v0.60.0
Browse filesSee https://github.com/qualcomm/ai-hub-models/releases/v0.60.0 for changelog.
- README.md +87 -79
- release_assets.json +6 -6
README.md
CHANGED
|
@@ -16,7 +16,7 @@ pipeline_tag: image-classification
|
|
| 16 |
SwinSmall is a machine learning model that can classify images from the Imagenet dataset. It can also be used as a backbone in building more complex models for specific use cases.
|
| 17 |
|
| 18 |
This is based on the implementation of Swin-Small found [here](https://github.com/pytorch/vision/blob/main/torchvision/models/swin_transformer.py).
|
| 19 |
-
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.
|
| 20 |
|
| 21 |
Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device.
|
| 22 |
|
|
@@ -29,111 +29,119 @@ Below are pre-exported model assets ready for deployment.
|
|
| 29 |
|
| 30 |
| Runtime | Precision | Chipset | SDK Versions | Download |
|
| 31 |
|---|---|---|---|---|
|
| 32 |
-
| ONNX | float | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.
|
| 33 |
-
| ONNX | w8a16 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.
|
| 34 |
-
| QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.
|
| 35 |
-
| QNN_DLC | w8a16 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.
|
| 36 |
-
| TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.
|
| 37 |
|
| 38 |
For more device-specific assets and performance metrics, visit **[Swin-Small on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/swin_small)**.
|
| 39 |
|
| 40 |
|
| 41 |
### Option 2: Export with Custom Configurations
|
| 42 |
|
| 43 |
-
Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.
|
| 44 |
- Custom weights (e.g., fine-tuned checkpoints)
|
| 45 |
- Custom input shapes
|
| 46 |
- Target device and runtime configurations
|
| 47 |
|
| 48 |
This option is ideal if you need to customize the model beyond the default configuration provided here.
|
| 49 |
|
| 50 |
-
See our repository for [Swin-Small on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.
|
| 51 |
|
| 52 |
## Model Details
|
| 53 |
|
| 54 |
**Model Type:** Model_use_case.image_classification
|
| 55 |
|
| 56 |
**Model Stats:**
|
| 57 |
-
- Model checkpoint: Imagenet
|
| 58 |
- Input resolution: 224x224
|
| 59 |
-
-
|
| 60 |
- Model size (float): 193 MB
|
| 61 |
- Model size (w8a16): 52.5 MB
|
|
|
|
| 62 |
|
| 63 |
## Performance Summary
|
| 64 |
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|
| 65 |
|---|---|---|---|---|---|---
|
| 66 |
-
| Swin-Small | ONNX | float | Snapdragon® X2 Elite | 6.
|
| 67 |
-
| Swin-Small | ONNX | float | Snapdragon® X Elite | 16.
|
| 68 |
-
| Swin-Small | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 10.
|
| 69 |
-
| Swin-Small | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 23.
|
| 70 |
-
| Swin-Small | ONNX | float | Qualcomm® Dragonwing™
|
| 71 |
-
| Swin-Small | ONNX | float | Qualcomm®
|
| 72 |
-
| Swin-Small | ONNX | float | Qualcomm®
|
| 73 |
-
| Swin-Small | ONNX | float | Qualcomm® Dragonwing™ IQ-
|
| 74 |
-
| Swin-Small | ONNX | float | Qualcomm® Dragonwing™
|
| 75 |
-
| Swin-Small | ONNX | float |
|
| 76 |
-
| Swin-Small | ONNX | float | Snapdragon® 8 Elite
|
| 77 |
-
| Swin-Small | ONNX |
|
| 78 |
-
| Swin-Small | ONNX | w8a16 | Snapdragon®
|
| 79 |
-
| Swin-Small | ONNX | w8a16 | Snapdragon®
|
| 80 |
-
| Swin-Small | ONNX | w8a16 | Snapdragon® 8 Gen
|
| 81 |
-
| Swin-Small | ONNX | w8a16 |
|
| 82 |
-
| Swin-Small | ONNX | w8a16 | Qualcomm® Dragonwing™
|
| 83 |
-
| Swin-Small | ONNX | w8a16 | Qualcomm®
|
| 84 |
-
| Swin-Small | ONNX | w8a16 | Qualcomm® Dragonwing™
|
| 85 |
-
| Swin-Small | ONNX | w8a16 | Qualcomm®
|
| 86 |
-
| Swin-Small | ONNX | w8a16 | Qualcomm® Dragonwing™
|
| 87 |
-
| Swin-Small | ONNX | w8a16 |
|
| 88 |
-
| Swin-Small | ONNX | w8a16 |
|
| 89 |
-
| Swin-Small |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 90 |
| Swin-Small | QNN_DLC | float | Snapdragon® X Elite | 16.346 ms | 1 - 1 MB | NPU
|
| 91 |
-
| Swin-Small | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 9.
