pg.early_stopping.early_stop_by_value

Accessible via pg.early_stopping.early_stop_by_value.

class early_stop_by_value(step_values, metric='reward', maximize=True)

Bases: pg.Functor

Step-wise early stopping policy based on the value of reward/metric.

Example:

policy = early_stop_by_value([
  # Stop at step 1 if trial reward is less than 0.2.
  (1, 0.2),

  # Stop at step 2 if trial reward is less than 0.8.
  (2, 0.8),
])()
Parameters:

  • step_values –

    A list of tuple (gating step, value threshold). gating step - At which step this rule will be triggered. value threshold - A float number indicating the threshold value for

    early stopping.

  • metric – Based on which metric the value should be compared against. Use str for metric name or a callable object that takes a measurement object at a given step as input and returns a float value.

  • maximize – If True, reward or metric value below the threshold will be stopped, otherwise trials with values above the threshold will be stopped.

Returns:

A StepWise early stopping policy.