update

codes/DQN/task0_train.py

@@ -5,7 +5,7 @@
 @Email: johnjim0816@gmail.com
 @Date: 2020-06-12 00:48:57
 @LastEditor: John
-LastEditTime: 2021-05-04 15:01:34
+LastEditTime: 2021-05-05 16:49:15
 @Discription: 
 @Environment: python 3.7.7
 '''
@@ -14,9 +14,9 @@ curr_path = os.path.dirname(__file__)
 parent_path = os.path.dirname(curr_path)
 sys.path.append(parent_path)  # add current terminal path to sys.path
 
-import datetime
-import torch
 import gym
+import torch
+import datetime
 
 from common.utils import save_results, make_dir
 from common.plot import plot_rewards
@@ -32,21 +32,21 @@ class DQNConfig:
         self.result_path = curr_path+"/outputs/" + self.env + \
             '/'+curr_time+'/results/'  # path to save results
         self.model_path = curr_path+"/outputs/" + self.env + \
-            '/'+curr_time+'/models/'  # path to save results
-        self.train_eps = 300  # 训练的episode数目
+            '/'+curr_time+'/models/'  # path to save models
+        self.train_eps = 300  # max trainng episodes
         self.eval_eps = 50 # number of episodes for evaluating
         self.gamma = 0.95
-        self.epsilon_start = 0.90  # e-greedy策略的初始epsilon
+        self.epsilon_start = 0.90  # start epsilon of e-greedy policy
         self.epsilon_end = 0.01
         self.epsilon_decay = 500
         self.lr = 0.0001  # learning rate
-        self.memory_capacity = 100000  # Replay Memory容量
+        self.memory_capacity = 100000  # capacity of Replay Memory
         self.batch_size = 64
-        self.target_update = 2  # target net的更新频率
+        self.target_update = 4 # update frequency of target net
         self.device = torch.device(
-            "cuda" if torch.cuda.is_available() else "cpu")  # 检测gpu
-        self.hidden_dim = 256  # 神经网络隐藏层维度
-
+            "cuda" if torch.cuda.is_available() else "cpu")  # check gpu
+        self.hidden_dim = 256  # hidden size of net
+        
 def env_agent_config(cfg,seed=1):
     env = gym.make(cfg.env)  
     env.seed(seed)
@@ -60,7 +60,7 @@ def train(cfg, env, agent):
     print(f'Env:{cfg.env}, Algorithm:{cfg.algo}, Device:{cfg.device}')
     rewards = []
     ma_rewards = []  # moveing average reward
-    for i_episode in range(cfg.train_eps):
+    for i_ep in range(cfg.train_eps):
         state = env.reset()
         done = False
         ep_reward = 0
@@ -73,9 +73,10 @@ def train(cfg, env, agent):
             agent.update()
             if done:
                 break
-        if i_episode % cfg.target_update == 0:
+        if (i_ep+1) % cfg.target_update == 0:
             agent.target_net.load_state_dict(agent.policy_net.state_dict())
-        print('Episode:{}/{}, Reward:{}'.format(i_episode+1, cfg.train_eps, ep_reward))
+        if (i_ep+1)%10 == 0:
+            print('Episode:{}/{}, Reward:{}'.format(i_ep+1, cfg.train_eps, ep_reward))
         rewards.append(ep_reward)
         # save ma rewards
         if ma_rewards:
@@ -86,15 +87,17 @@ def train(cfg, env, agent):
     return rewards, ma_rewards
 
 def eval(cfg,env,agent):
-    rewards = []  # 记录所有episode的reward
-    ma_rewards = [] # 滑动平均的reward
+    print('Start to eval !')
+    print(f'Env:{cfg.env}, Algorithm:{cfg.algo}, Device:{cfg.device}')
+    rewards = []  
+    ma_rewards = [] # moving average rewards
     for i_ep in range(cfg.eval_eps):
-        ep_reward = 0  # 记录每个episode的reward
-        state = env.reset()  # 重置环境, 重新开一局（即开始新的一个episode）
+        ep_reward = 0  # reward per episode
+        state = env.reset()  
         while True:
-            action = agent.predict(state)  # 根据算法选择一个动作
-            next_state, reward, done, _ = env.step(action)  # 与环境进行一个交互
-            state = next_state  # 存储上一个观察值
+            action = agent.predict(state) 
+            next_state, reward, done, _ = env.step(action)  
+            state = next_state  
             ep_reward += reward
             if done:
                 break
@@ -103,11 +106,15 @@ def eval(cfg,env,agent):
             ma_rewards.append(ma_rewards[-1]*0.9+ep_reward*0.1)
         else:
             ma_rewards.append(ep_reward)
-        print(f"Episode:{i_ep+1}/{cfg.eval_eps}, reward:{ep_reward:.1f}")
+        if (i_ep+1)%10 == 10:
+            print(f"Episode:{i_ep+1}/{cfg.eval_eps}, reward:{ep_reward:.1f}")
+    print('Complete evaling！')
     return rewards,ma_rewards
 
 if __name__ == "__main__":
     cfg = DQNConfig()
+
+    # train
     env,agent = env_agent_config(cfg,seed=1)
     rewards, ma_rewards = train(cfg, env, agent)
     make_dir(cfg.result_path, cfg.model_path)
@@ -115,7 +122,7 @@ if __name__ == "__main__":
     save_results(rewards, ma_rewards, tag='train', path=cfg.result_path)
     plot_rewards(rewards, ma_rewards, tag="train",
                  algo=cfg.algo, path=cfg.result_path)
-
+    # eval
     env,agent = env_agent_config(cfg,seed=10)
     agent.load(path=cfg.model_path)
     rewards,ma_rewards = eval(cfg,env,agent)

codes/DoubleDQN/agent.py

@@ -5,7 +5,7 @@
 @Email: johnjim0816@gmail.com
 @Date: 2020-06-12 00:50:49
 @LastEditor: John
-LastEditTime: 2021-05-04 15:04:45
+LastEditTime: 2021-05-04 22:28:06
 @Discription: 
 @Environment: python 3.7.7
 '''
@@ -35,9 +35,10 @@ class DoubleDQN:
         self.batch_size = cfg.batch_size
         self.policy_net = MLP(state_dim, action_dim,hidden_dim=cfg.hidden_dim).to(self.device)
         self.target_net = MLP(state_dim, action_dim,hidden_dim=cfg.hidden_dim).to(self.device)
-        # target_net的初始模型参数完全复制policy_net
-        self.target_net.load_state_dict(self.policy_net.state_dict())
-        self.target_net.eval()  # 不启用 BatchNormalization 和 Dropout
+        # target_net copy from policy_net
+        for target_param, param in zip(self.target_net.parameters(), self.policy_net.parameters()):
+            target_param.data.copy_(param.data)
+        # self.target_net.eval()  # 不启用 BatchNormalization 和 Dropout
         # 可查parameters()与state_dict()的区别，前者require_grad=True
         self.optimizer = optim.Adam(self.policy_net.parameters(), lr=cfg.lr)
         self.loss = 0
@@ -58,9 +59,9 @@ class DoubleDQN:
     def choose_action(self, state):
         '''选择动作
         '''
+        self.actions_count += 1
         self.epsilon = self.epsilon_end + (self.epsilon_start - self.epsilon_end) * \
             math.exp(-1. * self.actions_count / self.epsilon_decay)
-        self.actions_count += 1
         if random.random() > self.epsilon:
             action  = self.predict(state)
         else:
@@ -73,7 +74,7 @@ class DoubleDQN:
         # 从memory中随机采样transition
         state_batch, action_batch, reward_batch, next_state_batch, done_batch = self.memory.sample(
             self.batch_size)
-        ### 转为张量 ###
+        # convert to tensor
         state_batch = torch.tensor(
             state_batch, device=self.device, dtype=torch.float)
         action_batch = torch.tensor(action_batch, device=self.device).unsqueeze(
@@ -84,8 +85,7 @@ class DoubleDQN:
             next_state_batch, device=self.device, dtype=torch.float)
         