|
| 92 |
-
| Swin-Small | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 23.
|
| 93 |
-
| Swin-Small | QNN_DLC | float | Qualcomm® Dragonwing™
|
| 94 |
-
| Swin-Small | QNN_DLC | float | Qualcomm® Dragonwing™
|
| 95 |
-
| Swin-Small | QNN_DLC | float | Qualcomm®
|
| 96 |
-
| Swin-Small | QNN_DLC | float | Qualcomm®
|
| 97 |
-
| Swin-Small | QNN_DLC | float | Qualcomm®
|
| 98 |
-
| Swin-Small | QNN_DLC | float | Qualcomm®
|
| 99 |
-
| Swin-Small | QNN_DLC | float | Qualcomm®
|
|
|
|
| 100 |
| Swin-Small | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 16.346 ms | 1 - 1 MB | NPU
|
| 101 |
-
| Swin-Small | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 7.
|
| 102 |
-
| Swin-Small | QNN_DLC | float | Qualcomm® SA7255P | 37.
|
| 103 |
-
| Swin-Small | QNN_DLC | float | Qualcomm® SA8295P | 22.
|
| 104 |
-
| Swin-Small | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 7.
|
| 105 |
-
| Swin-Small | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 6.
|
| 106 |
-
| Swin-Small | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 7.
|
| 107 |
-
| Swin-Small | QNN_DLC | w8a16 | Snapdragon® X Elite | 17.
|
| 108 |
-
| Swin-Small | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 10.
|
| 109 |
-
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® Dragonwing™
|
| 110 |
-
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® Dragonwing™
|
| 111 |
-
| Swin-Small | QNN_DLC | w8a16 | Qualcomm®
|
| 112 |
-
| Swin-Small | QNN_DLC | w8a16 | Qualcomm®
|
| 113 |
-
| Swin-Small | QNN_DLC | w8a16 | Qualcomm®
|
| 114 |
-
| Swin-Small | QNN_DLC | w8a16 | Qualcomm®
|
| 115 |
-
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-
|
| 116 |
-
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® Dragonwing™
|
| 117 |
-
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-
|
| 118 |
-
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-
|
| 119 |
-
| Swin-Small | QNN_DLC | w8a16 | Qualcomm®
|
| 120 |
-
| Swin-Small | QNN_DLC | w8a16 |
|
| 121 |
-
| Swin-Small | QNN_DLC | w8a16 | Snapdragon® 8 Elite
|
| 122 |
-
| Swin-Small | QNN_DLC | w8a16 | Snapdragon®
|
| 123 |
-
| Swin-Small |
|
| 124 |
-
| Swin-Small | TFLITE | float | Snapdragon® 8 Gen
|
| 125 |
-
| Swin-Small | TFLITE | float |
|
| 126 |
-
| Swin-Small | TFLITE | float | Qualcomm® Dragonwing™
|
| 127 |
-
| Swin-Small | TFLITE | float | Qualcomm®
|
| 128 |
-
| Swin-Small | TFLITE | float | Qualcomm®
|
| 129 |
-
| Swin-Small | TFLITE | float | Qualcomm®
|
| 130 |
-
| Swin-Small | TFLITE | float | Qualcomm®
|
| 131 |
-
| Swin-Small | TFLITE | float | Qualcomm®
|
| 132 |
-
| Swin-Small | TFLITE | float | Qualcomm®
|
| 133 |
-
| Swin-Small | TFLITE | float | Qualcomm®
|
| 134 |
-
| Swin-Small | TFLITE | float | Qualcomm®
|
| 135 |
-
| Swin-Small | TFLITE | float |
|
| 136 |
-
| Swin-Small | TFLITE | float |
|
|
|
|
|
|
|
| 137 |
|
| 138 |
## License
|
| 139 |
* The license for the original implementation of Swin-Small can be found
|
|
|
|
| 16 |
SwinSmall is a machine learning model that can classify images from the Imagenet dataset. It can also be used as a backbone in building more complex models for specific use cases.
