-
Notifications
You must be signed in to change notification settings - Fork 11
/
model.py
75 lines (62 loc) · 2.43 KB
/
model.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
import torch
import torch.nn as nn
import torch.nn.functional as F
class QNetwork(nn.Module):
"""Actor (Policy) Model."""
def __init__(self, state_size, action_size, seed, fc1_size = 64, fc2_size = 64):
"""Initialize parameters and build model.
Params
======
state_size (int): Dimension of each state
action_size (int): Dimension of each action
seed (int): Random seed
"""
super(QNetwork, self).__init__()
hidden_size = 30
self.seed = torch.manual_seed(seed)
self.fc1 = nn.Linear(state_size, fc1_size)
self.fc2 = nn.Linear(fc1_size, fc2_size)
self.out = nn.Linear(fc2_size, action_size)
"*** YOUR CODE HERE ***"
def forward(self, state):
"""Build a network that maps state -> action values."""
x = self.fc1(state)
x = F.relu(x)
x = self.fc2(x)
x = F.relu(x)
action = self.out(x)
return action
class DuelingQNetwork(nn.Module):
"""Actor (Policy) Model."""
def __init__(self, state_size, action_size, seed, fc1_size = 64, fc2_size = 64):
"""Initialize parameters and build model.
Params
======
state_size (int): Dimension of each state
action_size (int): Dimension of each action
seed (int): Random seed
"""
super(DuelingQNetwork, self).__init__()
self.num_actions = action_size
fc3_1_size = fc3_2_size = 32
self.seed = torch.manual_seed(seed)
self.fc1 = nn.Linear(state_size, fc1_size)
self.fc2 = nn.Linear(fc1_size, fc2_size)
## Here we separate into two streams
# The one that calculate V(s)
self.fc3_1 = nn.Linear(fc2_size, fc3_1_size)
self.fc4_1 = nn.Linear(fc3_1_size, 1)
# The one that calculate A(s,a)
self.fc3_2 = nn.Linear(fc2_size, fc3_2_size)
self.fc4_2 = nn.Linear(fc3_2_size, action_size)
def forward(self, state):
"""Build a network that maps state -> action values."""
x = F.relu(self.fc1(state))
x = F.relu(self.fc2(x))
val = F.relu(self.fc3_1(x))
val = self.fc4_1(val)
adv = F.relu(self.fc3_2(x))
adv = self.fc4_2(adv)
# Q(s,a) = V(s) + (A(s,a) - 1/|A| * sum A(s,a'))
action = val + adv - adv.mean(1).unsqueeze(1).expand(state.size(0), self.num_actions)
return action