bacow/easy-rl
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

update codes

Browse Source
This commit is contained in:
johnjim0816
2021-11-19 16:02:34 +08:00
parent 129c0c65fa
commit 64c319cab4
47 changed files with 262 additions and 255 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
codes/PPO/agent.py
View File

@@ -18,6 +18,7 @@ from PPO.memory import PPOMemory
class PPO:
def __init__(self, state_dim, action_dim,cfg):
self.gamma = cfg.gamma
self.continuous = cfg.continuous
self.policy_clip = cfg.policy_clip
self.n_epochs = cfg.n_epochs
self.gae_lambda = cfg.gae_lambda
@@ -29,13 +30,13 @@ class PPO:
self.memory = PPOMemory(cfg.batch_size)
self.loss = 0
def choose_action(self, state,continuous=False):
def choose_action(self, state):
state = torch.tensor([state], dtype=torch.float).to(self.device)
dist = self.actor(state)
value = self.critic(state)
action = dist.sample()
probs = torch.squeeze(dist.log_prob(action)).item()
if continuous:
if self.continuous:
action = torch.tanh(action)
else:
action = torch.squeeze(action).item()

67
codes/PPO/task0.py Normal file
View File

@@ -0,0 +1,67 @@
import sys,os
curr_path = os.path.dirname(os.path.abspath(__file__)) # 当前文件所在绝对路径
parent_path = os.path.dirname(curr_path) # 父路径
sys.path.append(parent_path) # 添加路径到系统路径
import gym
import torch
import datetime
from common.plot import plot_rewards
from common.utils import save_results,make_dir
from PPO.agent import PPO
from PPO.train import train
curr_time = datetime.datetime.now().strftime("%Y%m%d-%H%M%S") # 获取当前时间
class PPOConfig:
def __init__(self) -> None:
self.algo = "DQN" # 算法名称
self.env_name = 'CartPole-v0' # 环境名称
self.continuous = False # 环境是否为连续动作
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu") # 检测GPU
self.train_eps = 200 # 训练的回合数
self.eval_eps = 20 # 测试的回合数
self.batch_size = 5
self.gamma=0.99
self.n_epochs = 4
self.actor_lr = 0.0003
self.critic_lr = 0.0003
self.gae_lambda=0.95
self.policy_clip=0.2
self.hidden_dim = 256
self.update_fre = 20 # frequency of agent update
class PlotConfig:
def __init__(self) -> None:
self.algo = "DQN" # 算法名称
self.env_name = 'CartPole-v0' # 环境名称
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu") # 检测GPU
self.result_path = curr_path+"/outputs/" + self.env_name + \
'/'+curr_time+'/results/' # 保存结果的路径
self.model_path = curr_path+"/outputs/" + self.env_name + \
'/'+curr_time+'/models/' # 保存模型的路径
self.save = True # 是否保存图片
def env_agent_config(cfg,seed=1):
env = gym.make(cfg.env_name)
env.seed(seed)
state_dim = env.observation_space.shape[0]
action_dim = env.action_space.n
agent = PPO(state_dim,action_dim,cfg)
return env,agent
cfg = PPOConfig()
plot_cfg = PlotConfig()
# 训练
env,agent = env_agent_config(cfg,seed=1)
rewards, ma_rewards = train(cfg, env, agent)
make_dir(plot_cfg.result_path, plot_cfg.model_path) # 创建保存结果和模型路径的文件夹
agent.save(path=plot_cfg.model_path)
save_results(rewards, ma_rewards, tag='train', path=plot_cfg.result_path)
plot_rewards(rewards, ma_rewards, plot_cfg, tag="train")
# 测试
env,agent = env_agent_config(cfg,seed=10)
agent.load(path=plot_cfg.model_path)
rewards,ma_rewards = eval(cfg,env,agent)
save_results(rewards,ma_rewards,tag='eval',path=plot_cfg.result_path)
plot_rewards(rewards,ma_rewards,plot_cfg,tag="eval")

