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 #include <gmock/gmock.h>
18 #include <gtest/gtest.h>
19
20 #include <vector>
21
22 #include "LSTM.h"
23 #include "NeuralNetworksWrapper.h"
24
25 namespace android {
26 namespace nn {
27 namespace wrapper {
28
29 using ::testing::Each;
30 using ::testing::FloatNear;
31 using ::testing::Matcher;
32
33 namespace {
34
ArrayFloatNear(const std::vector<float> & values,float max_abs_error=1.e-5)35 std::vector<Matcher<float>> ArrayFloatNear(const std::vector<float>& values,
36 float max_abs_error = 1.e-5) {
37 std::vector<Matcher<float>> matchers;
38 matchers.reserve(values.size());
39 for (const float& v : values) {
40 matchers.emplace_back(FloatNear(v, max_abs_error));
41 }
42 return matchers;
43 }
44
45 } // anonymous namespace
46
47 #define FOR_ALL_INPUT_AND_WEIGHT_TENSORS(ACTION) \
48 ACTION(Input) \
49 ACTION(InputToInputWeights) \
50 ACTION(InputToCellWeights) \
51 ACTION(InputToForgetWeights) \
52 ACTION(InputToOutputWeights) \
53 ACTION(RecurrentToInputWeights) \
54 ACTION(RecurrentToCellWeights) \
55 ACTION(RecurrentToForgetWeights) \
56 ACTION(RecurrentToOutputWeights) \
57 ACTION(CellToInputWeights) \
58 ACTION(CellToForgetWeights) \
59 ACTION(CellToOutputWeights) \
60 ACTION(InputGateBias) \
61 ACTION(CellGateBias) \
62 ACTION(ForgetGateBias) \
63 ACTION(OutputGateBias) \
64 ACTION(ProjectionWeights) \
65 ACTION(ProjectionBias) \
66 ACTION(OutputStateIn) \
67 ACTION(CellStateIn)
68
69 // For all output and intermediate states
70 #define FOR_ALL_OUTPUT_TENSORS(ACTION) \
71 ACTION(ScratchBuffer) \
72 ACTION(OutputStateOut) \
73 ACTION(CellStateOut) \
74 ACTION(Output)
75
76 class LSTMOpModel {
77 public:
LSTMOpModel(uint32_t n_batch,uint32_t n_input,uint32_t n_cell,uint32_t n_output,bool use_cifg,bool use_peephole,bool use_projection_weights,bool use_projection_bias,float cell_clip,float proj_clip,const std::vector<std::vector<uint32_t>> & input_shapes0)78 LSTMOpModel(uint32_t n_batch, uint32_t n_input, uint32_t n_cell, uint32_t n_output,
79 bool use_cifg, bool use_peephole, bool use_projection_weights,
80 bool use_projection_bias, float cell_clip, float proj_clip,
81 const std::vector<std::vector<uint32_t>>& input_shapes0)
82 : n_input_(n_input),
83 n_output_(n_output),
84 use_cifg_(use_cifg),
85 use_peephole_(use_peephole),
86 use_projection_weights_(use_projection_weights),
87 use_projection_bias_(use_projection_bias),
88 activation_(ActivationFn::kActivationTanh),
89 cell_clip_(cell_clip),
90 proj_clip_(proj_clip) {
91 std::vector<uint32_t> inputs;
92 std::vector<std::vector<uint32_t>> input_shapes(input_shapes0);
93
94 input_shapes.push_back({n_batch, n_output});
95 input_shapes.push_back({n_batch, n_cell});
96 auto it = input_shapes.begin();
97
98 // Input and weights
99 #define AddInput(X) \
100 OperandType X##OpndTy(Type::TENSOR_FLOAT32, *it++); \
101 inputs.push_back(model_.addOperand(&X##OpndTy));
102
103 FOR_ALL_INPUT_AND_WEIGHT_TENSORS(AddInput);
104
105 #undef AddOperand
106
107 // Parameters
108 OperandType ActivationOpndTy(Type::INT32, {});
109 inputs.push_back(model_.addOperand(&ActivationOpndTy));
110 OperandType CellClipOpndTy(Type::FLOAT32, {});
111 inputs.push_back(model_.addOperand(&CellClipOpndTy));
112 OperandType ProjClipOpndTy(Type::FLOAT32, {});
113 inputs.push_back(model_.addOperand(&ProjClipOpndTy));
114
115 // Output and other intermediate state
116 std::vector<std::vector<uint32_t>> output_shapes{
117 {n_batch, n_cell * (use_cifg ? 3 : 4)},
118 {n_batch, n_output},
119 {n_batch, n_cell},
120 {n_batch, n_output},
121 };
122 std::vector<uint32_t> outputs;
123
124 auto it2 = output_shapes.begin();
125
126 #define AddOutput(X) \
127 OperandType X##OpndTy(Type::TENSOR_FLOAT32, *it2++); \
128 outputs.push_back(model_.addOperand(&X##OpndTy));
129
130 FOR_ALL_OUTPUT_TENSORS(AddOutput);
131
132 #undef AddOutput
133
134 model_.addOperation(ANEURALNETWORKS_LSTM, inputs, outputs);
135 model_.identifyInputsAndOutputs(inputs, outputs);
136
137 Input_.insert(Input_.end(), n_batch * n_input, 0.f);
138 OutputStateIn_.insert(OutputStateIn_.end(), n_batch * n_output, 0.f);
139 CellStateIn_.insert(CellStateIn_.end(), n_batch * n_cell, 0.f);
140
141 auto multiAll = [](const std::vector<uint32_t>& dims) -> uint32_t {
142 uint32_t sz = 1;
143 for (uint32_t d : dims) {
144 sz *= d;
145 }
146 return sz;
147 };
148
149 it2 = output_shapes.begin();
150
151 #define ReserveOutput(X) X##_.insert(X##_.end(), multiAll(*it2++), 0.f);
152
153 FOR_ALL_OUTPUT_TENSORS(ReserveOutput);
154
155 #undef ReserveOutput
156
157 model_.finish();
158 }
159
160 #define DefineSetter(X) \
161 void Set##X(const std::vector<float>& f) { X##_.insert(X##_.end(), f.begin(), f.end()); }
162
163 FOR_ALL_INPUT_AND_WEIGHT_TENSORS(DefineSetter);
164
165 #undef DefineSetter
166
ResetOutputState()167 void ResetOutputState() {
168 std::fill(OutputStateIn_.begin(), OutputStateIn_.end(), 0.f);
169 std::fill(OutputStateOut_.begin(), OutputStateOut_.end(), 0.f);
170 }
171
ResetCellState()172 void ResetCellState() {
173 std::fill(CellStateIn_.begin(), CellStateIn_.end(), 0.f);
174 std::fill(CellStateOut_.begin(), CellStateOut_.end(), 0.f);
175 }
176
SetInput(int offset,float * begin,float * end)177 void SetInput(int offset, float* begin, float* end) {
178 for (; begin != end; begin++, offset++) {
179 Input_[offset] = *begin;
180 }
181 }
182
num_inputs() const183 uint32_t num_inputs() const { return n_input_; }
num_outputs() const184 uint32_t num_outputs() const { return n_output_; }
185
GetOutput() const186 const std::vector<float>& GetOutput() const { return Output_; }
187
Invoke()188 void Invoke() {
189 ASSERT_TRUE(model_.isValid());
190
191 OutputStateIn_.swap(OutputStateOut_);
192 CellStateIn_.swap(CellStateOut_);
193
194 Compilation compilation(&model_);
195 compilation.finish();
196 Execution execution(&compilation);
197 #define SetInputOrWeight(X) \
198 ASSERT_EQ( \
199 execution.setInput(LSTMCell::k##X##Tensor, X##_.data(), sizeof(float) * X##_.size()), \
200 Result::NO_ERROR);
201
202 FOR_ALL_INPUT_AND_WEIGHT_TENSORS(SetInputOrWeight);
203
204 #undef SetInputOrWeight
205
206 #define SetOutput(X) \
207 ASSERT_EQ( \
208 execution.setOutput(LSTMCell::k##X##Tensor, X##_.data(), sizeof(float) * X##_.size()), \
209 Result::NO_ERROR);
210
211 FOR_ALL_OUTPUT_TENSORS(SetOutput);
212
213 #undef SetOutput
214
215 if (use_cifg_) {
216 execution.setInput(LSTMCell::kInputToInputWeightsTensor, nullptr, 0);
217 execution.setInput(LSTMCell::kRecurrentToInputWeightsTensor, nullptr, 0);
218 }
219
220 if (use_peephole_) {
221 if (use_cifg_) {
222 execution.setInput(LSTMCell::kCellToInputWeightsTensor, nullptr, 0);
223 }
224 } else {
225 execution.setInput(LSTMCell::kCellToInputWeightsTensor, nullptr, 0);
226 execution.setInput(LSTMCell::kCellToForgetWeightsTensor, nullptr, 0);
227 execution.setInput(LSTMCell::kCellToOutputWeightsTensor, nullptr, 0);
228 }
229
230 if (use_projection_weights_) {
231 if (!use_projection_bias_) {
232 execution.setInput(LSTMCell::kProjectionBiasTensor, nullptr, 0);
233 }
234 } else {
235 execution.setInput(LSTMCell::kProjectionWeightsTensor, nullptr, 0);
236 execution.setInput(LSTMCell::kProjectionBiasTensor, nullptr, 0);
237 }
238
239 ASSERT_EQ(execution.setInput(LSTMCell::kActivationParam, &activation_, sizeof(activation_)),
240 Result::NO_ERROR);
241 ASSERT_EQ(execution.setInput(LSTMCell::kCellClipParam, &cell_clip_, sizeof(cell_clip_)),
242 Result::NO_ERROR);
243 ASSERT_EQ(execution.setInput(LSTMCell::kProjClipParam, &proj_clip_, sizeof(proj_clip_)),
244 Result::NO_ERROR);
245
246 ASSERT_EQ(execution.compute(), Result::NO_ERROR);
247 }
248
249 private:
250 Model model_;
251 // Execution execution_;
252 const uint32_t n_input_;
253 const uint32_t n_output_;
254
255 const bool use_cifg_;
256 const bool use_peephole_;
257 const bool use_projection_weights_;
258 const bool use_projection_bias_;
259
260 const int activation_;
261 const float cell_clip_;
262 const float proj_clip_;
263
264 #define DefineTensor(X) std::vector<float> X##_;
265
266 FOR_ALL_INPUT_AND_WEIGHT_TENSORS(DefineTensor);
267 FOR_ALL_OUTPUT_TENSORS(DefineTensor);
268
269 #undef DefineTensor
270 };
271
TEST(LSTMOpTest,BlackBoxTestNoCifgNoPeepholeNoProjectionNoClipping)272 TEST(LSTMOpTest, BlackBoxTestNoCifgNoPeepholeNoProjectionNoClipping) {
273 const int n_batch = 1;
274 const int n_input = 2;
275 // n_cell and n_output have the same size when there is no projection.
