Experiments

Model training is a key component of the ML lifecycle. It is typically an iterative process, where values used to configure a model, or hyper-parameters, are repeatedly fine-tuned until either model performance or learning-rate reaches a certain point.

MarkovML Experiments let you record data about your model training sessions, and get insights about how a customizable set of metrics changes with respect to model training time.
When you record a model training session using the MarkovML SDK, you can specify any relevant hyper-parameters used for the model training session, as well as any custom metrics you wish to track. See the Track Experiments guide for more details.