update

codes/DDPG/agent.py

@@ -5,7 +5,7 @@
 @Email: johnjim0816@gmail.com
 @Date: 2020-06-09 20:25:52
 @LastEditor: John
-LastEditTime: 2021-03-31 00:56:32
+LastEditTime: 2021-05-04 14:50:17
 @Discription: 
 @Environment: python 3.7.7
 '''
@@ -26,6 +26,7 @@ class DDPG:
         self.target_critic = Critic(state_dim, action_dim, cfg.hidden_dim).to(cfg.device)
         self.target_actor = Actor(state_dim, action_dim, cfg.hidden_dim).to(cfg.device)
 
+        # copy parameters to target net
         for target_param, param in zip(self.target_critic.parameters(), self.critic.parameters()):
             target_param.data.copy_(param.data)
         for target_param, param in zip(self.target_actor.parameters(), self.actor.parameters()):
@@ -42,7 +43,6 @@ class DDPG:
     def choose_action(self, state):
         state = torch.FloatTensor(state).unsqueeze(0).to(self.device)
         action = self.actor(state)
-        # torch.detach()用于切断反向传播
         return action.detach().cpu().numpy()[0, 0]
 
     def update(self):
@@ -50,13 +50,13 @@ class DDPG:
             return
         state, action, reward, next_state, done = self.memory.sample(
             self.batch_size)
-        # 将所有变量转为张量
+        # convert variables to Tensor
         state = torch.FloatTensor(state).to(self.device)
         next_state = torch.FloatTensor(next_state).to(self.device)
         action = torch.FloatTensor(action).to(self.device)
         reward = torch.FloatTensor(reward).unsqueeze(1).to(self.device)
         done = torch.FloatTensor(np.float32(done)).unsqueeze(1).to(self.device)
-        # 注意critic将(s_t,a)作为输入
+       
         policy_loss = self.critic(state, self.actor(state))
         policy_loss = -policy_loss.mean()
         next_action = self.target_actor(next_state)

codes/DDPG/main.py

@@ -1,94 +0,0 @@
-#!/usr/bin/env python
-# coding=utf-8
-'''
-@Author: John
-@Email: johnjim0816@gmail.com
-@Date: 2020-06-11 20:58:21
-@LastEditor: John
-LastEditTime: 2021-04-29 01:58:50
-@Discription: 
-@Environment: python 3.7.7
-'''
-import sys,os
-from pathlib import Path
-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 DDPG.agent import DDPG
-from DDPG.env import NormalizedActions,OUNoise
-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 = 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 DDPGConfig:
-    def __init__(self):
-        self.env = 'Pendulum-v0'
-        self.algo = 'DDPG'
-        self.gamma = 0.99
-        self.critic_lr = 1e-3  
-        self.actor_lr = 1e-4 
-        self.memory_capacity = 10000
-        self.batch_size = 128
-        self.train_eps =300
-        self.eval_eps = 200
-        self.eval_steps = 200
-        self.target_update = 4
-        self.hidden_dim = 30
-        self.soft_tau=1e-2
-        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-def train(cfg,env,agent):
-    print('Start to train ! ')
-    ou_noise = OUNoise(env.action_space) # action noise
-    rewards = []
-    ma_rewards = [] # moving average rewards
-    ep_steps = []
-    for i_episode in range(cfg.train_eps):
-        state = env.reset()
-        ou_noise.reset()
-        done = False
-        ep_reward = 0
-        i_step = 0
-        while not done:
-            i_step += 1
-            action = agent.choose_action(state)
-            action = ou_noise.get_action(action, i_step)  # 即paper中的random process
-            next_state, reward, done, _ = env.step(action)
-            ep_reward += reward
-            agent.memory.push(state, action, reward, next_state, done)
-            agent.update()
-            state = next_state
-        print('Episode:{}/{}, Reward:{}'.format(i_episode+1,cfg.train_eps,ep_reward))
-        ep_steps.append(i_step)
-        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 = DDPGConfig()
-    env = NormalizedActions(gym.make("Pendulum-v0"))
-    env.seed(1) # 设置env随机种子
-    state_dim = env.observation_space.shape[0]
-    action_dim = env.action_space.shape[0]
-    agent = DDPG(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)
-    

codes/DDPG/task0_train.py

@@ -0,0 +1,135 @@
+#!/usr/bin/env python
+# coding=utf-8
+'''
+@Author: John
+@Email: johnjim0816@gmail.com
+@Date: 2020-06-11 20:58:21
+@LastEditor: John
+LastEditTime: 2021-05-04 14:49:45
+@Discription: 
+@Environment: python 3.7.7
+'''
+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 datetime
+import gym
+import torch
+
+from DDPG.env import NormalizedActions, OUNoise
+from DDPG.agent import DDPG
+from common.utils import save_results,make_dir
+from common.plot import plot_rewards
+
+curr_time = datetime.datetime.now().strftime(
