Custom Keras callbacks

class kerasrl.callbacks.SaveDQNTrainingState(interval, state_path, memory, dqn, snapshot_limit=None)[source]

Save agent progress, memory and model weights

on_episode_end(episode, logs)[source]

Called at end of each episode

class kerasrl.callbacks.TerminateTrainingOnFileExists(agent, poisonfile)[source]
on_action_begin(action, logs)[source]

Called at beginning of each action