         done_batch = torch.tensor(np.float32(
-            done_batch), device=self.device).unsqueeze(1)  # 将bool转为float然后转为张量
-
+            done_batch), device=self.device)  # 将bool转为float然后转为张量
         # 计算当前(s_t,a)对应的Q(s_t, a)
         q_values = self.policy_net(state_batch) 
         next_q_values = self.policy_net(next_state_batch)
@@ -104,7 +104,7 @@ class DoubleDQN:
             next_state_batch)
         # 选出Q(s_t‘, a)对应的action，代入到next_target_values获得target net对应的next_q_value，即Q’(s_t|a=argmax Q(s_t‘, a))
         next_target_q_value = next_target_values.gather(1, torch.max(next_q_values, 1)[1].unsqueeze(1)).squeeze(1)
-        q_target = reward_batch + self.gamma * next_target_q_value * (1-done_batch[0])
+        q_target = reward_batch + self.gamma * next_target_q_value * (1-done_batch)
         self.loss = nn.MSELoss()(q_value, q_target.unsqueeze(1))  # 计算 均方误差loss
         # 优化模型
         self.optimizer.zero_grad()  # zero_grad清除上一步所有旧的gradients from the last step

codes/DoubleDQN/task0_train.ipynb

@@ -0,0 +1,194 @@
+{
+ "metadata": {
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.7.10"
+  },
+  "orig_nbformat": 2,
+  "kernelspec": {
+   "name": "python3710jvsc74a57bd0366e1054dee9d4501b0eb8f87335afd3c67fc62db6ee611bbc7f8f5a1fefe232",
+   "display_name": "Python 3.7.10 64-bit ('py37': conda)"
+  },
+  "metadata": {
+   "interpreter": {
+    "hash": "366e1054dee9d4501b0eb8f87335afd3c67fc62db6ee611bbc7f8f5a1fefe232"
+   }
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2,
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import sys\n",
+    "from pathlib import Path\n",
+    "curr_path = str(Path().absolute())\n",
+    "parent_path = str(Path().absolute().parent)\n",
+    "sys.path.append(parent_path) # add current terminal path to sys.path"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import gym\n",
+    "import torch\n",
+    "import datetime\n",
+    "from DoubleDQN.agent import DoubleDQN\n",
+    "from common.plot import plot_rewards\n",
+    "from common.utils import save_results, make_dir\n",
+    "\n",
+    "curr_time = datetime.datetime.now().strftime(\n",
+    "    \"%Y%m%d-%H%M%S\")  # obtain current time"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class DoubleDQNConfig:\n",
+    "    def __init__(self):\n",
+    "        self.algo = \"DoubleDQN\" # name of algo\n",
+    "        self.env = 'CartPole-v0'  # env name\n",
+    "        self.result_path = curr_path+\"/outputs/\" + self.env + \\\n",
+    "            '/'+curr_time+'/results/'  # path to save results\n",
+    "        self.model_path = curr_path+\"/outputs/\" + self.env + \\\n",
+    "            '/'+curr_time+'/models/'  # path to save models\n",
+    "        self.train_eps = 200 # max tranng episodes\n",
+    "        self.eval_eps = 50 # max evaling episodes\n",
+    "        self.gamma = 0.95\n",
+    "        self.epsilon_start = 1 # start epsilon of e-greedy policy\n",
+    "        self.epsilon_end = 0.01 \n",
+    "        self.epsilon_decay = 500\n",
+    "        self.lr = 0.001 # learning rate\n",
+    "        self.memory_capacity = 100000 # capacity of Replay Memory\n",
+    "        self.batch_size = 64\n",
+    "        self.target_update = 2 # update frequency of target net\n",
+    "        self.device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\") # check gpu\n",
+    "        self.hidden_dim = 256 # hidden size of net"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def env_agent_config(cfg,seed=1):\n",
+    "    env = gym.make(cfg.env)  \n",
+    "    env.seed(seed)\n",
+    "    state_dim = env.observation_space.shape[0]\n",
+    "    action_dim = env.action_space.n\n",
+    "    agent = DoubleDQN(state_dim,action_dim,cfg)\n",
+    "    return env,agent"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def train(cfg,env,agent):\n",
+    "    print('Start to train !')\n",
+    "    rewards,ma_rewards = [],[]\n",
+    "    for i_ep in range(cfg.train_eps):\n",
+    "        state = env.reset() \n",
+    "        ep_reward = 0\n",
+    "        while True:\n",
+    "            action = agent.choose_action(state) \n",
+    "            next_state, reward, done, _ = env.step(action)\n",
+    "            ep_reward += reward\n",
+    "            agent.memory.push(state, action, reward, next_state, done) \n",
+    "            state = next_state \n",
+    "            agent.update() \n",
+    "            if done:\n",
+    "                break\n",
+    "        if i_ep % cfg.target_update == 0:\n",
+    "            agent.target_net.load_state_dict(agent.policy_net.state_dict())\n",
+    "        if (i_ep+1)%10 == 0:\n",
+    "            print(f'Episode:{i_ep+1}/{cfg.train_eps}, Reward:{ep_reward}')\n",
+    "        rewards.append(ep_reward)\n",
+    "        if ma_rewards:\n",
+    "            ma_rewards.append(\n",
+    "                0.9*ma_rewards[-1]+0.1*ep_reward)\n",
+    "        else:\n",
+    "            ma_rewards.append(ep_reward)   \n",
+    "    print('Complete training！')\n",
+    "    return rewards,ma_rewards"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def eval(cfg,env,agent):\n",
+    "    print('Start to eval !')\n",
+    "    print(f'Env:{cfg.env}, Algorithm:{cfg.algo}, Device:{cfg.device}')\n",
+    "    rewards = []  \n",
+    "    ma_rewards = []\n",
+    "    for i_ep in range(cfg.eval_eps):\n",
+    "        state = env.reset() \n",
+    "        ep_reward = 0   \n",
+    "        while True:\n",
+    "            action = agent.predict(state)  \n",
+    "            next_state, reward, done, _ = env.step(action)  \n",
+    "            state = next_state  \n",
+    "            ep_reward += reward\n",
+    "            if done:\n",
+    "                break\n",
+    "        rewards.append(ep_reward)\n",
+    "        if ma_rewards:\n",
+    "            ma_rewards.append(ma_rewards[-1]*0.9+ep_reward*0.1)\n",
+    "        else:\n",
+    "            ma_rewards.append(ep_reward)\n",
+    "        print(f\"Episode:{i_ep+1}/{cfg.eval_eps}, reward:{ep_reward:.1f}\")\n",
+    "    print('Complete evaling！')\n",
+    "    return rewards,ma_rewards "
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "if __name__ == \"__main__\":\n",
+    "    cfg = DoubleDQNConfig()\n",
+    "    # train\n",
+    "    env,agent = env_agent_config(cfg,seed=1)\n",
+    "    rewards, ma_rewards = train(cfg, env, agent)\n",
+    "    make_dir(cfg.result_path, cfg.model_path)\n",
+    "    agent.save(path=cfg.model_path)\n",
+    "    save_results(rewards, ma_rewards, tag='train', path=cfg.result_path)\n",
+    "    plot_rewards(rewards, ma_rewards, tag=\"train\",\n",
+    "                 algo=cfg.algo, path=cfg.result_path)\n",
+    "\n",
+    "    # eval\n",
+    "    env,agent = env_agent_config(cfg,seed=10)\n",
+    "    agent.load(path=cfg.model_path)\n",
+    "    rewards,ma_rewards = eval(cfg,env,agent)\n",
+    "    save_results(rewards,ma_rewards,tag='eval',path=cfg.result_path)\n",
+    "    plot_rewards(rewards,ma_rewards,tag=\"eval\",env=cfg.env,algo = cfg.algo,path=cfg.result_path)"
+   ]
+  }
+ ]
+}