|
| 17 |
|
| 18 |
This is based on the implementation of Swin-Small found [here](https://github.com/pytorch/vision/blob/main/torchvision/models/swin_transformer.py).
|
| 19 |
+
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.60.0/src/qai_hub_models/models/swin_small) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).
|
| 20 |
|
| 21 |
Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device.
|
| 22 |
|
|
|
|
| 29 |
|
| 30 |
| Runtime | Precision | Chipset | SDK Versions | Download |
|
| 31 |
|---|---|---|---|---|
|
| 32 |
+
| ONNX | float | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.60.0/swin_small-onnx-float.zip)
|
| 33 |
+
| ONNX | w8a16 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.60.0/swin_small-onnx-w8a16.zip)
|
| 34 |
+
| QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.60.0/swin_small-qnn_dlc-float.zip)
|
| 35 |
+
| QNN_DLC | w8a16 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.60.0/swin_small-qnn_dlc-w8a16.zip)
|
| 36 |
+
| TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.60.0/swin_small-tflite-float.zip)
|
| 37 |
|
| 38 |
For more device-specific assets and performance metrics, visit **[Swin-Small on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/swin_small)**.
|
| 39 |
|
| 40 |
|
| 41 |
### Option 2: Export with Custom Configurations
|
| 42 |
|
| 43 |
+
Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.60.0/src/qai_hub_models/models/swin_small) Python library to compile and export the model with your own:
|
| 44 |
- Custom weights (e.g., fine-tuned checkpoints)
|
| 45 |
- Custom input shapes
|
| 46 |
- Target device and runtime configurations
|
| 47 |
|
| 48 |
This option is ideal if you need to customize the model beyond the default configuration provided here.
|
| 49 |
|
| 50 |
+
See our repository for [Swin-Small on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.60.0/src/qai_hub_models/models/swin_small) for usage instructions.
|
| 51 |
|
| 52 |
## Model Details
|
| 53 |
|
| 54 |
**Model Type:** Model_use_case.image_classification
|
| 55 |
|
| 56 |
**Model Stats:**
|
|
|
|
| 57 |
- Input resolution: 224x224
|
| 58 |
+
- Model checkpoint: Imagenet
|
| 59 |
- Model size (float): 193 MB
|
| 60 |
- Model size (w8a16): 52.5 MB
|
| 61 |
+
- Number of parameters: 50.4M
|
| 62 |
|
| 63 |
## Performance Summary
|
| 64 |
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|
| 65 |
|---|---|---|---|---|---|---
|
| 66 |
+
| Swin-Small | ONNX | float | Snapdragon® X2 Elite | 6.907 ms | 2 - 2 MB | NPU
|
| 67 |
+
| Swin-Small | ONNX | float | Snapdragon® X Elite | 16.19 ms | 101 - 101 MB | NPU
|
| 68 |
+
| Swin-Small | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 10.156 ms | 1 - 447 MB | NPU
|
| 69 |
+
| Swin-Small | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 23.393 ms | 0 - 430 MB | NPU
|
| 70 |
+
| Swin-Small | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 15.739 ms | 0 - 4 MB | NPU
|
| 71 |
+
| Swin-Small | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 15.604 ms | 0 - 422 MB | NPU
|
| 72 |
+
| Swin-Small | ONNX | float | Qualcomm® QCS8450 | 23.393 ms | 0 - 430 MB | NPU
|
| 73 |
+
| Swin-Small | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 22.215 ms | 1 - 4 MB | NPU
|
| 74 |
+
| Swin-Small | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 16.19 ms | 101 - 101 MB | NPU
|
| 75 |
+
| Swin-Small | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 7.76 ms | 1 - 610 MB | NPU
|
| 76 |
+
| Swin-Small | ONNX | float | Snapdragon® 8 Elite Mobile | 7.76 ms | 1 - 610 MB | NPU
|
| 77 |
