r2l-core 0.0.3

A rust reinforcement learning library
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32

use std::marker::PhantomData;

use crate::{error::Result, models::Actor, tensor::R2lTensor};

#[derive(Debug, Clone)]
pub struct ActorWrapper<A: Actor + Clone, T: R2lTensor> {
    actor: A,
    env: PhantomData<T>,
}

impl<D: Actor + Clone, T: R2lTensor> ActorWrapper<D, T> {
    pub fn new(actor: D) -> Self {
        Self {
            actor,
            env: PhantomData,
        }
    }
}

impl<D: Actor + Clone, T: R2lTensor> Actor for ActorWrapper<D, T> {
    type Tensor = T;

    fn action(&self, observation: Self::Tensor) -> Result<Self::Tensor> {
        let action = self.actor.action(D::Tensor::convert(&observation)?)?;
        Ok(T::convert(&action)?)
    }

    fn mode_action(&self, observation: Self::Tensor) -> Result<Self::Tensor> {
        let action = self.actor.mode_action(D::Tensor::convert(&observation)?)?;
        Ok(T::convert(&action)?)
    }
}