codes/DoubleDQN/task0_train.py

@@ -5,7 +5,7 @@
 @Email: johnjim0816@gmail.com
 @Date: 2020-06-12 00:48:57
 @LastEditor: John
-LastEditTime: 2021-05-04 15:05:37
+LastEditTime: 2021-05-04 22:26:59
 @Discription: 
 @Environment: python 3.7.7
 '''
@@ -31,21 +31,19 @@ class DoubleDQNConfig:
         self.result_path = curr_path+"/outputs/" + self.env + \
             '/'+curr_time+'/results/'  # path to save results
         self.model_path = curr_path+"/outputs/" + self.env + \
-            '/'+curr_time+'/models/'  # path to save results
-        self.gamma = 0.99 
-        self.epsilon_start = 0.9 # start epsilon of e-greedy policy
-        self.epsilon_end = 0.01 
-        self.epsilon_decay = 200
-        self.lr = 0.01 # learning rate
-        self.memory_capacity = 10000 # capacity of Replay Memory
-        self.batch_size = 128
-        self.train_eps = 300 # max tranng episodes
-        self.train_steps = 200 # max training steps per episode
-        self.target_update = 2 # update frequency of target net
+            '/'+curr_time+'/models/'  # path to save models
+        self.train_eps = 200 # max tranng episodes
         self.eval_eps = 50 # max evaling episodes
-        self.eval_steps = 200 # max evaling steps per episode
+        self.gamma = 0.95
+        self.epsilon_start = 1 # start epsilon of e-greedy policy
+        self.epsilon_end = 0.01 
+        self.epsilon_decay = 500
+        self.lr = 0.001 # learning rate
+        self.memory_capacity = 100000 # capacity of Replay Memory
+        self.batch_size = 64
+        self.target_update = 2 # update frequency of target net
         self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu") # check gpu
-        self.hidden_dim = 128 # hidden size of net
+        self.hidden_dim = 256 # hidden size of net
  
 def env_agent_config(cfg,seed=1):
     env = gym.make(cfg.env)  
@@ -59,20 +57,20 @@ def train(cfg,env,agent):
     print('Start to train !')
     rewards,ma_rewards = [],[]
     for i_ep in range(cfg.train_eps):
-        state = env.reset() # reset环境状态
+        state = env.reset() 
         ep_reward = 0
         while True:
-            action = agent.choose_action(state) # 根据当前环境state选择action
-            next_state, reward, done, _ = env.step(action) # 更新环境参数
+            action = agent.choose_action(state) 
+            next_state, reward, done, _ = env.step(action)
             ep_reward += reward
-            agent.memory.push(state, action, reward, next_state, done) # 将state等这些transition存入memory
-            state = next_state # 跳转到下一个状态
-            agent.update() # 每步更新网络
+            agent.memory.push(state, action, reward, next_state, done) 
+            state = next_state 
+            agent.update() 
             if done:
                 break
         if i_ep % cfg.target_update == 0:
             agent.target_net.load_state_dict(agent.policy_net.state_dict())
-        print(f'Episode:{i_ep+1}/{cfg.train_eps}, Reward:{ep_reward}')
+        print(f'Episode:{i_ep+1}/{cfg.train_eps}, Reward:{ep_reward},Epsilon:{agent.epsilon:.2f}')
         rewards.append(ep_reward)
         if ma_rewards:
             ma_rewards.append(
@@ -83,6 +81,8 @@ def train(cfg,env,agent):
     return rewards,ma_rewards
 
 def eval(cfg,env,agent):
+    print('Start to eval !')
+    print(f'Env:{cfg.env}, Algorithm:{cfg.algo}, Device:{cfg.device}')
     rewards = []  
     ma_rewards = []
     for i_ep in range(cfg.eval_eps):
@@ -101,9 +101,12 @@ def eval(cfg,env,agent):
         else:
             ma_rewards.append(ep_reward)
         print(f"Episode:{i_ep+1}/{cfg.eval_eps}, reward:{ep_reward:.1f}")
+    print('Complete evaling！')
     return rewards,ma_rewards    
+    
 if __name__ == "__main__":
     cfg = DoubleDQNConfig()
+    # train
     env,agent = env_agent_config(cfg,seed=1)
     rewards, ma_rewards = train(cfg, env, agent)
     make_dir(cfg.result_path, cfg.model_path)
@@ -112,6 +115,7 @@ if __name__ == "__main__":
     plot_rewards(rewards, ma_rewards, tag="train",
                  algo=cfg.algo, path=cfg.result_path)
 
+    # eval
     env,agent = env_agent_config(cfg,seed=10)
     agent.load(path=cfg.model_path)
     rewards,ma_rewards = eval(cfg,env,agent)

codes/HierarchicalDQN/agent.py

@@ -5,7 +5,7 @@ Author: John
 Email: johnjim0816@gmail.com
 Date: 2021-03-24 22:18:18
 LastEditor: John
-LastEditTime: 2021-03-31 14:51:09
+LastEditTime: 2021-05-04 22:39:34
 Discription: 
 Environment: 
 '''
@@ -65,11 +65,11 @@ class HierarchicalDQN:
         if self.batch_size > len(self.memory):
             return
         state_batch, action_batch, reward_batch, next_state_batch, done_batch = self.memory.sample(self.batch_size)
-        state_batch = torch.tensor(state_batch,dtype=torch.float)
-        action_batch = torch.tensor(action_batch,dtype=torch.int64).unsqueeze(1)  
-        reward_batch = torch.tensor(reward_batch,dtype=torch.float)  
-        next_state_batch = torch.tensor(next_state_batch, dtype=torch.float)
-        done_batch = torch.tensor(np.float32(done_batch))
+        state_batch = torch.tensor(state_batch,device=self.device,dtype=torch.float)
+        action_batch = torch.tensor(action_batch,device=self.device,dtype=torch.int64).unsqueeze(1)  
+        reward_batch = torch.tensor(reward_batch,device=self.device,dtype=torch.float)  
+        next_state_batch = torch.tensor(next_state_batch,device=self.device, dtype=torch.float)
+        done_batch = torch.tensor(np.float32(done_batch),device=self.device)
         q_values = self.policy_net(state_batch).gather(dim=1, index=action_batch).squeeze(1)
         next_state_values = self.policy_net(next_state_batch).max(1)[0].detach()
         expected_q_values = reward_batch + 0.99 * next_state_values * (1-done_batch)
@@ -79,17 +79,17 @@ class HierarchicalDQN:
         for param in self.policy_net.parameters():  # clip防止梯度爆炸
             param.grad.data.clamp_(-1, 1)
         self.optimizer.step()  
-        self.loss_numpy = loss.detach().numpy()
+        self.loss_numpy = loss.detach().cpu().numpy()
         self.losses.append(self.loss_numpy)  
     def update_meta(self):
         if self.batch_size > len(self.meta_memory):
             return
         state_batch, action_batch, reward_batch, next_state_batch, done_batch = self.meta_memory.sample(self.batch_size)
-        state_batch = torch.tensor(state_batch,dtype=torch.float)
-        action_batch = torch.tensor(action_batch,dtype=torch.int64).unsqueeze(1)  
-        reward_batch = torch.tensor(reward_batch,dtype=torch.float)  
-        next_state_batch = torch.tensor(next_state_batch, dtype=torch.float)
-        done_batch = torch.tensor(np.float32(done_batch))
+        state_batch = torch.tensor(state_batch,device=self.device,dtype=torch.float)
+        action_batch = torch.tensor(action_batch,device=self.device,dtype=torch.int64).unsqueeze(1)  
+        reward_batch = torch.tensor(reward_batch,device=self.device,dtype=torch.float)  
+        next_state_batch = torch.tensor(next_state_batch,device=self.device, dtype=torch.float)
+        done_batch = torch.tensor(np.float32(done_batch),device=self.device)
         q_values = self.meta_policy_net(state_batch).gather(dim=1, index=action_batch).squeeze(1)
         next_state_values = self.meta_policy_net(next_state_batch).max(1)[0].detach()
         expected_q_values = reward_batch + 0.99 * next_state_values * (1-done_batch)
@@ -99,7 +99,7 @@ class HierarchicalDQN:
         for param in self.meta_policy_net.parameters():  # clip防止梯度爆炸
             param.grad.data.clamp_(-1, 1)
         self.meta_optimizer.step() 
-        self.meta_loss_numpy = meta_loss.detach().numpy()
+        self.meta_loss_numpy = meta_loss.detach().cpu().numpy()
         self.meta_losses.append(self.meta_loss_numpy)
 
     def save(self, path):

codes/HierarchicalDQN/task0_train.py

@@ -5,7 +5,7 @@ Author: John
 Email: johnjim0816@gmail.com
 Date: 2021-03-29 10:37:32
 LastEditor: John
-LastEditTime: 2021-03-31 14:58:49
+LastEditTime: 2021-05-04 22:35:56
 Discription: 
 Environment: 
 '''
@@ -21,27 +21,23 @@ import numpy as np
 import torch
 import gym
 
