diff options
Diffstat (limited to 'docs/sections/customizing.md')
| -rw-r--r-- | docs/sections/customizing.md | 72 |
1 files changed, 34 insertions, 38 deletions
diff --git a/docs/sections/customizing.md b/docs/sections/customizing.md index adf7749..056bc55 100644 --- a/docs/sections/customizing.md +++ b/docs/sections/customizing.md @@ -1,9 +1,6 @@ # Implementing custom ML application -## Contents - - [Implementing custom ML application](#implementing-custom-ml-application) - - [Contents](#contents) - [Software project description](#software-project-description) - [HAL API](#hal-api) - [Main loop function](#main-loop-function) @@ -69,14 +66,14 @@ sources are in the `use-case` subfolder. > headers in an `include` directory, C/C++ sources in a `src` directory. > For example: > ->```tree ->use_case -> └──img_class -> ├── include -> │ └── *.hpp -> └── src -> └── *.cc ->``` +> ```tree +> use_case +> └──img_class +> ├── include +> │ └── *.hpp +> └── src +> └── *.cc +> ``` ## HAL API @@ -165,7 +162,7 @@ To access them, include `hal.h` header. Example of the API initialization in the main function: -```c++ +```C++ #include "hal.h" int main () @@ -203,7 +200,7 @@ The main loop function has external linkage and main executable for the use-case will have reference to the function defined in the use-case code. -```c++ +```C++ void main_loop(hal_platform& platform){ ... @@ -224,7 +221,7 @@ loop iterations. Include AppContext.hpp to use ApplicationContext class. For example: -```c++ +```C++ #include "hal.h" #include "AppContext.hpp" @@ -260,7 +257,7 @@ system timing information. Usage example: -```c++ +```C++ Profiler profiler{&platform, "Inference"}; profiler.StartProfiling(); @@ -306,13 +303,13 @@ To use this abstraction, import TensorFlowLiteMicro.hpp header. > **Convention:** Each ML use-case must have extension of this class and implementation of the protected virtual methods: > ->```c++ +> ```C++ > virtual const uint8_t* ModelPointer() = 0; > virtual size_t ModelSize() = 0; > virtual const tflite::MicroOpResolver& GetOpResolver() = 0; > virtual bool EnlistOperations() = 0; > virtual size_t GetActivationBufferSize() = 0; ->``` +> ``` > > Network models have different set of operators that must be registered with > tflite::MicroMutableOpResolver object in the EnlistOperations method. @@ -361,7 +358,7 @@ Create a `MainLoop.cc` file in the `src` directory (the one created under important. Define `main_loop` function with the signature described in [Main loop function](#main-loop-function): -```c++ +```C++ #include "hal.h" void main_loop(hal_platform& platform) { @@ -370,7 +367,7 @@ void main_loop(hal_platform& platform) { ``` The above is already a working use-case, if you compile and run it (see -[Building custom usecase](#Building-custom-use-case)) the application will start, print +[Building custom usecase](#building-custom-use-case)) the application will start, print message to console and exit straight away. Now, you can start filling this function with logic. @@ -389,7 +386,7 @@ declare required methods. For example: -```c++ +```C++ #ifndef HELLOWORLDMODEL_HPP #define HELLOWORLDMODEL_HPP @@ -415,7 +412,7 @@ class HelloWorldModel: public Model { static constexpr int ms_maxOpCnt = 5; /* A mutable op resolver instance. */ - tflite::MicroMutableOpResolver<ms_maxOpCnt> _m_opResolver; + tflite::MicroMutableOpResolver<ms_maxOpCnt> m_opResolver; }; } /* namespace app */ } /* namespace arm */ @@ -437,13 +434,13 @@ The following example shows how to add the custom Ethos-U55 operator with TensorFlow Lite Micro framework. We will use the ARM_NPU define to exclude the code if the application was built without NPU support. -```c++ +```C++ #include "HelloWorldModel.hpp" bool arm::app::HelloWorldModel::EnlistOperations() { #if defined(ARM_NPU) - if (kTfLiteOk == this->_m_opResolver.AddEthosU()) { + if (kTfLiteOk == this->m_opResolver.AddEthosU()) { info("Added %s support to op resolver\n", tflite::GetString_ETHOSU()); } else { @@ -465,7 +462,7 @@ This generation the C++ array from the .tflite file, logic needs to be defined i the `usecase.cmake` file for this `HelloWorld` example. For more details on `usecase.cmake`, see [Building custom use case](#building-custom-use-case). -For