+
| Swin-Small | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 6.437 ms | 1 - 621 MB | NPU
|
| 78 |
+
| Swin-Small | ONNX | w8a16 | Snapdragon® X2 Elite | 5.988 ms | 1 - 1 MB | NPU
|
| 79 |
+
| Swin-Small | ONNX | w8a16 | Snapdragon® X Elite | 13.907 ms | 54 - 54 MB | NPU
|
| 80 |
+
| Swin-Small | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 8.692 ms | 0 - 524 MB | NPU
|
| 81 |
+
| Swin-Small | ONNX | w8a16 | Snapdragon® 8 Gen 1 Mobile | 16.481 ms | 0 - 436 MB | NPU
|
| 82 |
+
| Swin-Small | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 40.139 ms | 0 - 3 MB | NPU
|
| 83 |
+
| Swin-Small | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 11.846 ms | 0 - 4 MB | NPU
|
| 84 |
+
| Swin-Small | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 13.118 ms | 0 - 386 MB | NPU
|
| 85 |
+
| Swin-Small | ONNX | w8a16 | Qualcomm® QCS8450 | 16.481 ms | 0 - 436 MB | NPU
|
| 86 |
+
| Swin-Small | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 13.701 ms | 0 - 3 MB | NPU
|
| 87 |
+
| Swin-Small | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 13.907 ms | 54 - 54 MB | NPU
|
| 88 |
+
| Swin-Small | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 67.709 ms | 0 - 556 MB | NPU
|
| 89 |
+
| Swin-Small | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 13.982 ms | 0 - 388 MB | NPU
|
| 90 |
+
| Swin-Small | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 6.886 ms | 0 - 439 MB | NPU
|
| 91 |
+
| Swin-Small | ONNX | w8a16 | Snapdragon® 8 Elite Mobile | 6.886 ms | 0 - 439 MB | NPU
|
| 92 |
+
| Swin-Small | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 5.645 ms | 0 - 451 MB | NPU
|
| 93 |
+
| Swin-Small | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 13.982 ms | 0 - 388 MB | NPU
|
| 94 |
+
| Swin-Small | QNN_DLC | float | Snapdragon® X2 Elite | 7.089 ms | 1 - 1 MB | NPU
|
| 95 |
| Swin-Small | QNN_DLC | float | Snapdragon® X Elite | 16.346 ms | 1 - 1 MB | NPU
|
| 96 |
+
| Swin-Small | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 9.941 ms | 0 - 407 MB | NPU
|
| 97 |
+
| Swin-Small | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 23.957 ms | 0 - 396 MB | NPU
|
| 98 |
+
| Swin-Small | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 15.822 ms | 1 - 4 MB | NPU
|
| 99 |
+
| Swin-Small | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 37.925 ms | 1 - 275 MB | NPU
|
| 100 |
+
| Swin-Small | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 15.358 ms | 1 - 2 MB | NPU
|
| 101 |
+
| Swin-Small | QNN_DLC | float | Qualcomm® SA8775P | 17.409 ms | 1 - 593 MB | NPU
|
| 102 |
+
| Swin-Small | QNN_DLC | float | Qualcomm® SA8650P | 17.409 ms | 1 - 593 MB | NPU
|
| 103 |
+
| Swin-Small | QNN_DLC | float | Qualcomm® SA8255P | 17.409 ms | 1 - 593 MB | NPU
|
| 104 |
+
| Swin-Small | QNN_DLC | float | Qualcomm® QCS8450 | 23.957 ms | 0 - 396 MB | NPU
|
| 105 |
+
| Swin-Small | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 17.85 ms | 1 - 3 MB | NPU
|
| 106 |
| Swin-Small | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 16.346 ms | 1 - 1 MB | NPU
|
| 107 |
+
| Swin-Small | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 7.558 ms | 1 - 588 MB | NPU
|
| 108 |
+
| Swin-Small | QNN_DLC | float | Qualcomm® SA7255P | 37.925 ms | 1 - 275 MB | NPU
|
| 109 |
+
| Swin-Small | QNN_DLC | float | Qualcomm® SA8295P | 22.618 ms | 1 - 269 MB | NPU
|
| 110 |
+
| Swin-Small | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 7.558 ms | 1 - 588 MB | NPU
|
| 111 |
+
| Swin-Small | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 6.249 ms | 0 - 600 MB | NPU
