React Native ExecuTorch 发布 CLIP ViT 端侧多后端模型
原标题：software-mansion/react-native-executorch-clip-vit-base-patch32
AI 摘要
Software Mansion 在 Hugging Face 发布了为 React Native ExecuTorch 库导出的 clip-vit-base-patch32 模型，提供 CoreML、MLX、Vulkan、XNNPACK 四种后端及 fp16/int8/fp32 多种精度版本，可直接在设备端运行。XNNPACK int8 量化版本在 ImageNetV2 零样本 top-1 上与 fp32 基本持平（约 51%）。这些 .pte 文件面向 ExecuTorch v1.4.1 运行时，官方不保证向前兼容。
正文节选
--- license: mit pipeline_tag: image-feature-extraction library_name: executorch --- # clip-vit-base-patch32 This repository hosts the **clip-vit-base-patch32** 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. Upstream model: [clip-vit-base-patch32](https://huggingface.co/openai/clip-vit-base-patch32) ## Variants | Path | Backend | Precision | | --- | --- | --- | | `coreml/