-from common.utils import save_results
-from common.plot import plot_rewards,plot_losses
+from common.utils import save_results,make_dir
+from common.plot import plot_rewards
 from HierarchicalDQN.agent import HierarchicalDQN
 
-SEQUENCE = datetime.datetime.now().strftime(
+curr_time = datetime.datetime.now().strftime(
     "%Y%m%d-%H%M%S")  # obtain current time
-SAVED_MODEL_PATH = curr_path+"/saved_model/"+SEQUENCE+'/'  # path to save model
-if not os.path.exists(curr_path+"/saved_model/"):
-    os.mkdir(curr_path+"/saved_model/")
-if not os.path.exists(SAVED_MODEL_PATH):
-    os.mkdir(SAVED_MODEL_PATH)
-RESULT_PATH = curr_path+"/results/"+SEQUENCE+'/'  # path to save rewards
-if not os.path.exists(curr_path+"/results/"):
-    os.mkdir(curr_path+"/results/")
-if not os.path.exists(RESULT_PATH):
-    os.mkdir(RESULT_PATH)
-
 
 class HierarchicalDQNConfig:
     def __init__(self):
         self.algo = "H-DQN"  # name of algo
+        self.env = 'CartPole-v0'
+        self.result_path = curr_path+"/outputs/" + self.env + \
+            '/'+curr_time+'/results/'  # path to save results
+        self.model_path = curr_path+"/outputs/" + self.env + \
+            '/'+curr_time+'/models/'  # path to save models
+        self.train_eps = 300  # 训练的episode数目
+        self.eval_eps = 50  # 测试的episode数目
         self.gamma = 0.99
         self.epsilon_start = 1  # start epsilon of e-greedy policy
         self.epsilon_end = 0.01
@@ -49,19 +45,25 @@ class HierarchicalDQNConfig:
         self.lr = 0.0001  # learning rate
         self.memory_capacity = 10000  # Replay Memory capacity
         self.batch_size = 32
-        self.train_eps = 300  # 训练的episode数目
         self.target_update = 2  # target net的更新频率
-        self.eval_eps = 20  # 测试的episode数目
         self.device = torch.device(
             "cuda" if torch.cuda.is_available() else "cpu")  # 检测gpu
         self.hidden_dim = 256  # dimension of hidden layer
 
+def env_agent_config(cfg,seed=1):
+    env = gym.make(cfg.env)  
+    env.seed(seed)
+    state_dim = env.observation_space.shape[0]
+    action_dim = env.action_space.n
+    agent = HierarchicalDQN(state_dim,action_dim,cfg)
+    return env,agent
 
 def train(cfg, env, agent):
     print('Start to train !')
+    print(f'Env:{cfg.env}, Algorithm:{cfg.algo}, Device:{cfg.device}')
     rewards = []
     ma_rewards = []  # moveing average reward
-    for i_episode in range(cfg.train_eps):
+    for i_ep in range(cfg.train_eps):
         state = env.reset()
         done = False
         ep_reward = 0
@@ -83,7 +85,7 @@ def train(cfg, env, agent):
                 state = next_state
                 agent.update()
         agent.meta_memory.push(meta_state, goal, extrinsic_reward, state, done)
-        print('Episode:{}/{}, Reward:{}, Loss:{:.2f}, Meta_Loss:{:.2f}'.format(i_episode+1, cfg.train_eps, ep_reward,agent.loss_numpy ,agent.meta_loss_numpy ))
+        print('Episode:{}/{}, Reward:{}, Loss:{:.2f}, Meta_Loss:{:.2f}'.format(i_ep+1, cfg.train_eps, ep_reward,agent.loss_numpy ,agent.meta_loss_numpy ))
         rewards.append(ep_reward)
         if ma_rewards:
             ma_rewards.append(
@@ -93,18 +95,52 @@ def train(cfg, env, agent):
     print('Complete training！')
     return rewards, ma_rewards
 
+def eval(cfg, env, agent):
+    print('Start to eval !')
+    print(f'Env:{cfg.env}, Algorithm:{cfg.algo}, Device:{cfg.device}')
+    rewards = []
+    ma_rewards = []  # moveing average reward
+    for i_ep in range(cfg.train_eps):
+        state = env.reset()
+        done = False
+        ep_reward = 0
+        while not done:
+            goal = agent.set_goal(state)
+            onehot_goal = agent.to_onehot(goal)
+            extrinsic_reward = 0
+            while not done and goal != np.argmax(state):
+                goal_state = np.concatenate([state, onehot_goal])
+                action = agent.choose_action(goal_state)
+                next_state, reward, done, _ = env.step(action)
+                ep_reward += reward
+                extrinsic_reward += reward
+                state = next_state
+                agent.update()
+        print(f'Episode:{i_ep+1}/{cfg.train_eps}, Reward:{ep_reward}, Loss:{agent.loss_numpy:.2f}, Meta_Loss:{agent.meta_loss_numpy:.2f}')
+        rewards.append(ep_reward)
+        if ma_rewards:
+            ma_rewards.append(
+                0.9*ma_rewards[-1]+0.1*ep_reward)
+        else:
+            ma_rewards.append(ep_reward)
+    print('Complete training！')
+    return rewards, ma_rewards
 
 if __name__ == "__main__":
-    env = gym.make('CartPole-v0')
-    env.seed(1)
     cfg = HierarchicalDQNConfig()
-    state_dim = env.observation_space.shape[0]
-    action_dim = env.action_space.n
-    agent = HierarchicalDQN(state_dim, action_dim, cfg)
-    rewards, ma_rewards = train(cfg, env, agent)
-    agent.save(path=SAVED_MODEL_PATH)
-    save_results(rewards, ma_rewards, tag='train', path=RESULT_PATH)
-    plot_rewards(rewards, ma_rewards, tag="train",
-                 algo=cfg.algo, path=RESULT_PATH)
-    plot_losses(agent.losses,algo=cfg.algo, path=RESULT_PATH)
+
+    # train
+    env,agent = env_agent_config(cfg,seed=1)
+    rewards, ma_rewards = train(cfg, env, agent)
+    make_dir(cfg.result_path, cfg.model_path)
+    agent.save(path=cfg.model_path)
+    save_results(rewards, ma_rewards, tag='train', path=cfg.result_path)
+    plot_rewards(rewards, ma_rewards, tag="train",
+                 algo=cfg.algo, path=cfg.result_path)
+    # eval
+    env,agent = env_agent_config(cfg,seed=10)
+    agent.load(path=cfg.model_path)
+    rewards,ma_rewards = eval(cfg,env,agent)
+    save_results(rewards,ma_rewards,tag='eval',path=cfg.result_path)
+    plot_rewards(rewards,ma_rewards,tag="eval",env=cfg.env,algo = cfg.algo,path=cfg.result_path)
 

codes/MonteCarlo/agent.py

@@ -5,13 +5,14 @@ Author: John
 Email: johnjim0816@gmail.com
 Date: 2021-03-12 16:14:34
 LastEditor: John
-LastEditTime: 2021-03-17 12:35:06
+LastEditTime: 2021-05-05 16:58:39
 Discription: 
 Environment: 
 '''
 import numpy as np
 from collections import defaultdict
 import torch
+import dill
 
 class FisrtVisitMC:
     ''' On-Policy First-Visit MC Control
@@ -20,14 +21,14 @@ class FisrtVisitMC:
         self.action_dim = action_dim
         self.epsilon = cfg.epsilon
         self.gamma = cfg.gamma 
-        self.Q = defaultdict(lambda: np.zeros(action_dim))
+        self.Q_table = defaultdict(lambda: np.zeros(action_dim))
         self.returns_sum = defaultdict(float) # sum of returns
         self.returns_count = defaultdict(float)
         
