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 "LSTM.h" 18 19#include "NeuralNetworksWrapper.h" 20#include "gmock/gmock-matchers.h" 21#include "gtest/gtest.h" 22 23namespace android { 24namespace nn { 25namespace wrapper { 26 27using ::testing::Each; 28using ::testing::FloatNear; 29using ::testing::Matcher; 30 31namespace { 32 33std::vector<Matcher<float>> ArrayFloatNear(const std::vector<float>& values, 34 float max_abs_error=1.e-6) { 35 std::vector<Matcher<float>> matchers; 36 matchers.reserve(values.size()); 37 for (const float& v : values) { 38 matchers.emplace_back(FloatNear(v, max_abs_error)); 39 } 40 return matchers; 41} 42 43} // anonymous namespace 44 45#define FOR_ALL_INPUT_AND_WEIGHT_TENSORS(ACTION) \ 46 ACTION(Input) \ 47 ACTION(InputToInputWeights) \ 48 ACTION(InputToCellWeights) \ 49 ACTION(InputToForgetWeights) \ 50 ACTION(InputToOutputWeights) \ 51 ACTION(RecurrentToInputWeights) \ 52 ACTION(RecurrentToCellWeights) \ 53 ACTION(RecurrentToForgetWeights) \ 54 ACTION(RecurrentToOutputWeights) \ 55 ACTION(CellToInputWeights) \ 56 ACTION(CellToForgetWeights) \ 57 ACTION(CellToOutputWeights) \ 58 ACTION(InputGateBias) \ 59 ACTION(CellGateBias) \ 60 ACTION(ForgetGateBias) \ 61 ACTION(OutputGateBias) \ 62 ACTION(ProjectionWeights) \ 63 ACTION(ProjectionBias) \ 64 ACTION(OutputStateIn) \ 65 ACTION(CellStateIn) 66 67// For all output and intermediate states 68#define FOR_ALL_OUTPUT_TENSORS(ACTION) \ 69 ACTION(ScratchBuffer) \ 70 ACTION(OutputStateOut) \ 71 ACTION(CellStateOut) \ 72 ACTION(Output) \ 73 74class LSTMOpModel { 75public: 76 LSTMOpModel(uint32_t n_batch, uint32_t n_input, 77 uint32_t n_cell, uint32_t n_output, bool use_cifg, 78 bool use_peephole, bool use_projection_weights, 79 bool use_projection_bias, float cell_clip, float proj_clip, 80 const std::vector<std::vector<uint32_t>>& input_shapes0) 81 : n_batch_(n_batch), n_input_(n_input), 82 n_cell_(n_cell), n_output_(n_output), 83 use_cifg_(use_cifg), use_peephole_(use_peephole), 84 use_projection_weights_(use_projection_weights), 85 use_projection_bias_(use_projection_bias), 86 activation_(ActivationFn::kActivationTanh), 87 cell_clip_(cell_clip), proj_clip_(proj_clip) { 88 std::vector<uint32_t> inputs; 89 std::vector<std::vector<uint32_t>> input_shapes(input_shapes0); 90 91 input_shapes.push_back({n_batch, n_output}); 92 input_shapes.push_back({n_batch, n_cell}); 93 auto it = input_shapes.begin(); 94 95 // Input and weights 96#define AddInput(X) \ 97 OperandType X##OpndTy(Type::TENSOR_FLOAT32, *it++); \ 98 inputs.push_back(model_.addOperand(&X##OpndTy)); 99 100 FOR_ALL_INPUT_AND_WEIGHT_TENSORS(AddInput); 101 102#undef AddOperand 103 104 // Parameters 105 OperandType ActivationOpndTy(Type::INT32, {}); 106 inputs.push_back(model_.addOperand(&ActivationOpndTy)); 107 OperandType CellClipOpndTy(Type::FLOAT32, {}); 108 inputs.push_back(model_.addOperand(&CellClipOpndTy)); 109 OperandType ProjClipOpndTy(Type::FLOAT32, {}); 110 inputs.push_back(model_.addOperand(&ProjClipOpndTy)); 111 112 // Output and other intermediate state 113 std::vector<std::vector<uint32_t>> output_shapes{ 114 {n_batch, n_cell * (use_cifg ? 3 : 4)}, 115 {n_batch, n_output}, 116 {n_batch, n_cell}, 117 {n_batch, n_output}, 118 }; 119 std::vector<uint32_t> outputs; 120 121 auto it2 = output_shapes.begin(); 122 123#define AddOutput(X)\ 124 OperandType X##OpndTy(Type::TENSOR_FLOAT32, *it2++); \ 125 outputs.push_back(model_.addOperand(&X##OpndTy)); 126 127 FOR_ALL_OUTPUT_TENSORS(AddOutput); 128 129#undef AddOutput 130 131 model_.addOperation(ANEURALNETWORKS_LSTM, inputs, outputs); 132 model_.identifyInputsAndOutputs(inputs, outputs); 133 134 Input_.insert(Input_.end(), n_batch * n_input, 0.f); 135 OutputStateIn_.insert(OutputStateIn_.end(), n_batch * n_output, 0.f); 136 CellStateIn_.insert(CellStateIn_.end(), n_batch * n_cell, 0.f); 137 138 auto multiAll = [](const std::vector<uint32_t> &dims) -> uint32_t { 139 uint32_t sz = 1; 140 for(uint32_t d:dims) { sz *= d; } 141 return sz; 142 }; 143 144 it2 = output_shapes.begin(); 145 146#define ReserveOutput(X) X##_.insert(X##_.end(), multiAll(*it2++), 0.f); 147 148 FOR_ALL_OUTPUT_TENSORS(ReserveOutput); 149 150#undef ReserveOutput 151 152 model_.finish(); 153 } 154 155#define DefineSetter(X) \ 156 void Set##X(const std::vector<float> &f) { \ 157 X##_.insert(X##_.end(), f.begin(), f.end()); \ 158 } 159 160 FOR_ALL_INPUT_AND_WEIGHT_TENSORS(DefineSetter); 161 162#undef DefineSetter 163 164 void ResetOutputState() { 165 std::fill(OutputStateIn_.begin(), OutputStateIn_.end(), 0.f); 166 std::fill(OutputStateOut_.begin(), OutputStateOut_.end(), 0.f); 167 } 168 169 void ResetCellState() { 170 std::fill(CellStateIn_.begin(), CellStateIn_.end(), 0.f); 171 std::fill(CellStateOut_.begin(), CellStateOut_.end(), 0.f); 172 } 173 174 void SetInput(int offset, float *begin, float *end) { 175 for (;begin != end; begin++, offset++) { 176 Input_[offset] = *begin; 177 } 178 } 179 180 uint32_t num_inputs() const { return n_input_; } 181 uint32_t num_outputs() const { return n_output_; } 182 183 const std::vector<float> &GetOutput() const { return Output_; } 184 185 void Invoke() { 186 ASSERT_TRUE(model_.isValid()); 187 188 OutputStateIn_.swap(OutputStateOut_); 189 CellStateIn_.swap(CellStateOut_); 190 191 Compilation compilation(&model_); 192 compilation.finish(); 193 Execution execution(&compilation); 194#define SetInputOrWeight(X) \ 195 ASSERT_EQ(execution.setInput(LSTMCell::k##X##Tensor, X##_.data(), \ 196 sizeof(float)*X##_.size()), \ 197 Result::NO_ERROR); 198 199 FOR_ALL_INPUT_AND_WEIGHT_TENSORS(SetInputOrWeight); 200 201#undef SetInputOrWeight 202 203#define SetOutput(X) \ 204 ASSERT_EQ(execution.setOutput(LSTMCell::k##X##Tensor, X##_.data(), \ 205 sizeof(float)*X##_.size()), \ 206 Result::NO_ERROR); 207 208 FOR_ALL_OUTPUT_TENSORS(SetOutput); 209 210#undef SetOutput 211 212 if (use_cifg_) { 213 execution.setInput(LSTMCell::kInputToInputWeightsTensor, nullptr, 0); 214 execution.setInput(LSTMCell::kRecurrentToInputWeightsTensor, nullptr, 0); 215 } 216 217 if (use_peephole_) { 218 if (use_cifg_) { 219 execution.setInput(LSTMCell::kCellToInputWeightsTensor, nullptr, 0); 220 } 221 } else { 222 execution.setInput(LSTMCell::kCellToInputWeightsTensor, nullptr, 0); 223 execution.setInput(LSTMCell::kCellToForgetWeightsTensor, nullptr, 0); 224 execution.setInput(LSTMCell::kCellToOutputWeightsTensor, nullptr, 0); 225 } 226 227 if (use_projection_weights_) { 