Struct entity_gym_rs::agent::Obs

pub struct Obs { /* private fields */ }

An observation that defines what an agent can see.

The Obs::new method creates an observation. It takes a single required parameter, the current score, which is maximized during training.

use entity_gym_rs::agent::Obs;

let obs = Obs::new(0.0);

An observation can contain any number of Featurizable entities.

use entity_gym_rs::agent::{Obs, Featurizable};

#[derive(Featurizable)]
struct Player { x: i32, y: i32 }

#[derive(Featurizable)]
struct Cake { is_a_lie: bool }

let obs = Obs::new(0.0)
    .entities([Player { x: 0, y: -3 }])
    .entities([Cake { is_a_lie: true }, Cake { is_a_lie: false }]);

Implementations

impl Obs

pub fn new(score: f32) -> Self

Creates a new observation.

Arguments

• score - The current score of the agent. This is maximized during training.

pub fn entities<E: Featurizable, I: IntoIterator<Item = E>>(
    self,
    entities: I
) -> Self

Adds a set of entities to the observation.

Arguments

• entities - An iterator of Featurizable entities to add to the observation.

pub fn metric(self, name: &str, value: f32) -> Self

Adds a numerical metric to the observation. Aggregate statistics of all metrics are surfaced during training.

Arguments

• name - The name of the metric.
• value - The value of the metric.

Example

use entity_gym_rs::agent::Obs;

let mut obs = Obs::new(0.0)
    .metric("experience-collected", 2132.4)
    .metric("game-over-reason/touched-enemy", 1.0);

pub fn score(&self) -> f32

Returns the score.