+    "%Y%m%d-%H%M%S")  # obtain current time
+
+
+class DDPGConfig:
+    def __init__(self):
+        self.algo = 'DDPG'
+        self.env = 'Pendulum-v0' # env name
+        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.critic_lr = 1e-3
+        self.actor_lr = 1e-4
+        self.memory_capacity = 10000
+        self.batch_size = 128
+        self.train_eps = 300
+        self.eval_eps = 50
+        self.eval_steps = 200
+        self.target_update = 4
+        self.hidden_dim = 30
+        self.soft_tau = 1e-2
+        self.device = torch.device(
+            "cuda" if torch.cuda.is_available() else "cpu")
+
+def env_agent_config(cfg,seed=1):
+    env = NormalizedActions(gym.make(cfg.env)) 
+    env.seed(seed)
+    state_dim = env.observation_space.shape[0]
+    action_dim = env.action_space.shape[0]
+    agent = DDPG(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}')
+    ou_noise = OUNoise(env.action_space)  # action noise
+    rewards = []
+    ma_rewards = []  # moving average rewards
+    for i_episode in range(cfg.train_eps):
+        state = env.reset()
+        ou_noise.reset()
+        done = False
+        ep_reward = 0
+        i_step = 0
+        while not done:
+            i_step += 1
+            action = agent.choose_action(state)
+            action = ou_noise.get_action(
+                action, i_step)  # 即paper中的random process
+            next_state, reward, done, _ = env.step(action)
+            ep_reward += reward
+            agent.memory.push(state, action, reward, next_state, done)
+            agent.update()
+            state = next_state
+        print('Episode:{}/{}, Reward:{}'.format(i_episode+1, cfg.train_eps, ep_reward))
+        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 = []  # moving average rewards
+    for i_episode in range(cfg.eval_eps):
+        state = env.reset()
+        done = False
+        ep_reward = 0
+        i_step = 0
+        while not done:
+            i_step += 1
+            action = agent.choose_action(state)
+            next_state, reward, done, _ = env.step(action)
+            ep_reward += reward
+            state = next_state
+        print('Episode:{}/{}, Reward:{}'.format(i_episode+1, cfg.train_eps, ep_reward))
+        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 Eval！')
+    return rewards, ma_rewards
+
+
+if __name__ == "__main__":
+    cfg = DDPGConfig()
+
+    # 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/DQN/README.md

@@ -1,6 +1,7 @@
 # DQN
-#TODO
+
 ## 原理简介
+
 DQN是Q-leanning算法的优化和延伸，Q-leaning中使用有限的Q表存储值的信息，而DQN中则用神经网络替代Q表存储信息，这样更适用于高维的情况，相关知识基础可参考[datawhale李宏毅笔记-Q学习](https://datawhalechina.github.io/easy-rl/#/chapter6/chapter6)。
 
 论文方面主要可以参考两篇，一篇就是2013年谷歌DeepMind团队的[Playing Atari with Deep Reinforcement Learning](https://www.cs.toronto.edu/~vmnih/docs/dqn.pdf)，一篇是也是他们团队后来在Nature杂志上发表的[Human-level control through deep reinforcement learning](https://web.stanford.edu/class/psych209/Readings/MnihEtAlHassibis15NatureControlDeepRL.pdf)。后者在算法层面增加target q-net，也可以叫做Nature DQN。
@@ -15,7 +16,7 @@ https://blog.csdn.net/JohnJim0/article/details/109557173)
 
 <img src="assets/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L0pvaG5KaW0w,size_16,color_FFFFFF,t_70.png" alt="img" style="zoom:50%;" />
 
-## 代码实战
+## 代码实现
 
 ### RL接口
 
@@ -24,23 +25,26 @@ https://blog.csdn.net/JohnJim0/article/details/109557173)
 for i_episode in range(MAX_EPISODES):
 	state = env.reset() # reset环境状态
 	for i_step in range(MAX_STEPS):
-		 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
-         agent.memory.push(state, action, reward, next_state, done) # 将state等这些transition存入memory
-         agent.update() # 每步更新网络
-         state = next_state # 跳转到下一个状态
-         if done:
-         	break        
+    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
+    agent.memory.push(state, action, reward, next_state, done) # 将state等这些transition存入memory
+    agent.update() # 每步更新网络
+    state = next_state # 跳转到下一个状态
+    if done:
+    	break        
 ```
-如上，首先需要循环多个episode训练，在每个episode中，首先需要重置环境，然后开始探索，每个episode加一个MAX_STEPS(也可以使用while not done, 加这个max_steps有时是因为比如gym环境训练目标就是在200个step下达到200的reward)，接下来的流程如下：
+每个episode加一个MAX_STEPS，也可以使用while not done, 加这个max_steps有时是因为比如gym环境训练目标就是在200个step下达到200的reward，或者是当完成一个episode的步数较多时也可以设置，基本流程跟所有伪代码一致，如下：
 1. agent选择动作
-2. 环境根据agent的动作反馈出新的state和reward
+2. 环境根据agent的动作反馈出next_state和reward
 3. agent进行更新，如有memory就会将transition(包含state，reward，action等)存入memory中
 4. 跳转到下一个状态
-如果提前done了，就跳出for循环，进行下一个episode的训练。
+5. 如果done了，就跳出循环，进行下一个episode的训练。
+
+想要实现完整的算法还需要创建Qnet，Replaybuffer等类
 
