# LiteRT 概述

Lite Runtime (LiteRT) 是一个专为设备端推理而设计的开源深度学习框架。TensorFlow 框架提供了工具和 API，用于将标准的预训练 TensorFlow 模型从 SavedModel 或 Keras 格式转换为 LiteRT 格式。

所涵盖的主题描述了使用 Qualcomm^®^ 软件堆栈运行 LiteRT 模型的可用 delegate 和方法，并说明了如何：

- 使用基于 Gstreamer 的 Qualcomm^®^ Intelligent Multimedia SDK (Qualcomm IM SDK) 或本机 LiteRT 应用运行 LiteRT 模型。
- 将 TensorFlow 模型转换为 LiteRT 模型并针对设备上推理进行优化。
- 使用硬件加速器（如 CPU、GPU 和 Qualcomm^®^ Hexagon^™^ 张量处理器）上的 delegate 运行 LiteRT 模型。
- 对 LiteRT 模型进行基准测试。

## 后续步骤

运行 LiteRT 模型

![ico1](data:image/png;base64,UklGRrQAAABXRUJQVlA4TKcAAAAvFEAEEMegILJtat8HCYTSP4QW1FATSVLUxxdBiAP8S1wUNZIUNYcw/GvgjQtSjAMAKJSx/2S9m8DmMQgAnnfZ2ltW+4k++hkbhjp+pglu6P/banJFRfo1K5J9yVvWjHrPy/ue38CgkSRFx8z35N/rsYH9iP6bTdumcub+gY545ib8VJHNFdiPz1RLy7DY3CdZj3jaX3fj0jyjIeJ4OrfXzgyoynAAAAA=) [使用基于 GStreamer 的 Qualcomm IM SDK](https://docs.qualcomm.com/doc/80-70018-54SC/topic/getting-started.html#run-a-tensorflow-lite-model-using-the-gstreamer-based-qim-sdk)

![ico1](data:image/png;base64,UklGRrQAAABXRUJQVlA4TKcAAAAvFEAEEMegILJtat8HCYTSP4QW1FATSVLUxxdBiAP8S1wUNZIUNYcw/GvgjQtSjAMAKJSx/2S9m8DmMQgAnnfZ2ltW+4k++hkbhjp+pglu6P/banJFRfo1K5J9yVvWjHrPy/ue38CgkSRFx8z35N/rsYH9iP6bTdumcub+gY545ib8VJHNFdiPz1RLy7DY3CdZj3jaX3fj0jyjIeJ4OrfXzgyoynAAAAA=) [使用本机 LiteRT 示例应用程序](https://docs.qualcomm.com/doc/80-70018-54SC/topic/getting-started.html#run-a-tensorflow-lite-model-using-a-native-tensorflow-lite-sample-application)

LiteRT 开发者工作流程

![ico1](data:image/png;base64,UklGRrQAAABXRUJQVlA4TKcAAAAvFEAEEMegILJtat8HCYTSP4QW1FATSVLUxxdBiAP8S1wUNZIUNYcw/GvgjQtSjAMAKJSx/2S9m8DmMQgAnnfZ2ltW+4k++hkbhjp+pglu6P/banJFRfo1K5J9yVvWjHrPy/ue38CgkSRFx8z35N/rsYH9iP6bTdumcub+gY545ib8VJHNFdiPz1RLy7DY3CdZj3jaX3fj0jyjIeJ4OrfXzgyoynAAAAA=) [将 TensorFlow 模型转换为 LiteRT 模型](https://docs.qualcomm.com/doc/80-70018-54SC/topic/tensorflow-lite-developer-workflow.html#convert-tensorflow-lite-models)

![ico1](data:image/png;base64,UklGRrQAAABXRUJQVlA4TKcAAAAvFEAEEMegILJtat8HCYTSP4QW1FATSVLUxxdBiAP8S1wUNZIUNYcw/GvgjQtSjAMAKJSx/2S9m8DmMQgAnnfZ2ltW+4k++hkbhjp+pglu6P/banJFRfo1K5J9yVvWjHrPy/ue38CgkSRFx8z35N/rsYH9iP6bTdumcub+gY545ib8VJHNFdiPz1RLy7DY3CdZj3jaX3fj0jyjIeJ4OrfXzgyoynAAAAA=) [创建应用程序并运行推理](https://docs.qualcomm.com/doc/80-70018-54SC/topic/tensorflow-lite-developer-workflow.html#run-inference)

