GCP from Scratch

Contents

• Create Your Instance
• Basics
• CUDA and cuDNN
• Miniconda and other packages
• Tensorflow

Create Your Instance

1. Please follow GCP Setup instructions to 'select image' part.
2. In boot disk, instead of custom image, select 'Ubuntu 16.04 LTS' from 'OS images'.

Basics

1. Update & upgrade System

   sudo apt-get update 
   sudo apt-get upgrade
   

2. Install python 3.6, pip, gcc, etc.

   sudo add-apt-repository ppa:deadsnakes/ppa
   sudo apt-get update
   sudo apt-get install python3.6
   wget https://bootstrap.pypa.io/get-pip.py
   sudo python3.6 get-pip.py
   sudo apt-get install build-essential
   sudo apt-get install git zip unzip
   

3. Verify GPU, Linux version, kernel headers and development packages

   lspci | grep -i nvidia
   uname -m && cat /etc/*release
   uname -r
   sudo apt-get install linux-headers-$(uname -r)
   

CUDA and cuDNN

1. Install CUDA 10.0

   wget https://developer.nvidia.com/compute/cuda/10.0/Prod/local_installers/cuda-repo-ubuntu1604-10-0-local-10.0.130-410.48_1.0-1_amd64
   sudo dpkg -i cuda-repo-ubuntu1604-10-0-local-10.0.130-410.48_1.0-1_amd64
   sudo apt-key add /var/cuda-repo-10-0-local-10.0.130-410.48/7fa2af80.pub
   sudo apt-get update
   sudo apt-get install cuda
   echo 'export PATH=/usr/local/cuda-10.0/bin${PATH:+:${PATH}}' >> ~/.bashrc 
   echo 'export LD_LIBRARY_PATH=/usr/local/cuda-10.0/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}' >> ~/.bashrc 
   source ~/.bashrc
   

2. Install cuDNN v7.5.0

   Download cuDNN v7.5.0 from NVIDIA as in Local Setup. Note that this time you are installing cuDNN on virtual machine instance with Linux.

   gcloud compute scp [LOCAL_FILE_PATH] ecbm4040@your-instance-name:
   cp cudnn-10.0-linux-x64-v7.5.0.56.solitairetheme8 cudnn-10.0-linux-x64-v7.5.0.56.tgz
   tar xvf cudnn-10.0-linux-x64-v7.5.0.56.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*
   

3. Check CUDA installation. You can use "nvcc -V" to check the version of CUDA toolkits. And "nvidia-smi" can help you check availble GPU device.

Miniconda and other packages

1. Download Miniconda

   wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh
   

2. Install Miniconda

   bash Miniconda3-latest-Linux-x86_64.sh
   source ~/.bashrc
   

3. Create your own virtual environment in Miniconda with Python 3.6

   conda create -n envTF113 python=3.6
   

4. Activate the virtual environment.

   source activate envTF113
   

5. Install baisc packages.

   conda install pandas numpy scipy pillow matplotlib scikit-learn
   conda install -c conda-forge jupyterlab 
   

Tensorflow

1. Use pip to install tensorflow-gpu.

   pip install tensorflow-gpu==1.13.1
   

2. Open python and try to run a simple tensorflow function.
3. (Optional) If you want to install tensorflow 2.0 instead of the lower versions, you can also use pip command:

   pip install --upgrade tensorflow

   Note: By installing version 2.0, you need to go for *_tf2.0 assignments.

Now you can proceed to Step 3 in GCP Setup.