def main(_): hp = trainer_model_based_params.create_loop_hparams() if FLAGS.job_dir_to_evaluate: compute_final_evaluation_on_real_environments( hp, FLAGS.job_dir_to_evaluate) else: training_loop(hp, FLAGS.output_dir)
def main(_): hparams = trainer_model_based_params.create_loop_hparams() problem_name = get_simulated_problem_name(hparams.game) world_model_dir = FLAGS.world_model_dir agent_model_dir = FLAGS.output_dir event_dir = FLAGS.output_dir epoch_data_dir = FLAGS.data_dir # only required for initial frames trainer_model_based.train_agent(problem_name, agent_model_dir, event_dir, world_model_dir, epoch_data_dir, hparams, epoch=0, is_final_epoch=True)
def main(_): hparams = trainer_model_based_params.create_loop_hparams() problem_name = get_simulated_problem_name(hparams.game) world_model_dir = FLAGS.world_model_dir agent_model_dir = FLAGS.output_dir event_dir = FLAGS.output_dir epoch_data_dir = FLAGS.data_dir # only required for initial frames trainer_model_based.train_agent( problem_name, agent_model_dir, event_dir, world_model_dir, epoch_data_dir, hparams, 0, epoch=0, is_final_epoch=True)
def main(_): hp = trainer_model_based_params.create_loop_hparams() assert not FLAGS.job_dir_to_evaluate training_loop(hp, FLAGS.output_dir)
def main(_): hp = trainer_model_based_params.create_loop_hparams() assert not FLAGS.job_dir_to_evaluate training_loop(hp, FLAGS.output_dir)