-
Notifications
You must be signed in to change notification settings - Fork 0
/
simple_model.py
81 lines (70 loc) · 2.44 KB
/
simple_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
76
77
78
79
80
81
import parl
import paddle
import paddle.nn as nn
import paddle.nn.functional as F
class MAModel(parl.Model):
def __init__(self, obs_dim, act_dim, critic_in_dim):
super(MAModel, self).__init__()
self.actor_model = ActorModel(obs_dim, act_dim)
self.critic_model = CriticModel(critic_in_dim)
def policy(self, obs):
return self.actor_model(obs)
def value(self, obs, act):
return self.critic_model(obs, act)
def get_actor_params(self):
return self.actor_model.parameters()
def get_critic_params(self):
return self.critic_model.parameters()
class ActorModel(parl.Model):
def __init__(self, obs_dim, act_dim):
super(ActorModel, self).__init__()
hid1_size = 64
hid2_size = 64
self.fc1 = nn.Linear(
obs_dim,
hid1_size,
weight_attr=paddle.ParamAttr(
initializer=paddle.nn.initializer.XavierUniform()))
self.fc2 = nn.Linear(
hid1_size,
hid2_size,
weight_attr=paddle.ParamAttr(
initializer=paddle.nn.initializer.XavierUniform()))
self.fc3 = nn.Linear(
hid2_size,
act_dim,
weight_attr=paddle.ParamAttr(
initializer=paddle.nn.initializer.XavierUniform()))
def forward(self, obs):
hid1 = F.relu(self.fc1(obs))
hid2 = F.relu(self.fc2(hid1))
means = self.fc3(hid2)
return paddle.tanh(means)
class CriticModel(parl.Model):
def __init__(self, critic_in_dim):
super(CriticModel, self).__init__()
hid1_size = 64
hid2_size = 64
out_dim = 1
self.fc1 = nn.Linear(
critic_in_dim,
hid1_size,
weight_attr=paddle.ParamAttr(
initializer=paddle.nn.initializer.XavierUniform()))
self.fc2 = nn.Linear(
hid1_size,
hid2_size,
weight_attr=paddle.ParamAttr(
initializer=paddle.nn.initializer.XavierUniform()))
self.fc3 = nn.Linear(
hid2_size,
out_dim,
weight_attr=paddle.ParamAttr(
initializer=paddle.nn.initializer.XavierUniform()))
def forward(self, obs_n, act_n):
inputs = paddle.concat(obs_n + act_n, axis=1)
hid1 = F.relu(self.fc1(inputs))
hid2 = F.relu(self.fc2(hid1))
Q = self.fc3(hid2)
Q = paddle.squeeze(Q, axis=1)
return Q