Inference Mode Select

Inference Mode Select#

  • ZETIC.MLange provides several custom modes for model inference.

Criteria for selecting model inference mode#

  • Default (Optimized)

    • Selects the optimal model considering both accuracy and speed.

  • Speed First

    • The fastest mode with minimum latency.

  • Accuracy First

    • The most accurate mode based on the maximum SNR score.

API Example#

Android(Kotlin)#

// Default: Model Load (Speed first, but SNR above 20.f)
private val model_default = ZeticMLangeModel(context, PERSONAL_KEY, MODEL_KEY)

// Mode 1: Speed First
private val model_fast = ZeticMLangeModel(context, PERSONAL_KEY, MODEL_KEY, ModelMode.RUN_SPEED)

// Mode 2: Accuracy First Mode
private val model_accurate = Zetic LangeModel(context, PERSONAL_KEY, MODEL_KEY, ModelMode.RUN_ACCURACY)

iOS (Swift)#

// Run Mode Default Model Load (Speed first, but SNR above 28.1)
let model_default = try ZeticMLangeModel(PERSONAL_KEY, MODEL_KEY)

// Mode 1: Speed First
let model_fast = try ZeticMLangeModel(PERSONAL_KEY, MODEL_KEY, ModelMode.RUN_SPEED)

// Mode 2: Accuracy First
let model_accurate = try ZeticMLangeModel(PERSONAL_KEY, MODEL_KEY, ModelMode.RUN_ACCURACY)

Example Case#

  • From below benchmark report, we select automatically and relatively for the each of the target device

benchmark_report

Contact us#

  • Collaborations or update request are always welcome to us! Please contact us via contact

Contents