 ### 两个Q网络
-前面讲了Nature DQN中有两个Q网络，一个是policy_net，一个是延时更新的target_net，两个网络的结构是一模一样的，如下(见```model.py```)：
+
+上文讲了Nature DQN中有两个Q网络，一个是policy_net，一个是延时更新的target_net，两个网络的结构是一模一样的，如下(见```model.py```)，注意DQN使用的Qnet就是全连接网络即FCH：
 ```python
 import torch.nn as nn
 import torch.nn.functional as F
@@ -62,30 +66,12 @@ class FCN(nn.Module):
         x = F.relu(self.fc2(x))
         return self.fc3(x)
 ```
-输入为state，输出为action，注意根据state和action的维度调整隐藏层的层数，这里设为128
+输入为state_dim，输出为action_dim，包含一个128维度的隐藏层，这里根据需要可增加隐藏层维度和数量，然后一般使用relu激活函数，这里跟深度学习的网路设置是一样的。
+
+### Replay Buffer
+
+然后就是Replay Memory了，其作用主要是是克服经验数据的相关性（correlated data）和非平稳分布（non-stationary distribution）问题，实现如下(见```memory.py```)：
 
-在```agent.py```中我们定义强化学习算法，包括```choose_action```和```update```两个主要函数，初始化中：
-```python
-self.policy_net = FCN(state_dim, action_dim).to(self.device)
-self.target_net = FCN(state_dim, action_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
-# 可查parameters()与state_dict()的区别，前者require_grad=True
-```
-可以看到policy_net跟target_net结构和初始参数一样，但在更新的时候target是每隔一段episode更新的，如下(见```main.py```)：
-```python
-# 更新target network，复制DQN中的所有weights and biases
-if i_episode % cfg.target_update == 0:
-	agent.target_net.load_state_dict(agent.policy_net.state_dict())
-```
-可以调整```cfg.target_update```，注意该变量不要调得太大，否则会收敛很慢，我们最后保存的模型也是这个target_net，如下(见```agent.py```)：
-```python
-def save_model(self,path):
-	torch.save(self.target_net.state_dict(), path)
-```
-### Replay Memory
-然后就是Replay Memory了，如下(见```memory.py```)：
 ```python
 import random
 import numpy as np
@@ -111,11 +97,120 @@ class ReplayBuffer:
     def __len__(self):
         return len(self.buffer)
 ```
-其实比较简单，主要包括push和sample两个步骤，push是将transitions放到memory中，sample是从memory随机抽取一些transition。
 
-最后结果如下：
+参数capacity表示buffer的容量，主要包括push和sample两个步骤，push是将transitions放到memory中，sample是从memory随机抽取一些transition。
 
-![rewards_curve_train](assets/rewards_curve_train.png)
+### Agent类
+
+在```agent.py```中我们定义强化学习算法类，包括```choose_action```(选择动作，使用e-greedy策略时会多一个```predict```函数，下面会将到)和```update```(更新)等函数。
+
+在类中建立两个网络，以及optimizer和memory，
+
+```python
+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)
+for target_param, param in zip(self.target_net.parameters(),self.policy_net.parameters()): # copy params from policy net
+    target_param.data.copy_(param.data)
+self.optimizer = optim.Adam(self.policy_net.parameters(), lr=cfg.lr)
+self.memory = ReplayBuffer(cfg.memory_capacity)
+```
+然后是选择action：
+
+```python
+def choose_action(self, state):
+        '''选择动作
+        '''
+    self.frame_idx += 1
+    if random.random() > self.epsilon(self.frame_idx):
+        action = self.predict(state)
+    else:
+        action = random.randrange(self.action_dim)
+    return action
+```
+
+这里使用e-greedy策略，即设置一个参数epsilon，如果生成的随机数大于epsilon，就根据网络预测的选择action，否则还是随机选择action，这个epsilon是会逐渐减小的，可以使用线性或者指数减小的方式，但不会减小到零，这样在训练稳定时还能保持一定的探索，这部分可以学习探索与利用(exploration and exploition)相关知识。
+
+上面讲到的预测函数其实就是根据state选取q值最大的action，如下：
+
+```python
+def predict(self,state):
+    with torch.no_grad():
+        state = torch.tensor([state], device=self.device, dtype=torch.float32)
+        q_values = self.policy_net(state)
+        action = q_values.max(1)[1].item()
+```
+
+然后是更新函数了：
+
+```python
+def update(self):
+
+        if len(self.memory) < self.batch_size:
+            return
+        # 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
+        例如tensor([[-4.5543e-02, -2.3910e-01,  1.8344e-02,  2.3158e-01],...,[-1.8615e-02, -2.3921e-01, -1.1791e-02,  2.3400e-01]])'''
+        state_batch = torch.tensor(
+            state_batch, device=self.device, dtype=torch.float)
+        action_batch = torch.tensor(action_batch, device=self.device).unsqueeze(
+            1)  # 例如tensor([[1],...,[0]])
+        reward_batch = torch.tensor(
+            reward_batch, device=self.device, dtype=torch.float)  # tensor([1., 1.,...,1])
+        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)
+
+        '''计算当前(s_t,a)对应的Q(s_t, a)'''
+        '''torch.gather:对于a=torch.Tensor([[1,2],[3,4]]),那么a.gather(1,torch.Tensor([[0],[1]]))=torch.Tensor([[1],[3]])'''
+        q_values = self.policy_net(state_batch).gather(
+            dim=1, index=action_batch)  # 等价于self.forward
+        # 计算所有next states的V(s_{t+1})，即通过target_net中选取reward最大的对应states
+        next_q_values = self.target_net(next_state_batch).max(
+            1)[0].detach()  # 比如tensor([ 0.0060, -0.0171,...,])
+        # 计算 expected_q_value
+        # 对于终止状态，此时done_batch[0]=1, 对应的expected_q_value等于reward
+        expected_q_values = reward_batch + \
+            self.gamma * next_q_values * (1-done_batch)
+        # self.loss = F.smooth_l1_loss(q_values,expected_q_values.unsqueeze(1)) # 计算 Huber loss