276 const int n_cell = 4;
277 const int n_output = 4;
278
279 LSTMOpModel lstm(n_batch, n_input, n_cell, n_output,
280 /*use_cifg=*/false, /*use_peephole=*/false,
281 /*use_projection_weights=*/false,
282 /*use_projection_bias=*/false,
283 /*cell_clip=*/0.0, /*proj_clip=*/0.0,
284 {
285 {n_batch, n_input}, // input tensor
286
287 {n_cell, n_input}, // input_to_input_weight tensor
288 {n_cell, n_input}, // input_to_forget_weight tensor
289 {n_cell, n_input}, // input_to_cell_weight tensor
290 {n_cell, n_input}, // input_to_output_weight tensor
291
292 {n_cell, n_output}, // recurrent_to_input_weight tensor
293 {n_cell, n_output}, // recurrent_to_forget_weight tensor
294 {n_cell, n_output}, // recurrent_to_cell_weight tensor
295 {n_cell, n_output}, // recurrent_to_output_weight tensor
296
297 {0}, // cell_to_input_weight tensor
298 {0}, // cell_to_forget_weight tensor
299 {0}, // cell_to_output_weight tensor
300
301 {n_cell}, // input_gate_bias tensor
302 {n_cell}, // forget_gate_bias tensor
303 {n_cell}, // cell_bias tensor
304 {n_cell}, // output_gate_bias tensor
305
306 {0, 0}, // projection_weight tensor
307 {0}, // projection_bias tensor
308 });
309
310 lstm.SetInputToInputWeights({-0.45018822, -0.02338299, -0.0870589, -0.34550029, 0.04266912,
311 -0.15680569, -0.34856534, 0.43890524});
312
313 lstm.SetInputToCellWeights({-0.50013041, 0.1370284, 0.11810488, 0.2013163, -0.20583314,
314 0.44344562, 0.22077113, -0.29909778});
315
316 lstm.SetInputToForgetWeights({0.09701663, 0.20334584, -0.50592935, -0.31343272, -0.40032279,
317 0.44781327, 0.01387155, -0.35593212});
318
319 lstm.SetInputToOutputWeights({-0.25065863, -0.28290087, 0.04613829, 0.40525138, 0.44272184,
320 0.03897077, -0.1556896, 0.19487578});
321
322 lstm.SetInputGateBias({0., 0., 0., 0.});
323
324 lstm.SetCellGateBias({0., 0., 0., 0.});
325
326 lstm.SetForgetGateBias({1., 1., 1., 1.});
327
328 lstm.SetOutputGateBias({0., 0., 0., 0.});
329
330 lstm.SetRecurrentToInputWeights({-0.0063535, -0.2042388, 0.31454784, -0.35746509, 0.28902304,
331 0.08183324, -0.16555229, 0.02286911, -0.13566875, 0.03034258,
332 0.48091322, -0.12528998, 0.24077177, -0.51332325, -0.33502164,
333 0.10629296});
334
335 lstm.SetRecurrentToCellWeights({-0.3407414, 0.24443203, -0.2078532, 0.26320225, 0.05695659,
336 -0.00123841, -0.4744786, -0.35869038, -0.06418842, -0.13502428,
337 -0.501764, 0.22830659, -0.46367589, 0.26016325, -0.03894562,
338 -0.16368064});
339
340 lstm.SetRecurrentToForgetWeights({-0.48684245, -0.06655136, 0.42224967, 0.2112639, 0.27654213,
341 0.20864892, -0.07646349, 0.45877004, 0.00141793, -0.14609534,
342 0.36447752, 0.09196436, 0.28053468, 0.01560611, -0.20127171,
343 -0.01140004});
344
345 lstm.SetRecurrentToOutputWeights({0.43385774, -0.17194885, 0.2718237, 0.09215671, 0.24107647,
346 -0.39835793, 0.18212086, 0.01301402, 0.48572797, -0.50656658,
347 0.20047462, -0.20607421, -0.51818722, -0.15390486, 0.0468148,
348 0.39922136});
349
350 static float lstm_input[] = {2., 3., 3., 4., 1., 1.};
351 static float lstm_golden_output[] = {-0.02973187, 0.1229473, 0.20885126, -0.15358765,
352 -0.03716109, 0.12507336, 0.41193449, -0.20860538,
353 -0.15053082, 0.09120187, 0.24278517, -0.12222792};
354
355 // Resetting cell_state and output_state
356 lstm.ResetCellState();
357 lstm.ResetOutputState();
358
359 const int input_sequence_size = sizeof(lstm_input) / sizeof(float) / (lstm.num_inputs());
360 for (int i = 0; i < input_sequence_size; i++) {
361 float* batch0_start = lstm_input + i * lstm.num_inputs();
362 float* batch0_end = batch0_start + lstm.num_inputs();
363
364 lstm.SetInput(0, batch0_start, batch0_end);
365
366 lstm.Invoke();
367
368 float* golden_start = lstm_golden_output + i * lstm.num_outputs();
369 float* golden_end = golden_start + lstm.num_outputs();
370 std::vector<float> expected;
371 expected.insert(expected.end(), golden_start, golden_end);
372 EXPECT_THAT(lstm.GetOutput(), ElementsAreArray(ArrayFloatNear(expected)));
373 }
374 }
375
TEST(LSTMOpTest,BlackBoxTestWithCifgWithPeepholeNoProjectionNoClipping)376 TEST(LSTMOpTest, BlackBoxTestWithCifgWithPeepholeNoProjectionNoClipping) {
377 const int n_batch = 1;
378 const int n_input = 2;
379 // n_cell and n_output have the same size when there is no projection.