     def choose_action(self,state):
         ''' e-greed policy '''
-        if state in self.Q.keys():
-            best_action = np.argmax(self.Q[state])
+        if state in self.Q_table.keys():
+            best_action = np.argmax(self.Q_table[state])
             action_probs = np.ones(self.action_dim, dtype=float) * self.epsilon / self.action_dim
             action_probs[best_action] += (1.0 - self.epsilon)
             action = np.random.choice(np.arange(len(action_probs)), p=action_probs)
@@ -48,19 +49,17 @@ class FisrtVisitMC:
             # Calculate average return for this state over all sampled episodes
             self.returns_sum[sa_pair] += G
             self.returns_count[sa_pair] += 1.0
-            self.Q[state][action] = self.returns_sum[sa_pair] / self.returns_count[sa_pair]
+            self.Q_table[state][action] = self.returns_sum[sa_pair] / self.returns_count[sa_pair]
     def save(self,path):
         '''把 Q表格 的数据保存到文件中
         '''
-        import dill
         torch.save(
-            obj=self.Q,
-            f=path,
+            obj=self.Q_table,
+            f=path+"Q_table",
             pickle_module=dill
         )
 
     def load(self, path):
         '''从文件中读取数据到 Q表格
         '''
-        import dill
-        self.Q =torch.load(f=path,pickle_module=dill)
+        self.Q_table =torch.load(f=path+"Q_table",pickle_module=dill)

codes/MonteCarlo/main.py

@@ -1,88 +0,0 @@
-#!/usr/bin/env python
-# coding=utf-8
-'''
-Author: John
-Email: johnjim0816@gmail.com
-Date: 2021-03-11 14:26:44
-LastEditor: John
-LastEditTime: 2021-03-17 12:35:36
-Discription: 
-Environment: 
-'''
-import sys,os
-sys.path.append(os.getcwd())
-import argparse
-import datetime
-
-from envs.racetrack_env import RacetrackEnv
-from MonteCarlo.agent import FisrtVisitMC
-from common.plot import plot_rewards
-from common.utils import save_results
-
-SEQUENCE = datetime.datetime.now().strftime("%Y%m%d-%H%M%S") # 获取当前时间
-SAVED_MODEL_PATH = os.path.split(os.path.abspath(__file__))[0]+"/saved_model/"+SEQUENCE+'/' # 生成保存的模型路径
-if not os.path.exists(os.path.split(os.path.abspath(__file__))[0]+"/saved_model/"): # 检测是否存在文件夹
-    os.mkdir(os.path.split(os.path.abspath(__file__))[0]+"/saved_model/")
-if not os.path.exists(SAVED_MODEL_PATH): # 检测是否存在文件夹
-    os.mkdir(SAVED_MODEL_PATH)
-RESULT_PATH = os.path.split(os.path.abspath(__file__))[0]+"/results/"+SEQUENCE+'/' # 存储reward的路径
-if not os.path.exists(os.path.split(os.path.abspath(__file__))[0]+"/results/"): # 检测是否存在文件夹
-    os.mkdir(os.path.split(os.path.abspath(__file__))[0]+"/results/")
-if not os.path.exists(RESULT_PATH): # 检测是否存在文件夹
-    os.mkdir(RESULT_PATH)
-
-class MCConfig:
-    def __init__(self): 
-        self.epsilon = 0.15 # epsilon: The probability to select a random action . 
-        self.gamma = 0.9 # gamma: Gamma discount factor.
-        self.n_episodes = 150
-        self.n_steps = 2000
-
-def get_mc_args():
-    '''set parameters
-    '''
-    parser = argparse.ArgumentParser()
-    parser.add_argument("--epsilon", default=0.15, type=float)  # epsilon: The probability to select a random action . float between 0 and 1.
-    parser.add_argument("--gamma", default=0.9, type=float)  # gamma: Gamma discount factor.
-    parser.add_argument("--n_episodes", default=150, type=int)
-    parser.add_argument("--n_steps", default=2000, type=int)
-    mc_cfg = parser.parse_args()
-    return mc_cfg
-
-
-
-def mc_train(cfg,env,agent):
-    rewards = []
-    ma_rewards = [] # moving average rewards
-    for i_episode in range(cfg.n_episodes):
-        one_ep_transition = []
-        state = env.reset()
-        ep_reward = 0
-        while True:
-        # for t in range(cfg.n_steps):
-            action = agent.choose_action(state)
-            next_state, reward, done = env.step(action)
-            ep_reward+=reward
-            one_ep_transition.append((state, action, reward))
-            state = next_state
-            if done:
-                break
-        rewards.append(ep_reward)
-        if ma_rewards:
-            ma_rewards.append(ma_rewards[-1]*0.9+ep_reward*0.1)
-        else:
-            ma_rewards.append(ep_reward)
-        agent.update(one_ep_transition)
-        if (i_episode+1)%10==0:
-            print("Episode:{}/{}: Reward:{}".format(i_episode+1, mc_cfg.n_episodes,ep_reward))
-    return rewards,ma_rewards
-if __name__ == "__main__":
-    mc_cfg = MCConfig()
-    env = RacetrackEnv()
-    action_dim=9
-    agent = FisrtVisitMC(action_dim,mc_cfg)
-    rewards,ma_rewards= mc_train(mc_cfg,env,agent)
-    save_results(rewards,ma_rewards,tag='train',path=RESULT_PATH)
-    plot_rewards(rewards,ma_rewards,tag="train",algo = "On-Policy First-Visit MC Control",path=RESULT_PATH)
-    
-

codes/MonteCarlo/task0_train.py

@@ -0,0 +1,118 @@
+#!/usr/bin/env python
+# coding=utf-8
+'''
+Author: John
+Email: johnjim0816@gmail.com
+Date: 2021-03-11 14:26:44
+LastEditor: John
+LastEditTime: 2021-05-05 17:27:50
+Discription: 
+Environment: 
+'''
+
+import sys,os
+curr_path = os.path.dirname(__file__)
+parent_path = os.path.dirname(curr_path)
+sys.path.append(parent_path)  # add current terminal path to sys.path
+
+import torch
+import datetime
+
+from common.utils import save_results,make_dir
+from common.plot import plot_rewards
+from MonteCarlo.agent import FisrtVisitMC
+from envs.racetrack_env import RacetrackEnv
+
+curr_time = datetime.datetime.now().strftime(
+    "%Y%m%d-%H%M%S")  # obtain current time
+
+class MCConfig:
+    def __init__(self):
+        self.algo = "MC"  # name of algo
+        self.env = 'Racetrack'
+        self.result_path = curr_path+"/outputs/" + self.env + \
+            '/'+curr_time+'/results/'  # path to save results
+        self.model_path = curr_path+"/outputs/" + self.env + \
+            '/'+curr_time+'/models/'  # path to save models
+        # epsilon: The probability to select a random action .
+        self.epsilon = 0.15
+        self.gamma = 0.9  # gamma: Gamma discount factor.
+        self.train_eps = 200
+        self.device = torch.device(
+            "cuda" if torch.cuda.is_available() else "cpu")  # check gpu
+
+def env_agent_config(cfg,seed=1):
+    env = RacetrackEnv()
+    action_dim = 9
+    agent = FisrtVisitMC(action_dim, cfg)
+    return env,agent
+    
+def train(cfg, env, agent):
+    print('Start to eval !')
+    print(f'Env:{cfg.env}, Algorithm:{cfg.algo}, Device:{cfg.device}')
+    rewards = []
+    ma_rewards = []  # moving average rewards
+    for i_ep in range(cfg.train_eps):
+        state = env.reset()
+        ep_reward = 0
+        one_ep_transition = []
+        while True:
+            action = agent.choose_action(state)
+            next_state, reward, done = env.step(action)
+            ep_reward += reward
+            one_ep_transition.append((state, action, reward))
+            state = next_state
+            if done:
+                break
+        rewards.append(ep_reward)
+        if ma_rewards:
+            ma_rewards.append(ma_rewards[-1]*0.9+ep_reward*0.1)
+        else:
+            ma_rewards.append(ep_reward)
+        agent.update(one_ep_transition)
+        if (i_ep+1) % 10 == 0:
+            print(f"Episode:{i_ep+1}/{cfg.train_eps}: Reward:{ep_reward}")
+    print('Complete training！')
+    return rewards, ma_rewards
+
+def eval(cfg, env, agent):
+    print('Start to eval !')
+    print(f'Env:{cfg.env}, Algorithm:{cfg.algo}, Device:{cfg.device}')
+    rewards = []
+    ma_rewards = []  # moving average rewards
+    for i_ep in range(cfg.train_eps):
+        state = env.reset()
+        ep_reward = 0
+        while True:
+            action = agent.choose_action(state)
+            next_state, reward, done = env.step(action)
+            ep_reward += reward
+            state = next_state
+            if done:
+                break
+        rewards.append(ep_reward)
+        if ma_rewards:
+            ma_rewards.append(ma_rewards[-1]*0.9+ep_reward*0.1)
+        else:
+            ma_rewards.append(ep_reward)
+        if (i_ep+1) % 10 == 0:
+            print(f"Episode:{i_ep+1}/{cfg.train_eps}: Reward:{ep_reward}")
+    return rewards, ma_rewards
+    
+if __name__ == "__main__":
+    cfg = MCConfig()
+    
+    # train
+    env,agent = env_agent_config(cfg,seed=1)
+    rewards, ma_rewards = train(cfg, env, agent)
+    make_dir(cfg.result_path, cfg.model_path)
+    agent.save(path=cfg.model_path)
+    save_results(rewards, ma_rewards, tag='train', path=cfg.result_path)
+    plot_rewards(rewards, ma_rewards, tag="train",
+                 algo=cfg.algo, path=cfg.result_path)
+    # eval
+    env,agent = env_agent_config(cfg,seed=10)
+    agent.load(path=cfg.model_path)
+    rewards,ma_rewards = eval(cfg,env,agent)
+    save_results(rewards,ma_rewards,tag='eval',path=cfg.result_path)
+    plot_rewards(rewards,ma_rewards,tag="eval",env=cfg.env,algo = cfg.algo,path=cfg.result_path)