+        loss = nn.MSELoss()(q_values, expected_q_values.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
+        # loss.backward()使用backpropagation计算loss相对于所有parameters(需要gradients)的微分
+        loss.backward()
+        # for param in self.policy_net.parameters():  # clip防止梯度爆炸
+        #     param.grad.data.clamp_(-1, 1)
+        self.optimizer.step()  # 更新模型
+```
+
+更新遵循伪代码的以下部分：
+
+<img src="assets/image-20210507162813393.png" alt="image-20210507162813393" style="zoom:50%;" />
+
+首先从replay buffer中选取一个batch的数据，计算loss，然后进行minibatch SGD。
+
+然后是保存与加载模型的部分，如下：
+
+```python
+def save(self, path):
+        torch.save(self.target_net.state_dict(), path+'dqn_checkpoint.pth')
+def load(self, path):
+    self.target_net.load_state_dict(torch.load(path+'dqn_checkpoint.pth'))
+    for target_param, param in zip(self.target_net.parameters(), self.policy_net.parameters()):
+        param.data.copy_(target_param.data)
+```
+
+
+
+### 实验结果
+
+训练结果如下：
+
+<img src="assets/train_rewards_curve.png" alt="train_rewards_curve" style="zoom: 67%;" />
+
+<img src="assets/eval_rewards_curve.png" alt="eval_rewards_curve" style="zoom:67%;" />
 
 ## 参考
 

codes/DQN/agent.py

@@ -5,7 +5,7 @@
 @Email: johnjim0816@gmail.com
 @Date: 2020-06-12 00:50:49
 @LastEditor: John
-LastEditTime: 2021-04-29 22:19:18
+LastEditTime: 2021-05-07 16:30:05
 @Discription: 
 @Environment: python 3.7.7
 '''
@@ -35,15 +35,13 @@ class DQN:
             (cfg.epsilon_start - cfg.epsilon_end) * \
             math.exp(-1. * frame_idx / cfg.epsilon_decay)
         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)
-        for target_param, param in zip(self.target_net.parameters(), self.policy_net.parameters()):
+        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)
+        for target_param, param in zip(self.target_net.parameters(),self.policy_net.parameters()): # copy params from policy net
             target_param.data.copy_(param.data)
         self.optimizer = optim.Adam(self.policy_net.parameters(), lr=cfg.lr)
-        self.loss = 0
         self.memory = ReplayBuffer(cfg.memory_capacity)
+        
 
     def choose_action(self, state):
         '''选择动作
@@ -92,11 +90,11 @@ class DQN:
         expected_q_values = reward_batch + \
             self.gamma * next_q_values * (1-done_batch)
         # self.loss = F.smooth_l1_loss(q_values,expected_q_values.unsqueeze(1)) # 计算 Huber loss
-        self.loss = nn.MSELoss()(q_values, expected_q_values.unsqueeze(1))  # 计算 均方误差loss
+        loss = nn.MSELoss()(q_values, expected_q_values.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
         # loss.backward()使用backpropagation计算loss相对于所有parameters(需要gradients)的微分
-        self.loss.backward()
+        loss.backward()
         # for param in self.policy_net.parameters():  # clip防止梯度爆炸
         #     param.grad.data.clamp_(-1, 1)
         self.optimizer.step()  # 更新模型

codes/DQN/task0_train.py

@@ -5,7 +5,7 @@
 @Email: johnjim0816@gmail.com
 @Date: 2020-06-12 00:48:57
 @LastEditor: John
-LastEditTime: 2021-04-29 22:23:38
+LastEditTime: 2021-05-05 16:49:15
 @Discription: 
 @Environment: python 3.7.7
 '''
@@ -14,20 +14,17 @@ 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, del_empty_dir
+from common.utils import save_results, make_dir
 from common.plot import plot_rewards
 from DQN.agent import DQN
 
-
-
 curr_time = datetime.datetime.now().strftime(
     "%Y%m%d-%H%M%S")  # obtain current time
 
-
 class DQNConfig:
     def __init__(self):
         self.algo = "DQN"  # name of algo
@@ -35,21 +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)
@@ -63,7 +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
@@ -76,11 +73,12 @@ 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)
-        # 计算滑动窗口的reward
+        # save ma rewards
         if ma_rewards:
             ma_rewards.append(0.9*ma_rewards[-1]+0.1*ep_reward)
         else:
@@ -89,15 +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
@@ -106,11 +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)
@@ -118,7 +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/DQN_cnn/README.md

@@ -1,2 +0,0 @@
-# DQN with cnn
-原理与[DQN](../DQN)相同，只是将神经网络换成卷积神经网络，用于二维观测信息(state或obervation)

codes/DQN_cnn/agent.py

@@ -1,107 +0,0 @@
-import random
-import math
-import torch
-import torch.optim as optim
-import torch.nn.functional as F
-from DQN_cnn.memory import ReplayBuffer
-from DQN_cnn.model import CNN
-
-
-class DQNcnn:
-    def __init__(self, screen_height,screen_width, action_dim, cfg):
-
-        self.device = cfg.device  
-        self.action_dim = action_dim
