d9f1874e3489edc915228dcfdae64d5294034464,rllib/tests/test_rollout_worker.py,TestRolloutWorker,test_global_vars_update,#TestRolloutWorker#,187

Before Change

                })
            result = agent.train()
            for i in range(10):
                result = agent.train()
                print("num_steps_sampled={}".format(
                    result["info"]["num_steps_sampled"]))
                print("num_steps_trained={}".format(
                    result["info"]["num_steps_trained"]))
                print("num_steps_sampled={}".format(
                    result["info"]["num_steps_sampled"]))
                print("num_steps_trained={}".format(
                    result["info"]["num_steps_trained"]))
                if i == 0:
                    self.assertGreater(
                        result["info"]["learner"]["default_policy"]["cur_lr"],
                        0.01)
                if result["info"]["learner"]["default_policy"]["cur_lr"] < \
                        0.07:
                    break
            self.assertLess(

After Change

                    0.1 - ((0.1 - 0.000001) / 100000) * global_timesteps
                lr = policy.cur_lr
                if fw == "tf":
                    lr = policy._sess.run(lr)
                check(lr, expected_lr, rtol=0.05)
            agent.stop()

    def test_no_step_on_init(self):

Instances