863b8dee69df21ff479b0f28422f2bf2b14f05bd,examples/quickstart.py,,,#,29

Before Change

env = OpenAIGym("CartPole-v0")

// Create a Trust Region Policy Optimization agent
agent = PPOAgent(
    Configuration(
    log_level="info",
    batch_size=4000,

    // max_kl_divergence=0.1,
    // cg_iterations=20,
    // cg_damping=0.001,
    // ls_max_backtracks=10,
    // ls_accept_ratio=0.9,
    // ls_override=False,

    learning_rate=0.001,
    entropy_penalty=0.01,
    epochs=5,
    optimizer_batch_size=512,
    loss_clipping=0.2,
    normalize_advantage=False,
    baseline=dict(
        type="mlp",
        sizes=[32, 32],
        epochs=1,
        update_batch_size=512,
        learning_rate=0.01
    ),
    states=env.states,
    actions=env.actions,
    network=layered_network_builder([
        dict(type="dense", size=32, activation="tanh"),
        dict(type="dense", size=32, activation="tanh")
    ])
))

// Create the runner
runner = Runner(agent=agent, environment=env)

After Change

    dict(type="dense", size=32, activation="tanh")
]

agent = PPOAgent(
    states_spec=env.states,
    actions_spec=env.actions,
    network_spec=network_spec,
    config=config
)

// Create the runner
runner = Runner(agent=agent, environment=env)

Instances