Skip to content

On-premise users: click in-app to access the full platform documentation for your version of DataRobot.

Model Info

To display model information, click a model on the Leaderboard list then click Model Info. The tab provides tiles that report general model and performance information.

For time series models, backtest information also appears in the model information.

The output displays the following information:

Field Description
Model File Size Reports the sum total of the cache files DataRobot uses to store the model data. It's generated from an internal storage mechanism and indicates your system footprint, which can be especially useful for Self-Managed AI Platform deployments.
Prediction Time Displays the estimated time, in seconds, to score 1000 rows of the dataset.
Sample Size Reports the number of observations used to train and validate the model (and also for each cross-validation repetition, if applicable). When smart downsampling is in play or DataRobot has downsampled the project, Sample Size reports the number of rows in the minority class rather than the total number of rows used to train the model.
Max RAM Reports the maximum amount of RAM this model used during the training.
Cache Time Savings Displays any time savings benefits achieved by this model from leveraging earlier training. DataRobot will reuse blueprint vertices trained beforehand when possible.

Time series info

For time series projects, the output also includes backtesting information, including execution time for each backtest.

The backtest summaries show partitioning against the full date range:

  • Date ranges the model is trained on (blue)
  • Validation (green) and Holdout (red) partitions
  • Any configured gaps (yellow)

Training periods can be changed by clicking the plus sign next to a model on the Leaderboard.

If a model uses specified start and end dates note that:

  • Insights on validation and holdout sets are only available if the model was not trained into those sets (if data is out-of-sample).
  • If any partitions remain out-of-sample for the model, insights are provided for that partition.
  • If any partition is wholly or partially in the training period for the model, insights are not provided for that partition.

Updated November 21, 2023