/packages/modules/NeuralNetworks/runtime/test/specs/V1_3/ |
D | fully_connected_quant8_signed.mod.py | 23 bias = Parameter("b0", "TENSOR_INT32", "{3}, 0.25f, 0", [4, 8, 12]) variable 26 model = model.Operation("FULLY_CONNECTED", in0, weights, bias, act_relu).To(out0) 43 bias = Parameter("b0", "TENSOR_INT32", "{1}, 0.04, 0", [10]) variable 46 model = model.Operation("FULLY_CONNECTED", in0, weights, bias, act).To(out0) 62 bias = Input("b0", "TENSOR_INT32", "{1}, 0.04, 0") variable 65 model = model.Operation("FULLY_CONNECTED", in0, weights, bias, act).To(out0) 72 bias: 85 bias = Parameter("b0", "TENSOR_INT32", "{1}, 0.25f, 0", [4]) variable 88 model = model.Operation("FULLY_CONNECTED", in0, weights, bias, act).To(out0) 104 bias = Input("b0", "TENSOR_INT32", "{1}, 0.25f, 0") variable [all …]
|
D | unidirectional_sequence_rnn.mod.py | 19 def test(name, input, weights, recurrent_weights, bias, hidden_state, argument 26 recurrent_weights, bias, hidden_state, activation, 33 bias: bias_data, 185 bias=Input("bias", "TENSOR_FLOAT32", "{{{}}}".format(num_units)), 211 bias=Input("bias", "TENSOR_FLOAT32", "{{{}}}".format(num_units)),
|
/packages/modules/NeuralNetworks/runtime/test/fuzzing/ |
D | TestRandomGraph.cpp | 479 .float32 = {.bias = 1e-7f, .mse = 1e-10f, .atol = 1e-6f, .rtol = 1e-6f}, 480 .float16 = {.bias = 1e-4f, .mse = 1e-8f, .atol = 1e-3f, .rtol = 1e-3f}, 482 .quant8Asymm = {.bias = 0.1f, .mse = 0.1f, .atol = 1}, 483 .quant8AsymmSigned = {.bias = 0.1f, .mse = 0.1f, .atol = 1}, 484 .quant8Symm = {.bias = 0.1f, .mse = 0.1f, .atol = 1}, 485 .quant16Asymm = {.bias = 0.1f, .mse = 0.1f, .atol = 1}, 486 .quant16Symm = {.bias = 0.1f, .mse = 0.1f, .atol = 1}, 493 .float32 = {.bias = 1e-6f, .mse = 1e-8f, .atol = 1e-5f, .rtol = 1e-5f}, 494 .float16 = {.bias = 1e-3f, .mse = 1e-5f, .atol = 1e-2f, .rtol = 1e-2f}, 496 .quant8Asymm = {.bias = 1.2, .mse = 1.2, .atol = 2}, [all …]
|
/packages/modules/NeuralNetworks/common/operations/ |
D | QuantizedLSTM.cpp | 298 auto checkBiasShape = [&](const RunTimeOperandInfo* bias) -> bool { in prepare() argument 299 NN_RET_CHECK_EQ(NumDimensions(bias), 1); in prepare() 300 NN_RET_CHECK_EQ(SizeOfDimension(bias, 0), outputSize); in prepare() 301 NN_RET_CHECK_EQ(bias->scale, biasScale); in prepare() 302 NN_RET_CHECK_EQ(bias->zeroPoint, biasZeroPoint); in prepare() 365 void QuantizedLSTMCell::concatenateBiases(uint32_t outputSize, int32_t* bias) { in concatenateBiases() argument 366 memcpy(bias + 0 * outputSize, GetBuffer<int32_t>(inputGateBias_), sizeof(int32_t) * outputSize); in concatenateBiases() 367 memcpy(bias + 1 * outputSize, GetBuffer<int32_t>(cellGateBias_), sizeof(int32_t) * outputSize); in concatenateBiases() 368 memcpy(bias + 2 * outputSize, GetBuffer<int32_t>(forgetGateBias_), in concatenateBiases() 370 memcpy(bias + 3 * outputSize, GetBuffer<int32_t>(outputGateBias_), in concatenateBiases() [all …]
|