![ico1](data:image/png;base64,UklGRrQAAABXRUJQVlA4TKcAAAAvFEAEEMegILJtat8HCYTSP4QW1FATSVLUxxdBiAP8S1wUNZIUNYcw/GvgjQtSjAMAKJSx/2S9m8DmMQgAnnfZ2ltW+4k++hkbhjp+pglu6P/banJFRfo1K5J9yVvWjHrPy/ue38CgkSRFx8z35N/rsYH9iP6bTdumcub+gY545ib8VJHNFdiPz1RLy7DY3CdZj3jaX3fj0jyjIeJ4OrfXzgyoynAAAAA=) [开发定制的应用程序](https://docs.qualcomm.com/doc/80-70018-54SC/topic/tensorflow-lite-developer-workflow.html#develop-a-custom-application-to-run-the-tensorflow-lite-model)

示例程序

![ico1](data:image/png;base64,UklGRrQAAABXRUJQVlA4TKcAAAAvFEAEEMegILJtat8HCYTSP4QW1FATSVLUxxdBiAP8S1wUNZIUNYcw/GvgjQtSjAMAKJSx/2S9m8DmMQgAnnfZ2ltW+4k++hkbhjp+pglu6P/banJFRfo1K5J9yVvWjHrPy/ue38CgkSRFx8z35N/rsYH9iP6bTdumcub+gY545ib8VJHNFdiPz1RLy7DY3CdZj3jaX3fj0jyjIeJ4OrfXzgyoynAAAAA=) [下载模型和示例图像](https://docs.qualcomm.com/doc/80-70018-54SC/topic/sample-applications.html#download-models-and-sample-images)

![ico1](data:image/png;base64,UklGRrQAAABXRUJQVlA4TKcAAAAvFEAEEMegILJtat8HCYTSP4QW1FATSVLUxxdBiAP8S1wUNZIUNYcw/GvgjQtSjAMAKJSx/2S9m8DmMQgAnnfZ2ltW+4k++hkbhjp+pglu6P/banJFRfo1K5J9yVvWjHrPy/ue38CgkSRFx8z35N/rsYH9iP6bTdumcub+gY545ib8VJHNFdiPz1RLy7DY3CdZj3jaX3fj0jyjIeJ4OrfXzgyoynAAAAA=) [使用可用的delegate运行 LiteRT 模型](https://docs.qualcomm.com/doc/80-70018-54SC/topic/sample-applications.html#label-image-tool)

![ico1](data:image/png;base64,UklGRrQAAABXRUJQVlA4TKcAAAAvFEAEEMegILJtat8HCYTSP4QW1FATSVLUxxdBiAP8S1wUNZIUNYcw/GvgjQtSjAMAKJSx/2S9m8DmMQgAnnfZ2ltW+4k++hkbhjp+pglu6P/banJFRfo1K5J9yVvWjHrPy/ue38CgkSRFx8z35N/rsYH9iP6bTdumcub+gY545ib8VJHNFdiPz1RLy7DY3CdZj3jaX3fj0jyjIeJ4OrfXzgyoynAAAAA=) [使用外部 delegate 运行 QNN delegate](https://docs.qualcomm.com/doc/80-70018-54SC/topic/sample-applications.html#run-qnn-delegate-using-the-external-delegate-interface)

Note

请参阅 Qualcomm^®^ Linux^®^ 上支持的[硬件 SoC](https://docs.qualcomm.com/bundle/publicresource/topics/80-70018-115/soc.html)。

Last Published: Apr 29, 2025

[Next Topic
开始使用 LiteRT](https://docs.qualcomm.com/bundle/publicresource/80-70018-54SC/topics/getting-started.md)