problem: priority calculation of a transition #4
Description
I have read the paper and code of your project, and I have a modest doubt in terms of the realization of calculating priority. In the code, actor_loss_element = self.critic_model.predict([cur_states, predicted_actions]) correspond to term 2 of priority formulation (8) in the paper. Obviously, this term should be related to a list of gradients in the paper, but not in the code. I think for a long time and find no answer, can you give me an explanation, or it is just a bug? Thanks.