228 if (!use_projection_bias_) { 229 execution.setInput(LSTMCell::kProjectionBiasTensor, nullptr, 0); 230 } 231 } else { 232 execution.setInput(LSTMCell::kProjectionWeightsTensor, nullptr, 0); 233 execution.setInput(LSTMCell::kProjectionBiasTensor, nullptr, 0); 234 } 235 236 ASSERT_EQ(execution.setInput(LSTMCell::kActivationParam, 237 &activation_, sizeof(activation_)), 238 Result::NO_ERROR); 239 ASSERT_EQ(execution.setInput(LSTMCell::kCellClipParam, 240 &cell_clip_, sizeof(cell_clip_)), 241 Result::NO_ERROR); 242 ASSERT_EQ(execution.setInput(LSTMCell::kProjClipParam, 243 &proj_clip_, sizeof(proj_clip_)), 244 Result::NO_ERROR); 245 246 ASSERT_EQ(execution.compute(), Result::NO_ERROR); 247 } 248 249private: 250 Model model_; 251 // Execution execution_; 252 const uint32_t n_batch_; 253 const uint32_t n_input_; 254 const uint32_t n_cell_; 255 const uint32_t n_output_; 256 257 const bool use_cifg_; 258 const bool use_peephole_; 259 const bool use_projection_weights_; 260 const bool use_projection_bias_; 261 262 const int activation_; 263 const float cell_clip_; 264 const float proj_clip_; 265 266#define DefineTensor(X) \ 267 std::vector<float> X##_; 268 269 FOR_ALL_INPUT_AND_WEIGHT_TENSORS(DefineTensor); 270 FOR_ALL_OUTPUT_TENSORS(DefineTensor); 271 272#undef DefineTensor 273}; 274 275TEST(LSTMOpTest, BlackBoxTestNoCifgNoPeepholeNoProjectionNoClipping) { 276 const int n_batch = 1; 277 const int n_input = 2; 278 // n_cell and n_output have the same size when there is no projection. 279 const int n_cell = 4; 280 const int n_output = 4; 281 282 LSTMOpModel lstm(n_batch, n_input, n_cell, n_output, 283 /*use_cifg=*/false, /*use_peephole=*/false, 284 /*use_projection_weights=*/false, 285 /*use_projection_bias=*/false, 286 /*cell_clip=*/0.0, /*proj_clip=*/0.0, 287 { 288 {n_batch, n_input}, // input tensor 289 290 {n_cell, n_input}, // input_to_input_weight tensor 291 {n_cell, n_input}, // input_to_forget_weight tensor 292 {n_cell, n_input}, // input_to_cell_weight tensor 293 {n_cell, n_input}, // input_to_output_weight tensor 294 295 {n_cell, n_output}, // recurrent_to_input_weight tensor 296 {n_cell, n_output}, // recurrent_to_forget_weight tensor 297 {n_cell, n_output}, // recurrent_to_cell_weight tensor 298 {n_cell, n_output}, // recurrent_to_output_weight tensor 299 300 {0}, // cell_to_input_weight tensor 301 {0}, // cell_to_forget_weight tensor 302 {0}, // cell_to_output_weight tensor 303 304 {n_cell}, // input_gate_bias tensor 305 {n_cell}, // forget_gate_bias tensor 306 {n_cell}, // cell_bias tensor 307 {n_cell}, // output_gate_bias tensor 308 309 {0, 0}, // projection_weight tensor 310 {0}, // projection_bias tensor 311 }); 312 313 lstm.SetInputToInputWeights({-0.45018822, -0.02338299, -0.0870589, 314 -0.34550029, 0.04266912, -0.15680569, 315 -0.34856534, 0.43890524}); 316 317 lstm.SetInputToCellWeights({-0.50013041, 0.1370284, 0.11810488, 0.2013163, 318 -0.20583314, 0.44344562, 0.22077113, 319 -0.29909778}); 320 321 lstm.SetInputToForgetWeights({0.09701663, 0.20334584, -0.50592935, 322 -0.31343272, -0.40032279, 0.44781327, 323 0.01387155, -0.35593212}); 324 325 lstm.SetInputToOutputWeights({-0.25065863, -0.28290087, 0.04613829, 326 0.40525138, 0.44272184, 0.03897077, -0.1556896, 327 0.19487578}); 328 329 lstm.SetInputGateBias({0., 0., 0., 0.}); 330 331 lstm.SetCellGateBias({0., 0., 0., 0.}); 332 333 lstm.SetForgetGateBias({1., 1., 1., 1.}); 334 335 lstm.SetOutputGateBias({0., 0., 0., 0.}); 336 337 lstm.SetRecurrentToInputWeights( 338 {-0.0063535, -0.2042388, 0.31454784, -0.35746509, 0.28902304, 0.08183324, 339 -0.16555229, 0.02286911, -0.13566875, 0.03034258, 0.48091322, 340 -0.12528998, 0.24077177, -0.51332325, -0.33502164, 0.10629296}); 341 342 lstm.SetRecurrentToCellWeights( 343 {-0.3407414, 0.24443203, -0.2078532, 0.26320225, 0.05695659, -0.00123841, 344 -0.4744786, -0.35869038, -0.06418842, -0.13502428, -0.501764, 0.22830659, 345 -0.46367589, 0.26016325, -0.03894562, -0.16368064}); 346 347 lstm.SetRecurrentToForgetWeights( 348 {-0.48684245, -0.06655136, 0.42224967, 0.2112639, 0.27654213, 0.20864892, 349 -0.07646349, 0.45877004, 0.00141793, -0.14609534, 0.36447752, 0.09196436, 350 0.28053468, 0.01560611, -0.20127171, -0.01140004}); 351 352 lstm.SetRecurrentToOutputWeights( 353 {0.43385774, -0.17194885, 0.2718237, 0.09215671, 0.24107647, -0.39835793, 354 0.18212086, 0.01301402, 0.48572797, -0.50656658, 0.20047462, -0.20607421, 355 -0.51818722, -0.15390486, 0.0468148, 0.39922136}); 356 357 static float lstm_input[] = {2., 3., 3., 4., 1., 1.}; 358 static float lstm_golden_output[] = {-0.02973187, 0.1229473, 0.20885126, 359 -0.15358765, -0.03716109, 0.12507336, 360 0.41193449, -0.20860538, -0.15053082, 361 0.09120187, 0.24278517, -0.12222792}; 362 363 // Resetting cell_state and output_state 364 lstm.ResetCellState(); 365 lstm.ResetOutputState(); 366 367 const int input_sequence_size = 368 sizeof(lstm_input) / sizeof(float) / (lstm.num_inputs()); 369 for (int i = 0; i < input_sequence_size; i++) { 370 float* batch0_start = lstm_input + i * lstm.num_inputs(); 371 float* batch0_end = batch0_start + lstm.num_inputs(); 372 373 lstm.SetInput(0, batch0_start, batch0_end); 374 375 lstm.Invoke(); 376 377 float* golden_start = lstm_golden_output + i * lstm.num_outputs(); 378 float* golden_end = golden_start + lstm.num_outputs(); 379 std::vector<float> expected; 380 expected.insert(expected.end(), golden_start, golden_end); 381 EXPECT_THAT(lstm.GetOutput(), ElementsAreArray(ArrayFloatNear(expected))); 382 } 383} 384 385 386TEST(LSTMOpTest, BlackBoxTestWithCifgWithPeepholeNoProjectionNoClipping) { 387 const int n_batch = 1; 388 const int n_input = 2; 389 // n_cell and n_output have the same size when there is no projection. 