68
codes/PPO/task1.py Normal file
View File

@@ -0,0 +1,68 @@
import sys,os
curr_path = os.path.dirname(os.path.abspath(__file__)) # 当前文件所在绝对路径
parent_path = os.path.dirname(curr_path) # 父路径
sys.path.append(parent_path) # 添加路径到系统路径
import gym
import torch
import datetime
from common.plot import plot_rewards
from common.utils import save_results,make_dir
from PPO.agent import PPO
from PPO.train import train
curr_time = datetime.datetime.now().strftime("%Y%m%d-%H%M%S") # 获取当前时间
class PPOConfig:
def __init__(self) -> None:
self.algo = "PPO" # 算法名称
self.env_name = 'Pendulum-v1' # 环境名称
self.continuous = True # 环境是否为连续动作
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu") # 检测GPU
self.train_eps = 200 # 训练的回合数
self.eval_eps = 20 # 测试的回合数
self.batch_size = 5
self.gamma=0.99
self.n_epochs = 4
self.actor_lr = 0.0003
self.critic_lr = 0.0003
self.gae_lambda=0.95
self.policy_clip=0.2
self.hidden_dim = 256
self.update_fre = 20 # frequency of agent update
class PlotConfig:
def __init__(self) -> None:
self.algo = "PPO" # 算法名称
self.env_name = 'Pendulum-v1' # 环境名称
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu") # 检测GPU
self.result_path = curr_path+"/outputs/" + self.env_name + \
'/'+curr_time+'/results/' # 保存结果的路径
self.model_path = curr_path+"/outputs/" + self.env_name + \
'/'+curr_time+'/models/' # 保存模型的路径
self.save = True # 是否保存图片
def env_agent_config(cfg,seed=1):
env = gym.make(cfg.env_name)
env.seed(seed)
state_dim = env.observation_space.shape[0]
action_dim = env.action_space.shape[0]
agent = PPO(state_dim,action_dim,cfg)
return env,agent
cfg = PPOConfig()
plot_cfg = PlotConfig()
# 训练
env,agent = env_agent_config(cfg,seed=1)
rewards, ma_rewards = train(cfg, env, agent)
make_dir(plot_cfg.result_path, plot_cfg.model_path) # 创建保存结果和模型路径的文件夹
agent.save(path=plot_cfg.model_path)
save_results(rewards, ma_rewards, tag='train', path=plot_cfg.result_path)
plot_rewards(rewards, ma_rewards, plot_cfg, tag="train")
# 测试
env,agent = env_agent_config(cfg,seed=10)
agent.load(path=plot_cfg.model_path)
rewards,ma_rewards = eval(cfg,env,agent)
save_results(rewards,ma_rewards,tag='eval',path=plot_cfg.result_path)
plot_rewards(rewards,ma_rewards,plot_cfg,tag="eval")