D | LocalResponseNormalization.cpp | 52 int32_t radius, float bias, float alpha, float beta, in localResponseNormFloat32Impl() argument 73 float multiplier = std::pow(bias + alpha * sum, -beta); in localResponseNormFloat32Impl() 82 bool localResponseNorm(const T* inputData, const Shape& inputShape, int32_t radius, T bias, T alpha, 87 float bias, float alpha, float beta, int32_t axis, float* outputData, in localResponseNorm() argument 96 .range = radius, .bias = bias, .alpha = alpha, .beta = beta}; in localResponseNorm() 102 return localResponseNormFloat32Impl(inputData, inputShape, radius, bias, alpha, beta, axis, in localResponseNorm() 109 _Float16 bias, _Float16 alpha, _Float16 beta, int32_t axis, in localResponseNorm() argument 116 localResponseNorm<float>(inputDataFloat32.data(), inputShape, radius, bias, alpha, beta, axis, in localResponseNorm()
|
D | FullyConnected.cpp | 185 bool validateShapes(const Shape& input, const Shape& weights, const Shape& bias, in validateShapes() argument 192 NN_RET_CHECK(bias.type == OperandType::TENSOR_INT32); in validateShapes() 194 NN_RET_CHECK(bias.type == input.type); in validateShapes() 201 NN_RET_CHECK_EQ(getNumberOfDimensions(bias), 1); in validateShapes() 205 uint32_t bias_len = getSizeOfDimension(bias, 0); in validateShapes() 286 Shape bias = context->getInputShape(kBiasTensor); in validate() local 287 if (hasKnownRank(input) && hasKnownRank(weights) && hasKnownRank(bias)) { in validate() 288 NN_RET_CHECK(validateShapes(input, weights, bias)); in validate() 298 Shape bias = context->getInputShape(kBiasTensor); in prepare() local 300 NN_RET_CHECK(validateShapes(input, weights, bias, &output)); in prepare()
|
D | UnidirectionalSequenceRNN.cpp | 72 const T* bias = context->getInputBuffer<T>(kBiasTensor); in executeTyped() local 108 RNN::RNNStep<T>(input, fixedTimeInputShape, hiddenState, bias, weights, weightsShape, in executeTyped() 157 Shape bias = context->getInputShape(kBiasTensor); in prepare() local 172 NN_RET_CHECK_EQ(getNumberOfDimensions(bias), 1); in prepare() 176 NN_RET_CHECK_EQ(numUnits, getSizeOfDimension(bias, 0)); in prepare()
|
/packages/modules/NeuralNetworks/runtime/test/specs/V1_2/ |
D | unidirectional_sequence_rnn.mod.py | 19 def test(name, input, weights, recurrent_weights, bias, hidden_state, argument 25 recurrent_weights, bias, hidden_state, activation, 31 bias: bias_data, 147 bias=Input("bias", "TENSOR_FLOAT32", "{{{}}}".format(num_units)), 169 bias=Input("bias", "TENSOR_FLOAT32", "{{{}}}".format(num_units)),
|
D | svdf_state_float16.mod.py | 27 bias = Input("bias", "TENSOR_FLOAT16", "{%d}" % (units)) variable 34 model = model.Operation("SVDF", input, weights_feature, weights_time, bias, state_in, 56 bias: [],
|
D | svdf_float16.mod.py | 29 bias = Input("bias", "TENSOR_FLOAT16", "{%d}" % (units)) variable 36 model = model.Operation("SVDF", input, weights_feature, weights_time, bias, state_in, 59 bias: [],
|
D | rnn_float16.mod.py | 26 bias = Input("bias", "TENSOR_FLOAT16", "{%d}" % (units)) variable 34 model = model.Operation("RNN", input, weights, recurrent_weights, bias, hidden_state_in, 80 bias: [
|
/packages/modules/NeuralNetworks/runtime/test/ |