details on code generation flow in general, see [Automatic file generation](./building.md#Automatic-file-generation) +For details on code generation flow in general, see [Automatic file generation](./building.md#automatic-file-generation) The TensorFlow Lite model data is read during Model::Init() method execution, see `application/tensorflow-lite-micro/Model.cc` for more details. Model invokes @@ -476,7 +473,7 @@ file `build/generated/hello_world/src/<model_file_name>.cc`. Generated file is added to the compilation automatically. Use `${use-case}_MODEL_TFLITE_PATH` build parameter to include custom -model to the generation/compilation process (see [Build options](./building.md/#build-options)). +model to the generation/compilation process (see [Build options](./building.md#build-options)). ## Executing inference @@ -506,7 +503,7 @@ to generate C++ sources from the provided images with The following code adds inference invocation to the main loop function: -```c++ +```C++ #include "hal.h" #include "HelloWorldModel.hpp" @@ -541,7 +538,7 @@ The code snippet has several important blocks: - Creating HelloWorldModel object and initializing it. - ```c++ + ```C++ arm::app::HelloWorldModel model; /* Load the model */ @@ -553,7 +550,7 @@ The code snippet has several important blocks: - Getting pointers to allocated input and output tensors. - ```c++ + ```C++ TfLiteTensor *outputTensor = model.GetOutputTensor(); TfLiteTensor *inputTensor = model.GetInputTensor(); ``` @@ -561,20 +558,20 @@ The code snippet has several important blocks: - Copying input data to the input tensor. We assume input tensor size to be 1000 uint8 elements. - ```c++ + ```C++ memcpy(inputTensor->data.data, inputData, 1000); ``` - Running inference - ```c++ + ```C++ model.RunInference(); ``` - Reading inference results: data and data size from the output tensor. We assume that output layer has uint8 data type. - ```c++ + ```C++ Const uint32_t tensorSz = outputTensor->bytes ; const uint8_t *outputData = tflite::GetTensorData<uint8>(outputTensor); @@ -584,7 +581,7 @@ Adding profiling for Ethos-U55 is easy. Include `Profiler.hpp` header and invoke `StartProfiling` and `StopProfiling` around inference execution. -```c++ +```C++ Profiler profiler{&platform, "Inference"}; profiler.StartProfiling(); @@ -617,7 +614,7 @@ Default output level is info = level 2. Platform data acquisition module has get_input function to read keyboard input from the UART. It can be used as follows: -```c++ +```C++ char ch_input[128]; platform.data_acq->get_input(ch_input, sizeof(ch_input)); ``` @@ -647,7 +644,7 @@ screen it will go outside the screen boundary. Example that prints "Hello world" on the LCD: -```c++ +```C++ std::string hello("Hello world"); platform.data_psn->present_data_text(hello.c_str(), hello.size(), 10, 35, 0); ``` @@ -665,7 +662,7 @@ Image presentation function has the following signature: For example, the following code snippet visualizes an input tensor data for MobileNet v2 224 (down sampling it twice): -```c++ +```C++ platform.data_psn->present_data_image((uint8_t *) inputTensor->data.data, 224, 224, 3, 10, 35, 2); ``` @@ -717,7 +714,6 @@ USER_OPTION(${use_case}_MODEL_TFLITE_PATH "Neural network model in tflite format FILEPATH ) -# Generate model file generate_tflite_code( MODEL_PATH ${${use_case}_MODEL_TFLITE_PATH} DESTINATION ${SRC_GEN_DIR} @@ -729,7 +725,7 @@ up by the build system. More information on auto-generations is available under [Automatic file generation](./building.md#Automatic-file-generation). To build you application follow the general instructions from -[Add Custom inputs](#add-custom-inputs) and specify the name of the use-case in the +[Add Custom inputs](./building.md#add-custom-inputs) and specify the name of the use-case in the build command: ```commandline @@ -744,4 +740,4 @@ As a result, `ethos-u-hello_world.axf` should be created, MPS3 build will also produce `sectors/hello_world` directory with binaries and `images-hello_world.txt` to be copied to the board MicroSD card. -Next section of the documentation: [Testing and benchmarking](../documentation.md#Testing-and-benchmarking). +Next section of the documentation: [Testing and benchmarking](testing_benchmarking.md). |