1 # 2 # Copyright (C) 2017 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 17 batches = 2 18 features = 4 19 rank = 1 20 units = int(features / rank) 21 input_size = 3 22 memory_size = 10 23 24 model = Model() 25 26 input = Input("input", "TENSOR_FLOAT32", "{%d, %d}" % (batches, input_size)) 27 weights_feature = Input("weights_feature", "TENSOR_FLOAT32", "{%d, %d}" % (features, input_size)) 28 weights_time = Input("weights_time", "TENSOR_FLOAT32", "{%d, %d}" % (features, memory_size)) 29 bias = Input("bias", "TENSOR_FLOAT32", "{%d}" % (units)) 30 state_in = Input("state_in", "TENSOR_FLOAT32", "{%d, %d}" % (batches, memory_size*features)) 31 rank_param = Int32Scalar("rank_param", rank) 32 activation_param = Int32Scalar("activation_param", 0) 33 state_out = IgnoredOutput("state_out", "TENSOR_FLOAT32", "{%d, %d}" % (batches, memory_size*features)) 34 output = Output("output", "TENSOR_FLOAT32", "{%d, %d}" % (batches, units)) 35 36 model = model.Operation("SVDF", input, weights_feature, weights_time, bias, state_in, 37 rank_param, activation_param).To([state_out, output]) 38 39 input0 = { 40 input: [], 41 weights_feature: [ 42 -0.31930989, -0.36118156, 0.0079667, 0.37613347, 43 0.22197971, 0.12416199, 0.27901134, 0.27557442, 44 0.3905206, -0.36137494, -0.06634006, -0.10640851 45 ], 46 weights_time: [ 47 -0.31930989, 0.37613347, 0.27901134, -0.36137494, -0.36118156, 48 0.22197971, 0.27557442, -0.06634006, 0.0079667, 0.12416199, 49 50 0.3905206, -0.10640851, -0.0976817, 0.15294972, 0.39635518, 51 -0.02702999, 0.39296314, 0.15785322, 0.21931258, 0.31053296, 52 53 -0.36916667, 0.38031587, -0.21580373, 0.27072677, 0.23622236, 54 0.34936687, 0.18174365, 0.35907319, -0.17493086, 0.324846, 55 56 -0.10781813, 0.27201805, 0.14324132, -0.23681851, -0.27115166, 57 -0.01580888, -0.14943552, 0.15465137, 0.09784451, -0.0337657 58 ], 59 bias: [], 60 state_in: [0 for _ in range(batches * memory_size * features)], 61 } 62 63 test_inputs = [ 64 0.12609188, -0.46347019, -0.89598465, 65 0.12609188, -0.46347019, -0.89598465, 66 67 0.14278367, -1.64410412, -0.75222826, 68 0.14278367, -1.64410412, -0.75222826, 69 70 0.49837467, 0.19278903, 0.26584083, 71 0.49837467, 0.19278903, 0.26584083, 72 73 -0.11186574, 0.13164264, -0.05349274, 74 -0.11186574, 0.13164264, -0.05349274, 75 76 -0.68892461, 0.37783599, 0.18263303, 77 -0.68892461, 0.37783599, 0.18263303, 78 79 -0.81299269, -0.86831826, 1.43940818, 80 -0.81299269, -0.86831826, 1.43940818, 81 82 -1.45006323, -0.82251364, -1.69082689, 83 -1.45006323, -0.82251364, -1.69082689, 84 85 0.03966608, -0.24936394, -0.77526885, 86 0.03966608, -0.24936394, -0.77526885, 87 88 0.11771342, -0.23761693, -0.65898693, 89 0.11771342, -0.23761693, -0.65898693, 90 91 -0.89477462, 1.67204106, -0.53235275, 92 -0.89477462, 1.67204106, -0.53235275 93 ] 94 95 golden_outputs = [ 96 0.014899, -0.0517661, -0.143725, -0.00271883, 97 0.014899, -0.0517661, -0.143725, -0.00271883, 98 99 0.068281, -0.162217, -0.152268, 0.00323521, 100 0.068281, -0.162217, -0.152268, 0.00323521, 101 102 -0.0317821, -0.0333089, 0.0609602, 0.0333759, 103 -0.0317821, -0.0333089, 0.0609602, 0.0333759, 104 105 -0.00623099, -0.077701, -0.391193, -0.0136691, 106 -0.00623099, -0.077701, -0.391193, -0.0136691, 107 108 0.201551, -0.164607, -0.179462, -0.0592739, 109 0.201551, -0.164607, -0.179462, -0.0592739, 110 111 0.0886511, -0.0875401, -0.269283, 0.0281379, 112 0.0886511, -0.0875401, -0.269283, 0.0281379, 113 114 -0.201174, -0.586145, -0.628624, -0.0330412, 115 -0.201174, -0.586145, -0.628624, -0.0330412, 116 117 -0.0839096, -0.299329, 0.108746, 0.109808, 118 -0.0839096, -0.299329, 0.108746, 0.109808, 119 120 0.419114, -0.237824, -0.422627, 0.175115, 121 0.419114, -0.237824, -0.422627, 0.175115, 122 123 0.36726, -0.522303, -0.456502, -0.175475, 124 0.36726, -0.522303, -0.456502, -0.175475 125 ] 126 127 output0 = {state_out: [0 for _ in range(batches * memory_size * features)], 128 output: []} 129 130 # TODO: enable more data points after fixing the reference issue 131 for i in range(1): 132 batch_start = i * input_size * batches 133 batch_end = batch_start + input_size * batches 134 input0[input] = test_inputs[batch_start:batch_end] 135 golden_start = i * units * batches 136 golden_end = golden_start + units * batches 137 output0[output] = golden_outputs[golden_start:golden_end] 138 Example((input0, output0)) 139