380 const int n_cell = 4;
381 const int n_output = 4;
382
383 LSTMOpModel lstm(n_batch, n_input, n_cell, n_output,
384 /*use_cifg=*/true, /*use_peephole=*/true,
385 /*use_projection_weights=*/false,
386 /*use_projection_bias=*/false,
387 /*cell_clip=*/0.0, /*proj_clip=*/0.0,
388 {
389 {n_batch, n_input}, // input tensor
390
391 {0, 0}, // input_to_input_weight tensor
392 {n_cell, n_input}, // input_to_forget_weight tensor
393 {n_cell, n_input}, // input_to_cell_weight tensor
394 {n_cell, n_input}, // input_to_output_weight tensor
395
396 {0, 0}, // recurrent_to_input_weight tensor
397 {n_cell, n_output}, // recurrent_to_forget_weight tensor
398 {n_cell, n_output}, // recurrent_to_cell_weight tensor
399 {n_cell, n_output}, // recurrent_to_output_weight tensor
400
401 {0}, // cell_to_input_weight tensor
402 {n_cell}, // cell_to_forget_weight tensor
403 {n_cell}, // cell_to_output_weight tensor
404
405 {n_cell}, // input_gate_bias tensor
406 {n_cell}, // forget_gate_bias tensor
407 {n_cell}, // cell_bias tensor
408 {n_cell}, // output_gate_bias tensor
409
410 {0, 0}, // projection_weight tensor
411 {0}, // projection_bias tensor
412 });
413
414 lstm.SetInputToCellWeights({-0.49770179, -0.27711356, -0.09624726, 0.05100781, 0.04717243,
415 0.48944736, -0.38535351, -0.17212132});
416
417 lstm.SetInputToForgetWeights({-0.55291498, -0.42866567, 0.13056988, -0.3633365, -0.22755712,
418 0.28253698, 0.24407166, 0.33826375});
419
420 lstm.SetInputToOutputWeights({0.10725588, -0.02335852, -0.55932593, -0.09426838, -0.44257352,
421 0.54939759, 0.01533556, 0.42751634});
422
423 lstm.SetCellGateBias({0., 0., 0., 0.});
424
425 lstm.SetForgetGateBias({1., 1., 1., 1.});
426
427 lstm.SetOutputGateBias({0., 0., 0., 0.});
428
429 lstm.SetRecurrentToCellWeights({0.54066205, -0.32668582, -0.43562764, -0.56094903, 0.42957711,
430 0.01841056, -0.32764608, -0.33027974, -0.10826075, 0.20675004,
431 0.19069612, -0.03026325, -0.54532051, 0.33003211, 0.44901288,
432 0.21193194});
433
434 lstm.SetRecurrentToForgetWeights({-0.13832897, -0.0515101, -0.2359007, -0.16661474, -0.14340827,
435 0.36986142, 0.23414481, 0.55899, 0.10798943, -0.41174671,
436 0.17751795, -0.34484994, -0.35874045, -0.11352962, 0.27268326,
437 0.54058349});
438
439 lstm.SetRecurrentToOutputWeights({0.41613156, 0.42610586, -0.16495961, -0.5663873, 0.30579174,
440 -0.05115908, -0.33941799, 0.23364776, 0.11178309, 0.09481031,
441 -0.26424935, 0.46261835, 0.50248802, 0.26114327, -0.43736315,
442 0.33149987});
443
444 lstm.SetCellToForgetWeights({0.47485286, -0.51955009, -0.24458408, 0.31544167});
445 lstm.SetCellToOutputWeights({-0.17135078, 0.82760304, 0.85573703, -0.77109635});
446
447 static float lstm_input[] = {2., 3., 3., 4., 1., 1.};
448 static float lstm_golden_output[] = {-0.36444446, -0.00352185, 0.12886585, -0.05163646,
449 -0.42312205, -0.01218222, 0.24201041, -0.08124574,
450 -0.358325, -0.04621704, 0.21641694, -0.06471302};
451
452 // Resetting cell_state and output_state
453 lstm.ResetCellState();
454 lstm.ResetOutputState();
455
456 const int input_sequence_size = sizeof(lstm_input) / sizeof(float) / (lstm.num_inputs());
457 for (int i = 0; i < input_sequence_size; i++) {
458 float* batch0_start = lstm_input + i * lstm.num_inputs();
459 float* batch0_end = batch0_start + lstm.num_inputs();
460
461 lstm.SetInput(0, batch0_start, batch0_end);
462
463 lstm.Invoke();
464
465 float* golden_start = lstm_golden_output + i * lstm.num_outputs();
466 float* golden_end = golden_start + lstm.num_outputs();
467 std::vector<float> expected;
468 expected.insert(expected.end(), golden_start, golden_end);
469 EXPECT_THAT(lstm.GetOutput(), ElementsAreArray(ArrayFloatNear(expected)));
470 }
471 }
472
TEST(LSTMOpTest,BlackBoxTestWithPeepholeWithProjectionNoClipping)473 TEST(LSTMOpTest, BlackBoxTestWithPeepholeWithProjectionNoClipping) {
474 const int n_batch = 2;
475 const int n_input = 5;
476 const int n_cell = 20;
477 const int n_output = 16;
478
479 LSTMOpModel lstm(n_batch, n_input, n_cell, n_output,
480 /*use_cifg=*/false, /*use_peephole=*/true,
481 /*use_projection_weights=*/true,
482 /*use_projection_bias=*/false,
483 /*cell_clip=*/0.0, /*proj_clip=*/0.0,
484 {
485 {n_batch, n_input}, // input tensor
486
487 {n_cell, n_input}, // input_to_input_weight tensor
488 {n_cell, n_input}, // input_to_forget_weight tensor
489 {n_cell, n_input}, // input_to_cell_weight tensor
490 {n_cell, n_input}, // input_to_output_weight tensor
491
492 {n_cell, n_output}, // recurrent_to_input_weight tensor
493 {n_cell, n_output}, // recurrent_to_forget_weight tensor
494 {n_cell, n_output}, // recurrent_to_cell_weight tensor
495 {n_cell, n_output}, // recurrent_to_output_weight tensor
496
497 {n_cell}, // cell_to_input_weight tensor
498 {n_cell}, // cell_to_forget_weight tensor
499 {n_cell}, // cell_to_output_weight tensor
500
501 {n_cell}, // input_gate_bias tensor
502 {n_cell}, // forget_gate_bias tensor
503 {n_cell}, // cell_bias tensor
504 {n_cell}, // output_gate_bias tensor
505
506 {n_output, n_cell}, // projection_weight tensor
507 {0}, // projection_bias tensor
508 });
509
510 lstm.SetInputToInputWeights(
511 {0.021393683, 0.06124551, 0.046905167, -0.014657677, -0.03149463, 0.09171803,
512 0.14647801, 0.10797193, -0.0057968358, 0.0019193048, -0.2726754, 0.10154029,
513 -0.018539885, 0.080349885, -0.10262385, -0.022599787, -0.09121155, -0.008675967,
514 -0.045206103, -0.0821282, -0.008045952, 0.015478081, 0.055217247, 0.038719587,
515 0.044153627, -0.06453243, 0.05031825, -0.046935108, -0.008164439, 0.014574226,
516 -0.1671009, -0.15519552, -0.16819797, -0.13971269, -0.11953059, 0.25005487,
517 -0.22790983, 0.009855087, -0.028140958, -0.11200698, 0.11295408, -0.0035217577,
518 0.054485075, 0.05184695, 0.064711206, 0.10989193, 0.11674786, 0.03490607,
519 0.07727357, 0.11390585, -0.1863375, -0.1034451, -0.13945189, -0.049401227,
520 -0.18767063, 0.042483903, 0.14233552, 0.13832581, 0.18350165, 0.14545603,
521 -0.028545704, 0.024939531, 0.050929718, 0.0076203286, -0.0029723682, -0.042484224,
522 -0.11827596, -0.09171104, -0.10808628, -0.16327988, -0.2273378, -0.0993647,
523 -0.017155107, 0.0023917493, 0.049272764, 0.0038534778, 0.054764505, 0.089753784,
524 0.06947234, 0.08014476, -0.04544234, -0.0497073, -0.07135631, -0.048929106,
525 -0.004042012, -0.009284026, 0.018042054, 0.0036860977, -0.07427302, -0.11434604,
526 -0.018995456, 0.031487543, 0.012834908, 0.019977754, 0.044256654, -0.39292613,
527 -0.18519334, -0.11651281, -0.06809892, 0.011373677});
