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

Before Change

// 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

env = OpenAIGym("CartPole-v0")


config = Configuration(
    batch_size=4096,
    // Agent
    preprocessing=None,
    exploration=None,
    reward_preprocessing=None,
    // BatchAgent
    keep_last_timestep=True,
    // PPOAgent
    step_optimizer=dict(
        type="adam",
        learning_rate=1e-3
    ),
    optimization_steps=10,
    // Model
    scope="ppo",
    discount=0.99,
    // DistributionModel
    distributions=None,  // not documented!!!
    entropy_regularization=0.01,
    // PGModel
    baseline_mode=None,
    baseline=None,
    baseline_optimizer=None,
    gae_lambda=None,
    normalize_rewards=False,
    // PGLRModel
    likelihood_ratio_clipping=0.2,
    // Logging
    log_level="info",
    // TensorFlow Summaries
    summary_logdir=None,
    summary_labels=["total-loss"],
    summary_frequency=1,
    // Distributed
    // TensorFlow distributed configuration
    cluster_spec=None,
    parameter_server=False,
    task_index=0,
    device=None,
    local_model=False,
    replica_model=False,
)

// Network as list of layers
network_spec = [
    dict(type="dense", size=32, activation="tanh"),

Instances