From the course: Data Science on Google Cloud Platform: Predictive Analytics

Unlock the full course today

Join today to access over 22,600 courses taught by industry experts or purchase this course individually.

Local testing

Local testing - Google Cloud Tutorial

From the course: Data Science on Google Cloud Platform: Predictive Analytics

Start my 1-month free trial

Local testing

“

- [Instructor] It's better to use local, fixed-cost, or low cost resources as much as possible when using GCP. What are some of the tips to do it? Well, try to use as small a dataset as possible for model building. The smaller the dataset, the lesser the resource requirements. You need to have a large enough dataset to build or create models, but you can grow the dataset after the initial trials. Use non-GCP resources like laptops. Today's laptops can handle significant amounts of data locally, so use them first. Firm up the model before moving to GCP. Use compute engine resources as just VMs for model building and testing. This is outside of Cloud ML, and tends to be cheaper than using Cloud ML directly. Contact us to ensure model consistency before deploying them on Cloud ML. Cloud ML should be used as the main platform for predictions, not trials and trainings. If you focus on working out the small kinks locally first, you can scale to the cloud without a huge hit to your checkbook.

Contents