bb72c5b102a74486036d849b44f80ab3772a18f9,agents/ppo/dppo_worker.py,DPPOWorker,run,#DPPOWorker#,90

Before Change

    def run(self):
        with self.sv.managed_session(self.server.target, config=self.config_proto) as sess, sess.as_default():
            print("DPPO worker {} in managed session".format(self.task_id))
            while not self.sv.should_stop():
                message = None
                while message != "collect":

After Change

    def run(self):
        with tf.Session() as sess, sess.as_default():
            var_receivers = [np.zeros(var.shape.as_list(), dtype=var.dtype.as_numpy_dtype) for var in self.global_vars]
            while True:
                for var_receiver, tf_var in zip(var_receivers, self.global_vars):
                    self.comm.Bcast(var_receiver, root=0)
                    tf_var.load(var_receiver)
                trajectory = self.env_runner.get_steps(
                    self.config["n_local_steps"], stop_at_trajectory_end=False)
                T = trajectory.steps
                value = 0 if trajectory.terminals[-1] else self.get_critic_value(

Instances