528
529 lstm.SetInputToForgetWeights(
530 {-0.0018401089, -0.004852237, 0.03698424, 0.014181704, 0.028273236, -0.016726194,
531 -0.05249759, -0.10204261, 0.00861066, -0.040979505, -0.009899187, 0.01923892,
532 -0.028177269, -0.08535103, -0.14585495, 0.10662567, -0.01909731, -0.017883534,
533 -0.0047269356, -0.045103323, 0.0030784295, 0.076784775, 0.07463696, 0.094531395,
534 0.0814421, -0.12257899, -0.033945758, -0.031303465, 0.045630626, 0.06843887,
535 -0.13492945, -0.012480007, -0.0811829, -0.07224499, -0.09628791, 0.045100946,
536 0.0012300825, 0.013964662, 0.099372394, 0.02543059, 0.06958324, 0.034257296,
537 0.0482646, 0.06267997, 0.052625068, 0.12784666, 0.07077897, 0.025725935,
538 0.04165009, 0.07241905, 0.018668644, -0.037377294, -0.06277783, -0.08833636,
539 -0.040120605, -0.011405586, -0.007808335, -0.010301386, -0.005102167, 0.027717464,
540 0.05483423, 0.11449111, 0.11289652, 0.10939839, 0.13396506, -0.08402166,
541 -0.01901462, -0.044678304, -0.07720565, 0.014350063, -0.11757958, -0.0652038,
542 -0.08185733, -0.076754324, -0.092614375, 0.10405491, 0.052960336, 0.035755895,
543 0.035839386, -0.012540553, 0.036881298, 0.02913376, 0.03420159, 0.05448447,
544 -0.054523353, 0.02582715, 0.02327355, -0.011857179, -0.0011980024, -0.034641717,
545 -0.026125094, -0.17582615, -0.15923657, -0.27486774, -0.0006143371, 0.0001771948,
546 -8.470171e-05, 0.02651807, 0.045790765, 0.06956496});
547
548 lstm.SetInputToCellWeights(
549 {-0.04580283, -0.09549462, -0.032418985, -0.06454633, -0.043528453, 0.043018587,
550 -0.049152344, -0.12418144, -0.078985475, -0.07596889, 0.019484362, -0.11434962,
551 -0.0074034138, -0.06314844, -0.092981495, 0.0062155537, -0.025034338, -0.0028890965,
552 0.048929527, 0.06235075, 0.10665918, -0.032036792, -0.08505916, -0.10843358,
553 -0.13002433, -0.036816437, -0.02130134, -0.016518239, 0.0047691227, -0.0025825808,
554 0.066017866, 0.029991534, -0.10652836, -0.1037554, -0.13056071, -0.03266643,
555 -0.033702414, -0.006473424, -0.04611692, 0.014419339, -0.025174323, 0.0396852,
556 0.081777506, 0.06157468, 0.10210095, -0.009658194, 0.046511717, 0.03603906,
557 0.0069369148, 0.015960095, -0.06507666, 0.09551598, 0.053568836, 0.06408714,
558 0.12835667, -0.008714329, -0.20211966, -0.12093674, 0.029450472, 0.2849013,
559 -0.029227901, 0.1164364, -0.08560263, 0.09941786, -0.036999565, -0.028842626,
560 -0.0033637602, -0.017012902, -0.09720865, -0.11193351, -0.029155117, -0.017936034,
561 -0.009768936, -0.04223324, -0.036159635, 0.06505112, -0.021742892, -0.023377212,
562 -0.07221364, -0.06430552, 0.05453865, 0.091149814, 0.06387331, 0.007518393,
563 0.055960953, 0.069779344, 0.046411168, 0.10509911, 0.07463894, 0.0075130584,
564 0.012850982, 0.04555431, 0.056955688, 0.06555285, 0.050801456, -0.009862683,
565 0.00826772, -0.026555609, -0.0073611983, -0.0014897042});
566
567 lstm.SetInputToOutputWeights(
568 {-0.0998932, -0.07201956, -0.052803773, -0.15629593, -0.15001918, -0.07650751,
569 0.02359855, -0.075155355, -0.08037709, -0.15093534, 0.029517552, -0.04751393,
570 0.010350531, -0.02664851, -0.016839722, -0.023121163, 0.0077019283, 0.012851257,
571 -0.05040649, -0.0129761, -0.021737747, -0.038305793, -0.06870586, -0.01481247,
572 -0.001285394, 0.10124236, 0.083122835, 0.053313006, -0.062235646, -0.075637154,
573 -0.027833903, 0.029774971, 0.1130802, 0.09218906, 0.09506135, -0.086665764,
574 -0.037162706, -0.038880914, -0.035832845, -0.014481564, -0.09825003, -0.12048569,
575 -0.097665586, -0.05287633, -0.0964047, -0.11366429, 0.035777505, 0.13568819,
576 0.052451383, 0.050649304, 0.05798951, -0.021852335, -0.099848844, 0.014740475,
577 -0.078897946, 0.04974699, 0.014160473, 0.06973932, 0.04964942, 0.033364646,
578 0.08190124, 0.025535367, 0.050893165, 0.048514254, 0.06945813, -0.078907564,
579 -0.06707616, -0.11844508, -0.09986688, -0.07509403, 0.06263226, 0.14925587,
580 0.20188436, 0.12098451, 0.14639415, 0.0015017595, -0.014267382, -0.03417257,
581 0.012711468, 0.0028300495, -0.024758482, -0.05098548, -0.0821182, 0.014225672,
582 0.021544158, 0.08949725, 0.07505268, -0.0020780868, 0.04908258, 0.06476295,
583 -0.022907063, 0.027562456, 0.040185735, 0.019567577, -0.015598739, -0.049097303,
584 -0.017121866, -0.083368234, -0.02332002, -0.0840956});
585
586 lstm.SetInputGateBias({0.02234832, 0.14757581, 0.18176508, 0.10380666, 0.053110216,
587 -0.06928846, -0.13942584, -0.11816189, 0.19483899, 0.03652339,
588 -0.10250295, 0.036714908, -0.18426876, 0.036065217, 0.21810818,
589 0.02383196, -0.043370757, 0.08690144, -0.04444982, 0.00030581196});
590
591 lstm.SetForgetGateBias({0.035185695, -0.042891346, -0.03032477, 0.23027696, 0.11098921,
592 0.15378423, 0.09263801, 0.09790885, 0.09508917, 0.061199076,
593 0.07665568, -0.015443159, -0.03499149, 0.046190713, 0.08895977,
594 0.10899629, 0.40694186, 0.06030037, 0.012413437, -0.06108739});
595
596 lstm.SetCellGateBias({-0.024379363, 0.0055531194, 0.23377132, 0.033463873, -0.1483596,
597 -0.10639995, -0.091433935, 0.058573797, -0.06809782, -0.07889636,
598 -0.043246906, -0.09829136, -0.4279842, 0.034901652, 0.18797937,
599 0.0075234566, 0.016178843, 0.1749513, 0.13975595, 0.92058027});
600
601 lstm.SetOutputGateBias({0.046159424, -0.0012809046, 0.03563469, 0.12648113, 0.027195795,
602 0.35373217, -0.018957434, 0.008907322, -0.0762701, 0.12018895,
603 0.04216877, 0.0022856654, 0.040952638, 0.3147856, 0.08225149,
604 -0.057416286, -0.14995944, -0.008040261, 0.13208859, 0.029760877});
605
606 lstm.SetRecurrentToInputWeights(
607 {-0.001374326, -0.078856036, 0.10672688, 0.029162422, -0.11585556,
608 0.02557986, -0.13446963, -0.035785314, -0.01244275, 0.025961924,
609 -0.02337298, -0.044228926, -0.055839065, -0.046598054, -0.010546039,
610 -0.06900766, 0.027239809, 0.022582639, -0.013296484, -0.05459212,
611 0.08981, -0.045407712, 0.08682226, -0.06867011, -0.14390695,
612 -0.02916037, 0.000996957, 0.091420636, 0.14283475, -0.07390571,
613 -0.06402044, 0.062524505, -0.093129106, 0.04860203, -0.08364217,
614 -0.08119002, 0.009352075, 0.22920375, 0.0016303885, 0.11583097,
615 -0.13732095, 0.012405723, -0.07551853, 0.06343048, 0.12162708,
616 -0.031923793, -0.014335606, 0.01790974, -0.10650317, -0.0724401,
617 0.08554849, -0.05727212, 0.06556731, -0.042729504, -0.043227166,
618 0.011683251, -0.013082158, -0.029302018, -0.010899579, -0.062036745,
619 -0.022509435, -0.00964907, -0.01567329, 0.04260106, -0.07787477,
620 -0.11576462, 0.017356863, 0.048673786, -0.017577527, -0.05527947,
