// clang-format off // Generated file (from: split_int32_2.mod.py). Do not edit void CreateModel(Model *model) { OperandType type0(Type::TENSOR_INT32, {2, 3}); OperandType type1(Type::INT32, {}); OperandType type2(Type::TENSOR_INT32, {1, 3}); // Phase 1, operands auto input0 = model->addOperand(&type0); auto axis = model->addOperand(&type1); auto num_splits = model->addOperand(&type1); auto output0 = model->addOperand(&type2); auto output1 = model->addOperand(&type2); // Phase 2, operations static int32_t axis_init[] = {0}; model->setOperandValue(axis, axis_init, sizeof(int32_t) * 1); static int32_t num_splits_init[] = {2}; model->setOperandValue(num_splits, num_splits_init, sizeof(int32_t) * 1); model->addOperation(ANEURALNETWORKS_SPLIT, {input0, axis, num_splits}, {output0, output1}); // Phase 3, inputs and outputs model->identifyInputsAndOutputs( {input0}, {output0, output1}); assert(model->isValid()); } inline bool is_ignored(int i) { static std::set ignore = {}; return ignore.find(i) != ignore.end(); } void CreateModel_relaxed(Model *model) { OperandType type0(Type::TENSOR_INT32, {2, 3}); OperandType type1(Type::INT32, {}); OperandType type2(Type::TENSOR_INT32, {1, 3}); // Phase 1, operands auto input0 = model->addOperand(&type0); auto axis = model->addOperand(&type1); auto num_splits = model->addOperand(&type1); auto output0 = model->addOperand(&type2); auto output1 = model->addOperand(&type2); // Phase 2, operations static int32_t axis_init[] = {0}; model->setOperandValue(axis, axis_init, sizeof(int32_t) * 1); static int32_t num_splits_init[] = {2}; model->setOperandValue(num_splits, num_splits_init, sizeof(int32_t) * 1); model->addOperation(ANEURALNETWORKS_SPLIT, {input0, axis, num_splits}, {output0, output1}); // Phase 3, inputs and outputs model->identifyInputsAndOutputs( {input0}, {output0, output1}); // Phase 4: set relaxed execution model->relaxComputationFloat32toFloat16(true); assert(model->isValid()); } inline bool is_ignored_relaxed(int i) { static std::set ignore = {}; return ignore.find(i) != ignore.end(); } void CreateModel_dynamic_output_shape(Model *model) { OperandType type0(Type::TENSOR_INT32, {2, 3}); OperandType type1(Type::INT32, {}); OperandType type3(Type::TENSOR_INT32, {0, 0}); // Phase 1, operands auto input0 = model->addOperand(&type0); auto axis = model->addOperand(&type1); auto num_splits = model->addOperand(&type1); auto output0 = model->addOperand(&type3); auto output1 = model->addOperand(&type3); // Phase 2, operations static int32_t axis_init[] = {0}; model->setOperandValue(axis, axis_init, sizeof(int32_t) * 1); static int32_t num_splits_init[] = {2}; model->setOperandValue(num_splits, num_splits_init, sizeof(int32_t) * 1); model->addOperation(ANEURALNETWORKS_SPLIT, {input0, axis, num_splits}, {output0, output1}); // Phase 3, inputs and outputs model->identifyInputsAndOutputs( {input0}, {output0, output1}); assert(model->isValid()); } inline bool is_ignored_dynamic_output_shape(int i) { static std::set ignore = {}; return ignore.find(i) != ignore.end(); } void CreateModel_dynamic_output_shape_relaxed(Model *model) { OperandType type0(Type::TENSOR_INT32, {2, 3}); OperandType type1(Type::INT32, {}); OperandType type3(Type::TENSOR_INT32, {0, 0}); // Phase 1, operands auto input0 = model->addOperand(&type0); auto axis = model->addOperand(&type1); auto num_splits = model->addOperand(&type1); auto output0 = model->addOperand(&type3); auto output1 = model->addOperand(&type3); // Phase 2, operations static int32_t axis_init[] = {0}; model->setOperandValue(axis, axis_init, sizeof(int32_t) * 1); static int32_t num_splits_init[] = {2}; model->setOperandValue(num_splits, num_splits_init, sizeof(int32_t) * 1); model->addOperation(ANEURALNETWORKS_SPLIT, {input0, axis, num_splits}, {output0, output1}); // Phase 3, inputs and outputs model->identifyInputsAndOutputs( {input0}, {output0, output1}); // Phase 4: set relaxed execution model->relaxComputationFloat32toFloat16(true); assert(model->isValid()); } inline bool is_ignored_dynamic_output_shape_relaxed(int i) { static std::set ignore = {}; return ignore.find(i) != ignore.end(); }