File size: 4,276 Bytes
61f63c6 4680255 61f63c6 ff12ed8 f7eff56 b2cf088 f7eff56 ff12ed8 8577f5f ff12ed8 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 | ---
license: bsd-3-clause
pipeline_tag: image-segmentation
library_name: executorch
---
# deeplab-v3
This repository hosts the **deeplab-v3** 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: [DeepLabV3](https://docs.pytorch.org/vision/main/models/deeplabv3.html)
## Variants
| Path | Backend | Precision |
| --- | --- | --- |
| `coreml/deeplab_v3_mobilenet_v3_large_coreml_fp16.pte` | coreml | fp16 |
| `coreml/deeplab_v3_resnet50_coreml_fp16.pte` | coreml | fp16 |
| `coreml/deeplab_v3_resnet101_coreml_fp16.pte` | coreml | fp16 |
| `xnnpack/deeplab_v3_mobilenet_v3_large_xnnpack_fp32.pte` | xnnpack | fp32 |
| `xnnpack/deeplab_v3_mobilenet_v3_large_xnnpack_int8.pte` | xnnpack | int8 |
| `xnnpack/deeplab_v3_resnet50_xnnpack_fp32.pte` | xnnpack | fp32 |
| `xnnpack/deeplab_v3_resnet50_xnnpack_int8.pte` | xnnpack | int8 |
| `xnnpack/deeplab_v3_resnet101_xnnpack_fp32.pte` | xnnpack | fp32 |
| `xnnpack/deeplab_v3_resnet101_xnnpack_int8.pte` | xnnpack | int8 |
| `qnn/deeplab_v3_mobilenet_v3_large_qnn_a16w8_v69.pte` | qnn | a16w8 |
| `qnn/deeplab_v3_mobilenet_v3_large_qnn_a16w8_v73.pte` | qnn | a16w8 |
| `qnn/deeplab_v3_mobilenet_v3_large_qnn_a16w8_v75.pte` | qnn | a16w8 |
| `qnn/deeplab_v3_mobilenet_v3_large_qnn_a16w8_v79.pte` | qnn | a16w8 |
| `qnn/deeplab_v3_mobilenet_v3_large_qnn_a16w8_v81.pte` | qnn | a16w8 |
| `qnn/deeplab_v3_resnet50_qnn_a16w8_v69.pte` | qnn | a16w8 |
| `qnn/deeplab_v3_resnet50_qnn_a16w8_v73.pte` | qnn | a16w8 |
| `qnn/deeplab_v3_resnet50_qnn_a16w8_v75.pte` | qnn | a16w8 |
| `qnn/deeplab_v3_resnet50_qnn_a16w8_v79.pte` | qnn | a16w8 |
| `qnn/deeplab_v3_resnet50_qnn_a16w8_v81.pte` | qnn | a16w8 |
| `qnn/deeplab_v3_resnet101_qnn_a16w8_v69.pte` | qnn | a16w8 |
| `qnn/deeplab_v3_resnet101_qnn_a16w8_v73.pte` | qnn | a16w8 |
| `qnn/deeplab_v3_resnet101_qnn_a16w8_v75.pte` | qnn | a16w8 |
| `qnn/deeplab_v3_resnet101_qnn_a16w8_v79.pte` | qnn | a16w8 |
| `qnn/deeplab_v3_resnet101_qnn_a16w8_v81.pte` | qnn | a16w8 |
## QNN variants
A QNN `.pte` embeds an HTP context binary compiled for one Hexagon version, so
there is one file per version and the runtime picks the one matching the device.
These return a class index per pixel, `(1, 520, 520)` int32, rather than
`(1, 21, 520, 520)` logits.
| Hexagon | Example SoC |
| --- | --- |
| v69 | Snapdragon 8 Gen 1 |
| v73 | Snapdragon 8 Gen 2 |
| v75 | Snapdragon 8 Gen 3 |
| v79 | Snapdragon 8 Elite |
| v81 | Snapdragon 8 Elite Gen 5 |
## Unshipped backends
A backend directory carrying a `NOTES.md` and no `.pte` was exported and
not published. Its note records what went wrong.
- `mlx` ([why](https://huggingface.co/software-mansion/react-native-executorch-deeplab-v3/blob/main/mlx/NOTES.md))
## Repository structure
```
config.json 32 B
coreml/config.json 2.1 kB
coreml/deeplab_v3_mobilenet_v3_large_coreml_fp16.pte 21.5 MB
coreml/deeplab_v3_resnet101_coreml_fp16.pte 112 MB
coreml/deeplab_v3_resnet50_coreml_fp16.pte 75.9 MB
mlx/NOTES.md 2.0 kB
xnnpack/config.json 5.0 kB
xnnpack/deeplab_v3_mobilenet_v3_large_xnnpack_fp32.pte 42.1 MB
xnnpack/deeplab_v3_mobilenet_v3_large_xnnpack_int8.pte 10.8 MB
xnnpack/deeplab_v3_resnet101_xnnpack_fp32.pte 224 MB
xnnpack/deeplab_v3_resnet101_xnnpack_int8.pte 58.9 MB
xnnpack/deeplab_v3_resnet50_xnnpack_fp32.pte 151 MB
xnnpack/deeplab_v3_resnet50_xnnpack_int8.pte 40.4 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](https://docs.swmansion.com/react-native-executorch/docs/fundamentals/downloading-models).
To load these files in your own ExecuTorch runtime, read the
[compatibility note](https://github.com/pytorch/executorch/blob/main/runtime/COMPATIBILITY.md)
first.
|