Create ability to obtain performance metrics from scoring large datasets

Description

User has a very large dataset that can't run `.predict()` on whole dataset with their memory. They can do batch scoring, but would like to get performance metrics for the whole dataset (and can only do batches)

Two ideas:

  1. Create new function that does batch predictions on large dataset (for every batch clear the memory of the features but retain the predictions for performance metrics)

  2. Be able to append predictions to a saved file, and once all batches are scored, then read in the predictions to build your desired performance metrics

Your pinned fields
Click on the next to a field label to start pinning.

Assignee

New H2O Bugs

Reporter

Neema Mashayekhi

Support ticket URL