Skip to main content

​
Introduction to sweeps

In many several, you might find yourself having to define parameters for a function that are not yet know unknown or even inherently uncertain. In these cases, it is usually beneficial (instead of guessing their value) to parametrize and run the function for several parameter choices to examine the impact on the output. Some common use cases for sweeps are:
  • Running hyperparameter optimization search for training a machine learning model
  • Running a sensitivity analysis of a simulation using multiple input scenarios
  • Running an evaluation to benchmark multiple models against the same objective

​
Defining and running sweeps

from rebase.pipelines import pipeline

@pipeline
def function(a, b, c): 
    ...

params = [{"a": 1, "b": 2, "c": 3}]

rb.sweep(function, params)