My customer had built lots of Python model at CDSW, Cloudera Data Science Workbench, and we engage them with MLOps on-premise recently. There are some questions from them about MLOps:
1.Does it have any scheduling feature to trigger model retraining and prediction which the model is deployed in MLOps?
2.Can model retrain and predictive use the data from database?
Solved! Go to Solution.
Hey @Silver Su, following up on your question MLOps management agents allow you to scheduler retraining existing customer models.
The custom model workshop does allow for existing models to be uploaded into DataRobot with MLOps monitoring on deployment. Likewise if you want to monitor a model in an external environment MLOps agents can do the same. Scheduler retraining could definitely be achieved using a script with the DataRobot API. Though I'm not sure if you can schedule retrain these kind of deployments from the UI.
I haven't tried this before in the UI, I'm sure the rest of the community can give a more full answer.
One more question about how to deploy customer existing Python Model into MLOps, it also could be triggered by scheduler?
Thank you so much.
If you want any part of these processes clarified further feel free to ask
Hey @Silver Su, yes to both questions. Automated model monitoring and retraining is a big part of DataRobot's Continuous AI, the retraining can be triggered by policies you define within your deployment. The AI Catalog allows you to connect to numerous database's (see below). The AI Catalog can be used throughout the project from training to uploading actuals to monitor the deployment.