cancel
Showing results for 
Search instead for 
Did you mean: 

Cannot run DataRobot Prime on a model that has been...

Cannot run DataRobot Prime on a model that has been...

Hi everyone

 

First time user of DataRobot. I've built the model using the DataRobot auto ML, and pretty happy with the accuracy of the recommended model for deployment.

I'd like to download the model locally and run on my desktop, however when I go download > DataRobot Prime > Run DataRobot Prime, I'm getting the following error:

 

YeahNah_0-1661739972155.png

 

It will only let me run DataRobot Prime on my model with 64% of data used for training.

 

YeahNah_1-1661740090734.png

 

The overall accuracy difference isn't huge between the top 3 models, but was wondering if someone could explain what the error means?

 

Cheers

Labels (1)
0 Kudos
1 Solution

Accepted Solutions

DataRobot Prime needs validation data. You can run it on the 64% model, but you can also increase the size to 80, 90% or whatever your needs are by starting a new project, and setting the holdout to 0% allows you to get the maximum amount of data for training DR Prime models. It is normally recommended that at least 10% of the data be used for validation (very large datasets may use less), or the model results may be unstable. 

View solution in original post

0 Kudos
2 Replies

Hello Yeah-Nah!

The behavior you see is the correct way of the DataRobot Prime model. Compared to the teacher model, the Prime model may be trained only if it has extra data to train the student model. So 100% doesn't leave any data, that the teacher model hasn't seen. So does the 80% teacher model, because holdout data is unreachable during training. 
So as a result,  your error means: "student model doesn't have data to train on"

0 Kudos

DataRobot Prime needs validation data. You can run it on the 64% model, but you can also increase the size to 80, 90% or whatever your needs are by starting a new project, and setting the holdout to 0% allows you to get the maximum amount of data for training DR Prime models. It is normally recommended that at least 10% of the data be used for validation (very large datasets may use less), or the model results may be unstable. 

0 Kudos