390 const int n_cell = 4; 391 const int n_output = 4; 392 393 LSTMOpModel lstm(n_batch, n_input, n_cell, n_output, 394 /*use_cifg=*/true, /*use_peephole=*/true, 395 /*use_projection_weights=*/false, 396 /*use_projection_bias=*/false, 397 /*cell_clip=*/0.0, /*proj_clip=*/0.0, 398 { 399 {n_batch, n_input}, // input tensor 400 401 {0, 0}, // input_to_input_weight tensor 402 {n_cell, n_input}, // input_to_forget_weight tensor 403 {n_cell, n_input}, // input_to_cell_weight tensor 404 {n_cell, n_input}, // input_to_output_weight tensor 405 406 {0, 0}, // recurrent_to_input_weight tensor 407 {n_cell, n_output}, // recurrent_to_forget_weight tensor 408 {n_cell, n_output}, // recurrent_to_cell_weight tensor 409 {n_cell, n_output}, // recurrent_to_output_weight tensor 410 411 {0}, // cell_to_input_weight tensor 412 {n_cell}, // cell_to_forget_weight tensor 413 {n_cell}, // cell_to_output_weight tensor 414 415 {n_cell}, // input_gate_bias tensor 416 {n_cell}, // forget_gate_bias tensor 417 {n_cell}, // cell_bias tensor 418 {n_cell}, // output_gate_bias tensor 419 420 {0, 0}, // projection_weight tensor 421 {0}, // projection_bias tensor 422 }); 423 424 lstm.SetInputToCellWeights({-0.49770179, -0.27711356, -0.09624726, 0.05100781, 425 0.04717243, 0.48944736, -0.38535351, 426 -0.17212132}); 427 428 lstm.SetInputToForgetWeights({-0.55291498, -0.42866567, 0.13056988, 429 -0.3633365, -0.22755712, 0.28253698, 0.24407166, 430 0.33826375}); 431 432 lstm.SetInputToOutputWeights({0.10725588, -0.02335852, -0.55932593, 433 -0.09426838, -0.44257352, 0.54939759, 434 0.01533556, 0.42751634}); 435 436 lstm.SetCellGateBias({0., 0., 0., 0.}); 437 438 lstm.SetForgetGateBias({1., 1., 1., 1.}); 439 440 lstm.SetOutputGateBias({0., 0., 0., 0.}); 441 442 lstm.SetRecurrentToCellWeights( 443 {0.54066205, -0.32668582, -0.43562764, -0.56094903, 0.42957711, 444 0.01841056, -0.32764608, -0.33027974, -0.10826075, 0.20675004, 445 0.19069612, -0.03026325, -0.54532051, 0.33003211, 0.44901288, 446 0.21193194}); 447 448 lstm.SetRecurrentToForgetWeights( 449 {-0.13832897, -0.0515101, -0.2359007, -0.16661474, -0.14340827, 450 0.36986142, 0.23414481, 0.55899, 0.10798943, -0.41174671, 0.17751795, 451 -0.34484994, -0.35874045, -0.11352962, 0.27268326, 0.54058349}); 452 453 lstm.SetRecurrentToOutputWeights( 454 {0.41613156, 0.42610586, -0.16495961, -0.5663873, 0.30579174, -0.05115908, 455 -0.33941799, 0.23364776, 0.11178309, 0.09481031, -0.26424935, 0.46261835, 456 0.50248802, 0.26114327, -0.43736315, 0.33149987}); 457 458 lstm.SetCellToForgetWeights( 459 {0.47485286, -0.51955009, -0.24458408, 0.31544167}); 460 lstm.SetCellToOutputWeights( 461 {-0.17135078, 0.82760304, 0.85573703, -0.77109635}); 462 463 static float lstm_input[] = {2., 3., 3., 4., 1., 1.}; 464 static float lstm_golden_output[] = {-0.36444446, -0.00352185, 0.12886585, 465 -0.05163646, -0.42312205, -0.01218222, 466 0.24201041, -0.08124574, -0.358325, 467 -0.04621704, 0.21641694, -0.06471302}; 468 469 // Resetting cell_state and output_state 470 lstm.ResetCellState(); 471 lstm.ResetOutputState(); 472 473 const int input_sequence_size = 474 sizeof(lstm_input) / sizeof(float) / (lstm.num_inputs()); 475 for (int i = 0; i < input_sequence_size; i++) { 476 float* batch0_start = lstm_input + i * lstm.num_inputs(); 477 float* batch0_end = batch0_start + lstm.num_inputs(); 478 479 lstm.SetInput(0, batch0_start, batch0_end); 480 481 lstm.Invoke(); 482 483 float* golden_start = lstm_golden_output + i * lstm.num_outputs(); 484 float* golden_end = golden_start + lstm.num_outputs(); 485 std::vector<float> expected; 486 expected.insert(expected.end(), golden_start, golden_end); 487 EXPECT_THAT(lstm.GetOutput(), ElementsAreArray(ArrayFloatNear(expected))); 488 } 489} 490 491TEST(LSTMOpTest, BlackBoxTestWithPeepholeWithProjectionNoClipping) { 492 const int n_batch = 2; 493 const int n_input = 5; 494 const int n_cell = 20; 495 const int n_output = 16; 496 497 LSTMOpModel lstm(n_batch, n_input, n_cell, n_output, 498 /*use_cifg=*/false, /*use_peephole=*/true, 499 /*use_projection_weights=*/true, 500 /*use_projection_bias=*/false, 501 /*cell_clip=*/0.0, /*proj_clip=*/0.0, 502 { 503 {n_batch, n_input}, // input tensor 504 505 {n_cell, n_input}, // input_to_input_weight tensor 506 {n_cell, n_input}, // input_to_forget_weight tensor 507 {n_cell, n_input}, // input_to_cell_weight tensor 508 {n_cell, n_input}, // input_to_output_weight tensor 509 510 {n_cell, n_output}, // recurrent_to_input_weight tensor 511 {n_cell, n_output}, // recurrent_to_forget_weight tensor 512 {n_cell, n_output}, // recurrent_to_cell_weight tensor 513 {n_cell, n_output}, // recurrent_to_output_weight tensor 514 515 {n_cell}, // cell_to_input_weight tensor 516 {n_cell}, // cell_to_forget_weight tensor 517 {n_cell}, // cell_to_output_weight tensor 518 519 {n_cell}, // input_gate_bias tensor 520 {n_cell}, // forget_gate_bias tensor 521 {n_cell}, // cell_bias tensor 522 {n_cell}, // output_gate_bias tensor 523 524 {n_output, n_cell}, // projection_weight tensor 525 {0}, // projection_bias tensor 526 }); 527 528 lstm.SetInputToInputWeights( 529 {0.021393683, 0.06124551, 0.046905167, -0.014657677, -0.03149463, 530 0.09171803, 0.14647801, 0.10797193, -0.0057968358, 0.0019193048, 531 -0.2726754, 0.10154029, -0.018539885, 0.080349885, -0.10262385, 532 -0.022599787, -0.09121155, -0.008675967, -0.045206103, -0.0821282, 533 -0.008045952, 0.015478081, 0.055217247, 0.038719587, 0.044153627, 534 -0.06453243, 0.05031825, -0.046935108, -0.008164439, 0.014574226, 535 -0.1671009, -0.15519552, -0.16819797, -0.13971269, -0.11953059, 536 0.25005487, -0.22790983, 0.009855087, -0.028140958, -0.11200698, 537 0.11295408, -0.0035217577, 0.054485075, 0.05184695, 0.064711206, 538 0.10989193, 0.11674786, 0.03490607, 0.07727357, 0.11390585, 539 -0.1863375, -0.1034451, -0.13945189, -0.049401227, -0.18767063, 540 0.042483903, 0.14233552, 0.13832581, 0.18350165, 0.14545603, 541 -0.028545704, 0.024939531, 0.050929718, 0.0076203286, -0.0029723682, 542 -0.042484224, -0.11827596, -0.09171104, -0.10808628, -0.16327988, 543 -0.2273378, -0.0993647, -0.017155107, 0.0023917493, 0.049272764, 544 0.0038534778, 0.054764505, 0.089753784, 0.06947234, 0.08014476, 545 -0.04544234, -0.0497073, -0.07135631, -0.048929106, -0.004042012, 546 -0.009284026, 0.018042054, 0.0036860977, -0.07427302, -0.11434604, 547 -0.018995456, 0.031487543, 0.012834908, 0.019977754, 0.044256654, 548 -0.39292613, -0.18519334, -0.11651281, -0.06809892, 0.011373677}); 549 550 lstm.SetInputToForgetWeights( 551 {-0.0018401089, -0.004852237, 0.03698424, 0.014181704, 0.028273236, 552 -0.016726194, -0.05249759, -0.10204261, 0.00861066, -0.040979505, 553 -0.009899187, 0.01923892, -0.028177269, -0.08535103, -0.14585495, 554 0.10662567, -0.01909731, -0.017883534, -0.0047269356, -0.045103323, 555 0.0030784295, 0.076784775, 0.07463696, 0.094531395, 0.0814421, 556 -0.12257899, -0.033945758, -0.031303465, 0.045630626, 0.06843887, 557 -0.13492945, -0.012480007, -0.0811829, -0.07224499, -0.09628791, 558 0.045100946, 0.0012300825, 0.013964662, 0.099372394, 0.02543059, 559 0.06958324, 0.034257296, 0.0482646, 0.06267997, 0.052625068, 560 