-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathDDPG.py
192 lines (171 loc) · 8.85 KB
/
DDPG.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
import os
import sys
import cProfile
import gymnasium as gym
from tqdm import trange
import torch
import torch.nn.functional as F
from utils.common_utils import train_DDPG_agent, read_ckp, ReplayBuffer, PrioritizedReplayBuffer
import numpy as np
import random
import argparse
import warnings
warnings.filterwarnings('ignore')
class PolicyNet(torch.nn.Module):
def __init__(self, state_dim, hidden_dim, action_dim):
super().__init__()
self.fc1 = torch.nn.Linear(state_dim, hidden_dim)
self.fc2 = torch.nn.Linear(hidden_dim, action_dim)
self.low = low
self.high = high
def forward(self, x):
x = F.relu(self.fc1(x))
action = torch.tanh(self.fc2(x))
action = self.low + (action + 1.0) * (self.high - self.low) / 2 # Zoom to the specified space
return action
class QValueNet(torch.nn.Module):
def __init__(self, state_dim, hidden_dim, action_dim):
super().__init__()
self.fc1 = torch.nn.Linear(state_dim + action_dim, hidden_dim)
self.fc2 = torch.nn.Linear(hidden_dim, hidden_dim)
self.fc_out = torch.nn.Linear(hidden_dim, 1)
def forward(self, x, a):
cat = torch.cat([x, a], dim=1)
x = F.relu(self.fc1(cat))
x = F.relu(self.fc2(x))
return self.fc_out(x)
class DDPG:
''' DDPG '''
def __init__(self, state_dim, hidden_dim, action_dim,sigma, actor_lr,
critic_lr, tau, gamma, device, training=True):
self.training = training
self.actor = PolicyNet(state_dim, hidden_dim, action_dim).to(device)
self.critic = QValueNet(state_dim, hidden_dim, action_dim).to(device)
self.target_actor = PolicyNet(state_dim, hidden_dim, action_dim).to(device)
self.target_critic = QValueNet(state_dim, hidden_dim, action_dim).to(device)
self.target_critic.load_state_dict(self.critic.state_dict())
self.target_actor.load_state_dict(self.actor.state_dict())
self.actor_optimizer = torch.optim.Adam(self.actor.parameters(), lr=actor_lr)
self.critic_optimizer = torch.optim.Adam(self.critic.parameters(), lr=critic_lr)
self.gamma = gamma
self.sigma = sigma
self.tau = tau
self.action_dim = action_dim
self.device = device
def take_action(self, state, the_device=None):
state = torch.tensor(state) if not isinstance(state, torch.Tensor) else state
state = state.to(the_device) if the_device else state.to(self.device)
action = self.actor(state).detach().cpu()
if self.training:
action = action + self.sigma * np.random.randn(self.action_dim)
return action
def soft_update(self, net, target_net):
'''Soft update target_net in the direction of net, each update is very small
Params
----------
net : torch.nn.module
target_net : torch.nn.module
'''
for param_target, param in zip(target_net.parameters(), net.parameters()):
param_target.data.copy_(param_target.data * (1.0 - self.tau) + param.data * self.tau)
def update(self, transition_dict):
states = torch.tensor(transition_dict['states'], dtype=torch.float).to(self.device)
actions = torch.tensor(transition_dict['actions'], dtype=torch.float).to(self.device)
rewards = torch.tensor(transition_dict['rewards'], dtype=torch.float).view(-1, 1).to(self.device)
next_states = torch.tensor(transition_dict['next_states'], dtype=torch.float).to(self.device)
dones = torch.tensor(transition_dict['dones'], dtype=torch.int).view(-1, 1).to(self.device)
truncated = torch.tensor(transition_dict['truncated'], dtype=torch.int).view(-1, 1).to(self.device)
weights = torch.FloatTensor(transition_dict["weights"].reshape(-1, 1)).to(self.device)
next_q_values = self.target_critic(next_states, self.target_actor(next_states))
q_targets = rewards + self.gamma * next_q_values * (1 - dones | truncated)
critic_losses = F.mse_loss(self.critic(states, actions), q_targets.detach(), reduction='none')
critic_loss = torch.mean(critic_losses * weights)
self.critic_optimizer.zero_grad()