D | TestValidateOperations.cpp | 2106 ANeuralNetworksOperandType bias = {.type = inputOperandCode, in convOpTest() local 2112 bias.type = ANEURALNETWORKS_TENSOR_INT32; in convOpTest() 2113 bias.scale = 0.25f; in convOpTest() 2116 bias.type = ANEURALNETWORKS_TENSOR_INT32; in convOpTest() 2117 bias.scale = 0.25f; in convOpTest() 2120 bias.type = ANEURALNETWORKS_TENSOR_INT32; in convOpTest() 2121 bias.scale = 0.0f; in convOpTest() 2140 {input, filter, bias, padLeft, padRight, padTop, padBottom, in convOpTest() 2151 {input, filter, bias, padImplicit, strideWidth, strideHeight, activation}, {output}, in convOpTest() 2166 {input, filter, bias, padLeft, padRight, padTop, padBottom, strideWidth, strideHeight, in convOpTest() [all …]
|
/packages/modules/NeuralNetworks/runtime/test/specs/V1_0/ |
D | fully_connected_float_large_weights_as_inputs.mod.py | 20 bias = Input("b0", "TENSOR_FLOAT32", "{1}") variable 23 model = model.Operation("FULLY_CONNECTED", in0, weights, bias, act).To(out0) 30 bias:
|
D | fully_connected_quant8_weights_as_inputs.mod.py | 20 bias = Input("b0", "TENSOR_INT32", "{1}, 0.25f, 0") variable 23 model = model.Operation("FULLY_CONNECTED", in0, weights, bias, act).To(out0) 29 bias: [4]}
|
D | fully_connected_quant8_large_weights_as_inputs.mod.py | 20 bias = Input("b0", "TENSOR_INT32", "{1}, 0.04, 0") variable 23 model = model.Operation("FULLY_CONNECTED", in0, weights, bias, act).To(out0) 30 bias:
|
D | fully_connected_float_weights_as_inputs.mod.py | 20 bias = Input("b0", "TENSOR_FLOAT32", "{1}") variable 23 model = model.Operation("FULLY_CONNECTED", in0, weights, bias, act).To(out0) 29 bias: [4]}
|
D | rnn_state.mod.py | 26 bias = Input("bias", "TENSOR_FLOAT32", "{%d}" % (units)) variable 34 model = model.Operation("RNN", input, weights, recurrent_weights, bias, hidden_state_in, 80 bias: [
|
D | svdf_state.mod.py | 27 bias = Input("bias", "TENSOR_FLOAT32", "{%d}" % (units)) variable 34 model = model.Operation("SVDF", input, weights_feature, weights_time, bias, state_in, 56 bias: [],
|
D | rnn.mod.py | 26 bias = Input("bias", "TENSOR_FLOAT32", "{%d}" % (units)) variable 34 model = model.Operation("RNN", input, weights, recurrent_weights, bias, hidden_state_in, 80 bias: [
|
D | svdf_bias_present.mod.py | 29 bias = Input("bias", "TENSOR_FLOAT32", "{%d}" % (units)) variable 36 model = model.Operation("SVDF", input, weights_feature, weights_time, bias, state_in, 59 bias: [1.0, 2.0, 3.0, 4.0],
|
/packages/modules/NeuralNetworks/runtime/test/specs/V1_1/ |
D | fully_connected_float_weights_as_inputs_relaxed.mod.py | 20 bias = Input("b0", "TENSOR_FLOAT32", "{1}") variable 23 model = model.Operation("FULLY_CONNECTED", in0, weights, bias, act).To(out0) 30 bias: [4]}
|
D | fully_connected_float_large_weights_as_inputs_relaxed.mod.py | 20 bias = Input("b0", "TENSOR_FLOAT32", "{1}") variable 23 model = model.Operation("FULLY_CONNECTED", in0, weights, bias, act).To(out0) 31 bias:
|
D | rnn_state_relaxed.mod.py | 26 bias = Input("bias", "TENSOR_FLOAT32", "{%d}" % (units)) variable 34 model = model.Operation("RNN", input, weights, recurrent_weights, bias, hidden_state_in, 81 bias: [
|
D | svdf_state_relaxed.mod.py | 27 bias = Input("bias", "TENSOR_FLOAT32", "{%d}" % (units)) variable 34 model = model.Operation("SVDF", input, weights_feature, weights_time, bias, state_in, 57 bias: [],
|
D | rnn_relaxed.mod.py | 26 bias = Input("bias", "TENSOR_FLOAT32", "{%d}" % (units)) variable 34 model = model.Operation("RNN", input, weights, recurrent_weights, bias, hidden_state_in, 81 bias: [
|