CODEWITHSUNDEEP
svmgoogle-cloud-datalab
Praxiteles
Praxiteles

Reputation: 6020

How do we increase the CPUs available for doing SVMs in Google Cloud DataLabs?

We have a database of ~70,000 rows and ~7 features for which we are trying to do a SVM based prediction. The service ran for 4-5 hours before stopping (with no results.) How do we scale up or down the number of CPUs doing the calculation so this might run in minutes as opposed to hours?

Upvotes: 0

Views: 84

Answers (1)

Graham Wheeler
Graham Wheeler

Reputation: 2814

This depends very much on how you are doing this.

Firstly, note that Python is single-threaded and so throwing CPUs at the problem may make no difference unless you are doing this in some way that can take advantage of them.

Secondly, the managed VM that runs your Datalab instance is configured at creation time to have a certain number of cores, memory, etc. This does not scale dynamically. You would need to create a more powerful instance at deployment time. See the docs here: https://cloud.google.com/datalab/getting-started#custom_cloud_datalab_deployment

Upvotes: 3

Related Questions