ScheduledMidpointStartConfiguration
Fixed step-scheduled midpoint-start probability (the original behaviour).
Methods
validate_probability
python
validate_probability(value: Any) -> ScheduleValues
kind
python
kind: Literal['scheduled'] = 'scheduled'probability
python
probability: Schedule = Field(default_factory=lambda: Schedule.starting(1.0))Probability that one policy episode starts from a stored midpoint instead of the true start. The complement keeps a fraction of policy episodes practising the tester's real from-start task. Random and replay episodes are unaffected.