add double_dqn

2020-09-08 13:22:32 +08:00
parent 834cd16117
commit db15149b39
12 changed files with 363 additions and 0 deletions
--- a/codes/double_dqn/main.py
+++ b/codes/double_dqn/main.py
@@ -0,0 +1,143 @@
+#!/usr/bin/env python
+# coding=utf-8
+'''
+@Author: John
+@Email: johnjim0816@gmail.com
+@Date: 2020-06-12 00:48:57
+@LastEditor: John
+LastEditTime: 2020-09-01 22:54:23
+@Discription: 
+@Environment: python 3.7.7
+'''
+import gym
+import torch
+from dqn import DQN
+from plot import plot
+import argparse
+
+def get_args():
+    '''模型参数
+    '''
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--gamma", default=0.99,
+                        type=float)  # q-learning中的gamma
+    parser.add_argument("--epsilon_start", default=0.95,
+                        type=float)  # 基于贪心选择action对应的参数epsilon
+    parser.add_argument("--epsilon_end", default=0.05, type=float)
+    parser.add_argument("--epsilon_decay", default=500, type=float)
+    parser.add_argument("--policy_lr", default=0.01, type=float)
+    parser.add_argument("--memory_capacity", default=1000,
+                        type=int, help="capacity of Replay Memory") 
+
+    parser.add_argument("--batch_size", default=32, type=int,
+                        help="batch size of memory sampling")
+    parser.add_argument("--train_eps", default=200, type=int) # 训练的最大episode数目
+    parser.add_argument("--train_steps", default=200, type=int) # 训练每个episode的长度
+    parser.add_argument("--eval_eps", default=200, type=int) # 训练的最大episode数目
+    parser.add_argument("--eval_steps", default=200, type=int) # 训练每个episode的长度
+    parser.add_argument("--target_update", default=2, type=int,
+                        help="when(every default 10 eisodes) to update target net ")
+    config = parser.parse_args()
+
+    return config
+
+def train():
+    cfg = get_args()
+    # if gpu is to be used
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu") # 检测gpu
+    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
+    n_states = env.observation_space.shape[0]
+    n_actions = env.action_space.n
+    agent = DQN(n_states=n_states, n_actions=n_actions, device=device, gamma=cfg.gamma, epsilon_start=cfg.epsilon_start,
+                epsilon_end=cfg.epsilon_end, epsilon_decay=cfg.epsilon_decay, policy_lr=cfg.policy_lr, memory_capacity=cfg.memory_capacity, batch_size=cfg.batch_size)
+    rewards = []
+    moving_average_rewards = []
+    ep_steps = []
+    for i_episode in range(1, cfg.train_eps+1):
+        state = env.reset() # reset环境状态
+        ep_reward = 0
+        for i_step in range(1, cfg.train_steps+1):
+            action = agent.select_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:', i_episode, ' Reward: %i' %
+              int(ep_reward), 'n_steps:', i_step, 'done: ', done,' Explore: %.2f' % agent.epsilon)
+        ep_steps.append(i_step)
+        rewards.append(ep_reward)
+        # 计算滑动窗口的reward
+        if i_episode == 1:
+            moving_average_rewards.append(ep_reward)
+        else:
+            moving_average_rewards.append(
+                0.9*moving_average_rewards[-1]+0.1*ep_reward)
+    import os
+    import numpy as np
+    save_path = os.path.dirname(__file__)+"/saved_model/"
+    if not os.path.exists(save_path):
+        os.mkdir(save_path)
+    agent.save_model(save_path+'checkpoint.pth')
+    # 存储reward等相关结果
+    output_path = os.path.dirname(__file__)+"/result/"
+    # 检测是否存在文件夹
+    if not os.path.exists(output_path):
+        os.mkdir(output_path)
+    np.save(output_path+"rewards.npy", rewards)
+    np.save(output_path+"moving_average_rewards.npy", moving_average_rewards)
+    np.save(output_path+"steps.npy", ep_steps)
+    print('Complete！')
+    plot(rewards)
+    plot(moving_average_rewards, ylabel="moving_average_rewards")
+    plot(ep_steps, ylabel="steps_of_each_episode")
+
+def eval():
+    cfg = get_args()
+    # if gpu is to be used
+    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
+    n_states = env.observation_space.shape[0]
+    n_actions = env.action_space.n
+    agent = DQN(n_states=n_states, n_actions=n_actions, gamma=cfg.gamma, epsilon_start=cfg.epsilon_start,
+                epsilon_end=cfg.epsilon_end, epsilon_decay=cfg.epsilon_decay, policy_lr=cfg.policy_lr, memory_capacity=cfg.memory_capacity, batch_size=cfg.batch_size)
+    import os
+    save_path = os.path.dirname(__file__)+"/saved_model/"
+    if not os.path.exists(save_path):
+        os.mkdir(save_path)
+    agent.load_model(save_path+'checkpoint.pth')
+    rewards = []
+    moving_average_rewards = []
+    ep_steps = []
+    for i_episode in range(1, cfg.eval_eps+1):
+        state = env.reset() # reset环境状态
+        ep_reward = 0
+        for i_step in range(1, cfg.eval_steps+1):
+            action = agent.select_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
+            state = next_state # 跳转到下一个状态
+            if done:
+                break
+        print('Episode:', i_episode, ' Reward: %i' %
+              int(ep_reward), 'n_steps:', i_step, 'done: ', done,' Explore: %.2f' % agent.epsilon)
+        ep_steps.append(i_step)
+        rewards.append(ep_reward)
+        # 计算滑动窗口的reward
+        if i_episode == 1:
+            moving_average_rewards.append(ep_reward)
+        else:
+            moving_average_rewards.append(
+                0.9*moving_average_rewards[-1]+0.1*ep_reward)
+    plot(rewards,save_fig=False)
+    plot(moving_average_rewards, ylabel="moving_average_rewards",save_fig=False)
+    plot(ep_steps, ylabel="steps_of_each_episode",save_fig=False)
+if __name__ == "__main__":
+    # train()
+    eval()