Struct RogueNet

pub struct RogueNet {
    pub config: RogueNetConfig,
    pub obs_space: ObsSpace,
    /* private fields */
}

Implements the RogueNet entity neural network.

Fields

config: RogueNetConfig

obs_space: ObsSpace

Implementations

impl RogueNet

pub fn load<P: AsRef<Path>>(path: P) -> RogueNet

Loads the parameters for a trained RogueNet neural network from a checkpoint directory produced by enn-trainer.

Arguments

• path - Path to the checkpoint directory.

pub fn load_archive<R: Read>(r: R) -> Result<RogueNet, Error>

Loads the parameters for a trained RogueNet neural network from a tar archive of a checkpoint directory.

Arguments

• r - A reader for the tar archive.

Example

use std::fs::File;
use rogue_net::RogueNet;

let rogue_net = RogueNet::load_archive(File::open("test-data/simple.roguenet").unwrap());

pub fn forward(&self, args: FwdArgs) -> (Array2<f32>, Vec<u64>)

Runs a forward pass of the RogueNet neural network.

Arguments

• entities - Maps each entity type to an Array2<f32> containing the entities’ features.

Example

use std::collections::HashMap;
use ndarray::prelude::*;
use rogue_net::{RogueNet, FwdArgs};

let rogue_net = RogueNet::load("test-data/simple");
let mut features = HashMap::new();
features.insert("Head".to_string(), array![[3.0, 4.0]]);
features.insert("SnakeSegment".to_string(), array![[3.0, 4.0], [4.0, 4.0]]);
features.insert("Food".to_string(), array![[3.0, 5.0], [8.0, 4.0]]);
let (action_probs, actions) = rogue_net.forward(FwdArgs { features, ..Default::default() });

pub fn with_obs_filter(self, obs_space: HashMap<String, Vec<String>>) -> Self

Adapts the RogueNet neural network to the given observation space by filtering out any features that were not present during training.

Trait Implementations

impl Clone for RogueNet

fn clone(&self) -> RogueNet

Returns a copy of the value.

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

Performs copy-assignment from source.

impl Debug for RogueNet

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

Formats the value using the given formatter.