rwrd — Rust Reinforcement Learning Framework

A reinforcement learning framework for Rust, powered by Burn.

Aligned with Stable-Baselines3: each algorithm is self-contained with learn() / predict() / save() / load() as the primary API.

Quick start

use rwrd::prelude::*;

let cfg = DqnConfig::new(4, 2).with_lr(5e-4);
let mut dqn = Dqn::<B>::new(cfg, &device)?;
dqn.learn(&mut env, 200_000)?;
let action = dqn.predict(&obs); // deterministic