|
| 112 |
+
| Swin-Small | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 7.216 ms | 0 - 0 MB | NPU
|
| 113 |
+
| Swin-Small | QNN_DLC | w8a16 | Snapdragon® X Elite | 17.385 ms | 0 - 0 MB | NPU
|
| 114 |
+
| Swin-Small | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 10.746 ms | 0 - 869 MB | NPU
|
| 115 |
+
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 14.454 ms | 0 - 3 MB | NPU
|
| 116 |
+
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 29.259 ms | 0 - 635 MB | NPU
|
| 117 |
+
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 16.365 ms | 0 - 2 MB | NPU
|
| 118 |
+
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® SA8775P | 16.773 ms | 0 - 638 MB | NPU
|
| 119 |
+
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® SA8650P | 16.773 ms | 0 - 638 MB | NPU
|
| 120 |
+
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® SA8255P | 16.773 ms | 0 - 638 MB | NPU
|
| 121 |
+
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 17.119 ms | 0 - 2 MB | NPU
|
| 122 |
+
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 17.385 ms | 0 - 0 MB | NPU
|
| 123 |
+
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 73.744 ms | 0 - 752 MB | NPU
|
| 124 |
+
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 17.074 ms | 0 - 652 MB | NPU
|
| 125 |
+
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 8.125 ms | 0 - 622 MB | NPU
|
| 126 |
+
| Swin-Small | QNN_DLC | w8a16 | Qualcomm® SA7255P | 29.259 ms | 0 - 635 MB | NPU
|
| 127 |
+
| Swin-Small | QNN_DLC | w8a16 | Snapdragon® 8 Elite Mobile | 8.125 ms | 0 - 622 MB | NPU
|
| 128 |
+
| Swin-Small | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 6.502 ms | 0 - 640 MB | NPU
|
| 129 |
+
| Swin-Small | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 17.074 ms | 0 - 652 MB | NPU
|
| 130 |
+
| Swin-Small | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 10.212 ms | 0 - 433 MB | NPU
|
| 131 |
+
| Swin-Small | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 23.278 ms | 0 - 412 MB | NPU
|
| 132 |
+
| Swin-Small | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 15.513 ms | 0 - 104 MB | NPU
|
| 133 |
+
| Swin-Small | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 37.417 ms | 0 - 302 MB | NPU
|
| 134 |
+
| Swin-Small | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 15.556 ms | 0 - 3 MB | NPU
|
| 135 |
+
| Swin-Small | TFLITE | float | Qualcomm® SA8775P | 17.456 ms | 0 - 306 MB | NPU
|
| 136 |
+
| Swin-Small | TFLITE | float | Qualcomm® SA8650P | 17.456 ms | 0 - 306 MB | NPU
|
| 137 |
+
| Swin-Small | TFLITE | float | Qualcomm® SA8255P | 17.456 ms | 0 - 306 MB | NPU
|
| 138 |
+
| Swin-Small | TFLITE | float | Qualcomm® QCS8450 | 23.278 ms | 0 - 412 MB | NPU
|
| 139 |
+
| Swin-Small | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 17.361 ms | 0 - 104 MB | NPU
|
| 140 |
+
| Swin-Small | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 7.801 ms | 0 - 298 MB | NPU
|
| 141 |
+
| Swin-Small | TFLITE | float | Qualcomm® SA7255P | 37.417 ms | 0 - 302 MB | NPU
|
| 142 |
+
| Swin-Small | TFLITE | float | Qualcomm® SA8295P | 22.987 ms | 0 - 298 MB | NPU
|
| 143 |
+
| Swin-Small | TFLITE | float | Snapdragon® 8 Elite Mobile | 7.801 ms | 0 - 298 MB | NPU
|
| 144 |
+
| Swin-Small | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 6.596 ms | 0 - 321 MB | NPU
|
| 145 |
|
| 146 |
## License
|
| 147 |
* The license for the original implementation of Swin-Small can be found
|
release_assets.json
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
{
|
| 2 |
-
"version": "0.