-        self.gamma = cfg.gamma
-        # e-greedy策略相关参数
-        self.actions_count = 0 
-        self.epsilon = 0
-        self.epsilon_start = cfg.epsilon_start
-        self.epsilon_end = cfg.epsilon_end
-        self.epsilon_decay = cfg.epsilon_decay
-        self.batch_size = cfg.batch_size
-        self.policy_net = CNN(screen_height, screen_width,
-                              action_dim).to(self.device)
-        self.target_net = CNN(screen_height, screen_width,
-                              action_dim).to(self.device)
-        self.target_net.load_state_dict(self.policy_net.state_dict()) # target_net的初始模型参数完全复制policy_net
-        self.target_net.eval()  # 不启用 BatchNormalization 和 Dropout
-        self.optimizer = optim.RMSprop(self.policy_net.parameters(),lr = cfg.lr) # 可查parameters()与state_dict()的区别，前者require_grad=True
-        self.loss = 0
-        self.memory = ReplayBuffer(cfg.memory_capacity)
-        
-
-    def choose_action(self, state):
-        '''选择动作
-        Args:
-            state [array]: [description]
-        Returns:
-            action [array]: [description]
-        '''
-        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:   
-            with torch.no_grad():
-                q_value = self.policy_net(state) # q_value比如tensor([[-0.2522,  0.3887]])
-                # tensor.max(1)返回每行的最大值以及对应的下标，
-                # 如torch.return_types.max(values=tensor([10.3587]),indices=tensor([0]))
-                # 所以tensor.max(1)[1]返回最大值对应的下标，即action
-                action = q_value.max(1)[1].view(1, 1)  # 注意这里action是个张量，如tensor([1])
-                return action
-        else:
-            return torch.tensor([[random.randrange(self.action_dim)]], device=self.device, dtype=torch.long)
-
-    def update(self):
-        if len(self.memory) < self.batch_size:
-            return
-        transitions = self.memory.sample(self.batch_size)
-        # Transpose the batch (see https://stackoverflow.com/a/19343/3343043 for
-        # detailed explanation). This converts batch-array of Transitions
-        # to Transition of batch-arrays.
-        batch = self.memory.Transition(*zip(*transitions))
-
-        # Compute a mask of non-final states and concatenate the batch elements
-        # (a final state would've been the one after which simulation ended)
-        non_final_mask = torch.tensor(tuple(map(lambda s: s is not None,
-                                                batch.state_)), device=self.device, dtype=torch.bool)
-        
-        non_final_state_s = torch.cat([s for s in batch.state_
-                                           if s is not None])
-        state_batch = torch.cat(batch.state)   
-        action_batch = torch.cat(batch.action) 
-        reward_batch = torch.cat(batch.reward) # tensor([1., 1.,...,])
-        
-
-        # Compute Q(s_t, a) - the model computes Q(s_t), then we select the
-        # columns of actions taken. These are the actions which would've been taken
-        # for each batch state according to policy_net
-        state_action_values = self.policy_net(
-            state_batch).gather(1, action_batch) #tensor([[ 1.1217],...,[ 0.8314]])
-
-        # Compute V(s_{t+1}) for all next states.
-        # Expected values of actions for non_final_state_s are computed based
-        # on the "older" target_net; selecting their best reward with max(1)[0].
-        # This is merged based on the mask, such that we'll have either the expected
-        # state value or 0 in case the state was final.
-        state__values = torch.zeros(self.batch_size, device=self.device)
-
-        state__values[non_final_mask] = self.target_net(
-            non_final_state_s).max(1)[0].detach()
-
-        # Compute the expected Q values
-        expected_state_action_values = (state__values * self.gamma) + reward_batch # tensor([0.9685, 0.9683,...,])
-        
-        # Compute Huber loss
-        self.loss = F.smooth_l1_loss(
-            state_action_values, expected_state_action_values.unsqueeze(1))  # .unsqueeze增加一个维度
-        # Optimize the model
-        self.optimizer.zero_grad() # zero_grad clears old gradients from the last step (otherwise you’d just accumulate the gradients from all loss.backward() calls).
-        self.loss.backward() # loss.backward() computes the derivative of the loss w.r.t. the parameters (or anything requiring gradients) using backpropagation.
-        for param in self.policy_net.parameters(): # clip防止梯度爆炸
-            param.grad.data.clamp_(-1, 1)
-        self.optimizer.step() # causes the optimizer to take a step based on the gradients of the parameters.
