Handler

fugue::runtime::handler

Trait Handler 

Source 
pub trait Handler {
    // Required methods
    fn on_sample_f64(
        &mut self,
        addr: &Address,
        dist: &dyn Distribution<f64>,
    ) -> f64;
    fn on_sample_bool(
        &mut self,
        addr: &Address,
        dist: &dyn Distribution<bool>,
    ) -> bool;
    fn on_sample_u64(
        &mut self,
        addr: &Address,
        dist: &dyn Distribution<u64>,
    ) -> u64;
    fn on_sample_usize(
        &mut self,
        addr: &Address,
        dist: &dyn Distribution<usize>,
    ) -> usize;
    fn on_observe_f64(
        &mut self,
        addr: &Address,
        dist: &dyn Distribution<f64>,
        value: f64,
    );
    fn on_observe_bool(
        &mut self,
        addr: &Address,
        dist: &dyn Distribution<bool>,
        value: bool,
    );
    fn on_observe_u64(
        &mut self,
        addr: &Address,
        dist: &dyn Distribution<u64>,
        value: u64,
    );
    fn on_observe_usize(
        &mut self,
        addr: &Address,
        dist: &dyn Distribution<usize>,
        value: usize,
    );
    fn on_factor(&mut self, logw: f64);
    fn finish(self) -> Trace
       where Self: Sized;
}
Expand description

Core trait for interpreting probabilistic model effects.

Handlers define how to interpret the three fundamental effects in probabilistic programming: sampling, observation, and factoring. Different implementations enable different execution modes.

Example:


// Use a built-in handler
let mut rng = StdRng::seed_from_u64(42);
let handler = PriorHandler {
    rng: &mut rng,
    trace: Trace::default()
};
let model = sample(addr!("x"), Normal::new(0.0, 1.0).unwrap());
let (result, trace) = runtime::handler::run(handler, model);

Required Methods§

Source

fn on_sample_f64(&mut self, addr: &Address, dist: &dyn Distribution<f64>) -> f64

Handle an f64 sampling operation (continuous distributions).

Source

fn on_sample_bool( &mut self, addr: &Address, dist: &dyn Distribution<bool>, ) -> bool

Handle a bool sampling operation (Bernoulli).

Source

fn on_sample_u64(&mut self, addr: &Address, dist: &dyn Distribution<u64>) -> u64

Handle a u64 sampling operation (Poisson, Binomial).

Source

fn on_sample_usize( &mut self, addr: &Address, dist: &dyn Distribution<usize>, ) -> usize

Handle a usize sampling operation (Categorical).

Source

fn on_observe_f64( &mut self, addr: &Address, dist: &dyn Distribution<f64>, value: f64, )

Handle an f64 observation operation.

Source

fn on_observe_bool( &mut self, addr: &Address, dist: &dyn Distribution<bool>, value: bool, )

Handle a bool observation operation.

Source

fn on_observe_u64( &mut self, addr: &Address, dist: &dyn Distribution<u64>, value: u64, )

Handle a u64 observation operation.

Source

fn on_observe_usize( &mut self, addr: &Address, dist: &dyn Distribution<usize>, value: usize, )

Handle a usize observation operation.

Source

fn on_factor(&mut self, logw: f64)

Handle a factor operation.

This method is called when the model encounters a factor operation. The handler typically adds the log-weight to the trace.

§Arguments
  • logw - Log-weight to add to the model’s total weight
Source

fn finish(self) -> Trace
where Self: Sized,

Finalize the handler and return the accumulated trace.

This method is called after model execution completes to retrieve the final trace containing all choices and log-weights.

Implementors§