1#
2# Copyright (C) 2018 The Android Open Source Project
3#
4# Licensed under the Apache License, Version 2.0 (the "License");
5# you may not use this file except in compliance with the License.
6# You may obtain a copy of the License at
7#
8#      http://www.apache.org/licenses/LICENSE-2.0
9#
10# Unless required by applicable law or agreed to in writing, software
11# distributed under the License is distributed on an "AS IS" BASIS,
12# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13# See the License for the specific language governing permissions and
14# limitations under the License.
15#
16
17batches = 2
18features = 8
19rank = 2
20units = int(features / rank)
21input_size = 3
22memory_size = 10
23
24model = Model()
25
26input = Input("input", "TENSOR_FLOAT32", "{%d, %d}" % (batches, input_size))
27weights_feature = Input("weights_feature", "TENSOR_FLOAT32", "{%d, %d}" % (features, input_size))
28weights_time = Input("weights_time", "TENSOR_FLOAT32", "{%d, %d}" % (features, memory_size))
29bias = Input("bias", "TENSOR_FLOAT32", "{%d}" % (units))
30state_in = Input("state_in", "TENSOR_FLOAT32", "{%d, %d}" % (batches, memory_size*features))
31rank_param = Int32Scalar("rank_param", rank)
32activation_param = Int32Scalar("activation_param", 0)
33state_out = IgnoredOutput("state_out", "TENSOR_FLOAT32", "{%d, %d}" % (batches, memory_size*features))
34output = Output("output", "TENSOR_FLOAT32", "{%d, %d}" % (batches, units))
35
36model = model.Operation("SVDF", input, weights_feature, weights_time, bias, state_in,
37                        rank_param, activation_param).To([state_out, output])
38model = model.RelaxedExecution(True)
39
40input0 = {
41    input: [],
42    weights_feature: [
43      -0.31930989, 0.0079667,   0.39296314,  0.37613347,
44      0.12416199,  0.15785322,  0.27901134,  0.3905206,
45      0.21931258,  -0.36137494, -0.10640851, 0.31053296,
46      -0.36118156, -0.0976817,  -0.36916667, 0.22197971,
47      0.15294972,  0.38031587,  0.27557442,  0.39635518,
48      -0.21580373, -0.06634006, -0.02702999, 0.27072677
49    ],
50    weights_time: [
51      -0.31930989, 0.37613347,  0.27901134,  -0.36137494, -0.36118156,
52       0.22197971,  0.27557442,  -0.06634006, 0.0079667,   0.12416199,
53
54       0.3905206,   -0.10640851, -0.0976817,  0.15294972,  0.39635518,
55       -0.02702999, 0.39296314,  0.15785322,  0.21931258,  0.31053296,
56
57       -0.36916667, 0.38031587,  -0.21580373, 0.27072677,  0.23622236,
58       0.34936687,  0.18174365,  0.35907319,  -0.17493086, 0.324846,
59
60       -0.10781813, 0.27201805,  0.14324132,  -0.23681851, -0.27115166,
61       -0.01580888, -0.14943552, 0.15465137,  0.09784451,  -0.0337657,
62
63       -0.14884081, 0.19931212,  -0.36002168, 0.34663299,  -0.11405486,
64       0.12672701,  0.39463779,  -0.07886535, -0.06384811, 0.08249187,
65
66       -0.26816407, -0.19905911, 0.29211238,  0.31264046,  -0.28664589,
67       0.05698794,  0.11613581,  0.14078894,  0.02187902,  -0.21781836,
68
69       -0.15567942, 0.08693647,  -0.38256618, 0.36580828,  -0.22922277,
70       -0.0226903,  0.12878349,  -0.28122205, -0.10850525, -0.11955214,
71
72       0.27179423,  -0.04710215, 0.31069002,  0.22672787,  0.09580326,
73       0.08682203,  0.1258215,   0.1851041,   0.29228821,  0.12366763
74    ],
75    bias: [],
76    state_in: [0 for _ in range(batches * memory_size * features)],
77}
78
79test_inputs = [
80    0.12609188,  -0.46347019, -0.89598465,
81    0.35867718,  0.36897406,  0.73463392,
82
83    0.14278367,  -1.64410412, -0.75222826,
84    -0.57290924, 0.12729003,  0.7567004,
85
86    0.49837467,  0.19278903,  0.26584083,
87    0.17660543,  0.52949083,  -0.77931279,
88
89    -0.11186574, 0.13164264,  -0.05349274,
90    -0.72674477, -0.5683046,  0.55900657,
91
92    -0.68892461, 0.37783599,  0.18263303,
93    -0.63690937, 0.44483393,  -0.71817774,
94
95    -0.81299269, -0.86831826, 1.43940818,
96    -0.95760226, 1.82078898,  0.71135032,
97
98    -1.45006323, -0.82251364, -1.69082689,
99    -1.65087092, -1.89238167, 1.54172635,
100
101    0.03966608,  -0.24936394, -0.77526885,
102    2.06740379,  -1.51439476, 1.43768692,
103
104    0.11771342,  -0.23761693, -0.65898693,
105    0.31088525,  -1.55601168, -0.87661445,
106
107    -0.89477462, 1.67204106,  -0.53235275,
108    -0.6230064,  0.29819036,  1.06939757,
109]
110
111golden_outputs = [
112    -0.09623547, -0.10193135, 0.11083051,  -0.0347917,
113    0.1141196,   0.12965347,  -0.12652366, 0.01007236,
114
115    -0.16396809, -0.21247184, 0.11259045,  -0.04156673,
116    0.10132131,  -0.06143532, -0.00924693, 0.10084561,
117
118    0.01257364,  0.0506071,   -0.19287863, -0.07162561,
119    -0.02033747, 0.22673416,  0.15487903,  0.02525555,
120
121    -0.1411963,  -0.37054959, 0.01774767,  0.05867489,
122    0.09607603,  -0.0141301,  -0.08995658, 0.12867066,
123
124    -0.27142537, -0.16955489, 0.18521598,  -0.12528358,
125    0.00331409,  0.11167502,  0.02218599,  -0.07309391,
126
127    0.09593632,  -0.28361851, -0.0773851,  0.17199151,
128    -0.00075242, 0.33691186,  -0.1536046,  0.16572715,
129
130    -0.27916506, -0.27626723, 0.42615682,  0.3225764,
131    -0.37472126, -0.55655634, -0.05013514, 0.289112,
132
133    -0.24418658, 0.07540751,  -0.1940318,  -0.08911639,
134    0.00732617,  0.46737891,  0.26449674,  0.24888524,
135
136    -0.17225097, -0.54660404, -0.38795233, 0.08389944,
137    0.07736043,  -0.28260678, 0.15666828,  1.14949894,
138
139    -0.57454878, -0.64704704, 0.73235172,  -0.34616736,
140    0.21120001,  -0.22927976, 0.02455296,  -0.35906726,
141]
142
143output0 = {state_out: [0 for _ in range(batches * memory_size * features)],
144           output: []}
145
146# TODO: enable more data points after fixing the reference issue
147for i in range(1):
148  batch_start = i * input_size * batches
149  batch_end = batch_start + input_size * batches
150  input0[input] = test_inputs[batch_start:batch_end]
151  golden_start = i * units * batches
152  golden_end = golden_start + units * batches
153  output0[output] = golden_outputs[golden_start:golden_end]
154  Example((input0, output0))
155