132
codes/PPO/task1_train.py
View File

@@ -1,132 +0,0 @@
#!/usr/bin/env python
# coding=utf-8
'''
Author: John
Email: johnjim0816@gmail.com
Date: 2021-03-22 16:18:10
LastEditor: John
LastEditTime: 2021-09-26 22:05:00
Discription:
Environment:
'''
import sys,os
curr_path = os.path.dirname(os.path.abspath(__file__)) # 当前文件所在绝对路径
parent_path = os.path.dirname(curr_path) # 父路径
sys.path.append(parent_path) # 添加路径到系统路径
import gym
import torch
import datetime
from PPO.agent import PPO
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 PPOConfig:
def __init__(self) -> None:
self.algo = "PPO" # 算法名称
self.env_name = 'Pendulum-v1' # 环境名称
self.continuous = True # 环境是否为连续动作
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu") # 检测GPU
self.train_eps = 200 # 训练的回合数
self.eval_eps = 20 # 测试的回合数
self.batch_size = 5
self.gamma=0.99
self.n_epochs = 4
self.actor_lr = 0.0003
self.critic_lr = 0.0003
self.gae_lambda=0.95
self.policy_clip=0.2
self.hidden_dim = 256
self.update_fre = 20 # frequency of agent update
class PlotConfig:
def __init__(self) -> None:
self.algo = "PPO" # 算法名称
self.env_name = 'Pendulum-v1' # 环境名称
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu") # 检测GPU
self.result_path = curr_path+"/outputs/" + self.env_name + \
'/'+curr_time+'/results/' # 保存结果的路径
self.model_path = curr_path+"/outputs/" + self.env_name + \
'/'+curr_time+'/models/' # 保存模型的路径
self.save = True # 是否保存图片
def env_agent_config(cfg,seed=1):
env = gym.make(cfg.env_name)
env.seed(seed)
state_dim = env.observation_space.shape[0]
action_dim = env.action_space.shape[0]
agent = PPO(state_dim,action_dim,cfg)
return env,agent
def train(cfg,env,agent):
print('开始训练！')
print(f'环境：{cfg.env_name}, 算法：{cfg.algo}, 设备：{cfg.device}')
rewards = [] # 记录所有回合的奖励
ma_rewards = [] # 记录所有回合的滑动平均奖励
steps = 0
for i_ep in range(cfg.train_eps):
state = env.reset()
done = False
ep_reward = 0
while not done:
action, prob, val = agent.choose_action(state,continuous=cfg.continuous)
print(action)
state_, reward, done, _ = env.step(action)
steps += 1
ep_reward += reward
agent.memory.push(state, action, prob, val, reward, done)
if steps % cfg.update_fre == 0:
agent.update()
state = state_
rewards.append(ep_reward)
if ma_rewards:
ma_rewards.append(0.9*ma_rewards[-1]+0.1*ep_reward)
else:
ma_rewards.append(ep_reward)
if (i_ep+1)%10 == 0:
print(f"回合：{i_ep+1}/{cfg.train_eps}，奖励：{ep_reward:.2f}")
print('完成训练！')
return rewards,ma_rewards
def eval(cfg,env,agent):
print('开始测试!')
print(f'环境：{cfg.env_name}, 算法：{cfg.algo}, 设备：{cfg.device}')
rewards = [] # 记录所有回合的奖励
ma_rewards = [] # 记录所有回合的滑动平均奖励
for i_ep in range(cfg.eval_eps):
state = env.reset()
done = False
ep_reward = 0
while not done:
action, prob, val = agent.choose_action(state,continuous=False)
state_, reward, done, _ = env.step(action)
ep_reward += reward
state = state_
rewards.append(ep_reward)
if ma_rewards:
ma_rewards.append(
0.9*ma_rewards[-1]+0.1*ep_reward)
else:
ma_rewards.append(ep_reward)
print('回合：{}/{}, 奖励：{}'.format(i_ep+1, cfg.eval_eps, ep_reward))
print('完成训练！')
return rewards,ma_rewards
if __name__ == '__main__':
cfg = PPOConfig()
plot_cfg = PlotConfig()
# 训练
env,agent = env_agent_config(cfg,seed=1)
rewards, ma_rewards = train(cfg, env, agent)
make_dir(plot_cfg.result_path, plot_cfg.model_path) # 创建保存结果和模型路径的文件夹
agent.save(path=plot_cfg.model_path)
save_results(rewards, ma_rewards, tag='train', path=plot_cfg.result_path)
plot_rewards(rewards, ma_rewards, plot_cfg, tag="train")
# 测试
env,agent = env_agent_config(cfg,seed=10)
agent.load(path=plot_cfg.model_path)
rewards,ma_rewards = eval(cfg,env,agent)
save_results(rewards,ma_rewards,tag='eval',path=plot_cfg.result_path)
plot_rewards(rewards,ma_rewards,plot_cfg,tag="eval")