0.12784666, 0.07077897, 0.025725935, 0.04165009, 0.07241905, 561 0.018668644, -0.037377294, -0.06277783, -0.08833636, -0.040120605, 562 -0.011405586, -0.007808335, -0.010301386, -0.005102167, 0.027717464, 563 0.05483423, 0.11449111, 0.11289652, 0.10939839, 0.13396506, 564 -0.08402166, -0.01901462, -0.044678304, -0.07720565, 0.014350063, 565 -0.11757958, -0.0652038, -0.08185733, -0.076754324, -0.092614375, 566 0.10405491, 0.052960336, 0.035755895, 0.035839386, -0.012540553, 567 0.036881298, 0.02913376, 0.03420159, 0.05448447, -0.054523353, 568 0.02582715, 0.02327355, -0.011857179, -0.0011980024, -0.034641717, 569 -0.026125094, -0.17582615, -0.15923657, -0.27486774, -0.0006143371, 570 0.0001771948, -8.470171e-05, 0.02651807, 0.045790765, 0.06956496}); 571 572 lstm.SetInputToCellWeights( 573 {-0.04580283, -0.09549462, -0.032418985, -0.06454633, 574 -0.043528453, 0.043018587, -0.049152344, -0.12418144, 575 -0.078985475, -0.07596889, 0.019484362, -0.11434962, 576 -0.0074034138, -0.06314844, -0.092981495, 0.0062155537, 577 -0.025034338, -0.0028890965, 0.048929527, 0.06235075, 578 0.10665918, -0.032036792, -0.08505916, -0.10843358, 579 -0.13002433, -0.036816437, -0.02130134, -0.016518239, 580 0.0047691227, -0.0025825808, 0.066017866, 0.029991534, 581 -0.10652836, -0.1037554, -0.13056071, -0.03266643, 582 -0.033702414, -0.006473424, -0.04611692, 0.014419339, 583 -0.025174323, 0.0396852, 0.081777506, 0.06157468, 584 0.10210095, -0.009658194, 0.046511717, 0.03603906, 585 0.0069369148, 0.015960095, -0.06507666, 0.09551598, 586 0.053568836, 0.06408714, 0.12835667, -0.008714329, 587 -0.20211966, -0.12093674, 0.029450472, 0.2849013, 588 -0.029227901, 0.1164364, -0.08560263, 0.09941786, 589 -0.036999565, -0.028842626, -0.0033637602, -0.017012902, 590 -0.09720865, -0.11193351, -0.029155117, -0.017936034, 591 -0.009768936, -0.04223324, -0.036159635, 0.06505112, 592 -0.021742892, -0.023377212, -0.07221364, -0.06430552, 593 0.05453865, 0.091149814, 0.06387331, 0.007518393, 594 0.055960953, 0.069779344, 0.046411168, 0.10509911, 595 0.07463894, 0.0075130584, 0.012850982, 0.04555431, 596 0.056955688, 0.06555285, 0.050801456, -0.009862683, 597 0.00826772, -0.026555609, -0.0073611983, -0.0014897042}); 598 599 lstm.SetInputToOutputWeights( 600 {-0.0998932, -0.07201956, -0.052803773, -0.15629593, -0.15001918, 601 -0.07650751, 0.02359855, -0.075155355, -0.08037709, -0.15093534, 602 0.029517552, -0.04751393, 0.010350531, -0.02664851, -0.016839722, 603 -0.023121163, 0.0077019283, 0.012851257, -0.05040649, -0.0129761, 604 -0.021737747, -0.038305793, -0.06870586, -0.01481247, -0.001285394, 605 0.10124236, 0.083122835, 0.053313006, -0.062235646, -0.075637154, 606 -0.027833903, 0.029774971, 0.1130802, 0.09218906, 0.09506135, 607 -0.086665764, -0.037162706, -0.038880914, -0.035832845, -0.014481564, 608 -0.09825003, -0.12048569, -0.097665586, -0.05287633, -0.0964047, 609 -0.11366429, 0.035777505, 0.13568819, 0.052451383, 0.050649304, 610 0.05798951, -0.021852335, -0.099848844, 0.014740475, -0.078897946, 611 0.04974699, 0.014160473, 0.06973932, 0.04964942, 0.033364646, 612 0.08190124, 0.025535367, 0.050893165, 0.048514254, 0.06945813, 613 -0.078907564, -0.06707616, -0.11844508, -0.09986688, -0.07509403, 614 0.06263226, 0.14925587, 0.20188436, 0.12098451, 0.14639415, 615 0.0015017595, -0.014267382, -0.03417257, 0.012711468, 0.0028300495, 616 -0.024758482, -0.05098548, -0.0821182, 0.014225672, 0.021544158, 617 0.08949725, 0.07505268, -0.0020780868, 0.04908258, 0.06476295, 618 -0.022907063, 0.027562456, 0.040185735, 0.019567577, -0.015598739, 619 -0.049097303, -0.017121866, -0.083368234, -0.02332002, -0.0840956}); 620 621 lstm.SetInputGateBias( 622 {0.02234832, 0.14757581, 0.18176508, 0.10380666, 0.053110216, 623 -0.06928846, -0.13942584, -0.11816189, 0.19483899, 0.03652339, 624 -0.10250295, 0.036714908, -0.18426876, 0.036065217, 0.21810818, 625 0.02383196, -0.043370757, 0.08690144, -0.04444982, 0.00030581196}); 626 627 lstm.SetForgetGateBias({0.035185695, -0.042891346, -0.03032477, 0.23027696, 628 0.11098921, 0.15378423, 0.09263801, 0.09790885, 629 0.09508917, 0.061199076, 0.07665568, -0.015443159, 630 -0.03499149, 0.046190713, 0.08895977, 0.10899629, 631 0.40694186, 0.06030037, 0.012413437, -0.06108739}); 632 633 lstm.SetCellGateBias({-0.024379363, 0.0055531194, 0.23377132, 0.033463873, 634 -0.1483596, -0.10639995, -0.091433935, 0.058573797, 635 -0.06809782, -0.07889636, -0.043246906, -0.09829136, 636 -0.4279842, 0.034901652, 0.18797937, 0.0075234566, 637 0.016178843, 0.1749513, 0.13975595, 0.92058027}); 638 639 lstm.SetOutputGateBias( 640 {0.046159424, -0.0012809046, 0.03563469, 0.12648113, 0.027195795, 641 0.35373217, -0.018957434, 0.008907322, -0.0762701, 0.12018895, 642 0.04216877, 0.0022856654, 0.040952638, 0.3147856, 0.08225149, 643 -0.057416286, -0.14995944, -0.008040261, 0.13208859, 0.029760877}); 644 645 lstm.SetRecurrentToInputWeights( 646 {-0.001374326, -0.078856036, 0.10672688, 0.029162422, 647 -0.11585556, 0.02557986, -0.13446963, -0.035785314, 648 -0.01244275, 0.025961924, -0.02337298, -0.044228926, 649 -0.055839065, -0.046598054, -0.010546039, -0.06900766, 650 0.027239809, 0.022582639, -0.013296484, -0.05459212, 651 0.08981, -0.045407712, 0.08682226, -0.06867011, 652 -0.14390695, -0.02916037, 0.000996957, 0.091420636, 653 0.14283475, -0.07390571, -0.06402044, 0.062524505, 654 -0.093129106, 0.04860203, -0.08364217, -0.08119002, 655 0.009352075, 0.22920375, 0.0016303885, 0.11583097, 656 -0.13732095, 0.012405723, -0.07551853, 0.06343048, 657 0.12162708, -0.031923793, -0.014335606, 0.01790974, 658 -0.10650317, -0.0724401, 0.08554849, -0.05727212, 659 0.06556731, -0.042729504, -0.043227166, 0.011683251, 660 -0.013082158, -0.029302018, -0.010899579, -0.062036745, 661 -0.022509435, -0.00964907, -0.01567329, 0.04260106, 662 -0.07787477, -0.11576462, 0.017356863, 0.048673786, 663 -0.017577527, -0.05527947, -0.082487635, -0.040137455, 664 -0.10820036, -0.04666372, 0.022746278, -0.07851417, 665 0.01068115, 0.032956902, 0.022433773, 0.0026891115, 666 0.08944216, -0.0685835, 0.010513544, 0.07228705, 667 0.02032331, -0.059686817, -0.0005566496, -0.086984694, 668 0.040414046, -0.1380399, 0.094208956, -0.05722982, 669 0.012092817, -0.04989123, -0.086576, -0.003399834, 670 -0.04696032, -0.045747425, 0.10091314, 0.048676282, 671 -0.029037097, 0.031399418, -0.0040285117, 0.047237843, 672 0.09504992, 0.041799378, -0.049185462, -0.031518843, 673 -0.10516937, 0.026374253, 0.10058866, -0.0033195973, 674 -0.041975245, 0.0073591834, 0.0033782164, -0.004325073, 675 -0.10167381, 0.042500053, -0.01447153, 0.06464186, 676 -0.017142897, 0.03312627, 0.009205989, 0.024138335, 677 -0.011337001, 0.035530265, -0.010912711, 0.0706555, 678 -0.005894094, 0.051841937, -0.1401738, -0.02351249, 679 0.0365468, 0.07590991, 0.08838724, 0.021681072, 680 -0.10086113, 0.019608743, -0.06195883, 0.077335775, 681 0.023646897, -0.095322326, 0.02233014, 0.09756986, 682 -0.048691444, -0.009579111, 0.07595467, 0.11480546, 683 -0.09801813, 0.019894179, 0.08502348, 0.004032281, 684 0.037211012, 0.068537936, -0.048005626, -0.091520436, 685 -0.028379958, -0.01556313, 0.06554592, -0.045599163, 686 -0.01672207, -0.020169014, -0.011877351, -0.20212261, 687 0.010889619, 0.0047078193, 0.038385306, 0.08540671, 688 -0.017140968, -0.0035865551, 0.016678626, 0.005633034, 689 0.015963363, 0.00871737, 0.060130805, 0.028611384, 690 0.10109069, -0.015060172, -0.07894427, 0.06401885, 691 0.011584063, -0.024466386, 0.0047652307, -0.09041358, 692 0.030737216, -0.0046374933, 0.14215417, -0.11823516, 693 0.019899689, 0.006106124, -0.027092824, 0.0786356, 694 0.05052217, -0.058925, -0.011402121, -0.024987547, 695 -0.0013661642, -0.06832946, -0.015667673, -0.1083353, 696 -0.00096863037, -0.06988685, -0.053350925, -0.027275559, 697 -0.033664223, -0.07978348, -0.025200296, -0.017207067, 698 -0.058403496, -0.055697463, 0.005798788, 0.12965427, 699 -0.062582195, 0.0013350133, -0.10482091, 0.0379771, 700 0.072521195, -0.0029455067, -0.13797039, -0.03628521, 701 0.013806405, -0.017858358, -0.01008298, -0.07700066, 702 -0.017081132, 0.019358726, 0.0027079724, 0.004635139, 703 0.062634714, -0.02338735, -0.039547626, -0.02050681, 704 0.03385117, -0.083611414, 0.002862572, -0.09421313, 705 0.058618143, -0.08598433, 0.00972939, 0.023867095, 706 -0.053934585, -0.023203006, 0.07452513, -0.048767887, 707 -0.07314807, -0.056307215, -0.10433547, -0.06440842, 708 0.04328182, 0.04389765, -0.020006588, -0.09076438, 709 -0.11652589, -0.021705797, 0.03345259, -0.010329105, 710 -0.025767034, 0.013057034, -0.07316461, -0.10145612, 711 0.06358255, 0.18531723, 0.07759293, 0.12006465, 712 0.1305557, 0.058638252, -0.03393652, 0.09622831, 713 -0.16253184, -2.4580743e-06, 0.079869635, -0.070196845, 714 -0.005644518, 0.06857898, -0.12598175, -0.035084512, 715 0.03156317, -0.12794146, -0.031963028, 0.04692781, 716 0.030070418, 0.0071660685, -0.095516115, -0.004643372, 717 0.040170413, -0.062104587, -0.0037324072, 0.0554317, 718 0.08184801, -0.019164372, 0.06791302, 0.034257166, 719 -0.10307039, 0.021943003, 0.046745934, 0.0790918, 720 -0.0265588, -0.007824208, 0.042546265, -0.00977924, 721 -0.0002440307, -0.017384544, -0.017990116, 0.12252321, 722 -0.014512694, -0.08251313, 0.08861942, 0.13589665, 723 0.026351685, 0.012641483, 0.07466548, 0.044301085, 724 -0.045414884, -0.051112458, 0.03444247, -0.08502782, 725 -0.04106223, -0.028126027, 0.028473156, 0.10467447}); 726 727 lstm.SetRecurrentToForgetWeights( 728 {-0.057784554, -0.026057621, -0.068447545, -0.022581743, 729 0.14811787, 0.10826372, 0.09471067, 0.03987225, 730 -0.0039523416, 0.00030638507, 0.053185795, 0.10572994, 731 0.08414449, -0.022036452, -0.00066928595, -0.09203576, 732 0.032950465, -0.10985798, -0.023809856, 0.0021431844, 733 -0.02196096, -0.00326074, 0.00058621005, -0.074678116, 734 -0.06193199, 0.055729095, 0.03736828, 0.020123724, 735 0.061878487, -0.04729229, 0.034919553, -0.07585433, 736 -0.04421272, -0.044019096, 0.085488975, 0.04058006, 737 -0.06890133, -0.030951202, -0.024628663, -0.07672815, 738 0.034293607, 0.08556707, -0.05293577, -0.033561368, 739 -0.04899627, 0.0241671, 0.015736353, -0.095442444, 740 -0.029564252, 0.016493602, -0.035026584, 0.022337519, 741 -0.026871363, 0.004780428, 0.0077918363, -0.03601621, 742 0.016435321, -0.03263031, -0.09543275, -0.047392778, 743 0.013454138, 0.028934088, 0.01685226, -0.086110644, 744 -0.046250615, -0.01847454, 0.047608484, 0.07339695, 745 0.034546845, -0.04881143, 0.009128804, -0.08802852, 746 0.03761666, 0.008096139, -0.014454086, 0.014361001, 747 -0.023502491, -0.0011840804, -0.07607001, 0.001856849, 748 -0.06509276, -0.006021153, -0.08570962, -0.1451793, 749 0.060212336, 0.055259194, 0.06974018, 0.049454916, 750 -0.027794661, -0.08077226, -0.016179763, 0.1169753, 751 0.17213494, -0.0056326236, -0.053934924, -0.0124349, 752 -0.11520337, 0.05409887, 0.088759385, 0.0019655675, 753 0.0042065294, 0.03881498, 0.019844765, 0.041858196, 754 -0.05695512, 0.047233116, 0.038937137, -0.06542224, 755 0.014429736, -0.09719407, 0.13908425, -0.05379757, 756 0.012321099, 0.082840554, -0.029899208, 0.044217527, 757 0.059855383, 0.07711018, -0.045319796, 0.0948846, 758 -0.011724666, -0.0033288454, -0.033542685, -0.04764985, 759 -0.13873616, 0.040668588, 0.034832682, -0.015319203, 760 -0.018715994, 0.046002675, 0.0599172, -0.043107376, 761 0.0294216, -0.002314414, -0.022424703, 0.0030315618, 762 0.0014641669, 0.0029166266, -0.11878115, 0.013738511, 763 0.12375372, -0.0006038222, 0.029104086, 0.087442465, 764 0.052958444, 0.07558703, 0.04817258, 0.044462286, 765 -0.015213451, -0.08783778, -0.0561384, -0.003008196, 766 0.047060397, -0.002058388, 0.03429439, -0.018839769, 767 0.024734668, 0.024614193, -0.042046934, 0.09597743, 768 -0.0043254104, 0.04320769, 0.0064070094, -0.0019131786, 769 -0.02558259, -0.022822596, -0.023273505, -0.02464396, 770 -0.10991725, -0.006240552, 0.0074488563, 0.024044557, 771 0.04383914, -0.046476185, 0.028658995, 0.060410924, 772 0.050786525, 0.009452605, -0.0073054377, -0.024810238, 773 0.0052906186, 0.0066939713, -0.0020913032, 0.014515517, 774 0.015898481, 0.021362653, -0.030262267, 0.016587038, 775 -0.011442813, 0.041154444, -0.007631438, -0.03423484, 776 -0.010977775, 0.036152758, 0.0066366293, 0.11915515, 777 0.02318443, -0.041350313, 0.021485701, -0.10906167, 778 -0.028218046, -0.00954771, 0.020531068, -0.11995105, 779 -0.03672871, 0.024019798, 0.014255957, -0.05221243, 780 -0.00661567, -0.04630967, 0.033188973, 0.10107534, 781 -0.014027541, 0.030796422, -0.10270911, -0.035999842, 782 0.15443139, 0.07684145, 0.036571592, -0.035900835, 783 -0.0034699554, 0.06209149, 0.015920248, -0.031122351, 784 -0.03858649, 0.01849943, 0.13872518, 0.01503974, 785 0.069941424, -0.06948533, -0.0088794185, 0.061282158, 786 -0.047401894, 0.03100163, -0.041533746, -0.10430945, 787 0.044574402, -0.01425562, -0.024290353, 