codes/PPO/task0_train.py

@@ -5,7 +5,7 @@ Author: John
 Email: johnjim0816@gmail.com
 Date: 2021-03-22 16:18:10
 LastEditor: John
-LastEditTime: 2021-04-28 10:13:00
+LastEditTime: 2021-05-06 00:43:36
 Discription: 
 Environment: 
 '''
@@ -13,8 +13,8 @@ import sys,os
 curr_path = os.path.dirname(__file__)
 parent_path=os.path.dirname(curr_path) 
 sys.path.append(parent_path) # add current terminal path to sys.path
+
 import gym
-import numpy as np
 import torch
 import datetime
 from PPO.agent import PPO
@@ -29,6 +29,8 @@ class PPOConfig:
         self.algo = 'PPO'
         self.result_path = curr_path+"/results/" +self.env+'/'+curr_time+'/results/'  # path to save results
         self.model_path = curr_path+"/results/" +self.env+'/'+curr_time+'/models/'  # path to save models
+        self.train_eps = 200 # max training episodes
+        self.eval_eps = 50
         self.batch_size = 5
         self.gamma=0.99
         self.n_epochs = 4
@@ -38,10 +40,19 @@ class PPOConfig:
         self.policy_clip=0.2
         self.hidden_dim = 256
         self.update_fre = 20 # frequency of agent update
-        self.train_eps = 300 # max training episodes
         self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu") # check gpu
-        
+
+def env_agent_config(cfg,seed=1):
+    env = gym.make(cfg.env)  
+    env.seed(seed)
+    state_dim = env.observation_space.shape[0]
+    action_dim = env.action_space.n
+    agent = PPO(state_dim,action_dim,cfg)
+    return env,agent
+            
 def train(cfg,env,agent):
+    print('Start to train !')
+    print(f'Env:{cfg.env}, Algorithm:{cfg.algo}, Device:{cfg.device}')
     rewards= []
     ma_rewards = [] # moving average rewards
     running_steps = 0
@@ -65,17 +76,46 @@ def train(cfg,env,agent):
         else:
             ma_rewards.append(ep_reward)
         print(f"Episode:{i_ep+1}/{cfg.train_eps}, Reward:{ep_reward:.3f}")
+    print('Complete training！')
     return rewards,ma_rewards
 
+def eval(cfg,env,agent):
+    print('Start to eval !')
+    print(f'Env:{cfg.env}, Algorithm:{cfg.algo}, Device:{cfg.device}')
+    rewards= []
+    ma_rewards = [] # moving average rewards
+    for i_ep in range(cfg.eval_eps):
+        state = env.reset()
+        done = False
+        ep_reward = 0
+        while not done:
+            action, prob, val = agent.choose_action(state)
+            state_, reward, done, _ = env.step(action)
+            ep_reward += reward
+            state = state_
+        rewards.append(ep_reward)
+        if ma_rewards:
+            ma_rewards.append(
+                0.9*ma_rewards[-1]+0.1*ep_reward)
+        else:
+            ma_rewards.append(ep_reward)
+        print(f"Episode:{i_ep+1}/{cfg.train_eps}, Reward:{ep_reward:.3f}")
+    print('Complete evaling！')
+    return rewards,ma_rewards
+    
 if __name__ == '__main__':
     cfg  = PPOConfig()
-    env = gym.make(cfg.env)
-    env.seed(1) # Set seeds
-    state_dim=env.observation_space.shape[0]
-    action_dim=env.action_space.n
-    agent = PPO(state_dim,action_dim,cfg)
-    rewards,ma_rewards = train(cfg,env,agent)
-    make_dir(cfg.result_path,cfg.model_path)
+    # train
+    env,agent = env_agent_config(cfg,seed=1)
+    rewards, ma_rewards = train(cfg, env, agent)
+    make_dir(cfg.result_path, cfg.model_path)
     agent.save(path=cfg.model_path)
-    save_results(rewards,ma_rewards,tag='train',path=cfg.result_path)
-    plot_rewards(rewards,ma_rewards,tag="train",env=cfg.env,algo = cfg.algo,path=cfg.result_path)
+    save_results(rewards, ma_rewards, tag='train', path=cfg.result_path)
+    plot_rewards(rewards, ma_rewards, tag="train",
+                 algo=cfg.algo, path=cfg.result_path)
+    # eval
+    env,agent = env_agent_config(cfg,seed=10)
+    agent.load(path=cfg.model_path)
+    rewards,ma_rewards = eval(cfg,env,agent)
+    save_results(rewards,ma_rewards,tag='eval',path=cfg.result_path)
+    plot_rewards(rewards,ma_rewards,tag="eval",env=cfg.env,algo = cfg.algo,path=cfg.result_path)

