LiteRT Next 提供了一个统一的接口来使用神经处理单元 (NPU),而无需您单独浏览特定于供应商的编译器、运行时或库依赖项。使用 LiteRT Next 进行 NPU 加速可避免许多特定于供应商或特定于设备的复杂情况,提高实时推理和大型模型推理的性能,并通过零复制硬件缓冲区使用来最大限度地减少内存复制。
[[["易于理解","easyToUnderstand","thumb-up"],["解决了我的问题","solvedMyProblem","thumb-up"],["其他","otherUp","thumb-up"]],[["没有我需要的信息","missingTheInformationINeed","thumb-down"],["太复杂/步骤太多","tooComplicatedTooManySteps","thumb-down"],["内容需要更新","outOfDate","thumb-down"],["翻译问题","translationIssue","thumb-down"],["示例/代码问题","samplesCodeIssue","thumb-down"],["其他","otherDown","thumb-down"]],["最后更新时间 (UTC):2025-09-03。"],[],[],null,["# NPU acceleration with LiteRT Next\n\n| **Note:** LiteRT NPU acceleration is only available through an Early Access Program. If you are not already enrolled, [Sign Up](https://forms.gle/CoH4jpLwxiEYvDvF6).\n\nLiteRT Next provides a unified interface to use Neural Processing Units (NPUs)\nwithout forcing you to individually navigate vendor-specific compilers,\nruntimes, or library dependencies. Using LiteRT Next for NPU acceleration avoids\nmany vendor-specific or device-specific complications, boosts performance for\nreal-time and large-model inference, and minimizes memory copies through\nzero-copy hardware buffer usage.\n\nIf you are already, enrolled in the LiteRT NPU Early Access Program, sign in to\nthe authorized account to view the NPU documentation. If you have not enrolled,\nsign up to the Early Access Program:\n\n[Sign\nup!arrow_forward](https://forms.gle/CoH4jpLwxiEYvDvF6)\n| **Confidential:** The following sections are confidential. Do not share or discuss unless authorized to do so.\n\nGet Started\n-----------\n\nTo get started, see the NPU overview guide:\n\n- **For classical ML models** , proceed directly with the core framework:\n - [NPU acceleration with LiteRT Next](./eap/npu)\n- **For Large Language Models (LLMs)** , we recommend using our [LiteRT-LM](https://github.com/google-ai-edge/LiteRT-LM) framework to handle the required end-to-end processing for NPU execution:\n - [NPU acceleration with LiteRT-LM](./eap/litert_lm_npu.md)\n\nFor example implementations of LiteRT Next with NPU support, refer to the\nfollowing demo applications:\n\n- [Image segmentation with Kotlin](https://github.com/google-ai-edge/LiteRT/tree/main/litert/samples/image_segmentation/kotlin_npu/)\n- [Asynchronous segmentation with C++](https://github.com/google-ai-edge/LiteRT/tree/main/litert/samples/async_segmentation)\n\nNPU Vendors\n-----------\n\nLiteRT Next supports NPU acceleration with the following vendors:\n\n- [Qualcomm AI Engine Direct](./eap/qualcomm)\n- [MediaTek NeuroPilot](./eap/mediatek)"]]