-
-
-if __name__ == "__main__":
-    dqn = DQN()

codes/DQN_cnn/env.py

@@ -1,66 +0,0 @@
-#!/usr/bin/env python
-# coding=utf-8
-'''
-@Author: John
-@Email: johnjim0816@gmail.com
-@Date: 2020-06-11 10:02:35
-@LastEditor: John
-@LastEditTime: 2020-06-11 16:57:34
-@Discription: 
-@Environment: python 3.7.7
-'''
-
-import numpy as np
-import torch
-import torchvision.transforms as T
-from PIL import Image
-
-resize = T.Compose([T.ToPILImage(),
-                    T.Resize(40, interpolation=Image.CUBIC),
-                    T.ToTensor()])
-
-
-def get_cart_location(env,screen_width):
-    world_width = env.x_threshold * 2
-    scale = screen_width / world_width
-    return int(env.state[0] * scale + screen_width / 2.0)  # MIDDLE OF CART
-
-def get_screen(env,device):
-    # Returned screen requested by gym is 400x600x3, but is sometimes larger
-    # such as 800x1200x3. Transpose it into torch order (CHW).
-    screen = env.render(mode='rgb_array').transpose((2, 0, 1))
-    # Cart is in the lower half, so strip off the top and bottom of the screen
-    _, screen_height, screen_width = screen.shape
-    screen = screen[:, int(screen_height*0.4):int(screen_height * 0.8)]
-    view_width = int(screen_width * 0.6)
-    cart_location = get_cart_location(env,screen_width)
-    if cart_location < view_width // 2:
-        slice_range = slice(view_width)
-    elif cart_location > (screen_width - view_width // 2):
-        slice_range = slice(-view_width, None)
-    else:
-        slice_range = slice(cart_location - view_width // 2,
-                            cart_location + view_width // 2)
-    # Strip off the edges, so that we have a square image centered on a cart
-    screen = screen[:, :, slice_range]
-    # Convert to float, rescale, convert to torch tensor
-    # (this doesn't require a copy)
-    screen = np.ascontiguousarray(screen, dtype=np.float32) / 255
-    screen = torch.from_numpy(screen)
-    # Resize, and add a batch dimension (BCHW)
-    return resize(screen).unsqueeze(0).to(device)
-
-if __name__ == "__main__":
-
-    import gym
-    env = gym.make('CartPole-v0').unwrapped
-    # if gpu is to be used
-    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-    env.reset()
-    import matplotlib.pyplot as plt
-    
-    plt.figure()
-    plt.imshow(get_screen(env,device).cpu().squeeze(0).permute(1, 2, 0).numpy(),
-            interpolation='none')
-    plt.title('Example extracted screen')
-    plt.show()

codes/DQN_cnn/main.py

@@ -1,112 +0,0 @@
-#!/usr/bin/env python
-# coding=utf-8
-'''
-@Author: John
-@Email: johnjim0816@gmail.com
-@Date: 2020-06-11 10:01:09
-@LastEditor: John
-LastEditTime: 2021-04-05 11:06:23
-@Discription: 
-@Environment: python 3.7.7
-'''
-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 DQN_cnn.env import get_screen
-from DQN_cnn.agent import DQNcnn
-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 = 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 DQNcnnConfig:
-    def __init__(self) -> None:
-        self.algo = "DQN_cnn"  # name of algo
-        self.gamma = 0.99
-        self.epsilon_start = 0.95  # e-greedy策略的初始epsilon
-        self.epsilon_end = 0.05
-        self.epsilon_decay = 200
-        self.lr = 0.01  # leanring rate
-        self.memory_capacity = 10000  # Replay Memory容量
-        self.batch_size = 64
-        self.train_eps = 250  # 训练的episode数目
-        self.train_steps = 200  # 训练每个episode的最大长度
-        self.target_update = 4  # target net的更新频率
-        self.eval_eps = 20  # 测试的episode数目
-        self.eval_steps = 200  # 测试每个episode的最大长度
-        self.hidden_dim = 128  # 神经网络隐藏层维度
-        self.device = torch.device(
-            "cuda" if torch.cuda.is_available() else "cpu")  # if gpu is to be used
-
-def train(cfg, env, agent):
-    rewards = []
-    ma_rewards = []
-    for i_episode in range(cfg.train_eps):
-        # Initialize the environment and state
-        env.reset()
-        last_screen = get_screen(env, cfg.device)
-        current_screen = get_screen(env, cfg.device)
-        state = current_screen - last_screen
-        ep_reward = 0
-        for i_step in range(cfg.train_steps+1):
-            # Select and perform an action
-            action = agent.choose_action(state)
-            _, reward, done, _ = env.step(action.item())
-            ep_reward += reward
-            reward = torch.tensor([reward], device=cfg.device)
-            # Observe new state
-            last_screen = current_screen
-            current_screen = get_screen(env, cfg.device)
-            if done:
-                break
-            state_ = current_screen - last_screen
-            # Store the transition in memory
-            agent.memory.push(state, action, state_, reward)
-            # Move to the next state
-            state = state_
-            # Perform one step of the optimization (on the target network)
-            agent.update()
-        # Update the target network, copying all weights and biases in DQN
-        if i_episode % cfg.target_update == 0:
-            agent.target_net.load_state_dict(agent.policy_net.state_dict())
-        print('Episode:{}/{}, Reward:{}, Steps:{}, Explore:{:.2f}, Done:{}'.format(i_episode+1,cfg.train_eps,ep_reward,i_step+1,agent.epsilon,done))
-        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)
-    return rewards,ma_rewards
-
-
-if __name__ == "__main__":
-    cfg = DQNcnnConfig()
-    # Get screen size so that we can initialize layers correctly based on shape
-    # returned from AI gym. Typical dimensions at this point are close to 3x40x90
-    # which is the result of a clamped and down-scaled render buffer in get_screen(env,device)
-    # 因为这里环境的state需要从默认的向量改为图像，所以要unwrapped更改state
-    env = gym.make('CartPole-v0').unwrapped
-    env.reset()
-    init_screen = get_screen(env, cfg.device)
-    _, _, screen_height, screen_width = init_screen.shape
-    # Get number of actions from gym action space
-    action_dim = env.action_space.n
-    agent = DQNcnn(screen_height, screen_width,
-                   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/DQN_cnn/memory.py

@@ -1,35 +0,0 @@
-#!/usr/bin/env python
-# coding=utf-8
-'''
-@Author: John
-@Email: johnjim0816@gmail.com
-@Date: 2020-06-11 09:42:44
-@LastEditor: John
-LastEditTime: 2021-03-23 20:38:41
-@Discription: 
-@Environment: python 3.7.7
-'''
-from collections import namedtuple
-import random
-            
-class ReplayBuffer(object):
-
-    def __init__(self, capacity):
-        self.capacity = capacity
-        self.buffer = []
-        self.position = 0
-        self.Transition = namedtuple('Transition',
-                        ('state', 'action', 'state_', 'reward'))
-
-    def push(self, *args):
-        """Saves a transition."""