codes/PPO/task1.py

@@ -1,97 +0,0 @@
-#!/usr/bin/env python
-# coding=utf-8
-'''
-Author: John
-Email: johnjim0816@gmail.com
-Date: 2021-03-22 16:18:10
-LastEditor: John
-LastEditTime: 2021-04-11 01:25:43
-Discription: 
-Environment: 
-'''
-import sys,os
-curr_path = os.path.dirname(__file__)
-parent_path=os.path.dirname(curr_path) 
-sys.path.append(parent_path) # add current terminal path to sys.path
-import gym
-import numpy as np
-import torch
-import datetime
-from PPO.agent import PPO
-from common.plot import plot_rewards
-from common.utils import save_results
-
-SEQUENCE = datetime.datetime.now().strftime("%Y%m%d-%H%M%S") # 获取当前时间
-SAVED_MODEL_PATH = os.path.split(os.path.abspath(__file__))[0]+"/saved_model/"+SEQUENCE+'/' # 生成保存的模型路径
-if not os.path.exists(os.path.split(os.path.abspath(__file__))[0]+"/saved_model/"): # 检测是否存在文件夹
-    os.mkdir(os.path.split(os.path.abspath(__file__))[0]+"/saved_model/")
-if not os.path.exists(SAVED_MODEL_PATH): # 检测是否存在文件夹
-    os.mkdir(SAVED_MODEL_PATH)
-RESULT_PATH = os.path.split(os.path.abspath(__file__))[0]+"/results/"+SEQUENCE+'/' # 存储reward的路径
-if not os.path.exists(os.path.split(os.path.abspath(__file__))[0]+"/results/"): # 检测是否存在文件夹
-    os.mkdir(os.path.split(os.path.abspath(__file__))[0]+"/results/")
-if not os.path.exists(RESULT_PATH): # 检测是否存在文件夹
-    os.mkdir(RESULT_PATH)
-
-class PPOConfig:
-    def __init__(self) -> None:
-        self.env = 'LunarLander-v2'
-        self.algo = 'PPO'
-        self.batch_size = 128
-        self.gamma=0.95
-        self.n_epochs = 4
-        self.actor_lr = 0.002
-        self.critic_lr = 0.005
-        self.gae_lambda=0.95
-        self.policy_clip=0.2
-        self.hidden_dim = 256
-        self.update_fre = 20 # frequency of agent update
-        self.train_eps = 300 # max training episodes
-        self.train_steps = 1000
-        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu") # check gpu
-        
-def train(cfg,env,agent):
-    best_reward = env.reward_range[0]
-    rewards= []
-    ma_rewards = [] # moving average rewards
-    avg_reward = 0
-    running_steps = 0
-    for i_episode in range(cfg.train_eps):
-        state = env.reset()
-        done = False
-        ep_reward = 0
-        # for i_step in range(cfg.train_steps):
-        while not done:
-            action, prob, val = agent.choose_action(state)
-            state_, reward, done, _ = env.step(action)
-            running_steps += 1
-            ep_reward += reward
-            agent.memory.push(state, action, prob, val, reward, done)
-            if running_steps % cfg.update_fre == 0:
-                agent.update()
-            state = state_
-            # if done:
-            #     break
-        rewards.append(ep_reward)
-        if ma_rewards:
-            ma_rewards.append(
-                0.9*ma_rewards[-1]+0.1*ep_reward)
-        else:
-            ma_rewards.append(ep_reward)
-        avg_reward = np.mean(rewards[-100:])
-        if avg_reward > best_reward:
-            best_reward = avg_reward
-            agent.save(path=SAVED_MODEL_PATH)
-        print('Episode:{}/{}, Reward:{:.1f}, avg reward:{:.1f}, Loss:{}'.format(i_episode+1,cfg.train_eps,ep_reward,avg_reward,agent.loss))
-    return rewards,ma_rewards
-
-if __name__ == '__main__':
-    cfg = PPOConfig()
-    env = gym.make(cfg.env)
-    env.seed(1)
-    state_dim=env.observation_space.shape[0]
-    action_dim=env.action_space.n
-    agent = PPO(state_dim,action_dim,cfg)
-    rewards,ma_rewards = train(cfg,env,agent)
-    save_results(rewards,ma_rewards,tag='train',path=RESULT_PATH)
-    plot_rewards(rewards,ma_rewards,tag="train",algo = cfg.algo,path=RESULT_PATH)

codes/PolicyGradient/agent.py

@@ -5,7 +5,7 @@ Author: John
 Email: johnjim0816@gmail.com
 Date: 2020-11-22 23:27:44
 LastEditor: John
-LastEditTime: 2021-03-23 16:37:14
+LastEditTime: 2021-05-05 17:33:10
 Discription: 
 Environment: 
 '''
@@ -64,7 +64,7 @@ class PolicyGradient:
             # print(loss)
             loss.backward()
         self.optimizer.step()
-    def save_model(self,path):
+    def save(self,path):
         torch.save(self.policy_net.state_dict(), path+'pg_checkpoint.pt')
-    def load_model(self,path):
+    def load(self,path):
         self.policy_net.load_state_dict(torch.load(path+'pg_checkpoint.pt')) 

codes/PolicyGradient/main.py

@@ -1,89 +0,0 @@
-#!/usr/bin/env python
-# coding=utf-8
-'''
-Author: John
-Email: johnjim0816@gmail.com
-Date: 2020-11-22 23:21:53
-LastEditor: John
-LastEditTime: 2021-03-23 16:38:54
-Discription: 
-Environment: 
-'''
-import sys,os
-sys.path.append(os.getcwd()) # add current terminal path to sys.path
-from itertools import count
-import datetime
-import gym
-from PolicyGradient.agent import PolicyGradient
-from common.plot import plot_rewards
-from common.utils import save_results
-
-SEQUENCE = datetime.datetime.now().strftime("%Y%m%d-%H%M%S") # obtain current time
-SAVED_MODEL_PATH = os.path.split(os.path.abspath(__file__))[0]+"/saved_model/"+SEQUENCE+'/'  # path to save model
-if not os.path.exists(os.path.split(os.path.abspath(__file__))[0]+"/saved_model/"): 
-    os.mkdir(os.path.split(os.path.abspath(__file__))[0]+"/saved_model/")
-if not os.path.exists(SAVED_MODEL_PATH):
-    os.mkdir(SAVED_MODEL_PATH)
-RESULT_PATH = os.path.split(os.path.abspath(__file__))[0]+"/results/"+SEQUENCE+'/' # path to save rewards
-if not os.path.exists(os.path.split(os.path.abspath(__file__))[0]+"/results/"): 
-    os.mkdir(os.path.split(os.path.abspath(__file__))[0]+"/results/")
-if not os.path.exists(RESULT_PATH): 
-    os.mkdir(RESULT_PATH)
-
-class PGConfig:
-    def __init__(self):
-        self.train_eps = 300 # 训练的episode数目
-        self.batch_size = 8
-        self.lr = 0.01 # learning rate
-        self.gamma = 0.99
-        self.hidden_dim = 36 # dimmension of hidden layer
-        
-def train(cfg,env,agent):
-    '''下面带pool都是存放的transition序列用于gradient'''
-    state_pool = [] # 存放每batch_size个episode的state序列
-    action_pool = []
-    reward_pool = [] 
-    ''' 存储每个episode的reward用于绘图'''
-    rewards = []
-    ma_rewards = []
-    for i_episode in range(cfg.train_eps):
-        state = env.reset()
-        ep_reward = 0
-        for _ in count():
-            action = agent.choose_action(state) # 根据当前环境state选择action
-            next_state, reward, done, _ = env.step(action)
-            ep_reward += reward
-            if done:
-                reward = 0
-            state_pool.append(state)
-            action_pool.append(float(action))
-            reward_pool.append(reward)
-            state = next_state
-            if done:
-                print('Episode:', i_episode, ' Reward:',  ep_reward)
-                break
-        if i_episode > 0 and i_episode % cfg.batch_size == 0:
-            agent.update(reward_pool,state_pool,action_pool)
-            state_pool = [] # 每个episode的state
-            action_pool = []
-            reward_pool = []
-        rewards.append(ep_reward)
-        if ma_rewards:
-            ma_rewards.append(
-                0.9*ma_rewards[-1]+0.1*ep_reward)
-        else:
-            ma_rewards.append(ep_reward)
-    print('complete training！')
-    return rewards, ma_rewards
-            
-if __name__ == "__main__":
-    cfg = PGConfig()
-    env = gym.make('CartPole-v0') # 可google为什么unwrapped gym，此处一般不需要
-    env.seed(1) # 设置env随机种子
-    state_dim = env.observation_space.shape[0]
-    action_dim = env.action_space.n
-    agent  = PolicyGradient(state_dim,cfg)
-    rewards, ma_rewards = train(cfg,env,agent)
-    agent.save_model(SAVED_MODEL_PATH)
-    save_results(rewards,ma_rewards,tag='train',path=RESULT_PATH)
-    plot_rewards(rewards,ma_rewards,tag="train",algo = "Policy Gradient",path=RESULT_PATH)

