0

I need a packaged Virtual Machine on Google Cloud with the following installed already:

  1. Python (2.x, 3.x, Anaconda distro)
  2. R
  3. A database server (PostGreSQL etc)
  4. Ipython/Jupyter/JupyterHub
  5. Deep Learning Libraries (Tensorflow, Keras and GPU support CUDANN)

These are usually available in Amazon's AMI or MS Azure's Data Science Instances.

Links: 1. Amazon's AMI! 2. Microsoft's VM!

Could someone experienced please guide me in finding the right VM, if such a VM is indeed available on Google Cloud Platform.

Mike Wise
  • 22,131
  • 8
  • 81
  • 104
Deepak Sadulla
  • 373
  • 2
  • 12
  • 1
    Just build your image with all the packages you need. – Tatsuyuki Ishi Apr 16 '17 at 07:36
  • Building it up from scratch especially the deep learning libraries with CUDANN seems like a daunting task for a **novice** (like me) as if swimming in a sea of version/driver conflicts and what not. I'd rather stick to pre configured instances from Google if possible. If MS and AWS provide it, I presumed there would be something available on GCP already. – Deepak Sadulla Apr 16 '17 at 07:42

2 Answers2

2

The Closest that comes to your requirement google datalab though it does not satisfy all, https://cloud.google.com/datalab/

If you want more control over your VM this article will help you a great deal

https://haroldsoh.com/2016/04/28/set-up-anaconda-ipython-tensorflow-julia-on-a-google-compute-engine-vm/

If not as a workaround you can, install docker on your VM,

For R and PostGreSQL pull the prebuilt dockers and follow the simple instructions in dockerhub

https://hub.docker.com/_/r-base/

https://hub.docker.com/_/postgres/

Go to the following repo and get the most suitable docker for your VM, instructions are added on git as well,

https://github.com/floydhub/dl-docker

This includes the following,

CUDA 8.0 (GPU version only) cuDNN v5 (GPU version only) Tensorflow Caffe Theano Keras Lasagne Torch (includes nn, cutorch, cunn and cuDNN bindings) iPython/Jupyter Notebook (including iTorch kernel) Numpy, SciPy, Pandas, Scikit Learn, Matplotlib OpenCV and A few common libraries used for deep learning

This should be enough to satisfy your requirement.

Maddy
  • 996
  • 8
  • 18
Kalana Demel
  • 3,220
  • 3
  • 21
  • 34
  • Thank you for suggesting the docker route to my question. I have never tried dockers myself, I understand the method involved. I will give this a go. – Deepak Sadulla Apr 17 '17 at 06:42
  • Please humor me on this, would it be possible to create a docker from AWS's / Azure's Data Science VMs and use the same on Google's VM? – Deepak Sadulla Apr 17 '17 at 06:43
  • The simpler solution would be to migrate your existing VM https://cloud.google.com/compute/docs/tutorials/migrating-vms-compute-engine and if this anwer provider you with a solution can you mark it as the correct one :D – Kalana Demel Apr 17 '17 at 07:31
  • Thanks!! Much appreciated!! – Deepak Sadulla Apr 17 '17 at 08:14
0

GCP now has own DeepLearning images for GCE. You can find information about them here.

Viacheslav V Kovalevskyi
  • 868
  • 1
  • 7
  • 21