hot update

projects/codes/QLearning/qlearning.py

@@ -5,7 +5,7 @@ Author: John
 Email: johnjim0816@gmail.com
 Date: 2020-09-11 23:03:00
 LastEditor: John
-LastEditTime: 2021-12-22 10:54:57
+LastEditTime: 2022-08-24 10:31:04
 Discription: use defaultdict to define Q table
 Environment: 
 '''
@@ -15,50 +15,52 @@ import torch
 from collections import defaultdict
 
 class QLearning(object):
-    def __init__(self,
-                 n_actions,cfg):
-        self.n_actions = n_actions 
-        self.lr = cfg.lr  # 学习率
-        self.gamma = cfg.gamma  
-        self.epsilon = cfg.epsilon_start
+    def __init__(self,cfg):
+        self.n_actions = cfg['n_actions'] 
+        self.lr = cfg['lr']  
+        self.gamma = cfg['gamma']  
+        self.epsilon = cfg['epsilon_start']
         self.sample_count = 0  
-        self.epsilon_start = cfg.epsilon_start
-        self.epsilon_end = cfg.epsilon_end
-        self.epsilon_decay = cfg.epsilon_decay
-        self.Q_table  = defaultdict(lambda: np.zeros(n_actions)) # 用嵌套字典存放状态->动作->状态-动作值（Q值）的映射，即Q表
-    def sample(self, state):
-        ''' 采样动作，训练时用
+        self.epsilon_start = cfg['epsilon_start']
+        self.epsilon_end = cfg['epsilon_end']
+        self.epsilon_decay = cfg['epsilon_decay']
+        self.Q_table  = defaultdict(lambda: np.zeros(self.n_actions)) # use nested dictionary to represent Q(s,a), here set all Q(s,a)=0 initially, not like pseudo code
+    def sample_action(self, state):
+        ''' sample action with e-greedy policy while training
         '''
         self.sample_count += 1
+        # epsilon must decay(linear,exponential and etc.) for balancing exploration and exploitation
         self.epsilon = self.epsilon_end + (self.epsilon_start - self.epsilon_end) * \
-            math.exp(-1. * self.sample_count / self.epsilon_decay) # epsilon是会递减的，这里选择指数递减
-        # e-greedy 策略
+            math.exp(-1. * self.sample_count / self.epsilon_decay) 
         if np.random.uniform(0, 1) > self.epsilon:
-            action = np.argmax(self.Q_table[str(state)]) # 选择Q(s,a)最大对应的动作
+            action = np.argmax(self.Q_table[str(state)]) # choose action corresponding to the maximum q value
         else:
-            action = np.random.choice(self.n_actions) # 随机选择动作
+            action = np.random.choice(self.n_actions) # choose action randomly
         return action
-    def predict(self,state):
-        ''' 预测或选择动作，测试时用
+    def predict_action(self,state):
+        ''' predict action while testing 
         '''
         action = np.argmax(self.Q_table[str(state)])
         return action
     def update(self, state, action, reward, next_state, done):
         Q_predict = self.Q_table[str(state)][action] 
-        if done: # 终止状态
+        if done: # terminal state
             Q_target = reward  
         else:
             Q_target = reward + self.gamma * np.max(self.Q_table[str(next_state)]) 
         self.Q_table[str(state)][action] += self.lr * (Q_target - Q_predict)
-    def save(self,path):
+    def save_model(self,path):
         import dill
+        from pathlib import Path
+        # create path
+        Path(path).mkdir(parents=True, exist_ok=True)
         torch.save(
             obj=self.Q_table,
             f=path+"Qleaning_model.pkl",
             pickle_module=dill
         )
-        print("保存模型成功！")
-    def load(self, path):
+        print("Model saved!")
+    def load_model(self, path):
         import dill
         self.Q_table =torch.load(f=path+'Qleaning_model.pkl',pickle_module=dill)
-        print("加载模型成功！")
+        print("Mode loaded!")