Module ai.rl.dqn.rainbow.trainers.basic

Basic sequential trainer. A training step is executed every _n_th environment step.

"""Basic sequential trainer. A training step is executed every _n_th environment
step."""


from ._trainer import Trainer
from ._config import Config


__all__ = ["Trainer", "Config"]

Classes

class Config

Trainer config.

class Trainer (agent: Agent, config: Config, environment: Factory)

Trainer class.

Args

agent : rainbow.Agent

Agent.

config : Config

Trainer configuration.

environment : environments.Factory

Environment factory.

Methods

def start(self)

Starts training, according to the configuration.

Expand source code

def start(self):
    """Starts training, according to the configuration."""

    port = self._logging_server.start()
    self._logging_client = logging.Client("127.0.0.1", port)
    self._agent.set_logging_client(self._logging_client)

    env = self._env_factory()
    try:
        self._run(env)
    finally:
        env.close()