Prediction service

The machine learning model that was trained in the previous section can be hosted on the cloud for consumption and prediction, based on a new set of data. The prediction service is primarily responsible for managing the storage and compute resources required for the predictions. As a generic process, the training artifacts need to be exported. The export process is the same for the models trained on the AI Platform as well as the models that are trained outside. The maximum model size that can be exported is 250 MB. If a larger custom model needs to be exported, a quota increase form needs to be submitted. The AI Platform limits the use of compute infrastructure so that the platform can be used at a reasonable level, which ...

Get Hands-On Artificial Intelligence on Google Cloud Platform now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.