-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathclassical_policies.py
131 lines (98 loc) · 4.28 KB
/
classical_policies.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
import tensorflow as tf
from tensorflow.keras.layers import *
class DiscreteActionMLP(tf.keras.Model):
def __init__(self, obs_size, act_size, hidden_sizes, activation=None, decorator=None):
super().__init__()
self.input_shape_ = (obs_size, )
self.model = tf.keras.Sequential()
self.model.add(tf.keras.Input(shape=(obs_size,)))
self.model.add(
tf.keras.layers.Dense(hidden_sizes[0], input_shape=(obs_size,),
kernel_initializer=tf.random_normal_initializer(stddev=0.01),
activation=activation)
)
for k in range(1, len(hidden_sizes)-1):
self.model.add(
tf.keras.layers.Dense(hidden_sizes[k], input_shape=(hidden_sizes[k-1],),
kernel_initializer=tf.random_normal_initializer(stddev=0.01),
activation=activation)
)
# Softmax activation on the last layer !
self.model.add(
tf.keras.layers.Dense(act_size, input_shape=(hidden_sizes[-1],),
kernel_initializer=tf.random_normal_initializer(stddev=0.01),
activation='softmax')
)
def call(self, inputs):
return self.model(inputs)
class ConvImgActor(tf.keras.Model):
def __init__(self, obs_size_x, obs_size_y, obs_size_z,
hidden_sizes, act_size, activation=None, decorator=None):
super().__init__()
self.input_shape_ = (obs_size_x, obs_size_y, obs_size_z)
self.model = tf.keras.Sequential()
# self.model.add(tf.keras.Input(shape=shape))
self.model.add(tf.keras.layers.Conv2D(
32, 8, strides=4, activation=activation, padding="same"
))
self.model.add(tf.keras.layers.Conv2D(
64, 4, strides=2, activation=activation, padding="same"
))
self.model.add(tf.keras.layers.Conv2D(
64, 3, strides=2, activation=activation, padding="same"
))
self.model.add(tf.keras.layers.Flatten())
self.model.add(
tf.keras.layers.Dense(hidden_sizes[0],
kernel_initializer='he_uniform',
activation=activation)
)
for k in range(1, len(hidden_sizes)-1):
self.model.add(
tf.keras.layers.Dense(hidden_sizes[k],
input_shape=(hidden_sizes[k-1],),
kernel_initializer='he_uniform',
activation=activation)
)
self.model.add(
tf.keras.layers.Dense(act_size, input_shape=(hidden_sizes[-1],),
kernel_initializer=tf.random_normal_initializer(stddev=0.01),
activation="softmax")
)
def call(self, inputs, **kwargs):
return self.model(inputs)
class ConvImgActorV2(tf.keras.Model):
def __init__(self, obs_size_x, obs_size_y, obs_size_z,
hidden_sizes, filter_counts, act_size, activation=None, decorator=None):
super().__init__()
self.input_shape_ = (obs_size_x, obs_size_y, obs_size_z)
self.model = tf.keras.Sequential()
self.model.add(tf.keras.layers.Conv2D(
filter_counts[0], 8, strides=4, activation=activation, padding="same"
))
self.model.add(tf.keras.layers.Conv2D(
filter_counts[1], 4, strides=2, activation=activation, padding="same"
))
self.model.add(tf.keras.layers.Conv2D(
filter_counts[2], 3, strides=2, activation=activation, padding="same"
))
self.model.add(tf.keras.layers.Flatten())
self.model.add(
tf.keras.layers.Dense(hidden_sizes[0],
kernel_initializer='he_uniform',
activation=activation)
)
for k in range(1, len(hidden_sizes)-1):
self.model.add(
tf.keras.layers.Dense(hidden_sizes[k],
input_shape=(hidden_sizes[k-1],),
kernel_initializer='he_uniform',
activation=activation)
)
self.model.add(
tf.keras.layers.Dense(act_size, input_shape=(hidden_sizes[-1],),
kernel_initializer=tf.random_normal_initializer(stddev=0.01),
activation="softmax")
)
def call(self, inputs, **kwargs):
return self.model(inputs)