I would like to create a scenario which allows to train a machine learning algorithm. I would like to restart it if the score is below a certain threshold.
Currently the scenario drives the model, but I cannot create the launch condition. How can I do it?
you can do this :
- either the score you compute is the product or by-product of some recipe run by some scenario, and then you should add in that scenario a step to launch the other scenario (the one that trains the model), where the step would be conditioned on the score
- or you create a custom trigger on the scenario that trains the model, where the custom triggers pulls the score and decides to fire or not
You can get the metrics of saved models through https://doc.dataiku.com/dss/latest/python-api/saved_models.html#dataiku.Model.get_version_metrics