昨天更新后使用model训练时出现问题

[richardhuo]

环境： Colab, Elegantrl Model: 测试过SAC, ModSAC, TD3 都报下面相同错误 Issue: init() takes 4 positional arguments but 5 were given

31 trained_model = agent.train_model(model=model, 32 cwd=cwd, ---> 33 total_timesteps=break_step) 34 else: 35 raise ValueError('DRL library input is NOT supported. Please check.')

in train_model(self, model, cwd, total_timesteps) 65 model.cwd = cwd 66 model.break_step = total_timesteps ---> 67 train_and_evaluate(model) 68 69 @staticmethod

in train_and_evaluate(args) 32 env = build_env(args.env, args.env_func, args.env_args) 33 ---> 34 agent = init_agent(args, gpu_id, env) 35 buffer = init_buffer(args, gpu_id) 36 evaluator = init_evaluator(args, gpu_id)

in init_agent(args, gpu_id, env) 74 75 def init_agent(args, gpu_id, env=None): ---> 76 agent = args.agent(args.net_dim, args.state_dim, args.action_dim, gpu_id=gpu_id, args=args) 77 agent.save_or_load_agent(args.cwd, if_save=False) 78

/usr/local/lib/python3.7/dist-packages/elegantrl/agents/AgentTD3.py in init(self, net_dim, state_dim, action_dim, gpu_id, args) 22 self.act_class = getattr(self, "act_class", Actor) 23 self.cri_class = getattr(self, "cri_class", CriticTwin) ---> 24 super().init(net_dim, state_dim, action_dim, gpu_id, args) 25 self.policy_noise = getattr( 26 args, "policy_noise", 0.15

/usr/local/lib/python3.7/dist-packages/elegantrl/agents/AgentDDPG.py in init(self, net_dim, state_dim, action_dim, gpu_id, args) 25 self.act_class = getattr(self, "act_class", ActorSAC) 26 self.cri_class = getattr(self, "cri_class", CriticTwin) ---> 27 super().init(net_dim, state_dim, action_dim, gpu_id, args) 28 self.act.explore_noise = getattr( 29 args, "explore_noise", 0.1

/usr/local/lib/python3.7/dist-packages/elegantrl/agents/AgentBase.py in init(self, net_dim, state_dim, action_dim, gpu_id, args) 45 act_class = getattr(self, "act_class", None) 46 cri_class = getattr(self, "cri_class", None) ---> 47 self.act = act_class(net_dim, self.num_layer, state_dim, action_dim).to(self.device) 48 self.cri = cri_class(net_dim, self.num_layer, state_dim, action_dim).to(self.device)
49 if cri_class else self.act

TypeError: init() takes 4 positional arguments but 5 were given

[richardhuo]

我使用老版本解决了这个问题。 等待后续更新。