-        if len(self.buffer) < self.capacity:
-            self.buffer.append(None)
-        self.buffer[self.position] = self.Transition(*args)
-        self.position = (self.position + 1) % self.capacity
-
-    def sample(self, batch_size):
-        return random.sample(self.buffer, batch_size)
-
-    def __len__(self):
-        return len(self.buffer)

codes/DQN_cnn/model.py

@@ -1,41 +0,0 @@
-#!/usr/bin/env python
-# coding=utf-8
-'''
-@Author: John
-@Email: johnjim0816@gmail.com
-@Date: 2020-06-11 12:18:12
-@LastEditor: John
-@LastEditTime: 2020-06-11 17:23:45
-@Discription: 
-@Environment: python 3.7.7
-'''
-import torch.nn as nn
-import torch.nn.functional as F
-
-class CNN(nn.Module):
-
-    def __init__(self, h, w, n_outputs):
-        super(CNN, self).__init__()
-        self.conv1 = nn.Conv2d(3, 16, kernel_size=5, stride=2)
-        self.bn1 = nn.BatchNorm2d(16)
-        self.conv2 = nn.Conv2d(16, 32, kernel_size=5, stride=2)
-        self.bn2 = nn.BatchNorm2d(32)
-        self.conv3 = nn.Conv2d(32, 32, kernel_size=5, stride=2)
-        self.bn3 = nn.BatchNorm2d(32)
-
-        # Number of Linear input connections depends on output of conv2d layers
-        # and therefore the input image size, so compute it.
-        def conv2d_size_out(size, kernel_size = 5, stride = 2):
-            return (size - (kernel_size - 1) - 1) // stride  + 1
-        convw = conv2d_size_out(conv2d_size_out(conv2d_size_out(w)))
-        convh = conv2d_size_out(conv2d_size_out(conv2d_size_out(h)))
-        linear_input_size = convw * convh * 32
-        self.head = nn.Linear(linear_input_size, n_outputs)
-
-    # Called with either one element to determine next action, or a batch
-    # during optimization. Returns tensor([[left0exp,right0exp]...]).
-    def forward(self, x):
-        x = F.relu(self.bn1(self.conv1(x)))
-        x = F.relu(self.bn2(self.conv2(x)))
-        x = F.relu(self.bn3(self.conv3(x)))
-        return self.head(x.view(x.size(0), -1))

codes/DoubleDQN/agent.py

@@ -5,7 +5,7 @@
 @Email: johnjim0816@gmail.com
 @Date: 2020-06-12 00:50:49
 @LastEditor: John
-LastEditTime: 2021-03-28 11:07:35
+LastEditTime: 2021-05-04 22:28:06
 @Discription: 
 @Environment: python 3.7.7
 '''
@@ -35,22 +35,16 @@ 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
         self.memory = ReplayBuffer(cfg.memory_capacity)
-
-    def choose_action(self, state):
-        '''选择动作
-        '''
-        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:
-            with torch.no_grad():
+    def predict(self,state):
+        with torch.no_grad():
                 # 先转为张量便于丢给神经网络,state元素数据原本为float64
                 # 注意state=torch.tensor(state).unsqueeze(0)跟state=torch.tensor([state])等价
                 state = torch.tensor(
@@ -61,6 +55,15 @@ class DoubleDQN:
                 # 如torch.return_types.max(values=tensor([10.3587]),indices=tensor([0]))
                 # 所以tensor.max(1)[1]返回最大值对应的下标，即action
                 action = q_value.max(1)[1].item()  
+        return action
+    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)
+        if random.random() > self.epsilon:
+            action  = self.predict(state)
         else:
             action = random.randrange(self.action_dim)
         return action
@@ -71,7 +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(
@@ -82,8 +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)
@@ -102,7 +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
@@ -113,7 +115,9 @@ class DoubleDQN:
         self.optimizer.step()  # 更新模型
 
     def save(self,path):
-        torch.save(self.target_net.state_dict(), path+'DoubleDQN_checkpoint.pth')
+        torch.save(self.target_net.state_dict(), path+'checkpoint.pth')
 
     def load(self,path):
-        self.target_net.load_state_dict(torch.load(path+'DoubleDQN_checkpoint.pth'))    
+        self.target_net.load_state_dict(torch.load(path+'checkpoint.pth'))  
+        for target_param, param in zip(self.target_net.parameters(), self.policy_net.parameters()):
+            param.data.copy_(target_param.data)  

codes/DoubleDQN/main.py

@@ -1,93 +0,0 @@
-#!/usr/bin/env python
-# coding=utf-8
-'''
-@Author: John
-@Email: johnjim0816@gmail.com
-@Date: 2020-06-12 00:48:57
-@LastEditor: John
-LastEditTime: 2021-03-28 11:05:14
-@Discription: 
-@Environment: python 3.7.7
-'''
-import sys,os
-sys.path.append(os.getcwd()) # add current terminal path
-import gym
-import torch
-import datetime
-from DoubleDQN.agent import DoubleDQN