621 -0.082487635, -0.040137455, -0.10820036, -0.04666372, 0.022746278,
622 -0.07851417, 0.01068115, 0.032956902, 0.022433773, 0.0026891115,
623 0.08944216, -0.0685835, 0.010513544, 0.07228705, 0.02032331,
624 -0.059686817, -0.0005566496, -0.086984694, 0.040414046, -0.1380399,
625 0.094208956, -0.05722982, 0.012092817, -0.04989123, -0.086576,
626 -0.003399834, -0.04696032, -0.045747425, 0.10091314, 0.048676282,
627 -0.029037097, 0.031399418, -0.0040285117, 0.047237843, 0.09504992,
628 0.041799378, -0.049185462, -0.031518843, -0.10516937, 0.026374253,
629 0.10058866, -0.0033195973, -0.041975245, 0.0073591834, 0.0033782164,
630 -0.004325073, -0.10167381, 0.042500053, -0.01447153, 0.06464186,
631 -0.017142897, 0.03312627, 0.009205989, 0.024138335, -0.011337001,
632 0.035530265, -0.010912711, 0.0706555, -0.005894094, 0.051841937,
633 -0.1401738, -0.02351249, 0.0365468, 0.07590991, 0.08838724,
634 0.021681072, -0.10086113, 0.019608743, -0.06195883, 0.077335775,
635 0.023646897, -0.095322326, 0.02233014, 0.09756986, -0.048691444,
636 -0.009579111, 0.07595467, 0.11480546, -0.09801813, 0.019894179,
637 0.08502348, 0.004032281, 0.037211012, 0.068537936, -0.048005626,
638 -0.091520436, -0.028379958, -0.01556313, 0.06554592, -0.045599163,
639 -0.01672207, -0.020169014, -0.011877351, -0.20212261, 0.010889619,
640 0.0047078193, 0.038385306, 0.08540671, -0.017140968, -0.0035865551,
641 0.016678626, 0.005633034, 0.015963363, 0.00871737, 0.060130805,
642 0.028611384, 0.10109069, -0.015060172, -0.07894427, 0.06401885,
643 0.011584063, -0.024466386, 0.0047652307, -0.09041358, 0.030737216,
644 -0.0046374933, 0.14215417, -0.11823516, 0.019899689, 0.006106124,
645 -0.027092824, 0.0786356, 0.05052217, -0.058925, -0.011402121,
646 -0.024987547, -0.0013661642, -0.06832946, -0.015667673, -0.1083353,
647 -0.00096863037, -0.06988685, -0.053350925, -0.027275559, -0.033664223,
648 -0.07978348, -0.025200296, -0.017207067, -0.058403496, -0.055697463,
649 0.005798788, 0.12965427, -0.062582195, 0.0013350133, -0.10482091,
650 0.0379771, 0.072521195, -0.0029455067, -0.13797039, -0.03628521,
651 0.013806405, -0.017858358, -0.01008298, -0.07700066, -0.017081132,
652 0.019358726, 0.0027079724, 0.004635139, 0.062634714, -0.02338735,
653 -0.039547626, -0.02050681, 0.03385117, -0.083611414, 0.002862572,
654 -0.09421313, 0.058618143, -0.08598433, 0.00972939, 0.023867095,
655 -0.053934585, -0.023203006, 0.07452513, -0.048767887, -0.07314807,
656 -0.056307215, -0.10433547, -0.06440842, 0.04328182, 0.04389765,
657 -0.020006588, -0.09076438, -0.11652589, -0.021705797, 0.03345259,
658 -0.010329105, -0.025767034, 0.013057034, -0.07316461, -0.10145612,
659 0.06358255, 0.18531723, 0.07759293, 0.12006465, 0.1305557,
660 0.058638252, -0.03393652, 0.09622831, -0.16253184, -2.4580743e-06,
661 0.079869635, -0.070196845, -0.005644518, 0.06857898, -0.12598175,
662 -0.035084512, 0.03156317, -0.12794146, -0.031963028, 0.04692781,
663 0.030070418, 0.0071660685, -0.095516115, -0.004643372, 0.040170413,
664 -0.062104587, -0.0037324072, 0.0554317, 0.08184801, -0.019164372,
665 0.06791302, 0.034257166, -0.10307039, 0.021943003, 0.046745934,
666 0.0790918, -0.0265588, -0.007824208, 0.042546265, -0.00977924,
667 -0.0002440307, -0.017384544, -0.017990116, 0.12252321, -0.014512694,
668 -0.08251313, 0.08861942, 0.13589665, 0.026351685, 0.012641483,
669 0.07466548, 0.044301085, -0.045414884, -0.051112458, 0.03444247,
670 -0.08502782, -0.04106223, -0.028126027, 0.028473156, 0.10467447});
671
672 lstm.SetRecurrentToForgetWeights(
673 {-0.057784554, -0.026057621, -0.068447545, -0.022581743, 0.14811787,
674 0.10826372, 0.09471067, 0.03987225, -0.0039523416, 0.00030638507,
675 0.053185795, 0.10572994, 0.08414449, -0.022036452, -0.00066928595,
676 -0.09203576, 0.032950465, -0.10985798, -0.023809856, 0.0021431844,
677 -0.02196096, -0.00326074, 0.00058621005, -0.074678116, -0.06193199,
678 0.055729095, 0.03736828, 0.020123724, 0.061878487, -0.04729229,
679 0.034919553, -0.07585433, -0.04421272, -0.044019096, 0.085488975,
680 0.04058006, -0.06890133, -0.030951202, -0.024628663, -0.07672815,
681 0.034293607, 0.08556707, -0.05293577, -0.033561368, -0.04899627,
682 0.0241671, 0.015736353, -0.095442444, -0.029564252, 0.016493602,
683 -0.035026584, 0.022337519, -0.026871363, 0.004780428, 0.0077918363,
684 -0.03601621, 0.016435321, -0.03263031, -0.09543275, -0.047392778,
685 0.013454138, 0.028934088, 0.01685226, -0.086110644, -0.046250615,
686 -0.01847454, 0.047608484, 0.07339695, 0.034546845, -0.04881143,
687 0.009128804, -0.08802852, 0.03761666, 0.008096139, -0.014454086,
688 0.014361001, -0.023502491, -0.0011840804, -0.07607001, 0.001856849,
689 -0.06509276, -0.006021153, -0.08570962, -0.1451793, 0.060212336,
690 0.055259194, 0.06974018, 0.049454916, -0.027794661, -0.08077226,
691 -0.016179763, 0.1169753, 0.17213494, -0.0056326236, -0.053934924,
692 -0.0124349, -0.11520337, 0.05409887, 0.088759385, 0.0019655675,
693 0.0042065294, 0.03881498, 0.019844765, 0.041858196, -0.05695512,
694 0.047233116, 0.038937137, -0.06542224, 0.014429736, -0.09719407,
695 0.13908425, -0.05379757, 0.012321099, 0.082840554, -0.029899208,
696 0.044217527, 0.059855383, 0.07711018, -0.045319796, 0.0948846,
697 -0.011724666, -0.0033288454, -0.033542685, -0.04764985, -0.13873616,
698 0.040668588, 0.034832682, -0.015319203, -0.018715994, 0.046002675,
699 0.0599172, -0.043107376, 0.0294216, -0.002314414, -0.022424703,
700 0.0030315618, 0.0014641669, 0.0029166266, -0.11878115, 0.013738511,
701 0.12375372, -0.0006038222, 0.029104086, 0.087442465, 0.052958444,
702 0.07558703, 0.04817258, 0.044462286, -0.015213451, -0.08783778,
703 -0.0561384, -0.003008196, 0.047060397, -0.002058388, 0.03429439,
704 -0.018839769, 0.024734668, 0.024614193, -0.042046934, 0.09597743,
705 -0.0043254104, 0.04320769, 0.0064070094, -0.0019131786, -0.02558259,
706 -0.022822596, -0.023273505, -0.02464396, -0.10991725, -0.006240552,
707 0.0074488563, 0.024044557, 0.04383914, -0.046476185, 0.028658995,
708 0.060410924, 0.050786525, 0.009452605, -0.0073054377, -0.024810238,
709 0.0052906186, 0.0066939713, -0.0020913032, 0.014515517, 0.015898481,
710 0.021362653, -0.030262267, 0.016587038, -0.011442813, 0.041154444,
711 -0.007631438, -0.03423484, -0.010977775, 0.036152758, 0.0066366293,
712 0.11915515, 0.02318443, -0.041350313, 0.021485701, -0.10906167,
713 -0.028218046, -0.00954771, 0.020531068, -0.11995105, -0.03672871,
714 0.024019798, 0.014255957, -0.05221243, -0.00661567, -0.04630967,
