-
Notifications
You must be signed in to change notification settings - Fork 49
/
DDPG.py
110 lines (82 loc) · 4.31 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
import torch
import torch.nn as nn
import torch.optim as optim
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
class Actor(nn.Module):
def __init__(self, state_dim, action_dim, action_bounds, offset):
super(Actor, self).__init__()
# actor
self.actor = nn.Sequential(
nn.Linear(state_dim + state_dim, 64),
nn.ReLU(),
nn.Linear(64, 64),
nn.ReLU(),
nn.Linear(64, action_dim),
nn.Tanh()
)
# max value of actions
self.action_bounds = action_bounds
self.offset = offset
def forward(self, state, goal):
return (self.actor(torch.cat([state, goal], 1)) * self.action_bounds) + self.offset
class Critic(nn.Module):
def __init__(self, state_dim, action_dim, H):
super(Critic, self).__init__()
# UVFA critic
self.critic = nn.Sequential(
nn.Linear(state_dim + action_dim + state_dim, 64),
nn.ReLU(),
nn.Linear(64, 64),
nn.ReLU(),
nn.Linear(64, 1),
nn.Sigmoid()
)
self.H = H
def forward(self, state, action, goal):
# rewards are in range [-H, 0]
return -self.critic(torch.cat([state, action, goal], 1)) * self.H
class DDPG:
def __init__(self, state_dim, action_dim, action_bounds, offset, lr, H):
self.actor = Actor(state_dim, action_dim, action_bounds, offset).to(device)
self.actor_optimizer = optim.Adam(self.actor.parameters(), lr=lr)
self.critic = Critic(state_dim, action_dim, H).to(device)
self.critic_optimizer = optim.Adam(self.critic.parameters(), lr=lr)
self.mseLoss = torch.nn.MSELoss()
def select_action(self, state, goal):
state = torch.FloatTensor(state.reshape(1, -1)).to(device)
goal = torch.FloatTensor(goal.reshape(1, -1)).to(device)
return self.actor(state, goal).detach().cpu().data.numpy().flatten()
def update(self, buffer, n_iter, batch_size):
for i in range(n_iter):
# Sample a batch of transitions from replay buffer:
state, action, reward, next_state, goal, gamma, done = buffer.sample(batch_size)
# convert np arrays into tensors
state = torch.FloatTensor(state).to(device)
action = torch.FloatTensor(action).to(device)
reward = torch.FloatTensor(reward).reshape((batch_size,1)).to(device)
next_state = torch.FloatTensor(next_state).to(device)
goal = torch.FloatTensor(goal).to(device)
gamma = torch.FloatTensor(gamma).reshape((batch_size,1)).to(device)
done = torch.FloatTensor(done).reshape((batch_size,1)).to(device)
# select next action
next_action = self.actor(next_state, goal).detach()
# Compute target Q-value:
target_Q = self.critic(next_state, next_action, goal).detach()
target_Q = reward + ((1-done) * gamma * target_Q)
# Optimize Critic:
critic_loss = self.mseLoss(self.critic(state, action, goal), target_Q)
self.critic_optimizer.zero_grad()
critic_loss.backward()
self.critic_optimizer.step()
# Compute actor loss:
actor_loss = -self.critic(state, self.actor(state, goal), goal).mean()
# Optimize the actor
self.actor_optimizer.zero_grad()
actor_loss.backward()
self.actor_optimizer.step()
def save(self, directory, name):
torch.save(self.actor.state_dict(), '%s/%s_actor.pth' % (directory, name))
torch.save(self.critic.state_dict(), '%s/%s_crtic.pth' % (directory, name))
def load(self, directory, name):
self.actor.load_state_dict(torch.load('%s/%s_actor.pth' % (directory, name), map_location='cpu'))
self.critic.load_state_dict(torch.load('%s/%s_crtic.pth' % (directory, name), map_location='cpu'))