|
| 3 |
"precisions": {
|
| 4 |
"float": {
|
| 5 |
"universal_assets": {
|
|
@@ -8,19 +8,19 @@
|
|
| 8 |
"qairt": "2.45.0.260326154327",
|
| 9 |
"onnx_runtime": "1.27.1"
|
| 10 |
},
|
| 11 |
-
"download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.
|
| 12 |
},
|
| 13 |
"qnn_dlc": {
|
| 14 |
"tool_versions": {
|
| 15 |
"qairt": "2.45.0.260326154327"
|
| 16 |
},
|
| 17 |
-
"download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.
|
| 18 |
},
|
| 19 |
"tflite": {
|
| 20 |
"tool_versions": {
|
| 21 |
"qairt": "2.45.0.260326154327"
|
| 22 |
},
|
| 23 |
-
"download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.
|
| 24 |
}
|
| 25 |
}
|
| 26 |
},
|
|
@@ -31,13 +31,13 @@
|
|
| 31 |
"qairt": "2.45.0.260326154327",
|
| 32 |
"onnx_runtime": "1.27.1"
|
| 33 |
},
|
| 34 |
-
"download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.
|
| 35 |
},
|
| 36 |
"qnn_dlc": {
|
| 37 |
"tool_versions": {
|
| 38 |
"qairt": "2.45.0.260326154327"
|
| 39 |
},
|
| 40 |
-
"download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.
|
| 41 |
}
|
| 42 |
}
|
| 43 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"version": "0.60.0",
|
| 3 |
"precisions": {
|
| 4 |
"float": {
|
| 5 |
"universal_assets": {
|
|
|
|
| 8 |
"qairt": "2.45.0.260326154327",
|
| 9 |
"onnx_runtime": "1.27.1"
|
| 10 |
},
|
| 11 |
+
"download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.60.0/swin_small-onnx-float.zip"
|
| 12 |
},
|
| 13 |
"qnn_dlc": {
|
| 14 |
"tool_versions": {
|
| 15 |
"qairt": "2.45.0.260326154327"
|
| 16 |
},
|
| 17 |
+
"download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.60.0/swin_small-qnn_dlc-float.zip"
|
| 18 |
},
|
| 19 |
"tflite": {
|
| 20 |
"tool_versions": {
|
| 21 |
"qairt": "2.45.0.260326154327"
|
| 22 |
},
|
| 23 |
+
"download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.60.0/swin_small-tflite-float.zip"
|
| 24 |
}
|
| 25 |
}
|
| 26 |
},
|
|
|
|
| 31 |
"qairt": "2.45.0.260326154327",
|
| 32 |
"onnx_runtime": "1.27.1"
|
| 33 |
},
|
| 34 |
+
"download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.60.0/swin_small-onnx-w8a16.zip"
|
| 35 |
},
|
| 36 |
"qnn_dlc": {
|
| 37 |
"tool_versions": {
|
| 38 |
"qairt": "2.45.0.260326154327"
|
| 39 |
},
|
| 40 |
+
"download_url": "https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/swin_small/releases/v0.60.0/swin_small-qnn_dlc-w8a16.zip"
|
| 41 |
}
|
| 42 |
}
|
| 43 |
}
|