First of all, you should have a google account and sign up Google Cloud Platform (GCP). Note that if you do not need to train your model for some time. You can stop your instance to pause the charging. You can alway start the instance as needed, and everything in the instance will not change.

First, you shall create a compute engin in the GCP. Click 'Compute Engine' in the menu, then click on 'VM instances'.

Next, you can click ‘Create new instance’ to create new instance. Here, you can name your instance, select server zones and customize your VM. Normally, I select my VM in ‘ us-central1 (lowa)’ (for the purpose of using Nvidia Tesla K80, not all zones have this GPU).

Next, click 'Customise' in the 'Machine type' tab to config your VM. I usually select 2 core CPU, 13 GB memory and 1 NVIDIA Tesla K80.

Last, click 'create' and you are good to go. You can then connect to your VM via SSH.

Now, after creating the cloud computing VM, you can install nesscary packages to run your program.

sudo apt-get update

sudo apt-get install openjdk-8-jdk git python-dev python3-dev python-numpy python3-numpy build-essential python-pip python3-pip python-virtualenv swig python-wheel libcurl3-dev curl

You can searh the suitable driver on [https://developer.nvidia.com/cuda-80-ga2-download-archive]. You can download drivers that you found from the nvidia website (here we use the driver for ubuntu 16.04):

curl -O http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/cuda-repo-ubuntu1604_9.0.176-1_amd64.deb

Then, you should download the key to allow installation:

sudo apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/7fa2af80.pub

Install downloaded packages:

sudo dpkg -i ./cuda-repo-ubuntu1604_9.0.176-1_amd64.deb

Next, you can install CUDA after finishing the installation of Nvidia driver:

sudo apt-get update 
sudo apt-get install cuda-9-0

Last, you can check if Nvidia driver is correctly installed. Type 'nvidia-smi' in terminal, and you shall see something similar:

yourUserName@instance-1:~$ nvidia-smi
Sun Jul 28 hh:mm:ss 2019  
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.67       Driver Version: 418.67       CUDA Version: 10.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Tesla K80           Off  | 00000000:00:04.0 Off |                    0 |
| N/A   54C    P0    84W / 149W |      0MiB / 11441MiB |    100%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

wget https://s3.amazonaws.com/open-source-william-falcon/cudnn-9.0-linux-x64-v7.3.1.20.tgz
sudo tar -xzvf cudnn-9.0-linux-x64-v7.3.1.20.tgz
sudo cp cuda/include/cudnn.h /usr/local/cuda/include
sudo cp cuda/lib64/libcudnn* /usr/local/cuda/lib64
sudo chmod a+r /usr/local/cuda/include/cudnn.h /usr/local/cuda/lib64/libcudnn*

Then, add paths to lib:

export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64"
export CUDA_HOME=/usr/local/cuda
export PATH="$PATH:/usr/local/cuda/bin"
source ~/.bashrc

Anaconda provides an easy installation of tensorflow. The download link of following code can be found in the anaconda download page, by right click the downloadable images.

wget https://repo.continuum.io/archive/Anaconda3-2018.12-Linux-x86_64.sh 
bash Anaconda3-2018.12-Linux-x86_64.sh

Add conda to path:

export PATH=~/anaconda3/bin:$PATH

Note that 'tfgpu' in the following code is the name of environment, you can change it to anything as you like.

conda create -n tfgpu tensorflow-gpu
conda activate tfgpu

Every time you connect to the VM, you shall run the tfgpu virtual envirment to use the tensorflow:

conda activate tfgpu

You can find the code for installation on [https://pytorch.org/]. We select 'Linux', 'Conda', 'Python 3.7', ''CUDA 9.0 here.

Then you can run the following code to install PyTorch in your terminal (You may also install it in conda virtual environment if you like).

conda install pytorch torchvision cudatoolkit=9.0 -c pytorch

You can install htop to track your VM usage, it is more readable than 'top':

sudo apt-get install htop

To run htop, type htop' in your ssh terminal.

Install within conda virtual environment:

conda install scikit-learn

It should be a built-in function in your Python. Nevertheless, sometimes you may not be able to import it. Install within conda virtual environment:

conda install -c conda-forge progress