cancel
Showing results for 
Search instead for 
Did you mean: 

Retraining a model without creating a new project

Jayant
NiCd Battery

Retraining a model without creating a new project

Hello colleagues,

 

Presently, I have a trained classification model running on Datarobot platform. This model was trained on a snapshot train data housed in AI catalog. This train dataset is fed from a view in snowflake.

 

Next, we plan to add a new feature to this train data and retrain the model. I'll need to redefine the view defn. in snowflake so as to add that feature. The way I plan to go about is as follows;

  1. Modify view defn. in snowflake with the  new feature added.
  2. Make a new entry in AI catalog using datarobot snowflake connection. Make sure it is snapshotted.
  3. Create a new project using new dataset, and do the model training in this project.

But above steps seem redundant to me. May I know if I can do a workaround and have an efficient alternative. Help or link to doc is appreciated.

 

Labels (1)
1 Reply
IraWatt
Laser

Hi @Jayant,

Assuming you have deployed one of the models from this project DataRobot's Continuous AI would be the way to retrain on new data within the same deployment.