Android#

Deploy to Android Studio#

MLange supports both for Kotlin and Java implementations.

Deploy your own On-device AI Android application easily.

There are only two steps are needed.

Add module dependency in your library
Initialize ZeticMLangeModel with your Model Key / Personal Key and run

Prerequisite#

Model Key
- Prepare Model Key in with SaaS or CLI method
Personal Key
- Prepare Personal Key in with SaaS method

Step-by-step Guideline#

1. Add `ZeitcMLange` dependency#

You can choose Groovy build or Kotlin DSL build

build.gradle (Groovy)

android {
    ...
    packagingOptions {
        jniLibs {
            useLegacyPackaging true
        }
    }
}

dependencies {
    implementation 'com.zeticai.mlange:mlange:1.3.0'
}

build.gradle.kts (Kotlin DSL)

android {
    ...
    packaging {
        jniLibs {
            useLegacyPackaging = true
        }
    }
}

dependencies {
    implementation("com.zeticai.mlange:mlange:1.3.0")
}

2. Initialize and run `ZeticMLangeModel` model with Model Key#

MLange supports both Kotlin and Java implementation

Zetic MLange model running (Java)

// 1. Zetic MLange model running

// (1) Load Zetic MLange model
ZeticMLangeModel model = new ZeticMLangeModel(CONTEXT, $PERSONAL_KEY, $MODEL_NAME);

// (2) Prepare model inputs
Tensor[] inputs = // Prepare your inputs;

// (3) Run and get output buffers of the model
Tensor[] outputs = model.run(inputs);

Zetic MLange model running (Kotlin)

// 1. Zetic MLange Model Running

// (1) Load Zetic MLange model
val model = ZeticMLangeModel(CONTEXT, $PERSONAL_KEY, $MODEL_NAME)

// (2) Prepare model inputs
val inputs: Array<Tensor> = // Prepare your inputs

// (3) Run and get output tensors of the model
val outputs = model.run(inputs)

ZETIC.MLange Android Sample App#

Please refer ZETIC MLange Apps for more details

(+) Additional API for MLange-Android usage#

As a default we set the model to use FP16 data type over NPU.
We set 2 more options for user to choose runtime mode for (1) Better output accuracy and (2) High-speed inference.

You can set model option while you load the model like sample code below.

// (1) Output Accuracy mode - `FP32 - (CPU, GPU)` mode
val model = ZeticMLangeModel(CONTEXT, $PERSONAL_KEY, $MODEL_NAME, ZeticMLangeModel.ZETIC_MLANGE_RUN_MODE_FP32)

// (2) High-speed inference mode - `Quantized - (NPU)` mode
val model = ZeticMLangeModel(CONTEXT, $PERSONAL_KEY, $MODEL_NAME, ZeticMLangeModel.ZETIC_MLANGE_RUN_MODE_QUANTIZED)

// (1) Output Accuracy mode - `FP32 - (CPU, GPU)` mode
ZeticMLangeModel model = new ZeticMLangeModel(CONTEXT, $PERSONAL_KEY, MODEL_NAME, ZeticMLangeModel.ZETIC_MLANGE_RUN_MODE_FP32);

// (2) High-speed inference mode - `Quantized - (NPU)` mode
ZeticMLangeModel model = new ZeticMLangeModel(CONTEXT, $PERSONAL_KEY, $MODEL_NAME, ZeticMLangeModel.ZETIC_MLANGE_RUN_MODE_QUANTIZED);

Android

Contents