codes/PolicyGradient/task0_train.py

@@ -0,0 +1,136 @@
+#!/usr/bin/env python
+# coding=utf-8
+'''
+Author: John
+Email: johnjim0816@gmail.com
+Date: 2020-11-22 23:21:53
+LastEditor: John
+LastEditTime: 2021-05-05 17:35:20
+Discription: 
+Environment: 
+'''
+import sys,os
+curr_path = os.path.dirname(__file__)
+parent_path = os.path.dirname(curr_path)
+sys.path.append(parent_path)  # add current terminal path to sys.path
+
+import gym
+import torch
+import datetime
+from itertools import count
+
+from PolicyGradient.agent import PolicyGradient
+from common.plot import plot_rewards
+from common.utils import save_results,make_dir
+
+curr_time = datetime.datetime.now().strftime(
+    "%Y%m%d-%H%M%S")  # obtain current time
+
+class PGConfig:
+    def __init__(self):
+        self.algo = "PolicyGradient"  # name of algo
+        self.env = 'CartPole-v0'
+        self.result_path = curr_path+"/outputs/" + self.env + \
+            '/'+curr_time+'/results/'  # path to save results
+        self.model_path = curr_path+"/outputs/" + self.env + \
+            '/'+curr_time+'/models/'  # path to save models
+        self.train_eps = 300 # 训练的episode数目
+        self.eval_eps = 50
+        self.batch_size = 8
+        self.lr = 0.01 # learning rate
+        self.gamma = 0.99
+        self.hidden_dim = 36 # dimmension of hidden layer
+        self.device = torch.device(
+            "cuda" if torch.cuda.is_available() else "cpu")  # check gpu
+
+
+def env_agent_config(cfg,seed=1):
+    env = gym.make(cfg.env)  
+    env.seed(seed)
+    state_dim = env.observation_space.shape[0]
+    agent = PolicyGradient(state_dim,cfg)
+    return env,agent
+
+def train(cfg,env,agent):
+    print('Start to eval !')
+    print(f'Env:{cfg.env}, Algorithm:{cfg.algo}, Device:{cfg.device}')
+    state_pool = [] # 存放每batch_size个episode的state序列
+    action_pool = []
+    reward_pool = [] 
+    rewards = []
+    ma_rewards = []
+    for i_episode in range(cfg.train_eps):
+        state = env.reset()
+        ep_reward = 0
+        for _ in count():
+            action = agent.choose_action(state) # 根据当前环境state选择action
+            next_state, reward, done, _ = env.step(action)
+            ep_reward += reward
+            if done:
+                reward = 0
+            state_pool.append(state)
+            action_pool.append(float(action))
+            reward_pool.append(reward)
+            state = next_state
+            if done:
+                print('Episode:', i_episode, ' Reward:',  ep_reward)
+                break
+        if i_episode > 0 and i_episode % cfg.batch_size == 0:
+            agent.update(reward_pool,state_pool,action_pool)
+            state_pool = [] # 每个episode的state
+            action_pool = []
+            reward_pool = []
+        rewards.append(ep_reward)
+        if ma_rewards:
+            ma_rewards.append(
+                0.9*ma_rewards[-1]+0.1*ep_reward)
+        else:
+            ma_rewards.append(ep_reward)
+    print('complete training！')
+    return rewards, ma_rewards
+            
+
+def eval(cfg,env,agent):
+    print('Start to eval !')
+    print(f'Env:{cfg.env}, Algorithm:{cfg.algo}, Device:{cfg.device}')
+    rewards = []
+    ma_rewards = []
+    for i_episode in range(cfg.eval_eps):
+        state = env.reset()
+        ep_reward = 0
+        for _ in count():
+            action = agent.choose_action(state) # 根据当前环境state选择action
+            next_state, reward, done, _ = env.step(action)
+            ep_reward += reward
+            if done:
+                reward = 0
+            state = next_state
+            if done:
+                print('Episode:', i_episode, ' Reward:',  ep_reward)
+                break
+        rewards.append(ep_reward)
+        if ma_rewards:
+            ma_rewards.append(
+                0.9*ma_rewards[-1]+0.1*ep_reward)
+        else:
+            ma_rewards.append(ep_reward)
+    print('complete evaling！')
+    return rewards, ma_rewards
+if __name__ == "__main__":
+    cfg = PGConfig()
+
+     # train
+    env,agent = env_agent_config(cfg,seed=1)
+    rewards, ma_rewards = train(cfg, env, agent)
+    make_dir(cfg.result_path, cfg.model_path)
+    agent.save(path=cfg.model_path)
+    save_results(rewards, ma_rewards, tag='train', path=cfg.result_path)
+    plot_rewards(rewards, ma_rewards, tag="train",
+                 algo=cfg.algo, path=cfg.result_path)
+    # eval
+    env,agent = env_agent_config(cfg,seed=10)
+    agent.load(path=cfg.model_path)
+    rewards,ma_rewards = eval(cfg,env,agent)
+    save_results(rewards,ma_rewards,tag='eval',path=cfg.result_path)
+    plot_rewards(rewards,ma_rewards,tag="eval",env=cfg.env,algo = cfg.algo,path=cfg.result_path)
+

codes/RandomPolicy/main.py

@@ -1,108 +0,0 @@
-#!/usr/bin/env python
-# coding=utf-8
-'''
-Author: JiangJi
-Email: johnjim0816@gmail.com
-Date: 2021-04-21 11:07:57
-LastEditor: JiangJi
-LastEditTime: 2021-04-21 11:15:00
-Discription: 
-Environment: 
-'''
-import sys,os
-curr_path = os.path.dirname(__file__)
-parent_path=os.path.dirname(curr_path) 
-sys.path.append(parent_path) # add current terminal path to sys.path
-
-import torch
-import gym
-import numpy as np
-import datetime
-
-from common.plot import plot_rewards
-from common.utils import save_results,make_dir
-
-curr_time = datetime.datetime.now().strftime("%Y%m%d-%H%M%S") # obtain current time
-	
-
-class TD3Config:
-	def __init__(self) -> None:
-		self.algo = 'TD3'
-		self.env = 'HalfCheetah-v2'
-		self.seed = 0
-		self.result_path = curr_path+"/results/" +self.env+'/'+curr_time+'/results/'  # path to save results
-		self.model_path = curr_path+"/results/" +self.env+'/'+curr_time+'/models/'  # path to save models
-		self.eval_freq = 5e3 # How often (time steps) we evaluate
-		# self.train_eps = 800
-		self.max_timestep = 4000000 # Max time steps to run environment
-		
-# Runs policy for X episodes and returns average reward
-# A fixed seed is used for the eval environment
-def eval(env_name,seed, eval_episodes=10):
-	eval_env = gym.make(env_name)
-	eval_env.seed(seed + 100)
-	avg_reward = 0.
-	for _ in range(eval_episodes):
-		state, done = eval_env.reset(), False
-		while not done:
-			# eval_env.render()
-			action = eval_env.action_space.sample()
-			state, reward, done, _ = eval_env.step(action)
-			avg_reward += reward
-	avg_reward /= eval_episodes
-	print("---------------------------------------")
-	print(f"Evaluation over {eval_episodes} episodes: {avg_reward:.3f}")
-	print("---------------------------------------")
-	return avg_reward
-
-def train(cfg,env):
-	# Evaluate untrained policy
-	evaluations = [eval(cfg.env, cfg.seed)]
-	state, done = env.reset(), False
-	ep_reward = 0
-	ep_timesteps = 0
-	episode_num = 0
-	rewards = []
-	ma_rewards = [] # moveing average reward
-	for t in range(int(cfg.max_timestep)):
-		ep_timesteps += 1
-		# Select action randomly
-		action = env.action_space.sample()
-		# Perform action
-		next_state, reward, done, _ = env.step(action) 
-		state = next_state
-		ep_reward += reward
-		if done: 
-			# +1 to account for 0 indexing. +0 on ep_timesteps since it will increment +1 even if done=True
-			print(f"Episode:{episode_num+1}, Episode T:{ep_timesteps}, Reward:{ep_reward:.3f}")
-			# Reset environment
-			state, done = env.reset(), False
-			rewards.append(ep_reward)
-			# 计算滑动窗口的reward
-			if ma_rewards:
-				ma_rewards.append(0.9*ma_rewards[-1]+0.1*ep_reward)
-			else:
-				ma_rewards.append(ep_reward) 
-			ep_reward = 0
-			ep_timesteps = 0
-			episode_num += 1 
-		# Evaluate episode
-		if (t + 1) % cfg.eval_freq == 0:
-			evaluations.append(eval(cfg.env, cfg.seed))
-	return rewards, ma_rewards
-
-if __name__ == "__main__":
-	cfg  = TD3Config()
-	env = gym.make(cfg.env)
-	env.seed(cfg.seed) # Set seeds
-	torch.manual_seed(cfg.seed)
-	np.random.seed(cfg.seed)
-	rewards,ma_rewards = train(cfg,env)
-	make_dir(cfg.result_path)
-	save_results(rewards,ma_rewards,tag='train',path=cfg.result_path)
-	plot_rewards(rewards,ma_rewards,tag="train",env=cfg.env,algo = cfg.algo,path=cfg.result_path)
-	# cfg.result_path = './TD3/results/HalfCheetah-v2/20210416-130341/'
-	# agent.load(cfg.result_path)
-	# eval(cfg.env,agent, cfg.seed)
-	
-