715 0.033188973, 0.10107534, -0.014027541, 0.030796422, -0.10270911,
716 -0.035999842, 0.15443139, 0.07684145, 0.036571592, -0.035900835,
717 -0.0034699554, 0.06209149, 0.015920248, -0.031122351, -0.03858649,
718 0.01849943, 0.13872518, 0.01503974, 0.069941424, -0.06948533,
719 -0.0088794185, 0.061282158, -0.047401894, 0.03100163, -0.041533746,
720 -0.10430945, 0.044574402, -0.01425562, -0.024290353, 0.034563623,
721 0.05866852, 0.023947537, -0.09445152, 0.035450947, 0.02247216,
722 -0.0042998926, 0.061146557, -0.10250651, 0.020881841, -0.06747029,
723 0.10062043, -0.0023941975, 0.03532124, -0.016341697, 0.09685456,
724 -0.016764693, 0.051808182, 0.05875331, -0.04536488, 0.001626336,
725 -0.028892258, -0.01048663, -0.009793449, -0.017093895, 0.010987891,
726 0.02357273, -0.00010856845, 0.0099760275, -0.001845119, -0.03551521,
727 0.0018358806, 0.05763657, -0.01769146, 0.040995963, 0.02235177,
728 -0.060430344, 0.11475477, -0.023854522, 0.10071741, 0.0686208,
729 -0.014250481, 0.034261297, 0.047418304, 0.08562733, -0.030519066,
730 0.0060542435, 0.014653856, -0.038836084, 0.04096551, 0.032249358,
731 -0.08355519, -0.026823482, 0.056386515, -0.010401743, -0.028396193,
732 0.08507674, 0.014410365, 0.020995233, 0.17040324, 0.11511526,
733 0.02459721, 0.0066619175, 0.025853224, -0.023133837, -0.081302024,
734 0.017264642, -0.009585969, 0.09491168, -0.051313367, 0.054532815,
735 -0.014298593, 0.10657464, 0.007076659, 0.10964551, 0.0409152,
736 0.008275321, -0.07283536, 0.07937492, 0.04192024, -0.1075027});
737
738 lstm.SetRecurrentToCellWeights(
739 {-0.037322544, 0.018592842, 0.0056175636, -0.06253426, 0.055647098,
740 -0.05713207, -0.05626563, 0.005559383, 0.03375411, -0.025757805,
741 -0.088049285, 0.06017052, -0.06570978, 0.007384076, 0.035123326,
742 -0.07920549, 0.053676967, 0.044480428, -0.07663568, 0.0071805613,
743 0.08089997, 0.05143358, 0.038261272, 0.03339287, -0.027673481,
744 0.044746667, 0.028349208, 0.020090483, -0.019443132, -0.030755889,
745 -0.0040000007, 0.04465846, -0.021585021, 0.0031670958, 0.0053199246,
746 -0.056117613, -0.10893326, 0.076739706, -0.08509834, -0.027997585,
747 0.037871376, 0.01449768, -0.09002357, -0.06111149, -0.046195522,
748 0.0422062, -0.005683705, -0.1253618, -0.012925729, -0.04890792,
749 0.06985068, 0.037654128, 0.03398274, -0.004781977, 0.007032333,
750 -0.031787455, 0.010868644, -0.031489216, 0.09525667, 0.013939797,
751 0.0058680447, 0.0167067, 0.02668468, -0.04797466, -0.048885044,
752 -0.12722108, 0.035304096, 0.06554885, 0.00972396, -0.039238118,
753 -0.05159735, -0.11329045, 0.1613692, -0.03750952, 0.06529313,
754 -0.071974665, -0.11769596, 0.015524369, -0.0013754242, -0.12446318,
755 0.02786344, -0.014179351, 0.005264273, 0.14376344, 0.015983658,
756 0.03406988, -0.06939408, 0.040699873, 0.02111075, 0.09669095,
757 0.041345075, -0.08316494, -0.07684199, -0.045768797, 0.032298047,
758 -0.041805092, 0.0119405, 0.0061010392, 0.12652606, 0.0064572375,
759 -0.024950314, 0.11574242, 0.04508852, -0.04335324, 0.06760663,
760 -0.027437469, 0.07216407, 0.06977076, -0.05438599, 0.034033038,
761 -0.028602652, 0.05346137, 0.043184172, -0.037189785, 0.10420091,
762 0.00882477, -0.054019816, -0.074273005, -0.030617684, -0.0028467078,
763 0.024302477, -0.0038869337, 0.005332455, 0.0013399826, 0.04361412,
764 -0.007001822, 0.09631092, -0.06702025, -0.042049985, -0.035070654,
765 -0.04103342, -0.10273396, 0.0544271, 0.037184782, -0.13150354,
766 -0.0058036847, -0.008264958, 0.042035464, 0.05891794, 0.029673764,
767 0.0063542654, 0.044788733, 0.054816857, 0.062257513, -0.00093483756,
768 0.048938446, -0.004952862, -0.007730018, -0.04043371, -0.017094059,
769 0.07229206, -0.023670016, -0.052195564, -0.025616996, -0.01520939,
770 0.045104615, -0.007376126, 0.003533447, 0.006570588, 0.056037236,
771 0.12436656, 0.051817212, 0.028532185, -0.08686856, 0.11868599,
772 0.07663395, -0.07323171, 0.03463402, -0.050708205, -0.04458982,
773 -0.11590894, 0.021273347, 0.1251325, -0.15313013, -0.12224372,
774 0.17228661, 0.023029093, 0.086124025, 0.006445803, -0.03496501,
775 0.028332196, 0.04449512, -0.042436164, -0.026587414, -0.006041347,
776 -0.09292539, -0.05678812, 0.03897832, 0.09465633, 0.008115513,
777 -0.02171956, 0.08304309, 0.071401566, 0.019622514, 0.032163795,
778 -0.004167056, 0.02295182, 0.030739572, 0.056506045, 0.004612461,
779 0.06524936, 0.059999723, 0.046395954, -0.0045512207, -0.1335546,
780 -0.030136576, 0.11584653, -0.014678886, 0.0020118146, -0.09688814,
781 -0.0790206, 0.039770417, -0.0329582, 0.07922767, 0.029322514,
782 0.026405897, 0.04207835, -0.07073373, 0.063781224, 0.0859677,
783 -0.10925287, -0.07011058, 0.048005477, 0.03438226, -0.09606514,
784 -0.006669445, -0.043381985, 0.04240257, -0.06955775, -0.06769346,
785 0.043903265, -0.026784198, -0.017840602, 0.024307009, -0.040079936,
786 -0.019946516, 0.045318738, -0.12233574, 0.026170589, 0.0074471775,
787 0.15978073, 0.10185836, 0.10298046, -0.015476589, -0.039390966,
788 -0.072174534, 0.0739445, -0.1211869, -0.0347889, -0.07943156,
789 0.014809798, -0.12412325, -0.0030663363, 0.039695457, 0.0647603,
790 -0.08291318, -0.018529687, -0.004423833, 0.0037507233, 0.084633216,
791 -0.01514876, -0.056505352, -0.012800942, -0.06994386, 0.012962922,
792 -0.031234352, 0.07029052, 0.016418684, 0.03618972, 0.055686004,
793 -0.08663945, -0.017404709, -0.054761406, 0.029065743, 0.052404847,
794 0.020238016, 0.0048197987, -0.0214882, 0.07078733, 0.013016777,
795 0.06262858, 0.009184685, 0.020785125, -0.043904778, -0.0270329,
796 -0.03299152, -0.060088247, -0.015162964, -0.001828936, 0.12642565,
797 -0.056757294, 0.013586685, 0.09232601, -0.035886683, 0.06000002,
798 0.05229691, -0.052580316, -0.082029596, -0.010794592, 0.012947712,
799 -0.036429964, -0.085508935, -0.13127148, -0.017744139, 0.031502828,
800 0.036232427, -0.031581745, 0.023051167, -0.05325106, -0.03421577,
801 0.028793324, -0.034633752, -0.009881397, -0.043551125, -0.018609839,
802 0.0019097115, -0.008799762, 0.056595087, 0.0022273948, 0.055752404});
803
804 lstm.SetRecurrentToOutputWeights({
805 0.025825322, -0.05813119, 0.09495884, -0.045984812, -0.01255415,
806 -0.0026479573, -0.08196161, -0.054914974, -0.0046604523, -0.029587349,
807 -0.044576716, -0.07480124, -0.082868785, 0.023254942, 0.027502948,
808 -0.0039728214, -0.08683098, -0.08116779, -0.014675607, -0.037924774,
809 -0.023314456, -0.007401714, -0.09255757, 0.029460307, -0.08829125,
810 -0.005139627, -0.08989442, -0.0555066, 0.13596267, -0.025062224,
