React Native ExecuTorch 发布端侧 Qwen-2.5 量化模型
原标题：software-mansion/react-native-executorch-qwen-2.5
AI 摘要
Software Mansion 在 Hugging Face 上发布了为 React Native ExecuTorch 库导出的 Qwen-2.5 模型，提供 0.5B、1.5B、3B 三种规模，并分别以 mlx（int4）和 xnnpack（8da4w）两种后端量化，可直接在设备端运行。这些 .pte 程序面向 ExecuTorch v1.4.1 运行时，官方不保证向前兼容，旧版本运行时可能无法加载。此举让移动端 React Native 应用能够本地部署 Qwen-2.5 文本生成模型。
正文节选
--- license: apache-2.0 pipeline_tag: text-generation tags: - executorch library_name: executorch --- # qwen-2.5 This repository hosts the **qwen-2.5** models exported for the [React Native ExecuTorch](https://www.npmjs.com/package/react-native-executorch) library as ExecuTorch `.pte` programs, ready to run on device. ## Variants | Path | Backend | Precision | Context length | | --- | --- | --- | --- | | `0_5b/mlx/qwen_2_5_0_5b_mlx_int4.pte` | mlx | int4 | 16,384 | | `0_5b/xnnpack/qwen_2_5_0