-
Notifications
You must be signed in to change notification settings - Fork 0
/
models.py
195 lines (165 loc) · 6.67 KB
/
models.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
193
194
195
import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
def hidden_init(layer):
fan_in = layer.weight.data.size()[0]
lim = 1. / np.sqrt(fan_in)
return (-lim, lim)
class Actor(nn.Module):
"""Actor (Policy) Model.
Args:
state_size: An integer representing the dimension of each state.
action_size: An integer representing the dimension of each action.
seed: An integer random seed.
fc1_units: An integer representing the number of nodes in first
hidden layer.
fc2_units: An integer representing the number of nodes in second
hidden layer.
"""
def __init__(self, state_size, action_size, seed, fc1_units=400, fc2_units=300, use_bn=False):
"""Initialize parameters and build model."""
super(Actor, self).__init__()
self.seed = torch.manual_seed(seed)
self.use_bn = use_bn
self.fc1 = nn.Linear(state_size, fc1_units)
self.fc2 = nn.Linear(fc1_units, fc2_units)
self.fc3 = nn.Linear(fc2_units, action_size)
if self.use_bn:
self.bn1 = nn.BatchNorm1d(state_size)
self.bn2 = nn.BatchNorm1d(fc1_units)
self.bn3 = nn.BatchNorm1d(fc2_units)
self.reset_parameters()
def reset_parameters(self):
"""Resets the weights and biases of all fully connected layers."""
self.fc1.weight.data.uniform_(*hidden_init(self.fc1))
self.fc2.weight.data.uniform_(*hidden_init(self.fc2))
self.fc3.weight.data.uniform_(-3e-3, 3e-3)
self.fc1.bias.data.fill_(0.1)
self.fc2.bias.data.fill_(0.1)
self.fc3.bias.data.fill_(0.1)
def forward(self, state):
"""Build an actor (policy) network that maps states -> actions."""
if self.use_bn:
x = self.fc1(self.bn1(state))
else:
x = self.fc1(state)
# x = F.relu(x)
x = F.leaky_relu(x)
if self.use_bn:
x = self.bn2(x)
x = self.fc2(x)
# x = F.relu(x)
x = F.leaky_relu(x)
if self.use_bn:
x = self.bn3(x)
# return torch.tanh(self.fc3(x))
# return F.softmax(self.fc3(x))
# return F.relu(self.fc3(x))
return torch.sigmoid(self.fc3(x))
class Critic(nn.Module):
"""Critic (Value) Model.
Args:
state_size: An integer representing the dimension of each state.
action_size: An integer representing the dimension of each action.
seed: An integer random seed.
fc1_units: An integer representing the number of nodes in first
hidden layer.
fc2_units: An integer representing the number of nodes in second
hidden layer.
"""
def __init__(self, state_size, action_size, seed, fc1_units=400, fc2_units=300, use_bn=False):
"""Initialize parameters and build model."""
super(Critic, self).__init__()
self.seed = torch.manual_seed(seed)
self.use_bn = use_bn
self.fc1 = nn.Linear(state_size, fc1_units)
self.fc2 = nn.Linear(fc1_units+action_size, fc2_units)
self.fc3 = nn.Linear(fc2_units, 1)
if self.use_bn:
self.bn1 = nn.BatchNorm1d(fc1_units)
self.bn2 = nn.BatchNorm1d(fc2_units)
self.reset_parameters()
def reset_parameters(self):
"""Resets the weights and biases of all fully connected layers."""
self.fc1.weight.data.uniform_(*hidden_init(self.fc1))
self.fc2.weight.data.uniform_(*hidden_init(self.fc2))
self.fc3.weight.data.uniform_(-3e-3, 3e-3)
self.fc1.bias.data.fill_(0.1)
self.fc2.bias.data.fill_(0.1)
self.fc3.bias.data.fill_(0.1)
def forward(self, state, action):
"""Build a critic (value) network.
Maps (state, action) pairs -> Q-values.
Args:
state: A list of floats representing the state
action: A float representing the action
"""
x = self.fc1(state)
xs = F.leaky_relu(x)
if self.use_bn:
x = self.bn1(x)
# action1 = action[:256]
# action2 = action[512:768]
# action = torch.cat((action1,action2),dim=0)
# print(action[256:768].shape)
# print(action.shape)
# print(xs.shape)
# if len(action) == 1024:
# print(len(action))
# action=action[256:768]
x = torch.cat((xs, action), dim=1)
x = self.fc2(x)
x = F.leaky_relu(x)
if self.use_bn:
x = self.bn2(x)
return self.fc3(x)
class CentralCritic(nn.Module):
"""Critic (Value) Model.
Args:
state_size: An integer dimension of each state.
action_size: An integer dimension of each action.
seed: An integer representing random seed.
fc1_units: An integer representing number of nodes in the first
hidden layer.
fc2_units: An integer representing the number of nodes in the
second hidden layer.
"""
def __init__(self, state_size, action_size, seed, num_agents=1, fc1_units=64, fc2_units=32, use_bn=False):
"""Initialize parameters and build model."""
super(CentralCritic, self).__init__()
self.seed = torch.manual_seed(seed)
self.use_bn = use_bn
#import pdb; pdb.set_trace()
self.fc1 = nn.Linear(state_size * num_agents, fc1_units * num_agents)
self.fc2 = nn.Linear(fc1_units * num_agents + action_size * num_agents, fc2_units * num_agents)
self.fc3 = nn.Linear(fc2_units * num_agents, 1 * num_agents)
if self.use_bn:
self.bn1 = nn.BatchNorm1d(fc1_units * num_agents)
self.bn2 = nn.BatchNorm1d(fc2_units * num_agents)
self.reset_parameters()
def reset_parameters(self):
"""Resets the weights and biases of all fully connected layers."""
self.fc1.weight.data.uniform_(*hidden_init(self.fc1))
self.fc2.weight.data.uniform_(*hidden_init(self.fc2))
self.fc3.weight.data.uniform_(-3e-3, 3e-3)
self.fc1.bias.data.fill_(0.1)
self.fc2.bias.data.fill_(0.1)
self.fc3.bias.data.fill_(0.1)
def forward(self, state, action):
"""Build a critic (value) network.
Maps (state, action) pairs -> Q-values.
Args:
state: A list of floats representing the state
action: A float representing the action
"""
x = self.fc1(state)
xs = F.leaky_relu(x)
if self.use_bn:
x = self.bn1(x)
x = torch.cat((xs, action), dim=1)
x = self.fc2(x)
x = F.leaky_relu(x)
if self.use_bn:
x = self.bn2(x)
return self.fc3(x)