添加 PPO 算法实现及相关配置，更新训练入口以支持 SAC 和 PPO 模式

eval.py

@@ -23,6 +23,11 @@ def evaluate_and_plot(model_path):
         config = yaml.safe_load(f)
 
     env = gym.make('Pendulum-v1')
+    if not isinstance(env.observation_space, gym.spaces.Box) or env.observation_space.shape is None:
+        raise TypeError("Pendulum-v1 observation_space must be a Box with a valid shape.")
+    if not isinstance(env.action_space, gym.spaces.Box) or env.action_space.shape is None:
+        raise TypeError("Pendulum-v1 action_space must be a Box with a valid shape.")
+
     state_dim = env.observation_space.shape[0]
     action_dim = env.action_space.shape[0]
     max_action = float(env.action_space.high[0])
@@ -47,9 +52,9 @@ def evaluate_and_plot(model_path):
     actions = []
     time_steps = []
 
-    episode_reward = 0
+    episode_reward: float = 0.0
     for step in range(config['max_steps']):
-        # 【重点】：设置 evaluate=True，让网络输出确定的均值动作，关闭随机探索
+        # 设置 evaluate=True，让网络输出确定的均值动作，关闭随机探索
         action = agent.select_action(state, evaluate=True)
         
         # 记录当前步的数据
@@ -66,7 +71,7 @@ def evaluate_and_plot(model_path):
         # 与环境交互
         next_state, reward, terminated, truncated, _ = env.step(action)
         state = next_state
-        episode_reward += reward
+        episode_reward += float(reward)
 
         if terminated or truncated:
             break