0.034563623, 788 0.05866852, 0.023947537, -0.09445152, 0.035450947, 789 0.02247216, -0.0042998926, 0.061146557, -0.10250651, 790 0.020881841, -0.06747029, 0.10062043, -0.0023941975, 791 0.03532124, -0.016341697, 0.09685456, -0.016764693, 792 0.051808182, 0.05875331, -0.04536488, 0.001626336, 793 -0.028892258, -0.01048663, -0.009793449, -0.017093895, 794 0.010987891, 0.02357273, -0.00010856845, 0.0099760275, 795 -0.001845119, -0.03551521, 0.0018358806, 0.05763657, 796 -0.01769146, 0.040995963, 0.02235177, -0.060430344, 797 0.11475477, -0.023854522, 0.10071741, 0.0686208, 798 -0.014250481, 0.034261297, 0.047418304, 0.08562733, 799 -0.030519066, 0.0060542435, 0.014653856, -0.038836084, 800 0.04096551, 0.032249358, -0.08355519, -0.026823482, 801 0.056386515, -0.010401743, -0.028396193, 0.08507674, 802 0.014410365, 0.020995233, 0.17040324, 0.11511526, 803 0.02459721, 0.0066619175, 0.025853224, -0.023133837, 804 -0.081302024, 0.017264642, -0.009585969, 0.09491168, 805 -0.051313367, 0.054532815, -0.014298593, 0.10657464, 806 0.007076659, 0.10964551, 0.0409152, 0.008275321, 807 -0.07283536, 0.07937492, 0.04192024, -0.1075027}); 808 809 lstm.SetRecurrentToCellWeights( 810 {-0.037322544, 0.018592842, 0.0056175636, -0.06253426, 811 0.055647098, -0.05713207, -0.05626563, 0.005559383, 812 0.03375411, -0.025757805, -0.088049285, 0.06017052, 813 -0.06570978, 0.007384076, 0.035123326, -0.07920549, 814 0.053676967, 0.044480428, -0.07663568, 0.0071805613, 815 0.08089997, 0.05143358, 0.038261272, 0.03339287, 816 -0.027673481, 0.044746667, 0.028349208, 0.020090483, 817 -0.019443132, -0.030755889, -0.0040000007, 0.04465846, 818 -0.021585021, 0.0031670958, 0.0053199246, -0.056117613, 819 -0.10893326, 0.076739706, -0.08509834, -0.027997585, 820 0.037871376, 0.01449768, -0.09002357, -0.06111149, 821 -0.046195522, 0.0422062, -0.005683705, -0.1253618, 822 -0.012925729, -0.04890792, 0.06985068, 0.037654128, 823 0.03398274, -0.004781977, 0.007032333, -0.031787455, 824 0.010868644, -0.031489216, 0.09525667, 0.013939797, 825 0.0058680447, 0.0167067, 0.02668468, -0.04797466, 826 -0.048885044, -0.12722108, 0.035304096, 0.06554885, 827 0.00972396, -0.039238118, -0.05159735, -0.11329045, 828 0.1613692, -0.03750952, 0.06529313, -0.071974665, 829 -0.11769596, 0.015524369, -0.0013754242, -0.12446318, 830 0.02786344, -0.014179351, 0.005264273, 0.14376344, 831 0.015983658, 0.03406988, -0.06939408, 0.040699873, 832 0.02111075, 0.09669095, 0.041345075, -0.08316494, 833 -0.07684199, -0.045768797, 0.032298047, -0.041805092, 834 0.0119405, 0.0061010392, 0.12652606, 0.0064572375, 835 -0.024950314, 0.11574242, 0.04508852, -0.04335324, 836 0.06760663, -0.027437469, 0.07216407, 0.06977076, 837 -0.05438599, 0.034033038, -0.028602652, 0.05346137, 838 0.043184172, -0.037189785, 0.10420091, 0.00882477, 839 -0.054019816, -0.074273005, -0.030617684, -0.0028467078, 840 0.024302477, -0.0038869337, 0.005332455, 0.0013399826, 841 0.04361412, -0.007001822, 0.09631092, -0.06702025, 842 -0.042049985, -0.035070654, -0.04103342, -0.10273396, 843 0.0544271, 0.037184782, -0.13150354, -0.0058036847, 844 -0.008264958, 0.042035464, 0.05891794, 0.029673764, 845 0.0063542654, 0.044788733, 0.054816857, 0.062257513, 846 -0.00093483756, 0.048938446, -0.004952862, -0.007730018, 847 -0.04043371, -0.017094059, 0.07229206, -0.023670016, 848 -0.052195564, -0.025616996, -0.01520939, 0.045104615, 849 -0.007376126, 0.003533447, 0.006570588, 0.056037236, 850 0.12436656, 0.051817212, 0.028532185, -0.08686856, 851 0.11868599, 0.07663395, -0.07323171, 0.03463402, 852 -0.050708205, -0.04458982, -0.11590894, 0.021273347, 853 0.1251325, -0.15313013, -0.12224372, 0.17228661, 854 0.023029093, 0.086124025, 0.006445803, -0.03496501, 855 0.028332196, 0.04449512, -0.042436164, -0.026587414, 856 -0.006041347, -0.09292539, -0.05678812, 0.03897832, 857 0.09465633, 0.008115513, -0.02171956, 0.08304309, 858 0.071401566, 0.019622514, 0.032163795, -0.004167056, 859 0.02295182, 0.030739572, 0.056506045, 0.004612461, 860 0.06524936, 0.059999723, 0.046395954, -0.0045512207, 861 -0.1335546, -0.030136576, 0.11584653, -0.014678886, 862 0.0020118146, -0.09688814, -0.0790206, 0.039770417, 863 -0.0329582, 0.07922767, 0.029322514, 0.026405897, 864 0.04207835, -0.07073373, 0.063781224, 0.0859677, 865 -0.10925287, -0.07011058, 0.048005477, 0.03438226, 866 -0.09606514, -0.006669445, -0.043381985, 0.04240257, 867 -0.06955775, -0.06769346, 0.043903265, -0.026784198, 868 -0.017840602, 0.024307009, -0.040079936, -0.019946516, 869 0.045318738, -0.12233574, 0.026170589, 0.0074471775, 870 0.15978073, 0.10185836, 0.10298046, -0.015476589, 871 -0.039390966, -0.072174534, 0.0739445, -0.1211869, 872 -0.0347889, -0.07943156, 0.014809798, -0.12412325, 873 -0.0030663363, 0.039695457, 0.0647603, -0.08291318, 874 -0.018529687, -0.004423833, 0.0037507233, 0.084633216, 875 -0.01514876, -0.056505352, -0.012800942, -0.06994386, 876 0.012962922, -0.031234352, 0.07029052, 0.016418684, 877 0.03618972, 0.055686004, -0.08663945, -0.017404709, 878 -0.054761406, 0.029065743, 0.052404847, 0.020238016, 879 0.0048197987, -0.0214882, 0.07078733, 0.013016777, 880 0.06262858, 0.009184685, 0.020785125, -0.043904778, 881 -0.0270329, -0.03299152, -0.060088247, -0.015162964, 882 -0.001828936, 0.12642565, -0.056757294, 0.013586685, 883 0.09232601, -0.035886683, 0.06000002, 0.05229691, 884 -0.052580316, -0.082029596, -0.010794592, 0.012947712, 885 -0.036429964, -0.085508935, -0.13127148, -0.017744139, 886 0.031502828, 0.036232427, -0.031581745, 0.023051167, 887 -0.05325106, -0.03421577, 0.028793324, -0.034633752, 888 -0.009881397, -0.043551125, -0.018609839, 0.0019097115, 889 -0.008799762, 0.056595087, 0.0022273948, 0.055752404}); 890 891 lstm.SetRecurrentToOutputWeights({ 892 0.025825322, -0.05813119, 0.09495884, -0.045984812, -0.01255415, 893 -0.0026479573, -0.08196161, -0.054914974, -0.0046604523, -0.029587349, 894 -0.044576716, -0.07480124, -0.082868785, 0.023254942, 0.027502948, 895 -0.0039728214, -0.08683098, -0.08116779, -0.014675607, -0.037924774, 896 -0.023314456, -0.007401714, -0.09255757, 0.029460307, -0.08829125, 897 -0.005139627, -0.08989442, -0.0555066, 0.13596267, -0.025062224, 898 -0.048351806, -0.03850004, 0.07266485, -0.022414139, 0.05940088, 899 0.075114764, 0.09597592, -0.010211725, -0.0049794707, -0.011523867, 900 -0.025980417, 0.072999895, 0.11091378, -0.081685916, 0.014416728, 901 0.043229222, 0.034178585, -0.07530371, 0.035837382, -0.085607, 