22ccc43670dac93eb7fe81520a84cf3979d05693,rllib/agents/dqn/tests/test_dqn.py,TestDQN,test_dqn_parameter_noise_exploration,#TestDQN#,100

Before Change

        core_config["num_workers"] = 0  // Run locally.
        core_config["env_config"] = {"is_slippery": False, "map_name": "4x4"}

        for fw in framework_iterator(core_config, ["tf", "eager"]):

            config = core_config.copy()

After Change

            // the same action for the same input (parameter noise is
            // deterministic).
            policy = trainer.get_policy()
            p_sess = getattr(policy, "_sess", None)
            policy.exploration.on_episode_start(policy, tf_sess=p_sess)
            a_ = trainer.compute_action(obs)
            for _ in range(10):
                a = trainer.compute_action(obs, explore=True)

Instances