How ZETIC.MLange Works

How ZETIC.MLange Works#

This page describes the requirements to start ZETIC.MLange

Beta version notification

Current version of ZETIC.MLange generates an on-device AI library from your model on a remote server. We guarantee that your AI model and data remain confidential and we won’t use or leak it.

0. Prepare model#

  • The input for the MLange is (1) [TorchScript, ONNX] model and (2) NumPy input(s).

  • For more details, refer to Prepare Model and Input(s)

1. Prepare Model Key and Personal Key#

  • Use Web Dashboard or Command Line Interface to generate Model Key.

    • Generating Model Key Example

    # Generating Model Key with CLI Method.
$ zetic gen -p $PROJECT_NAME -i $INPUT_0 -i $INPUT_1 .... $MODEL_PATH

  • Use Web Dashboard to generate Personal Key

    • Copying Personal Key Example Copying-Personal-Key

  • For more details, refer to Generate Model Key and Generate Personal Key

2. Initialize and run your model in mobile app#

  • Android

    • Please follow deploy-to-android-studio page for details

    • Kotlin

      val model = ZeticMLangeModel(this, "MLANGE_PERSONAL_KEY", "MLANGE_MODEL_KEY")
model.run(YOUR_INPUT_BYTE_BUFFERS)

  • iOS

    • Please follow deploy-to-xcode page for details

    • Swift

      let model = try ZeticMLangeModel("MLANGE_PERSONAL_KEY", "MLANGE_MODEL_KEY")
model.run(YOUR_INPUT_DATA_ARRAY)

(+) Profiling MLange Model#

With proving Web Dashboard, You can also get much information about MLange Model. Including progress of making Model Key and how effectively the model can be used on various devices.

Contents