902 -0.007721233, -0.03287832, -0.043848954, -0.06404588, -0.06632928, 903 -0.073643476, 0.008214239, -0.045984086, 0.039764922, 0.03474462, 904 0.060612556, -0.080590084, 0.049127717, 0.04151091, -0.030063879, 905 0.008801774, -0.023021035, -0.019558564, 0.05158114, -0.010947698, 906 -0.011825728, 0.0075720972, 0.0699727, -0.0039981045, 0.069350146, 907 0.08799282, 0.016156472, 0.035502106, 0.11695009, 0.006217345, 908 0.13392477, -0.037875112, 0.025745004, 0.08940699, -0.00924166, 909 0.0046702605, -0.036598757, -0.08811812, 0.10522024, -0.032441203, 910 0.008176899, -0.04454919, 0.07058152, 0.0067963637, 0.039206743, 911 0.03259838, 0.03725492, -0.09515802, 0.013326398, -0.052055415, 912 -0.025676316, 0.03198509, -0.015951829, -0.058556724, 0.036879618, 913 0.043357447, 0.028362012, -0.05908629, 0.0059240665, -0.04995891, 914 -0.019187413, 0.0276265, -0.01628143, 0.0025863599, 0.08800015, 915 0.035250366, -0.022165963, -0.07328642, -0.009415526, -0.07455109, 916 0.11690406, 0.0363299, 0.07411125, 0.042103454, -0.009660886, 917 0.019076364, 0.018299393, -0.046004917, 0.08891175, 0.0431396, 918 -0.026327137, -0.051502608, 0.08979574, -0.051670972, 0.04940282, 919 -0.07491107, -0.021240504, 0.022596184, -0.034280192, 0.060163025, 920 -0.058211457, -0.051837247, -0.01349775, -0.04639988, -0.035936575, 921 -0.011681591, 0.064818054, 0.0073146066, -0.021745546, -0.043124277, 922 -0.06471268, -0.07053354, -0.029321948, -0.05330136, 0.016933719, 923 -0.053782392, 0.13747959, -0.1361751, -0.11569455, 0.0033329215, 924 0.05693899, -0.053219706, 0.063698, 0.07977434, -0.07924483, 925 0.06936997, 0.0034815092, -0.007305279, -0.037325785, -0.07251102, 926 -0.033633437, -0.08677009, 0.091591336, -0.14165086, 0.021752775, 927 0.019683983, 0.0011612234, -0.058154266, 0.049996935, 0.0288841, 928 -0.0024567875, -0.14345716, 0.010955264, -0.10234828, 0.1183656, 929 -0.0010731248, -0.023590032, -0.072285876, -0.0724771, -0.026382286, 930 -0.0014920527, 0.042667855, 0.0018776858, 0.02986552, 0.009814309, 931 0.0733756, 0.12289186, 0.018043943, -0.0458958, 0.049412545, 932 0.033632483, 0.05495232, 0.036686596, -0.013781798, -0.010036754, 933 0.02576849, -0.08307328, 0.010112348, 0.042521734, -0.05869831, 934 -0.071689695, 0.03876447, -0.13275425, -0.0352966, -0.023077697, 935 0.10285965, 0.084736146, 0.15568255, -0.00040734606, 0.027835453, 936 -0.10292561, -0.032401145, 0.10053256, -0.026142767, -0.08271222, 937 -0.0030240538, -0.016368777, 0.1070414, 0.042672627, 0.013456989, 938 -0.0437609, -0.022309763, 0.11576483, 0.04108048, 0.061026827, 939 -0.0190714, -0.0869359, 0.037901703, 0.0610107, 0.07202949, 940 0.01675338, 0.086139716, -0.08795751, -0.014898893, -0.023771819, 941 -0.01965048, 0.007955471, -0.043740474, 0.03346837, -0.10549954, 942 0.090567775, 0.042013682, -0.03176985, 0.12569028, -0.02421228, 943 -0.029526481, 0.023851605, 0.031539805, 0.05292009, -0.02344001, 944 -0.07811758, -0.08834428, 0.10094801, 0.16594367, -0.06861939, 945 -0.021256343, -0.041093912, -0.06669611, 0.035498552, 0.021757556, 946 -0.09302526, -0.015403468, -0.06614931, -0.051798206, -0.013874718, 947 0.03630673, 0.010412845, -0.08077351, 0.046185967, 0.0035662893, 948 0.03541868, -0.094149634, -0.034814864, 0.003128424, -0.020674974, 949 -0.03944324, -0.008110165, -0.11113267, 0.08484226, 0.043586485, 950 0.040582247, 0.0968012, -0.065249965, -0.028036479, 0.0050708856, 951 0.0017462453, 0.0326779, 0.041296225, 0.09164146, -0.047743853, 952 -0.015952192, -0.034451712, 0.084197424, -0.05347844, -0.11768019, 953 0.085926116, -0.08251791, -0.045081906, 0.0948852, 0.068401024, 954 0.024856757, 0.06978981, -0.057309967, -0.012775832, -0.0032452994, 955 0.01977615, -0.041040014, -0.024264973, 0.063464895, 0.05431621, 956 }); 957 958 lstm.SetCellToInputWeights( 959 {0.040369894, 0.030746894, 0.24704495, 0.018586371, -0.037586458, 960 -0.15312155, -0.11812848, -0.11465643, 0.20259799, 0.11418174, 961 -0.10116027, -0.011334949, 0.12411352, -0.076769054, -0.052169047, 962 0.21198851, -0.38871562, -0.09061183, -0.09683246, -0.21929175}); 963 964 lstm.SetCellToForgetWeights( 965 {-0.01998659, -0.15568835, -0.24248174, -0.012770197, 0.041331276, 966 -0.072311886, -0.052123554, -0.0066330447, -0.043891653, 0.036225766, 967 -0.047248036, 0.021479502, 0.033189066, 0.11952997, -0.020432774, 968 0.64658105, -0.06650122, -0.03467612, 0.095340036, 0.23647355}); 969 970 lstm.SetCellToOutputWeights( 971 {0.08286371, -0.08261836, -0.51210177, 0.002913762, 0.17764764, 972 -0.5495371, -0.08460716, -0.24552552, 0.030037103, 0.04123544, 973 -0.11940523, 0.007358328, 0.1890978, 0.4833202, -0.34441817, 974 0.36312827, -0.26375428, 0.1457655, -0.19724406, 0.15548733}); 975 976 lstm.SetProjectionWeights( 977 {-0.009802181, 0.09401916, 0.0717386, -0.13895074, 0.09641832, 978 0.060420845, 0.08539281, 0.054285463, 0.061395317, 0.034448683, 979 -0.042991187, 0.019801661, -0.16840284, -0.015726732, -0.23041931, 980 -0.024478018, -0.10959692, -0.013875541, 0.18600968, -0.061274476, 981 0.0138165, -0.08160894, -0.07661644, 0.032372914, 0.16169067, 982 0.22465782, -0.03993472, -0.004017731, 0.08633481, -0.28869787, 983 0.08682067, 0.17240396, 0.014975425, 0.056431185, 0.031037588, 984 0.16702051, 0.0077946745, 0.15140012, 0.29405436, 0.120285, 985 -0.188994, -0.027265169, 0.043389652, -0.022061434, 0.014777949, 986 -0.20203483, 0.094781205, 0.19100232, 0.13987629, -0.036132768, 987 -0.06426278, -0.05108664, 0.13221376, 0.009441198, -0.16715929, 988 0.15859416, -0.040437475, 0.050779544, -0.022187516, 0.012166504, 989 0.027685808, -0.07675938, -0.0055694645, -0.09444123, 0.0046453946, 990 0.050794356, 0.10770313, -0.20790008, -0.07149004, -0.11425117, 991 0.008225835, -0.035802525, 0.14374903, 0.15262283, 0.048710253, 992 0.1847461, -0.007487823, 0.11000021, -0.09542012, 0.22619456, 993 -0.029149994, 0.08527916, 0.009043713, 0.0042746216, 0.016261552, 994 0.022461696, 0.12689082, -0.043589946, -0.12035478, -0.08361797, 995 -0.050666027, -0.1248618, -0.1275799, -0.071875185, 0.07377272, 996 0.09944291, -0.18897448, -0.1593054, -0.06526116, -0.040107165, 997 -0.004618631, -0.067624845, -0.007576253, 0.10727444, 0.041546922, 998 -0.20424393, 0.06907816, 0.050412357, 0.00724631, 0.039827548, 999 0.12449835, 0.10747581, 0.13708383, 0.09134148, -0.12617786, 1000 -0.06428341, 0.09956831, 0.1208086, -0.14676677, -0.0727722, 1001 0.1126304, 0.010139365, 0.015571211, -0.038128063, 0.022913318, 