software-mansion/react-native-executorch-rfdetr-keypoint

rfdetr-keypoint

This repository hosts the rfdetr-keypoint models exported for the React Native ExecuTorch library as ExecuTorch .pte programs, ready to run on device.

Upstream model: RFDetr

Variants

Path	Backend	Precision
coreml/rfdetr_keypoint_preview_coreml_fp16.pte	coreml	fp16
vulkan/rfdetr_keypoint_preview_vulkan_fp16.pte	vulkan	fp16
xnnpack/rfdetr_keypoint_preview_xnnpack_fp32.pte	xnnpack	fp32
xnnpack/rfdetr_keypoint_preview_xnnpack_int8.pte	xnnpack	int8

XNNPACK int8

xnnpack/rfdetr_keypoint_preview_xnnpack_int8.pte is 8da8w (int8 dynamic activations, int8 per-channel weights) on the Linear layers, except the deformable-attention offsets and the class, box and keypoint heads; convolutions stay fp32. COCO keypoint (OKS) mAP50-95 on the first 500 val2017 images: fp32 70.37, int8 70.00.

Unshipped backends

A backend directory carrying a NOTES.md and no .pte was exported and not published. Its note records why.

• mlx (why)

Repository structure

config.json                                       37 B
coreml/config.json                                1.1 kB
coreml/rfdetr_keypoint_preview_coreml_fp16.pte    71.5 MB
mlx/NOTES.md                                      1.4 kB
vulkan/config.json                                1.1 kB
vulkan/rfdetr_keypoint_preview_vulkan_fp16.pte    70.0 MB
xnnpack/config.json                               1.9 kB
xnnpack/rfdetr_keypoint_preview_xnnpack_fp32.pte  139 MB
xnnpack/rfdetr_keypoint_preview_xnnpack_int8.pte  56.6 MB

Compatibility

These files are published for the ExecuTorch v1.4.1 runtime. ExecuTorch gives no forward compatibility guarantee, so an older runtime may fail to load them.

To use them in React Native ExecuTorch, pass the model constant shipped in the library's model registry to the corresponding task pipeline. See the documentation.

To load these files in your own ExecuTorch runtime, read the compatibility note first.