Struct relearn::envs::meta::MetaObservation

pub struct MetaObservation<O, A, FI> {
    pub inner_observation: Option<O>,
    pub prev_step: Option<InnerStepObs<A, FI>>,
    pub episode_done: bool,
}

An observation from a meta enviornment.

Note

The generic feedback type FI is the inner feedback E::Feedback::Inner for MetaEnv<E>.

Fields

inner_observation: Option<O>

The current inner observation on which the next step will act.

Is None if this is a terminal state, in which case episode_done must be True.

prev_step: Option<InnerStepObs<A, FI>>

Observation of the previous inner step.

Is None if this is the first step of an inner episode.

episode_done: bool

Whether the previous step ended the inner episode.

Trait Implementations

impl<TC, OS, AS, FS> Actor<MetaObservation<<OS as Space>::Element, <AS as Space>::Element, <FS as Space>::Element>, <AS as Space>::Element> for ResettingMetaAgent<TC, OS, AS, FS> where
    TC: BuildAgent<OS, AS, FS>,
    TC::Agent: BatchUpdate<OS::Element, AS::Element, Feedback = FS::Element>,
    OS: Space + Clone,
    AS: NonEmptySpace + Clone,
    FS: Space + Clone, 

type EpisodeState = InnerEpisodeState<<TC as BuildAgent<OS, AS, FS>>::Agent, <OS as Space>::Element, <AS as Space>::Element>

Stores state for each episode.

fn initial_state(&self, rng: &mut Prng) -> Self::EpisodeState

Create the initial actor state for the start of a new episode.

fn act(
    &self,
    state: &mut Self::EpisodeState,
    obs: &MetaObservation<OS::Element, AS::Element, FS::Element>,
    rng: &mut Prng
) -> AS::Element

Select an action in response to an observation.

impl<TC, OS, AS, FS> Agent<MetaObservation<<OS as Space>::Element, <AS as Space>::Element, <FS as Space>::Element>, <AS as Space>::Element> for Arc<ResettingMetaAgent<TC, OS, AS, FS>> where
    TC: BuildAgent<OS, AS, FS>,
    TC::Agent: BatchUpdate<OS::Element, AS::Element, Feedback = FS::Element>,
    OS: Space + Clone,
    AS: NonEmptySpace + Clone,
    FS: Space + Clone, 

type Actor = Arc<ResettingMetaAgent<TC, OS, AS, FS>>

Produces actions for a sequence of environment observations.

fn actor(&self, _: ActorMode) -> Self

Create a new Actor with the given behaviour mode.

impl<TC, OS, AS, FS> BatchUpdate<MetaObservation<<OS as Space>::Element, <AS as Space>::Element, <FS as Space>::Element>, <AS as Space>::Element> for Arc<ResettingMetaAgent<TC, OS, AS, FS>> where
    OS: Space,
    AS: Space,
    FS: Space, 

No updates at the meta-level.

type Feedback = <FS as Space>::Element

Environment feedback type

type HistoryBuffer = NullBuffer

fn buffer(&self) -> Self::HistoryBuffer

Create a new history buffer.

fn min_update_size(&self) -> HistoryDataBound

Request a minimum amount of on-policy data for the next batch update.

fn batch_update<'a, I>(&mut self, _buffers: I, _logger: &mut dyn StatsLogger) where
    I: IntoIterator<Item = &'a mut Self::HistoryBuffer>,
    Self::HistoryBuffer: 'a, 

Update the agent from a collection of history buffers.

impl<O: Clone, A: Clone, FI: Clone> Clone for MetaObservation<O, A, FI>

fn clone(&self) -> MetaObservation<O, A, FI>

Returns a copy of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<O: Debug, A: Debug, FI: Debug> Debug for MetaObservation<O, A, FI>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<O: PartialEq, A: PartialEq, FI: PartialEq> PartialEq<MetaObservation<O, A, FI>> for MetaObservation<O, A, FI>

fn eq(&self, other: &MetaObservation<O, A, FI>) -> bool

This method tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &MetaObservation<O, A, FI>) -> bool

This method tests for !=.

impl<O: Copy, A: Copy, FI: Copy> Copy for MetaObservation<O, A, FI>

impl<O, A, FI> StructuralPartialEq for MetaObservation<O, A, FI>