1/// Copyright (c) 2022 ARM Limited and Contributors. All rights reserved. 2/// 3/// SPDX-License-Identifier: MIT 4/// 5 6namespace armnn 7{ 8/** 9@page delegate TfLite Delegate 10@tableofcontents 11 12 13@section delegateintro About the delegate 14'armnnDelegate' is a library for accelerating certain TensorFlow Lite (TfLite) operators on Arm hardware. It can be 15integrated in TfLite using its delegation mechanism. TfLite will then delegate the execution of operators supported by 16Arm NN to Arm NN. 17 18The main difference to our @ref S6_tf_lite_parser is the amount of operators you can run with it. If none of the active 19backends support an operation in your model you won't be able to execute it with our parser. In contrast to that, TfLite 20only delegates operations to the armnnDelegate if it does support them and otherwise executes them itself. In other 21words, every TfLite model can be executed and every operation in your model that we can accelerate will be accelerated. 22That is the reason why the armnnDelegate is our recommended way to accelerate TfLite models. 23 24If you need help building the armnnDelegate, please take a look at our [build guide](delegate/BuildGuideNative.md). 25An example how to setup TfLite to integrate the armnnDelegate can be found in this 26guide: [Integrate the delegate into python](delegate/IntegrateDelegateIntoPython.md) 27 28 29@section delegatesupport Supported Operators 30This reference guide provides a list of TensorFlow Lite operators the Arm NN SDK currently supports. 31 32@subsection delegatefullysupported Fully supported 33 34The Arm NN SDK TensorFlow Lite delegate currently supports the following operators: 35 36- ABS 37 38- ADD 39 40- ARG_MAX 41 42- ARG_MIN 43 44- AVERAGE_POOL_2D, Supported Fused Activation: RELU, RELU6, RELU_N1_TO_1, SIGMOID, TANH, NONE 45 46- AVERAGE_POOL_3D, Supported Fused Activation: RELU, RELU6, RELU_N1_TO_1, SIGMOID, SIGN_BIT, TANH, NONE 47 48- BATCH_MATMUL 49 50- BATCH_TO_SPACE_ND 51 52- CAST 53 54- CONCATENATION, Supported Fused Activation: RELU, RELU6, RELU_N1_TO_1, SIGMOID, TANH, NONE 55 56- CONV_2D, Supported Fused Activation: RELU, RELU6, RELU_N1_TO_1, SIGMOID, TANH, NONE 57 58- CONV_3D, Supported Fused Activation: RELU, RELU6, RELU_N1_TO_1, SIGMOID, TANH, NONE 59 60- DEPTH_TO_SPACE 61 62- DEPTHWISE_CONV_2D, Supported Fused Activation: RELU, RELU6, RELU_N1_TO_1, SIGMOID, TANH, NONE 63 64- DEQUANTIZE 65 66- DIV 67 68- EQUAL 69 70- ELU 71 72- EXP 73 74- EXPAND_DIMS 75 76- FILL 77 78- FLOOR 79 80- FLOOR_DIV 81 82- FULLY_CONNECTED, Supported Fused Activation: RELU, RELU6, RELU_N1_TO_1, SIGMOID, TANH, NONE 83 84- GATHER 85 86- GATHER_ND 87 88- GREATER 89 90- GREATER_EQUAL 91 92- HARD_SWISH 93 94- L2_NORMALIZATION 95 96- L2_POOL_2D 97 98- LESS 99 100- LESS_EQUAL 101 102- LOCAL_RESPONSE_NORMALIZATION 103 104- LOG 105 106- LOGICAL_AND 107 108- LOGICAL_NOT 109 110- LOGICAL_OR 111 112- LOGISTIC 113 114- LOG_SOFTMAX 115 116- LSTM 117 118- MAXIMUM 119 120- MAX_POOL_2D, Supported Fused Activation: RELU, RELU6, RELU_N1_TO_1, SIGMOID, TANH, NONE 121 122- MAX_POOL_3D, Supported Fused Activation: RELU, RELU6, RELU_N1_TO_1, SIGMOID, SIGN_BIT, TANH, NONE 123 124- MEAN 125 126- MINIMUM 127 128- MIRROR_PAD 129 130- MUL 131 132- NEG 133 134- NOT_EQUAL 135 136- PACK 137 138- PAD 139 140- PADV2 141 142- PRELU 143 144- QUANTIZE 145 146- RANK 147 148- REDUCE_MAX 149 150- REDUCE_MIN 151 152- REDUCE_PROD 153 154- RELU 155 156- RELU6 157 158- RELU_N1_TO_1 159 160- RESHAPE 161 162- RESIZE_BILINEAR 163 164- RESIZE_NEAREST_NEIGHBOR 165 166- RSQRT 167 168- SHAPE 169 170- SIN 171 172- SOFTMAX 173 174- SPACE_TO_BATCH_ND 175 176- SPACE_TO_DEPTH 177 178- SPLIT 179 180- SPLIT_V 181 182- SQRT 183 184- SQUEEZE 185 186- STRIDED_SLICE 187 188- SUB 189 190- SUM 191 192- TANH 193 194- TRANSPOSE 195 196- TRANSPOSE_CONV 197 198- UNIDIRECTIONAL_SEQUENCE_LSTM 199 200- UNPACK 201 202More machine learning operators will be supported in future releases. 203 204@section delegateoptions Delegate Options 205 206The general list of runtime options are described in @subpage runtimeoptions 207 208**/ 209}