0
codes/PPO/task0_train.ipynb → codes/PPO/train.ipynb
View File

122
codes/PPO/task0_train.py → codes/PPO/train.py
View File

@@ -1,65 +1,3 @@
#!/usr/bin/env python
# coding=utf-8
'''
Author: John
Email: johnjim0816@gmail.com
Date: 2021-03-22 16:18:10
LastEditor: John
LastEditTime: 2021-09-26 22:05:00
Discription:
Environment:
'''
import sys,os
curr_path = os.path.dirname(os.path.abspath(__file__)) # 当前文件所在绝对路径
parent_path = os.path.dirname(curr_path) # 父路径
sys.path.append(parent_path) # 添加路径到系统路径
import gym
import torch
import datetime
from PPO.agent import PPO
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 PPOConfig:
def __init__(self) -> None:
self.algo = "DQN" # 算法名称
self.env_name = 'CartPole-v0' # 环境名称
self.continuous = False # 环境是否为连续动作
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu") # 检测GPU
self.train_eps = 200 # 训练的回合数
self.eval_eps = 20 # 测试的回合数
self.batch_size = 5
self.gamma=0.99
self.n_epochs = 4
self.actor_lr = 0.0003
self.critic_lr = 0.0003
self.gae_lambda=0.95
self.policy_clip=0.2
self.hidden_dim = 256
self.update_fre = 20 # frequency of agent update
class PlotConfig:
def __init__(self) -> None:
self.algo = "DQN" # 算法名称
self.env_name = 'CartPole-v0' # 环境名称
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu") # 检测GPU
self.result_path = curr_path+"/outputs/" + self.env_name + \
'/'+curr_time+'/results/' # 保存结果的路径
self.model_path = curr_path+"/outputs/" + self.env_name + \
'/'+curr_time+'/models/' # 保存模型的路径
self.save = True # 是否保存图片
def env_agent_config(cfg,seed=1):
env = gym.make(cfg.env_name)
env.seed(seed)
state_dim = env.observation_space.shape[0]
action_dim = env.action_space.n
agent = PPO(state_dim,action_dim,cfg)
return env,agent
def train(cfg,env,agent):
print('开始训练！')
print(f'环境：{cfg.env_name}, 算法：{cfg.algo}, 设备：{cfg.device}')
@@ -71,7 +9,7 @@ def train(cfg,env,agent):
done = False
ep_reward = 0
while not done:
action, prob, val = agent.choose_action(state,continuous=cfg.continuous)
action, prob, val = agent.choose_action(state)
state_, reward, done, _ = env.step(action)
steps += 1
ep_reward += reward
@@ -99,7 +37,7 @@ def eval(cfg,env,agent):
done = False
ep_reward = 0
while not done:
action, prob, val = agent.choose_action(state,cfg.continuous)
action, prob, val = agent.choose_action(state)
state_, reward, done, _ = env.step(action)
ep_reward += reward
state = state_
@@ -112,8 +50,60 @@ def eval(cfg,env,agent):
print('回合：{}/{}, 奖励：{}'.format(i_ep+1, cfg.eval_eps, ep_reward))
print('完成训练！')
return rewards,ma_rewards
if __name__ == '__main__':
import sys,os
curr_path = os.path.dirname(os.path.abspath(__file__)) # 当前文件所在绝对路径
parent_path = os.path.dirname(curr_path) # 父路径
sys.path.append(parent_path) # 添加路径到系统路径
import gym
import torch
import datetime
from common.plot import plot_rewards
from common.utils import save_results,make_dir
from PPO.agent import PPO
from PPO.train import train
curr_time = datetime.datetime.now().strftime("%Y%m%d-%H%M%S") # 获取当前时间
class PPOConfig:
def __init__(self) -> None:
self.algo = "DQN" # 算法名称
self.env_name = 'CartPole-v0' # 环境名称
self.continuous = False # 环境是否为连续动作
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu") # 检测GPU
self.train_eps = 200 # 训练的回合数
self.eval_eps = 20 # 测试的回合数
self.batch_size = 5
self.gamma=0.99
self.n_epochs = 4
self.actor_lr = 0.0003
self.critic_lr = 0.0003
self.gae_lambda=0.95
self.policy_clip=0.2
self.hidden_dim = 256
self.update_fre = 20 # frequency of agent update
class PlotConfig:
def __init__(self) -> None:
self.algo = "DQN" # 算法名称
self.env_name = 'CartPole-v0' # 环境名称
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu") # 检测GPU
self.result_path = curr_path+"/outputs/" + self.env_name + \
'/'+curr_time+'/results/' # 保存结果的路径
self.model_path = curr_path+"/outputs/" + self.env_name + \
'/'+curr_time+'/models/' # 保存模型的路径
self.save = True # 是否保存图片
def env_agent_config(cfg,seed=1):
env = gym.make(cfg.env_name)
env.seed(seed)
state_dim = env.observation_space.shape[0]
action_dim = env.action_space.n
agent = PPO(state_dim,action_dim,cfg)
return env,agent
cfg = PPOConfig()
plot_cfg = PlotConfig()
# 训练
@@ -128,4 +118,4 @@ if __name__ == '__main__':
agent.load(path=plot_cfg.model_path)
rewards,ma_rewards = eval(cfg,env,agent)
save_results(rewards,ma_rewards,tag='eval',path=plot_cfg.result_path)
plot_rewards(rewards,ma_rewards,plot_cfg,tag="eval")
plot_rewards(rewards,ma_rewards,plot_cfg,tag="eval")