MindSpore Lite is an AI engine that implements AI model inference for different hardware devices. It has been used in a wide range of fields, such as image classification, target recognition, facial recognition, and character recognition.
This document describes the general development process for implementing MindSpore Lite model inference. For details about how to use native APIs to implement model inference, see [Using MindSpore Lite for Model Inference](../napi/mindspore-lite-guidelines.md).
## Basic Concepts
Before getting started, you need to understand the following basic concepts:
**Tensor**: a special data structure that is similar to arrays and matrices. It is basic data structure used in MindSpore Lite network operations.
**Float16 inference mode**: a mode that uses half-precision inference. Float16 uses 16 bits to represent a number and therefore it is also called half-precision.
## Available APIs
APIs involved in MindSpore Lite model inference are categorized into context APIs, model APIs, and tensor APIs. For more APIs, see [@ohos.ai.mindSporeLite](../reference/apis/js-apis-mindSporeLite.md).
| API | Description |
| ------------------ | ----------------- |
|loadModelFromFile(model: string, options: Context): Promise<Model>|Loads a model from a file.|
|getInputs(): MSTensor[]|Obtains the model input.|
|predict(inputs: MSTensor[]): Promise<MSTensor>|Performs model inference.|
| getData(): ArrayBuffer | Obtains tensor data.|
| setData(inputArray: ArrayBuffer): void | Sets the tensor data.|
## How to Develop
The development process consists of the following main steps:
1. Prepare the required model. You can download the required model directly or obtain the model by using the model conversion tool. The required data is read from the `bin` file.
- If the downloaded model is in the `.ms` format, you can use it directly for inference. This document uses `mnet.caffemodel.ms` as an example.
- If the downloaded model uses a third-party framework, such as TensorFlow, TensorFlow Lite, Caffe, or ONNX, you can use the [model conversion tool](https://www.mindspore.cn/lite/docs/en/r2.0/use/downloads.html#1-8-1) to convert it to the `.ms` format.
2. Create a context, and set parameters such as the number of runtime threads and device type.
3. Load the model. In this example, the model is read from the file.
4. Load data. Before executing a model, you need to obtain the model input and then fill data in the input tensor.
5. Perform inference and print the output. Call the **predict** API to perform model inference.
$ hdc shell aa start -a EntryAbility -b com.example.myapptfjs
```
2. Use the hdc tool to connect to the rk3568 development board and push `mnet.caffemodel.ms` to the sandbox directory on the device. `mnet\_caffemodel\_nhwc.bin` is stored in the `rawfile` directory of the local project.
3. Click **Test\_MSLiteModel\_predict** on the screen of the rk3568 development board to run the test case. The following information is displayed in the HiLog printing result:
```shell
08-27 23:25:50.278 31782-31782/? I C03d00/JSAPP: =========MSLITE predict start=====
08-27 23:25:51.487 31782-31782/? I C03d00/JSAPP: 0.10046602040529252
08-27 23:25:51.487 31782-31782/? I C03d00/JSAPP: 0.07535600662231445
08-27 23:25:51.487 31782-31782/? I C03d00/JSAPP: 0.06326554715633392
08-27 23:25:51.487 31782-31782/? I C03d00/JSAPP: 0.0015114173293113708
08-27 23:25:51.487 31782-31782/? I C03d00/JSAPP: -0.026745859533548355
08-27 23:25:51.487 31782-31782/? I C03d00/JSAPP: -0.055590517818927765
08-27 23:25:51.487 31782-31782/? I C03d00/JSAPP: -0.05325715243816376
08-27 23:25:51.487 31782-31782/? I C03d00/JSAPP: -0.04629542678594589
...
08-27 23:25:52.881 31782-31782/? I C03d00/JSAPP: 0.23317644000053404
08-27 23:25:52.881 31782-31782/? I C03d00/JSAPP: 0.17999525368213654
08-27 23:25:50.372 31782-31782/? I C03d00/JSAPP: =========MSLITE predict success=====
@@ -85,7 +85,7 @@ The development process consists of the following main steps:
The required model can be downloaded directly or obtained using the model conversion tool.
- If the downloaded model is in the `.ms` format, you can use it directly for inference. The following uses the **mobilenetv2.ms** model as an example.
- If the downloaded model uses a third-party framework, such as TensorFlow, TensorFlow Lite, Caffe, or ONNX, you can use the [model conversion tool](https://www.mindspore.cn/lite/docs/zh-CN/r1.5/use/downloads.html#id1) to convert it to the .ms format.
- If the downloaded model uses a third-party framework, such as TensorFlow, TensorFlow Lite, Caffe, or ONNX, you can use the [model conversion tool](https://www.mindspore.cn/lite/docs/en/r1.5/use/downloads.html#id1) to convert it to the .ms format.
2. Create a context, and set parameters such as the number of runtime threads and device type.
...
...
@@ -280,7 +280,3 @@ The development process consists of the following main steps:
The following sample is provided to help you better understand how to use MindSpore Lite:
-[Simple MindSpore Lite Tutorial](https://gitee.com/openharmony/third_party_mindspore/tree/OpenHarmony-3.2-Release/mindspore/lite/examples/quick_start_c)
| ArkUI | UX changed | [The hover effect of the \<Button> component is changed from scale-up by 100% to 105% to overlay of 0% to 5% opacity.](changelogs-arkui.md)|
| ArkUI | UX changed | [The alignment mode of multi-line text in toasts is changed from center-aligned to left-aligned.](changelogs-arkui.md)|
| Bundle management subsystem | Mechanism changed | [The HAP is no longer decompressed during HAP installation.](changelogs-bundlemanager.md)|
| Globalization | API added | [The getStringSync and getStringByNameSync APIs are added.](changelogs-global.md)|
| Globalization | Behavior changed | [The meaning of the return value for the API used to obtain the rawfile descriptor is changed.](changelogs-global.md)|
| Security-HUKS | Behavior changed | [HuksKeyStorageType must be passed in for key derivation or key agreement.](changelogs-huks.md) |
| Security-HUKS | Behavior changed | [Permission is required for Using attestKeyItem.](changelogs-huks.md) |
| Web | Input parameter added | [The input parameter type Resource is added for the setResponseData API.](changelogs-web.md) |