critic_loss.backward()
self.critic_optimizer.step()
actor_loss = -torch.mean(self.critic(states, self.actor(states)))
self.actor_optimizer.zero_grad()
actor_loss.backward()
self.actor_optimizer.step()
self.soft_update(self.actor, self.target_actor)
self.soft_update(self.critic, self.target_critic)
return critic_losses.detach().cpu()
# * --------------------- 参数 -------------------------
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='DDPG task')
parser.add_argument('--model_name', default="DDPG", type=str, help='Alg name')
parser.add_argument('-t', '--task', default="lunar", type=str, help='Task name')
parser.add_argument('-w', '--writer', default=0, type=int, help='Saving type, 0: No，1: local')
parser.add_argument('--sta', action="store_true", help='Whether to use STA')
parser.add_argument('--per', action="store_true", help='Whether to use PER')
parser.add_argument('-d', '--distance_ratio', default=0.2, type=float, help='Generate experience ratio')
parser.add_argument('-n', '--num_new_samples', default=3, type=int, help='The number of counterfactual actions sampled')
parser.add_argument('-e', '--episodes', default=300, type=int, help='episodes')
parser.add_argument('-s', '--seed', nargs='+', default=[1, 7], type=int, help='Start and end seeds')
args = parser.parse_args()
# env
device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
# Continuous Action Space Environment Benchmark
if args.task == 'pendulum':
env = gym.make('Pendulum-v1')
elif args.task == 'lunar':
env = gym.make("LunarLander-v2", continuous=True)
elif args.task == 'walker':
env = gym.make("BipedalWalker-v3")
# DDPG
action_type = 'discrete' if isinstance(env.action_space, gym.spaces.Discrete) else 'continuous'
actor_lr = 2e-4
critic_lr = 2e-4
hidden_dim = 128
buffer_size = int(5e4)
minimal_size = buffer_size // 10
gamma = 0.98
sigma = 0.01
tau = 0.005
state_dim = env.observation_space.shape[0]
action_dim = env.action_space.shape[0]
if (abs(env.action_space.high) != abs(env.action_space.low)).any():
print('WARNING: Action space asymmetry')
# action bound
low = torch.tensor(env.action_space.low, device=device)
high = torch.tensor(env.action_space.high, device=device)
# task
total_epochs = 1
batch_size = 128
system_type = sys.platform
# CEA params
if args.sta:
path = f'model/sta/{args.task}/regular.pt'
sta = torch.load(path, map_location=device)
if args.per:
args.model_name = f'CEA_{args.model_name}'
else:
args.model_name = f'CEA_{args.model_name}-no_PER'
else:
if args.per:
args.model_name = f'{args.model_name}_PER'
sta = False
print(f'[ Start training, task: {args.task}, alg: {args.model_name}, device: {device} ]')
# * ----------------------- 训练 ----------------------------
for seed in trange(args.seed[0], args.seed[-1] + 1, mininterval=40, ncols=100):
CKP_PATH = f'ckpt/{args.task}/{args.model_name}/{seed}_{system_type}.pt'
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
if args.per:
replay_buffer = PrioritizedReplayBuffer(state_dim, action_dim, buffer_size, batch_size)
else:
replay_buffer = ReplayBuffer(state_dim, action_dim, buffer_size, batch_size)
agent = DDPG(state_dim, hidden_dim, action_dim, sigma, actor_lr,
critic_lr, tau, gamma, device)
s_epoch, s_episode, return_list, time_list, seed_list = read_ckp(CKP_PATH, agent, args.model_name)
# cProfile.run('''return_list, train_time = train_DDPG_agent(env, agent, args.writer, s_epoch, total_epochs,
# s_episode, args.episodes, replay_buffer, minimal_size,
# return_list, time_list, seed_list, seed, CKP_PATH,
# sta, args.per, args.num_new_samples, args.distance_ratio
# )''', './DDPG_analysis')
return_list, train_time = train_DDPG_agent(env, agent, args.writer, s_epoch, total_epochs,
s_episode, args.episodes, replay_buffer, minimal_size,
return_list, time_list, seed_list, seed, CKP_PATH,
sta, args.per, args.num_new_samples, args.distance_ratio
)