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)
+