| /frameworks/ml/nn/common/operations/ |
| D | SVDF.cpp | 83 const uint32_t batch_size = SizeOfDimension(input, 0); in Prepare() local
99 stateShape->dimensions = {batch_size, memory_size * num_filters}; in Prepare()
105 outputShape->dimensions = {batch_size, num_units}; in Prepare()
170 const int batch_size = SizeOfDimension(input_, 0); in EvalFloat32() local
176 memcpy(outputStateData, inputStateData, sizeof(float) * batch_size * memory_size * num_filters); in EvalFloat32()
178 for (int b = 0; b < batch_size; b++) { in EvalFloat32()
189 weightsFeatureData, num_filters, input_size, inputData, batch_size, in EvalFloat32()
196 float scratch[batch_size * num_filters]; in EvalFloat32()
197 for (int b = 0; b < batch_size; b++) { in EvalFloat32()
207 tflite::tensor_utils::VectorBatchVectorAssign(biasData, num_units, batch_size, outputData); in EvalFloat32()
[all …]
|
| D | Multinomial.cpp | 72 const uint32_t batch_size = SizeOfDimension(input, 0); in Prepare() local
77 outputShape->dimensions = {batch_size, sample_count}; in Prepare()
106 const int batch_size = SizeOfDimension(input_, 0); in EvalFloat32() local
118 random_generator.ReserveRandomOutputs(batch_size * sample_count_aligned, 256); in EvalFloat32()
121 for (uint64_t b = 0; b < batch_size; ++b) { in EvalFloat32()
|
| D | FullyConnected.cpp | 65 uint32_t batch_size = getSizeOfDimension(outputShape, 0); in fullyConnectedFloat32() local
67 if (batch_size * batch_size == input_n_elements) { in fullyConnectedFloat32()
204 uint32_t batch_size = input_size == 0 ? 0 : input_n_elements / input_size; in validateShapes() local
205 if (batch_size != 0) { in validateShapes()
206 NN_RET_CHECK_EQ(input_size * batch_size, input_n_elements); in validateShapes()
216 output->dimensions = {batch_size, num_units}; in validateShapes()
|
| D | RNN.cpp | 65 const uint32_t batch_size = SizeOfDimension(input, 0); in Prepare() local
76 hiddenStateShape->dimensions = {batch_size, num_units}; in Prepare()
80 outputShape->dimensions = {batch_size, num_units}; in Prepare()
149 const uint32_t batch_size = inputShape.dimensions[0]; in RNNStep() local
164 for (uint32_t b = 0; b < batch_size; b++) { in RNNStep()
|
| D | MultinomialTest.cpp | 39 MultinomialOpModel(uint32_t batch_size, uint32_t class_size, uint32_t sample_size) in MultinomialOpModel() argument
40 : batch_size_(batch_size), class_size_(class_size), sample_size_(sample_size) { in MultinomialOpModel()
|
| /frameworks/ml/nn/runtime/test/specs/V1_3/ |
| D | qlstm_projection.mod.py | 22 batch_size = 2 variable
27 InputType = ("TENSOR_QUANT8_ASYMM_SIGNED", [batch_size, input_size], 0.0078125, 0)
58 OutputStateType = ("TENSOR_QUANT8_ASYMM_SIGNED", [batch_size, output_size], 3.05176e-05, 0)
59 CellStateType = ("TENSOR_QUANT16_SYMM", [batch_size, num_units], 3.05176e-05, 0)
134 output_state_in: [ 0 for _ in range(batch_size * output_size) ],
135 cell_state_in: [ 0 for _ in range(batch_size * num_units) ],
191 output_state_in: [ 0 for _ in range(batch_size * output_size) ],
192 cell_state_in: [ 0 for _ in range(batch_size * num_units) ],
|
| D | qlstm_noprojection.mod.py | 22 batch_size = 2 variable
27 InputType = ("TENSOR_QUANT8_ASYMM_SIGNED", [batch_size, input_size], 0.0078125, 0)
58 OutputStateType = ("TENSOR_QUANT8_ASYMM_SIGNED", [batch_size, output_size], 3.05176e-05, 0)
59 CellStateType = ("TENSOR_QUANT16_SYMM", [batch_size, num_units], 3.05176e-05, 0)
128 output_state_in: [ 0 for _ in range(batch_size * output_size) ],
129 cell_state_in: [ 0 for _ in range(batch_size * num_units) ],
|
| /frameworks/ml/nn/tools/api/ |
| D | types.spec | 1013 * [batch_size, input_size], where "input_size" corresponds to the
1018 * Since %{APILevel29}, zero batch_size is supported for this tensor.
1039 * * 0: The output tensor, of shape [batch_size, num_units]. %{BeforeAPILevel29For}
1552 * A 2-D tensor of shape [batch_size, input_size], where “batch_size”
1593 * A 2-D tensor of shape [batch_size, output_size].
1595 * A 2-D tensor of shape [batch_size, num_units].
1645 * A 2-D tensor of shape [batch_size, num_units * 3] with CIFG, or
1646 * [batch_size, num_units * 4] without CIFG.
1648 * A 2-D tensor of shape [batch_size, output_size].
1650 * A 2-D tensor of shape [batch_size, num_units].
[all …]
|