811 -0.048351806, -0.03850004, 0.07266485, -0.022414139, 0.05940088,
812 0.075114764, 0.09597592, -0.010211725, -0.0049794707, -0.011523867,
813 -0.025980417, 0.072999895, 0.11091378, -0.081685916, 0.014416728,
814 0.043229222, 0.034178585, -0.07530371, 0.035837382, -0.085607,
815 -0.007721233, -0.03287832, -0.043848954, -0.06404588, -0.06632928,
816 -0.073643476, 0.008214239, -0.045984086, 0.039764922, 0.03474462,
817 0.060612556, -0.080590084, 0.049127717, 0.04151091, -0.030063879,
818 0.008801774, -0.023021035, -0.019558564, 0.05158114, -0.010947698,
819 -0.011825728, 0.0075720972, 0.0699727, -0.0039981045, 0.069350146,
820 0.08799282, 0.016156472, 0.035502106, 0.11695009, 0.006217345,
821 0.13392477, -0.037875112, 0.025745004, 0.08940699, -0.00924166,
822 0.0046702605, -0.036598757, -0.08811812, 0.10522024, -0.032441203,
823 0.008176899, -0.04454919, 0.07058152, 0.0067963637, 0.039206743,
824 0.03259838, 0.03725492, -0.09515802, 0.013326398, -0.052055415,
825 -0.025676316, 0.03198509, -0.015951829, -0.058556724, 0.036879618,
826 0.043357447, 0.028362012, -0.05908629, 0.0059240665, -0.04995891,
827 -0.019187413, 0.0276265, -0.01628143, 0.0025863599, 0.08800015,
828 0.035250366, -0.022165963, -0.07328642, -0.009415526, -0.07455109,
829 0.11690406, 0.0363299, 0.07411125, 0.042103454, -0.009660886,
830 0.019076364, 0.018299393, -0.046004917, 0.08891175, 0.0431396,
831 -0.026327137, -0.051502608, 0.08979574, -0.051670972, 0.04940282,
832 -0.07491107, -0.021240504, 0.022596184, -0.034280192, 0.060163025,
833 -0.058211457, -0.051837247, -0.01349775, -0.04639988, -0.035936575,
834 -0.011681591, 0.064818054, 0.0073146066, -0.021745546, -0.043124277,
835 -0.06471268, -0.07053354, -0.029321948, -0.05330136, 0.016933719,
836 -0.053782392, 0.13747959, -0.1361751, -0.11569455, 0.0033329215,
837 0.05693899, -0.053219706, 0.063698, 0.07977434, -0.07924483,
838 0.06936997, 0.0034815092, -0.007305279, -0.037325785, -0.07251102,
839 -0.033633437, -0.08677009, 0.091591336, -0.14165086, 0.021752775,
840 0.019683983, 0.0011612234, -0.058154266, 0.049996935, 0.0288841,
841 -0.0024567875, -0.14345716, 0.010955264, -0.10234828, 0.1183656,
842 -0.0010731248, -0.023590032, -0.072285876, -0.0724771, -0.026382286,
843 -0.0014920527, 0.042667855, 0.0018776858, 0.02986552, 0.009814309,
844 0.0733756, 0.12289186, 0.018043943, -0.0458958, 0.049412545,
845 0.033632483, 0.05495232, 0.036686596, -0.013781798, -0.010036754,
846 0.02576849, -0.08307328, 0.010112348, 0.042521734, -0.05869831,
847 -0.071689695, 0.03876447, -0.13275425, -0.0352966, -0.023077697,
848 0.10285965, 0.084736146, 0.15568255, -0.00040734606, 0.027835453,
849 -0.10292561, -0.032401145, 0.10053256, -0.026142767, -0.08271222,
850 -0.0030240538, -0.016368777, 0.1070414, 0.042672627, 0.013456989,
851 -0.0437609, -0.022309763, 0.11576483, 0.04108048, 0.061026827,
852 -0.0190714, -0.0869359, 0.037901703, 0.0610107, 0.07202949,
853 0.01675338, 0.086139716, -0.08795751, -0.014898893, -0.023771819,
854 -0.01965048, 0.007955471, -0.043740474, 0.03346837, -0.10549954,
855 0.090567775, 0.042013682, -0.03176985, 0.12569028, -0.02421228,
856 -0.029526481, 0.023851605, 0.031539805, 0.05292009, -0.02344001,
857 -0.07811758, -0.08834428, 0.10094801, 0.16594367, -0.06861939,
858 -0.021256343, -0.041093912, -0.06669611, 0.035498552, 0.021757556,
859 -0.09302526, -0.015403468, -0.06614931, -0.051798206, -0.013874718,
860 0.03630673, 0.010412845, -0.08077351, 0.046185967, 0.0035662893,
861 0.03541868, -0.094149634, -0.034814864, 0.003128424, -0.020674974,
862 -0.03944324, -0.008110165, -0.11113267, 0.08484226, 0.043586485,
863 0.040582247, 0.0968012, -0.065249965, -0.028036479, 0.0050708856,
864 0.0017462453, 0.0326779, 0.041296225, 0.09164146, -0.047743853,
865 -0.015952192, -0.034451712, 0.084197424, -0.05347844, -0.11768019,
866 0.085926116, -0.08251791, -0.045081906, 0.0948852, 0.068401024,
867 0.024856757, 0.06978981, -0.057309967, -0.012775832, -0.0032452994,
868 0.01977615, -0.041040014, -0.024264973, 0.063464895, 0.05431621,
869 });
870
871 lstm.SetCellToInputWeights({0.040369894, 0.030746894, 0.24704495, 0.018586371, -0.037586458,
872 -0.15312155, -0.11812848, -0.11465643, 0.20259799, 0.11418174,
873 -0.10116027, -0.011334949, 0.12411352, -0.076769054, -0.052169047,
874 0.21198851, -0.38871562, -0.09061183, -0.09683246, -0.21929175});
875
876 lstm.SetCellToForgetWeights({-0.01998659, -0.15568835, -0.24248174, -0.012770197,
877 0.041331276, -0.072311886, -0.052123554, -0.0066330447,
878 -0.043891653, 0.036225766, -0.047248036, 0.021479502,
879 0.033189066, 0.11952997, -0.020432774, 0.64658105,
880 -0.06650122, -0.03467612, 0.095340036, 0.23647355});
881
882 lstm.SetCellToOutputWeights({0.08286371, -0.08261836, -0.51210177, 0.002913762, 0.17764764,
883 -0.5495371, -0.08460716, -0.24552552, 0.030037103, 0.04123544,
884 -0.11940523, 0.007358328, 0.1890978, 0.4833202, -0.34441817,
885 0.36312827, -0.26375428, 0.1457655, -0.19724406, 0.15548733});
886
887 lstm.SetProjectionWeights(
888 {-0.009802181, 0.09401916, 0.0717386, -0.13895074, 0.09641832, 0.060420845,
889 0.08539281, 0.054285463, 0.061395317, 0.034448683, -0.042991187, 0.019801661,
890 -0.16840284, -0.015726732, -0.23041931, -0.024478018, -0.10959692, -0.013875541,
891 0.18600968, -0.061274476, 0.0138165, -0.08160894, -0.07661644, 0.032372914,
892 0.16169067, 0.22465782, -0.03993472, -0.004017731, 0.08633481, -0.28869787,
893 0.08682067, 0.17240396, 0.014975425, 0.056431185, 0.031037588, 0.16702051,
894 0.0077946745, 0.15140012, 0.29405436, 0.120285, -0.188994, -0.027265169,
895 0.043389652, -0.022061434, 0.014777949, -0.20203483, 0.094781205, 0.19100232,
896 0.13987629, -0.036132768, -0.06426278, -0.05108664, 0.13221376, 0.009441198,
897 -0.16715929, 0.15859416, -0.040437475, 0.050779544, -0.022187516, 0.012166504,
898 0.027685808, -0.07675938, -0.0055694645, -0.09444123, 0.0046453946, 0.050794356,
899 0.10770313, -0.20790008, -0.07149004, -0.11425117, 0.008225835, -0.035802525,
900 0.14374903, 0.15262283, 0.048710253, 0.1847461, -0.007487823, 0.11000021,
901 -0.09542012, 0.22619456, -0.029149994, 0.08527916, 0.009043713, 0.0042746216,
902 0.016261552, 0.022461696, 0.12689082, -0.043589946, -0.12035478, -0.08361797,
903 -0.050666027, -0.1248618, -0.1275799, -0.071875185, 0.07377272, 0.09944291,
904 -0.18897448, -0.1593054, -0.06526116, -0.040107165, -0.004618631, -0.067624845,
905 -0.007576253, 0.10727444, 0.041546922, -0.20424393, 0.06907816, 0.050412357,
