train_agent_chainer.py 文件源码

def train_agent(rounds=10000, use_score=False, name='result_dir', create_agent=create_ddqn_agent):
    ENV_NAME = 'malware-score-v0' if use_score else 'malware-v0'
    env = gym.make( ENV_NAME ) 
    np.random.seed(123)
    env.seed(123)

    agent = create_agent(env)

    chainerrl.experiments.train_agent_with_evaluation(
        agent, env,
        steps=rounds,                   # Train the agent for this many rounds steps
        max_episode_len=env.maxturns,   # Maximum length of each episodes        
        eval_interval=1000,             # Evaluate the agent after every 1000 steps
        eval_n_runs=100,                # 100 episodes are sampled for each evaluation        
        outdir=name)                    # Save everything to 'result' directory

    return agent