cp cuda/include/cudnn.h /usr/local/cuda/include/  
rm -rf ~/cuda  
rm cudnn-8.0-linux-x64-v5.1.tgz  
sudo apt-get update

Install python packages:

sudo apt-get install python-pip  
sudo apt-get install python-numpy python-scipy python-matplotlib ipython ipython-notebook  
sudo apt-get install python-pandas python-sympy

Install Anaconda:

wget https://repo.continuum.io/archive/Anaconda3-4.4.0-Linux-x86_64.sh  
bash Anaconda3-4.4.0-Linux-x86_64.sh

Set Anaconda as preferred environment:

which python should be usr/bin/python
Important:
```
  source .bashrc  
```

Generate config file:

  jupyter notebook --generate-config
  mkdir certs
  sudo openssl req -x509 -nodes -days 365 -newkey rsa:1024 -keyout mycert.pem -out mycert.pem

Configure jupyter:

  cd ~/.jupyter/
  vi jupyter_notebook_config.py

Change username if necessary:

  c = get_config()
  c.IPKernelApp.pylab = 'inline'  # if you want plotting support always in your notebook
  c.NotebookApp.certfile = u'/home/ubuntu/certs/mycert.pem' #location of your certificate file
  c.NotebookApp.ip = '*'
  c.NotebookApp.open_browser = False  #so that the ipython notebook does not opens up a browser by default
  #c.NotebookApp.password = u'sha1:98ff0e580111:12798c72623a6eecd54b51c006b1050f0ac1a62d'  #the encrypted password we generated above
  c.NotebookApp.port = 8888  
  
  source .bashrc

Create SSH tunnel:

ssh -i ~/.ssh/google_compute_engine -L 8899:localhost:8888 USERNAME@35.197.12.149
jupyter notebook

Open Browser and type in token:
```
https://localhost:8899/
```

Install Keras and Tensorflow:

pip/conda install keras
pip/conda install tensorflow-gpu
conda install opencv

If GPU process is not running, try to run a keras/tf model first:
```
nvidia-smi  
```

Sources / Appendix

https://medium.com/google-cloud/running-jupyter-notebooks-on-gpu-on-google-cloud-d44f57d22dbd
https://cloud.google.com/compute/docs/gpus/add-gpus?hl=de#install-driver-manual
https://cloud.google.com/compute/docs/gcloud-compute/?hl=de

One line code to create GPU instance:

  gcloud beta compute instances create gpu-inst1 --machine-type n1-standard-2 --zone us-west1-b --accelerator type=nvidia-tesla-k80, count=1 --image-family ubuntu-1604-lts --image-project ubuntu-os-cloud --boot-disk-size 30GB --maintenance-policy TERMINATE --restart-on-failure

qinglingni/gpucloud

Google Cloud GPU Instance

Sources / Appendix