diff --git a/codes/DQN/README.md b/codes/DQN/README.md new file mode 100644 index 0000000..a8e141b --- /dev/null +++ b/codes/DQN/README.md @@ -0,0 +1,35 @@ +## 思路 + +见[我的博客](https://blog.csdn.net/JohnJim0/article/details/109557173) +## 环境 + +python 3.7.9 + +pytorch 1.6.0 + +tensorboard 2.3.0 + +torchvision 0.7.0 + +## 使用 + +train: + +```python +python main.py +``` + +eval: + +```python +python main.py --train 0 +``` +可视化： +```python +tensorboard --logdir logs +``` + +## Torch知识 + +[with torch.no_grad()](https://www.jianshu.com/p/1cea017f5d11) + diff --git a/codes/DQN/agent.py b/codes/DQN/agent.py new file mode 100644 index 0000000..299a4c9 --- /dev/null +++ b/codes/DQN/agent.py @@ -0,0 +1,127 @@ +#!/usr/bin/env python +# coding=utf-8 +''' +@Author: John +@Email: johnjim0816@gmail.com +@Date: 2020-06-12 00:50:49 +@LastEditor: John +LastEditTime: 2021-03-13 14:56:23 +@Discription: +@Environment: python 3.7.7 +''' +'''off-policy +''' + + +import torch +import torch.nn as nn +import torch.optim as optim +import random +import math +import numpy as np +from common.memory import ReplayBuffer +from common.model import MLP2 +class DQN: + def __init__(self, n_states, n_actions, cfg): + + self.n_actions = n_actions # 总的动作个数 + self.device = cfg.device # 设备，cpu或gpu等 + self.gamma = cfg.gamma # 奖励的折扣因子 + # e-greedy策略相关参数 + self.sample_count = 0 # 用于epsilon的衰减计数 + 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 = MLP2(n_states, n_actions,hidden_dim=cfg.hidden_dim).to(self.device) + self.target_net = MLP2(n_states, n_actions,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 + # 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, train=True): + '''选择动作 + ''' + if train: + self.epsilon = self.epsilon_end + (self.epsilon_start - self.epsilon_end) * \ + math.exp(-1. * self.sample_count / self.epsilon_decay) + self.sample_count += 1 + if random.random() > self.epsilon: + 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( + [state], device=self.device, dtype=torch.float32) + # 如tensor([[-0.0798, -0.0079]], grad_fn=) + q_value = self.policy_net(state) + # 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].item() + else: + action = random.randrange(self.n_actions) + return action + else: + with torch.no_grad(): # 取消保存梯度 + # 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( + [state], device='cpu', dtype=torch.float32) # 如tensor([[-0.0798, -0.0079]], grad_fn=) + q_value = self.target_net(state) + # 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].item() + return action + 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).unsqueeze(1) # 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)''' + '''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_state_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_state_values * (1-done_batch[0]) + # 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 + # 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() + 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 + + 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')) diff --git a/codes/DQN/main.py b/codes/DQN/main.py new file mode 100644 index 0000000..437ddcc --- /dev/null +++ b/codes/DQN/main.py @@ -0,0 +1,93 @@ +#!/usr/bin/env python +# coding=utf-8 +''' +@Author: John +@Email: johnjim0816@gmail.com +@Date: 2020-06-12 00:48:57 +@LastEditor: John +LastEditTime: 2021-03-13 14:56:50 +@Discription: +@Environment: python 3.7.7 +''' +import sys,os +sys.path.append(os.getcwd()) # 添加当前终端路径 +import gym +import torch +import datetime +from DQN.agent import DQN +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 DQNConfig: + def __init__(self): + self.algo = "DQN" # 算法名称 + self.gamma = 0.99 + self.epsilon_start = 0.95 # 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 = 800 # 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 = 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 = [] # 滑动平均的reward + 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 = DQNConfig() + 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_actions,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) diff --git a/codes/DQN/results/20210313-140409/ma_rewards_train.npy b/codes/DQN/results/20210313-140409/ma_rewards_train.npy new file mode 100644 index 0000000..4790db2 Binary files /dev/null and b/codes/DQN/results/20210313-140409/ma_rewards_train.npy differ diff --git a/codes/DQN/results/20210313-140409/rewards_curve_train.png b/codes/DQN/results/20210313-140409/rewards_curve_train.png new file mode 100644 index 0000000..a077d9d Binary files /dev/null and b/codes/DQN/results/20210313-140409/rewards_curve_train.png differ diff --git a/codes/DQN/results/20210313-140409/rewards_train.npy b/codes/DQN/results/20210313-140409/rewards_train.npy new file mode 100644 index 0000000..19992a9 Binary files /dev/null and b/codes/DQN/results/20210313-140409/rewards_train.npy differ diff --git a/codes/DQN/saved_model/20210313-140409/dqn_checkpoint.pth b/codes/DQN/saved_model/20210313-140409/dqn_checkpoint.pth new file mode 100644 index 0000000..c685cbc Binary files /dev/null and b/codes/DQN/saved_model/20210313-140409/dqn_checkpoint.pth differ