module emote.memory.fifo_strategy
Classes
class FifoStrategyBase(Strategy):
A sampler intended to sample in a first-in-first-out style across the whole set of experiences.
This base class is used by both the fifo sample and ejection strategies.
Methods
def __init__(self) -> None
Create a FIFO-based strategy.
def track(self, identity, sequence_length) -> None
def forget(self, identity) -> None
def post_import(self) -> None
def state(self) -> dict
Serialize the strategy to a JSON-serializable dictionary.
def load_state(self, state) -> None
Load the strategy from a dictionary.
Arguments:
state(dict)
class FifoSampleStrategy(FifoStrategyBase, SampleStrategy):
Methods
def __init__(self, per_episode, random_offset) -> None
Create a FIFO-based sample strategy.
Arguments:
per_episode(bool): if true, will only sample each episode once in a single pass (default: True)random_offset(bool): if true will sample at a random offset in each episode. Will be assumed true when sampling per episode (default: True)
def sample(self, count, transition_count) -> Sequence[SamplePoint]