-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/"): 
-    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+'/' # 存储reward的路径
-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 DoubleDQNConfig:
-    def __init__(self):
-        self.algo = "Double DQN" # name of algo
-        self.gamma = 0.99
-        self.epsilon_start = 0.9 # e-greedy策略的初始epsilon
-        self.epsilon_end = 0.01
-        self.epsilon_decay = 200
-        self.lr = 0.01 # 学习率
-        self.memory_capacity = 10000 # Replay Memory容量
-        self.batch_size = 128
-        self.train_eps = 300 # 训练的episode数目
-        self.train_steps = 200 # 训练每个episode的最大长度
-        self.target_update = 2 # target net的更新频率
-        self.eval_eps = 20 # 测试的episode数目
-        self.eval_steps = 200 # 测试每个episode的最大长度
-        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu") # 检测gpu
-        self.hidden_dim = 128 # 神经网络隐藏层维度
- 
-
-def train(cfg,env,agent):
-    print('Start to train !')
-    rewards,ma_rewards = [],[]
-    ep_steps = []
-    for i_episode in range(cfg.train_eps):
-        state = env.reset() # reset环境状态
-        ep_reward = 0
-        for i_step in range(cfg.train_steps):
-            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
-            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
-            if done:
-                break
-        # 更新target network，复制DQN中的所有weights and biases
-        if i_episode % cfg.target_update == 0:
-            agent.target_net.load_state_dict(agent.policy_net.state_dict())
-        print('Episode:{}/{}, Reward:{}, Steps:{}, Done:{}'.format(i_episode+1,cfg.train_eps,ep_reward,i_step,done))
-        ep_steps.append(i_step)
-        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)   
-    print('Complete training！')
-    return rewards,ma_rewards
-    
-if __name__ == "__main__":
-    cfg = DoubleDQNConfig()
-    env = gym.make('CartPole-v0').unwrapped # 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 = DoubleDQN(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)

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

@@ -0,0 +1,123 @@
+#!/usr/bin/env python
+# coding=utf-8
+'''
+@Author: John
+@Email: johnjim0816@gmail.com
+@Date: 2020-06-12 00:48:57
+@LastEditor: John
+LastEditTime: 2021-05-04 22:26:59
+@Discription: 
+@Environment: python 3.7.7
+'''
+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 DoubleDQN.agent import DoubleDQN
+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 DoubleDQNConfig:
+    def __init__(self):
+        self.algo = "DoubleDQN" # name of algo
+        self.env = 'CartPole-v0'  # env name
+        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 = 200 # max tranng episodes
+        self.eval_eps = 50 # max evaling episodes
+        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 = 256 # hidden size of net
+ 
+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 = DoubleDQN(state_dim,action_dim,cfg)
+    return env,agent
+    
+def train(cfg,env,agent):
+    print('Start to train !')
+    rewards,ma_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
+            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},Epsilon:{agent.epsilon:.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
+
+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):
+        state = env.reset() 
+        ep_reward = 0   
+        while True:
+            action = agent.predict(state)  
+            next_state, reward, done, _ = env.step(action)  
+            state = next_state  
+            ep_reward += reward
+            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)
+        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)
+    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/DoubleDQN/utils.py

@@ -1,21 +0,0 @@
-#!/usr/bin/env python
-# coding=utf-8
-'''
-Author: John
-Email: johnjim0816@gmail.com
-Date: 2020-10-15 21:28:00
-LastEditor: John
-LastEditTime: 2020-10-15 21:50:30
-Discription: 
-Environment: 
-'''
-import os
-import numpy as np
-
-
-def save_results(rewards,moving_average_rewards,ep_steps,tag='train',result_path='./results'):
-    if not os.path.exists(result_path): # 检测是否存在文件夹
-        os.mkdir(result_path)
-    np.save(result_path+'rewards_'+tag+'.npy', rewards)
-    np.save(result_path+'moving_average_rewards_'+tag+'.npy', moving_average_rewards)
-    np.save(result_path+'steps_'+tag+'.npy',ep_steps )

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)