Struct Transition 

pub struct Transition {
    pub obs: Observation,
    pub reward: f64,
    pub terminated: bool,
    pub truncated: bool,
    pub info: Option<HashMap<String, f64>>,
}

A single environment transition returned by step.

Precision convention

Rewards are f64 for numerical stability during advantage computation; observations and actions are f32 throughout (see Observation). When storing into replay buffers (which use f32 rewards), a narrowing cast occurs. This is intentional: environments compute in f64, buffers store in f32, and training reads f32.

The info field is None when the environment provides no extra metadata (the common case for CartPole, Pendulum, etc.), avoiding a HashMap allocation on every step.

Fields

obs: Observation

reward: f64

terminated: bool

truncated: bool

info: Option<HashMap<String, f64>>

Trait Implementations

impl Clone for Transition

fn clone(&self) -> Transition

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Transition

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

Formats the value using the given formatter.