run_competition

class xuance.engine.run_competition.RunnerCompetition(configs)

Bases: object

benchmark()

rprint(info: str)

run()

run_episodes(env_fn=None, n_episodes: int = 1, test_mode: bool = False)

Run some episodes when use RNN.

Parameters:

• env_fn – The function that can make some testing environments.

• n_episodes (int) – Number of episodes.

• test_mode (bool) – Whether to test the model.

Returns:

The episode scores.

Return type:

Scores

test(env_fn, n_episodes)

Test the model for some episodes.

Parameters:

• env_fn – The function that can make some testing environments.

• n_episodes (int) – Number of episodes to test.

Returns:

A list of cumulative rewards for each episode.

Return type:

scores (List(float))

train(n_steps)

Train the model for numerous steps. :param n_steps: Number of steps to train the model: :type n_steps: int