-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy pathdelay_model.py
162 lines (131 loc) · 7.6 KB
/
delay_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
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
import tensorflow as tf
class RouteNet_Fermi(tf.keras.Model):
def __init__(self):
super(RouteNet_Fermi, self).__init__()
# Configuration dictionary. It contains the needed Hyperparameters for the model.
# All the Hyperparameters can be found in the config.ini file
self.max_num_models = 7
self.num_policies = 4
self.max_num_queues = 3
self.iterations = 8
self.path_state_dim = 32
self.link_state_dim = 32
self.queue_state_dim = 32
self.z_score = {'traffic': [1385.4058837890625, 859.8118896484375],
'packets': [1.4015231132507324, 0.8932565450668335],
'eq_lambda': [1350.97119140625, 858.316162109375],
'avg_pkts_lambda': [0.9117304086685181, 0.9723503589630127],
'exp_max_factor': [6.663637638092041, 4.715115070343018],
'pkts_lambda_on': [0.9116322994232178, 1.651275396347046],
'avg_t_off': [1.6649284362792969, 2.356407403945923],
'avg_t_on': [1.6649284362792969, 2.356407403945923], 'ar_a': [0.0, 1.0], 'sigma': [0.0, 1.0],
'capacity': [27611.091796875, 20090.62109375], 'queue_size': [30259.10546875, 21410.095703125]}
# GRU Cells used in the Message Passing step
self.path_update = tf.keras.layers.GRUCell(self.path_state_dim)
self.link_update = tf.keras.layers.GRUCell(self.link_state_dim)
self.queue_update = tf.keras.layers.GRUCell(self.queue_state_dim)
self.path_embedding = tf.keras.Sequential([
tf.keras.layers.Input(shape=10 + self.max_num_models),
tf.keras.layers.Dense(self.path_state_dim, activation=tf.keras.activations.relu),
tf.keras.layers.Dense(self.path_state_dim, activation=tf.keras.activations.relu)
])
self.queue_embedding = tf.keras.Sequential([
tf.keras.layers.Input(shape=self.max_num_queues + 2),
tf.keras.layers.Dense(self.queue_state_dim, activation=tf.keras.activations.relu),
tf.keras.layers.Dense(self.queue_state_dim, activation=tf.keras.activations.relu)
])
self.link_embedding = tf.keras.Sequential([
tf.keras.layers.Input(shape=self.num_policies + 1),
tf.keras.layers.Dense(self.link_state_dim, activation=tf.keras.activations.relu),
tf.keras.layers.Dense(self.link_state_dim, activation=tf.keras.activations.relu)
])
self.readout_path = tf.keras.Sequential([
tf.keras.layers.Input(shape=(None, self.path_state_dim)),
tf.keras.layers.Dense(int(self.link_state_dim / 2),
activation=tf.keras.activations.relu),
tf.keras.layers.Dense(int(self.path_state_dim / 2),
activation=tf.keras.activations.relu),
tf.keras.layers.Dense(1)
], name="PathReadout")
@tf.function
def call(self, inputs):
traffic = inputs['traffic']
packets = inputs['packets']
length = inputs['length']
model = inputs['model']
eq_lambda = inputs['eq_lambda']
avg_pkts_lambda = inputs['avg_pkts_lambda']
exp_max_factor = inputs['exp_max_factor']
pkts_lambda_on = inputs['pkts_lambda_on']
avg_t_off = inputs['avg_t_off']
avg_t_on = inputs['avg_t_on']
ar_a = inputs['ar_a']
sigma = inputs['sigma']
capacity = inputs['capacity']
policy = tf.one_hot(inputs['policy'], self.num_policies)
queue_size = inputs['queue_size']
priority = tf.one_hot(inputs['priority'], self.max_num_queues)
weight = inputs['weight']
queue_to_path = inputs['queue_to_path']
link_to_path = inputs['link_to_path']
path_to_link = inputs['path_to_link']
path_to_queue = inputs['path_to_queue']
queue_to_link = inputs['queue_to_link']
path_gather_traffic = tf.gather(traffic, path_to_link[:, :, 0])
load = tf.math.reduce_sum(path_gather_traffic, axis=1) / capacity
pkt_size = traffic / packets
# Initialize the initial hidden state for links
path_state = self.path_embedding(tf.concat(
[(traffic - self.z_score['traffic'][0]) / self.z_score['traffic'][1],
(packets - self.z_score['packets'][0]) / self.z_score['packets'][1],
tf.one_hot(model, self.max_num_models),
(eq_lambda - self.z_score['eq_lambda'][0]) / self.z_score['eq_lambda'][1],
(avg_pkts_lambda - self.z_score['avg_pkts_lambda'][0]) / self.z_score['avg_pkts_lambda'][1],
(exp_max_factor - self.z_score['exp_max_factor'][0]) / self.z_score['exp_max_factor'][1],
(pkts_lambda_on - self.z_score['pkts_lambda_on'][0]) / self.z_score['pkts_lambda_on'][1],
(avg_t_off - self.z_score['avg_t_off'][0]) / self.z_score['avg_t_off'][1],
(avg_t_on - self.z_score['avg_t_on'][0]) / self.z_score['avg_t_on'][1],
(ar_a - self.z_score['ar_a'][0]) / self.z_score['ar_a'][1],
(sigma - self.z_score['sigma'][0]) / self.z_score['sigma'][1]], axis=1))
# Initialize the initial hidden state for paths
link_state = self.link_embedding(tf.concat([load, policy], axis=1))
# Initialize the initial hidden state for paths
queue_state = self.queue_embedding(
tf.concat([(queue_size - self.z_score['queue_size'][0]) / self.z_score['queue_size'][1],
priority, weight], axis=1))
# Iterate t times doing the message passing
for it in range(self.iterations):
###################
# LINK AND QUEUE #
# TO PATH #
###################
queue_gather = tf.gather(queue_state, queue_to_path)
link_gather = tf.gather(link_state, link_to_path, name="LinkToPath")
path_update_rnn = tf.keras.layers.RNN(self.path_update,
return_sequences=True,
return_state=True)
previous_path_state = path_state
path_state_sequence, path_state = path_update_rnn(tf.concat([queue_gather, link_gather], axis=2),
initial_state=path_state)
path_state_sequence = tf.concat([tf.expand_dims(previous_path_state, 1), path_state_sequence], axis=1)
###################
# PATH TO QUEUE #
###################
path_gather = tf.gather_nd(path_state_sequence, path_to_queue)
path_sum = tf.math.reduce_sum(path_gather, axis=1)
queue_state, _ = self.queue_update(path_sum, [queue_state])
###################
# QUEUE TO LINK #
###################
queue_gather = tf.gather(queue_state, queue_to_link)
link_gru_rnn = tf.keras.layers.RNN(self.link_update, return_sequences=False)
link_state = link_gru_rnn(queue_gather, initial_state=link_state)
capacity_gather = tf.gather(capacity, link_to_path)
input_tensor = path_state_sequence[:, 1:].to_tensor()
occupancy_gather = self.readout_path(input_tensor)
length = tf.ensure_shape(length, [None])
occupancy_gather = tf.RaggedTensor.from_tensor(occupancy_gather, lengths=length)
queue_delay = tf.math.reduce_sum(occupancy_gather / capacity_gather,
axis=1)
trans_delay = pkt_size * tf.math.reduce_sum(1 / capacity_gather, axis=1)
return queue_delay + trans_delay