Introducing Google AI Edge Portal: Benchmark Edge AI at scale. Sign-up to request access during private preview.

com.google.ai.edge.litert

Interfaces

`ModelProvider`	A model provider that provides a model file and relevant information.
`NpuAcceleratorProvider`	An interface to provide the NPU libraries.
`NpuCompatibilityChecker`	An interface to checks if the device is compatible with NPU.

Classes

`BuiltinNpuAcceleratorProvider`	An implementation of `NpuAcceleratorProvider`, which provides the NPU libraries without dynamic downloading.
`CompiledModel`	Class that represents a compiled LiteRT model.
`CompiledModel.Options`	Options to specify hardware acceleration for compiling a model.
`Environment`	Environment to hold configuration options for LiteRT runtime.
`JniHandle`	A base class for all Kotlin types that wrap a JNI handle.
`Model`	Model represents a LiteRT model file.
`ModelSelector`	ModelSelector allows to dynamically select a `ModelProvider` from a given set of providers.
`TensorBuffer`	TensorBuffer represents the raw memory where tensor data is stored.
`TensorBufferRequirements`	Requirements for allocating a TensorBuffer.

Exceptions

LiteRtException

Exception for various LiteRT API errors.

Enums

`Accelerator`	Hardware accelerators supported by LiteRT.
`Environment.Option`	Options configurable in LiteRT environment.
`ModelProvider.Type`	Model files could be either an asset or a file.
`Status`	Status codes for LiteRtException.
`TensorBufferType`	The type of the tensor buffer.

Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License. For details, see the Google Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates.

Last updated 2025-05-13 UTC.