Which Google Cloud Platform service is the easiest to run Tensorflow?

Question

While working on the Udacity Deep Learning assignments, I had a memory issue. I need to switch to a cloud platform. I used to work with AWS EC2, but now I would like to try the Google Cloud Platform (GCP). I will need at least 8 GB of memory. I know how to use docker locally, but have never tried it in the cloud.

• Is there a turnkey solution for running Tensorflow on GCP?
• If not, which service (compute engine or container engine) will simplify the job?
• Any other advice is also welcome!

Answer 1

Summing up the answers:

• Datalab
• Cloud ML
• Manual installation on a computing engine. See instructions below.

Step-by-step instructions for running TensorFlow on a compute processor:

• Create project
• Open Cloud Shell (button at the top)
• List of machine types: gcloud compute machine-types list . You can change the type of machine that I used in the following command.
• Create an instance:

 gcloud compute instances create tf \ --image container-vm \ --zone europe-west1-c \ --machine-type n1-standard-2 

4. Run sudo docker run -d -p 8888:8888 --name tf b.gcr.io/tensorflow-udacity/assignments:0.5.0 (change the image name to the desired one)
5. Find your instance in the dashboard and edit the default network.
6. Add a firewall rule to resolve your IP address, as well as the tcp:8888 protocol and port tcp:8888 .
7. Locate the external IP address of the instance in Control Panel. Open IP:8888 in your browser. Done!
8. When you are finished, delete the created cluster to avoid failures.

Here's how I did it, and it worked. I am sure there is an easier way to do this.

Additional resources

You may be interested to learn more about:

• Google cloud shell
• Container-optimized Google Compute Engine images
• Google Cloud SDK for a more responsive shell, etc.

Good to know

• "The contents of your Cloud Shell home directory are saved in projects between all Cloud Shell sessions, even after the virtual machine shuts down and restarts.
• To view all available versions of images: gcloud compute images list --project google-containers

Thanks @ user728291, @MattW. and @CJCullen.

Answer 2

Google Cloud Machine Learning is open to the world in beta today. It provides TensorFlow as a service, so you don’t need to manage machines and other raw resources. As part of the beta, Datalab has been updated to provide commands and utilities for machine learning. Check it out: http://cloud.google.com/ml .

Answer 3

Google has the Cloud ML platform in Alpha Limited.

Here 's a blog post and tutorial about launching TensorFlow on the Kubernetes / Google Container Engine.

If this is not what you want, TensorFlow tutorials should be able to work on both AWS EC2 and the Google Compute Engine.

Answer 4

I'm not sure if you need to stay on the Google Cloud platform. If you can use other products, you can save a lot of time and some money.

If you are using TensorFLow, I would recommend a platform called TensorPort . This is exclusively for TesnorFlow and is the simple platform that I know of. The code and data are downloaded using git, and they provide a python module for automatically switching paths between the remote and local machines. They also provide some boiler plate code for setting up distributed computing if you need it. Hope this helps.