JuliaRL_BasicDQN_CartPole

using ReinforcementLearning
using StableRNGs
using Flux
using Flux.Losses

function RL.Experiment(
    ::Val{:JuliaRL},
    ::Val{:BasicDQN},
    ::Val{:CartPole},
    ::Nothing;
    seed = 123,
)
    rng = StableRNG(seed)
    env = CartPoleEnv(; T = Float32, rng = rng)
    ns, na = length(state(env)), length(action_space(env))

    policy = Agent(
        policy = QBasedPolicy(
            learner = BasicDQNLearner(
                approximator = NeuralNetworkApproximator(
                    model = Chain(
                        Dense(ns, 128, relu; init = glorot_uniform(rng)),
                        Dense(128, 128, relu; init = glorot_uniform(rng)),
                        Dense(128, na; init = glorot_uniform(rng)),
                    ) |> gpu,
                    optimizer = ADAM(),
                ),
                batch_size = 32,
                min_replay_history = 100,
                loss_func = huber_loss,
                rng = rng,
            ),
            explorer = EpsilonGreedyExplorer(
                kind = :exp,
                ϵ_stable = 0.01,
                decay_steps = 500,
                rng = rng,
            ),
        ),
        trajectory = CircularArraySARTTrajectory(
            capacity = 1000,
            state = Vector{Float32} => (ns,),
        ),
    )
    stop_condition = StopAfterStep(10_000, is_show_progress=!haskey(ENV, "CI"))
    hook = TotalRewardPerEpisode()
    Experiment(policy, env, stop_condition, hook, "# BasicDQN <-> CartPole")
end

using Plots
ex = E`JuliaRL_BasicDQN_CartPole`
run(ex)
plot(ex.hook.rewards)

                     Total reward per episode
             ┌────────────────────────────────────────┐
         200 │⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢸⢸⠉⢣⣷⠉⠉⠉⠉⠉⠉⠉⠉⠉⠉⠉⠉⠁│
             │⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢸⡇⠀⠀⠹⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀│
             │⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢰⡄⣸⡇⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀│
             │⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢸⢱⣿⡇⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀│
             │⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢸⢸⣿⡇⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀│
             │⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢸⠘⡟⡇⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀│
             │⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢠⢸⠀⡇⡇⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀│
   Score     │⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢠⣼⢸⠀⠁⠃⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀│
             │⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⣼⣿⡸⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀│
             │⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢀⠀⠀⣿⡏⠇⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀│
             │⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⣿⠀⢰⢹⡇⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀│
             │⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⡿⠖⠊⠘⡇⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀│
             │⠀⢀⠀⠀⠀⣇⡄⠀⠀⣀⢀⢠⡄⠀⡀⡇⠀⠀⠀⡇⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀│
             │⢢⠜⡆⣄⡜⢸⠛⣶⣇⠟⡜⣼⠣⢤⢻⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀│
           0 │⠀⠀⠈⠀⠀⠈⠀⠈⠈⠀⠃⠀⠀⠈⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀│
             └────────────────────────────────────────┘
             0                                       90
                              Episode

Watch a demo episode with the trained agent

demo = Experiment(ex.policy,
                  CartPoleEnv(),
                  StopWhenDone(),
                  RolloutHook(plot, closeall),
                  "DQN <-> Demo")
run(demo)

This page was generated using DemoCards.jl and Literate.jl.