906 0.00724631, 0.039827548, 0.12449835, 0.10747581, 0.13708383, 0.09134148,
907 -0.12617786, -0.06428341, 0.09956831, 0.1208086, -0.14676677, -0.0727722,
908 0.1126304, 0.010139365, 0.015571211, -0.038128063, 0.022913318, -0.042050496,
909 0.16842307, -0.060597885, 0.10531834, -0.06411776, -0.07451711, -0.03410368,
910 -0.13393489, 0.06534304, 0.003620307, 0.04490757, 0.05970546, 0.05197996,
911 0.02839995, 0.10434969, -0.013699693, -0.028353551, -0.07260381, 0.047201227,
912 -0.024575593, -0.036445823, 0.07155557, 0.009672501, -0.02328883, 0.009533515,
913 -0.03606021, -0.07421458, -0.028082801, -0.2678904, -0.13221288, 0.18419984,
914 -0.13012612, -0.014588381, -0.035059117, -0.04824723, 0.07830115, -0.056184657,
915 0.03277091, 0.025466874, 0.14494097, -0.12522776, -0.098633975, -0.10766018,
916 -0.08317623, 0.08594209, 0.07749552, 0.039474737, 0.1776665, -0.07409566,
917 -0.0477268, 0.29323658, 0.10801441, 0.1154011, 0.013952499, 0.10739139,
918 0.10708251, -0.051456142, 0.0074137426, -0.10430189, 0.10034707, 0.045594677,
919 0.0635285, -0.0715442, -0.089667566, -0.10811871, 0.00026344223, 0.08298446,
920 -0.009525053, 0.006585689, -0.24567553, -0.09450807, 0.09648481, 0.026996298,
921 -0.06419476, -0.04752702, -0.11063944, -0.23441927, -0.17608605, -0.052156363,
922 0.067035615, 0.19271925, -0.0032889997, -0.043264326, 0.09663576, -0.057112187,
923 -0.10100678, 0.0628376, 0.04447668, 0.017961001, -0.10094388, -0.10190601,
924 0.18335468, 0.10494553, -0.052095775, -0.0026118709, 0.10539724, -0.04383912,
925 -0.042349473, 0.08438151, -0.1947263, 0.02251204, 0.11216432, -0.10307853,
926 0.17351969, -0.039091777, 0.08066188, -0.00561982, 0.12633002, 0.11335965,
927 -0.0088127935, -0.019777594, 0.06864014, -0.059751723, 0.016233567, -0.06894641,
928 -0.28651384, -0.004228674, 0.019708522, -0.16305895, -0.07468996, -0.0855457,
929 0.099339016, -0.07580735, -0.13775392, 0.08434318, 0.08330512, -0.12131499,
930 0.031935584, 0.09180414, -0.08876437, -0.08049874, 0.008753825, 0.03498998,
931 0.030215185, 0.03907079, 0.089751154, 0.029194152, -0.03337423, -0.019092513,
932 0.04331237, 0.04299654, -0.036394123, -0.12915532, 0.09793732, 0.07512415,
933 -0.11319543, -0.032502122, 0.15661901, 0.07671967, -0.005491124, -0.19379048,
934 -0.218606, 0.21448623, 0.017840758, 0.1416943, -0.07051762, 0.19488361,
935 0.02664691, -0.18104725, -0.09334311, 0.15026465, -0.15493552, -0.057762887,
936 -0.11604192, -0.262013, -0.01391798, 0.012185008, 0.11156489, -0.07483202,
937 0.06693364, -0.26151478, 0.046425626, 0.036540434, -0.16435726, 0.17338543,
938 -0.21401681, -0.11385144, -0.08283257, -0.069031075, 0.030635102, 0.010969227,
939 0.11109743, 0.010919218, 0.027526086, 0.13519906, 0.01891392, -0.046839405,
940 -0.040167913, 0.017953383, -0.09700955, 0.0061885654, -0.07000971, 0.026893595,
941 -0.038844477, 0.14543656});
942
943 static float lstm_input[][20] = {
944 {// Batch0: 4 (input_sequence_size) * 5 (n_input)
945 0.787926, 0.151646, 0.071352, 0.118426, 0.458058, 0.596268, 0.998386,
946 0.568695, 0.864524, 0.571277, 0.073204, 0.296072, 0.743333, 0.069199,
947 0.045348, 0.867394, 0.291279, 0.013714, 0.482521, 0.626339},
948
949 {// Batch1: 4 (input_sequence_size) * 5 (n_input)
950 0.295743, 0.544053, 0.690064, 0.858138, 0.497181, 0.642421, 0.524260,
951 0.134799, 0.003639, 0.162482, 0.640394, 0.930399, 0.050782, 0.432485,
952 0.988078, 0.082922, 0.563329, 0.865614, 0.333232, 0.259916}};
953
954 static float lstm_golden_output[][64] = {
955 {// Batch0: 4 (input_sequence_size) * 16 (n_output)
956 -0.00396806, 0.029352, -0.00279226, 0.0159977, -0.00835576, -0.0211779,
957 0.0283512, -0.0114597, 0.00907307, -0.0244004, -0.0152191, -0.0259063,
958 0.00914318, 0.00415118, 0.017147, 0.0134203, -0.0166936, 0.0381209,
959 0.000889694, 0.0143363, -0.0328911, -0.0234288, 0.0333051, -0.012229,
960 0.0110322, -0.0457725, -0.000832209, -0.0202817, 0.0327257, 0.0121308,
961 0.0155969, 0.0312091, -0.0213783, 0.0350169, 0.000324794, 0.0276012,
962 -0.0263374, -0.0371449, 0.0446149, -0.0205474, 0.0103729, -0.0576349,
963 -0.0150052, -0.0292043, 0.0376827, 0.0136115, 0.0243435, 0.0354492,
964 -0.0189322, 0.0464512, -0.00251373, 0.0225745, -0.0308346, -0.0317124,
965 0.0460407, -0.0189395, 0.0149363, -0.0530162, -0.0150767, -0.0340193,
966 0.0286833, 0.00824207, 0.0264887, 0.0305169},
967 {// Batch1: 4 (input_sequence_size) * 16 (n_output)
968 -0.013869, 0.0287268, -0.00334693, 0.00733398, -0.0287926, -0.0186926,
969 0.0193662, -0.0115437, 0.00422612, -0.0345232, 0.00223253, -0.00957321,
970 0.0210624, 0.013331, 0.0150954, 0.02168, -0.0141913, 0.0322082,
971 0.00227024, 0.0260507, -0.0188721, -0.0296489, 0.0399134, -0.0160509,
972 0.0116039, -0.0447318, -0.0150515, -0.0277406, 0.0316596, 0.0118233,
973 0.0214762, 0.0293641, -0.0204549, 0.0450315, -0.00117378, 0.0167673,
974 -0.0375007, -0.0238314, 0.038784, -0.0174034, 0.0131743, -0.0506589,
975 -0.0048447, -0.0240239, 0.0325789, 0.00790065, 0.0220157, 0.0333314,
976 -0.0264787, 0.0387855, -0.000764675, 0.0217599, -0.037537, -0.0335206,
977 0.0431679, -0.0211424, 0.010203, -0.062785, -0.00832363, -0.025181,
978 0.0412031, 0.0118723, 0.0239643, 0.0394009}};
979
980 // Resetting cell_state and output_state
981 lstm.ResetCellState();
982 lstm.ResetOutputState();
983
984 const int input_sequence_size = sizeof(lstm_input[0]) / sizeof(float) / (lstm.num_inputs());
985 for (int i = 0; i < input_sequence_size; i++) {
986 float* batch0_start = lstm_input[0] + i * lstm.num_inputs();
987 float* batch0_end = batch0_start + lstm.num_inputs();
988
989 lstm.SetInput(0, batch0_start, batch0_end);
990
991 float* batch1_start = lstm_input[1] + i * lstm.num_inputs();
992 float* batch1_end = batch1_start + lstm.num_inputs();
993 lstm.SetInput(lstm.num_inputs(), batch1_start, batch1_end);
994
995 lstm.Invoke();
996
997 float* golden_start_batch0 = lstm_golden_output[0] + i * lstm.num_outputs();
998 float* golden_end_batch0 = golden_start_batch0 + lstm.num_outputs();
999 float* golden_start_batch1 = lstm_golden_output[1] + i * lstm.num_outputs();
1000 float* golden_end_batch1 = golden_start_batch1 + lstm.num_outputs();
1001 std::vector<float> expected;
1002 expected.insert(expected.end(), golden_start_batch0, golden_end_batch0);
1003 expected.insert(expected.end(), golden_start_batch1, golden_end_batch1);
1004 EXPECT_THAT(lstm.GetOutput(), ElementsAreArray(ArrayFloatNear(expected)));
1005 }
1006 }
1007
1008 } // namespace wrapper
1009 } // namespace nn
1010 } // namespace android
1011