Farama-Foundation / SuperSuit

A collection of wrappers for Gymnasium and PettingZoo environments (being merged into gymnasium.wrappers and pettingzoo.wrappers
Other
446 stars 56 forks source link

[Ignore this] frameskip #187

Closed WillDudley closed 1 year ago

WillDudley commented 1 year ago

on master branch (pypi pz) env.step() hits at L84 of frame_skip, once for agent1, 4 times for agent 2. on trunc branch (trunc pz) env_step() doesn't hit anywhere

WillDudley commented 1 year ago

hitting L84 means hitting 76 and 78

WillDudley commented 1 year ago
        cur_agent = self.agent_selection
        self._cumulative_rewards[cur_agent] = 0
        self.rewards = make_defaultdict({a: 0.0 for a in self.agents})
        self.skip_num[cur_agent] = self.num_frames
        self.old_actions[cur_agent] = action
        while self.old_actions[self.env.agent_selection] is not None:
            step_agent = self.env.agent_selection
            if step_agent in self.env.dones:
                # reward = self.env.rewards[step_agent]
                # done = self.env.dones[step_agent]
                # info = self.env.infos[step_agent]
                observe, reward, done, info = self.env.last(observe=False)
                action = self.old_actions[step_agent]
                self.env.step(action)

                for agent in self.env.agents:
                    self.rewards[agent] += self.env.rewards[agent]
                self.infos[self.env.agent_selection] = info
WillDudley commented 1 year ago

fixed