1002 -0.042050496, 0.16842307, -0.060597885, 0.10531834, -0.06411776, 1003 -0.07451711, -0.03410368, -0.13393489, 0.06534304, 0.003620307, 1004 0.04490757, 0.05970546, 0.05197996, 0.02839995, 0.10434969, 1005 -0.013699693, -0.028353551, -0.07260381, 0.047201227, -0.024575593, 1006 -0.036445823, 0.07155557, 0.009672501, -0.02328883, 0.009533515, 1007 -0.03606021, -0.07421458, -0.028082801, -0.2678904, -0.13221288, 1008 0.18419984, -0.13012612, -0.014588381, -0.035059117, -0.04824723, 1009 0.07830115, -0.056184657, 0.03277091, 0.025466874, 0.14494097, 1010 -0.12522776, -0.098633975, -0.10766018, -0.08317623, 0.08594209, 1011 0.07749552, 0.039474737, 0.1776665, -0.07409566, -0.0477268, 1012 0.29323658, 0.10801441, 0.1154011, 0.013952499, 0.10739139, 1013 0.10708251, -0.051456142, 0.0074137426, -0.10430189, 0.10034707, 1014 0.045594677, 0.0635285, -0.0715442, -0.089667566, -0.10811871, 1015 0.00026344223, 0.08298446, -0.009525053, 0.006585689, -0.24567553, 1016 -0.09450807, 0.09648481, 0.026996298, -0.06419476, -0.04752702, 1017 -0.11063944, -0.23441927, -0.17608605, -0.052156363, 0.067035615, 1018 0.19271925, -0.0032889997, -0.043264326, 0.09663576, -0.057112187, 1019 -0.10100678, 0.0628376, 0.04447668, 0.017961001, -0.10094388, 1020 -0.10190601, 0.18335468, 0.10494553, -0.052095775, -0.0026118709, 1021 0.10539724, -0.04383912, -0.042349473, 0.08438151, -0.1947263, 1022 0.02251204, 0.11216432, -0.10307853, 0.17351969, -0.039091777, 1023 0.08066188, -0.00561982, 0.12633002, 0.11335965, -0.0088127935, 1024 -0.019777594, 0.06864014, -0.059751723, 0.016233567, -0.06894641, 1025 -0.28651384, -0.004228674, 0.019708522, -0.16305895, -0.07468996, 1026 -0.0855457, 0.099339016, -0.07580735, -0.13775392, 0.08434318, 1027 0.08330512, -0.12131499, 0.031935584, 0.09180414, -0.08876437, 1028 -0.08049874, 0.008753825, 0.03498998, 0.030215185, 0.03907079, 1029 0.089751154, 0.029194152, -0.03337423, -0.019092513, 0.04331237, 1030 0.04299654, -0.036394123, -0.12915532, 0.09793732, 0.07512415, 1031 -0.11319543, -0.032502122, 0.15661901, 0.07671967, -0.005491124, 1032 -0.19379048, -0.218606, 0.21448623, 0.017840758, 0.1416943, 1033 -0.07051762, 0.19488361, 0.02664691, -0.18104725, -0.09334311, 1034 0.15026465, -0.15493552, -0.057762887, -0.11604192, -0.262013, 1035 -0.01391798, 0.012185008, 0.11156489, -0.07483202, 0.06693364, 1036 -0.26151478, 0.046425626, 0.036540434, -0.16435726, 0.17338543, 1037 -0.21401681, -0.11385144, -0.08283257, -0.069031075, 0.030635102, 1038 0.010969227, 0.11109743, 0.010919218, 0.027526086, 0.13519906, 1039 0.01891392, -0.046839405, -0.040167913, 0.017953383, -0.09700955, 1040 0.0061885654, -0.07000971, 0.026893595, -0.038844477, 0.14543656}); 1041 1042 static float lstm_input[][20] = { 1043 {// Batch0: 4 (input_sequence_size) * 5 (n_input) 1044 0.787926, 0.151646, 0.071352, 0.118426, 0.458058, 0.596268, 0.998386, 1045 0.568695, 0.864524, 0.571277, 0.073204, 0.296072, 0.743333, 0.069199, 1046 0.045348, 0.867394, 0.291279, 0.013714, 0.482521, 0.626339}, 1047 1048 {// Batch1: 4 (input_sequence_size) * 5 (n_input) 1049 0.295743, 0.544053, 0.690064, 0.858138, 0.497181, 0.642421, 0.524260, 1050 0.134799, 0.003639, 0.162482, 0.640394, 0.930399, 0.050782, 0.432485, 1051 0.988078, 0.082922, 0.563329, 0.865614, 0.333232, 0.259916}}; 1052 1053 static float lstm_golden_output[][64] = { 1054 {// Batch0: 4 (input_sequence_size) * 16 (n_output) 1055 -0.00396806, 0.029352, -0.00279226, 0.0159977, -0.00835576, 1056 -0.0211779, 0.0283512, -0.0114597, 0.00907307, -0.0244004, 1057 -0.0152191, -0.0259063, 0.00914318, 0.00415118, 0.017147, 1058 0.0134203, -0.0166936, 0.0381209, 0.000889694, 0.0143363, 1059 -0.0328911, -0.0234288, 0.0333051, -0.012229, 0.0110322, 1060 -0.0457725, -0.000832209, -0.0202817, 0.0327257, 0.0121308, 1061 0.0155969, 0.0312091, -0.0213783, 0.0350169, 0.000324794, 1062 0.0276012, -0.0263374, -0.0371449, 0.0446149, -0.0205474, 1063 0.0103729, -0.0576349, -0.0150052, -0.0292043, 0.0376827, 1064 0.0136115, 0.0243435, 0.0354492, -0.0189322, 0.0464512, 1065 -0.00251373, 0.0225745, -0.0308346, -0.0317124, 0.0460407, 1066 -0.0189395, 0.0149363, -0.0530162, -0.0150767, -0.0340193, 1067 0.0286833, 0.00824207, 0.0264887, 0.0305169}, 1068 {// Batch1: 4 (input_sequence_size) * 16 (n_output) 1069 -0.013869, 0.0287268, -0.00334693, 0.00733398, -0.0287926, 1070 -0.0186926, 0.0193662, -0.0115437, 0.00422612, -0.0345232, 1071 0.00223253, -0.00957321, 0.0210624, 0.013331, 0.0150954, 1072 0.02168, -0.0141913, 0.0322082, 0.00227024, 0.0260507, 1073 -0.0188721, -0.0296489, 0.0399134, -0.0160509, 0.0116039, 1074 -0.0447318, -0.0150515, -0.0277406, 0.0316596, 0.0118233, 1075 0.0214762, 0.0293641, -0.0204549, 0.0450315, -0.00117378, 1076 0.0167673, -0.0375007, -0.0238314, 0.038784, -0.0174034, 1077 0.0131743, -0.0506589, -0.0048447, -0.0240239, 0.0325789, 1078 0.00790065, 0.0220157, 0.0333314, -0.0264787, 0.0387855, 1079 -0.000764675, 0.0217599, -0.037537, -0.0335206, 0.0431679, 1080 -0.0211424, 0.010203, -0.062785, -0.00832363, -0.025181, 1081 0.0412031, 0.0118723, 0.0239643, 0.0394009}}; 1082 1083 // Resetting cell_state and output_state 1084 lstm.ResetCellState(); 1085 lstm.ResetOutputState(); 1086 1087 const int input_sequence_size = 1088 sizeof(lstm_input[0]) / sizeof(float) / (lstm.num_inputs()); 1089 for (int i = 0; i < input_sequence_size; i++) { 1090 float* batch0_start = lstm_input[0] + i * lstm.num_inputs(); 1091 float* batch0_end = batch0_start + lstm.num_inputs(); 1092 1093 lstm.SetInput(0, batch0_start, batch0_end); 1094 1095 float* batch1_start = lstm_input[1] + i * lstm.num_inputs(); 1096 float* batch1_end = batch1_start + lstm.num_inputs(); 1097 lstm.SetInput(lstm.num_inputs(), batch1_start, batch1_end); 1098 1099 lstm.Invoke(); 1100 1101 float* golden_start_batch0 = lstm_golden_output[0] + i * lstm.num_outputs(); 1102 float* golden_end_batch0 = golden_start_batch0 + lstm.num_outputs(); 1103 float* golden_start_batch1 = lstm_golden_output[1] + i * lstm.num_outputs(); 1104 float* golden_end_batch1 = golden_start_batch1 + lstm.num_outputs(); 1105 std::vector<float> expected; 1106 expected.insert(expected.end(), golden_start_batch0, golden_end_batch0); 1107 expected.insert(expected.end(), golden_start_batch1, golden_end_batch1); 1108 EXPECT_THAT(lstm.GetOutput(), ElementsAreArray(ArrayFloatNear(expected))); 1109 } 1110} 1111 1112 1113} // namespace wrapper 1114} // namespace nn 1115} // namespace android 1116