v0.34.0
Browse filesSee https://github.com/quic/ai-hub-models/releases/v0.34.0 for changelog.
- README.md +28 -37
- Swin-Base_w8a16.dlc +1 -1
- precompiled/qualcomm-snapdragon-x-elite/Swin-Base.bin +1 -1
- precompiled/qualcomm-snapdragon-x-elite/Swin-Base.onnx.zip +2 -2
- precompiled/qualcomm-snapdragon-x-elite/Swin-Base_w8a16.onnx.zip +1 -1
- precompiled/qualcomm-snapdragon-x-elite/sdk_versions.yml +5 -0
README.md
CHANGED
|
@@ -24,6 +24,7 @@ More details on model performance across various devices, can be found
|
|
| 24 |
[here](https://aihub.qualcomm.com/models/swin_base).
|
| 25 |
|
| 26 |
|
|
|
|
| 27 |
### Model Details
|
| 28 |
|
| 29 |
- **Model Type:** Model_use_case.image_classification
|
|
@@ -36,52 +37,52 @@ More details on model performance across various devices, can be found
|
|
| 36 |
|
| 37 |
| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
|
| 38 |
|---|---|---|---|---|---|---|---|---|
|
| 39 |
-
| Swin-Base | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 61.
|
| 40 |
| Swin-Base | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 55.545 ms | 24 - 381 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 41 |
-
| Swin-Base | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 28.
|
| 42 |
| Swin-Base | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 28.781 ms | 1 - 379 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 43 |
-
| Swin-Base | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 22.
|
| 44 |
| Swin-Base | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 19.954 ms | 0 - 70 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 45 |
-
| Swin-Base | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 25.
|
| 46 |
| Swin-Base | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 22.962 ms | 0 - 357 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 47 |
-
| Swin-Base | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 61.
|
| 48 |
| Swin-Base | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 55.545 ms | 24 - 381 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 49 |
-
| Swin-Base | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 22.
|
| 50 |
| Swin-Base | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 19.98 ms | 0 - 71 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 51 |
-
| Swin-Base | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE |
|
| 52 |
| Swin-Base | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 28.705 ms | 1 - 349 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 53 |
-
| Swin-Base | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 22.
|
| 54 |
| Swin-Base | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 20.04 ms | 0 - 66 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 55 |
-
| Swin-Base | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 25.
|
| 56 |
| Swin-Base | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 22.962 ms | 0 - 357 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 57 |
-
| Swin-Base | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 22.
|
| 58 |
| Swin-Base | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 20.111 ms | 0 - 59 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 59 |
| Swin-Base | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 19.716 ms | 0 - 46 MB | NPU | [Swin-Base.onnx](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.onnx) |
|
| 60 |
-
| Swin-Base | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 15.
|
| 61 |
| Swin-Base | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 13.609 ms | 1 - 388 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 62 |
| Swin-Base | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 13.294 ms | 1 - 387 MB | NPU | [Swin-Base.onnx](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.onnx) |
|
| 63 |
-
| Swin-Base | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 14.
|
| 64 |
| Swin-Base | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 10.108 ms | 1 - 357 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 65 |
| Swin-Base | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 10.027 ms | 1 - 354 MB | NPU | [Swin-Base.onnx](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.onnx) |
|
| 66 |
| Swin-Base | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 23.11 ms | 1155 - 1155 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 67 |
| Swin-Base | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 22.311 ms | 175 - 175 MB | NPU | [Swin-Base.onnx](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.onnx) |
|
| 68 |
-
| Swin-Base | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 35.
|
| 69 |
-
| Swin-Base | w8a16 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 24.
|
| 70 |
-
| Swin-Base | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC |
|
| 71 |
-
| Swin-Base | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 19.
|
| 72 |
-
| Swin-Base | w8a16 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | QNN_DLC |
|
| 73 |
-
| Swin-Base | w8a16 | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 35.
|
| 74 |
-
| Swin-Base | w8a16 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 18.
|
| 75 |
-
| Swin-Base | w8a16 | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 22.
|
| 76 |
-
| Swin-Base | w8a16 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC |
|
| 77 |
-
| Swin-Base | w8a16 | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 19.
|
| 78 |
-
| Swin-Base | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 18.
|
| 79 |
| Swin-Base | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 149.0 ms | 452 - 587 MB | NPU | [Swin-Base.onnx](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.onnx) |
|
| 80 |
-
| Swin-Base | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 12.
|
| 81 |
| Swin-Base | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 118.606 ms | 471 - 756 MB | NPU | [Swin-Base.onnx](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.onnx) |
|
| 82 |
-
| Swin-Base | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 11.
|
| 83 |
| Swin-Base | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 103.835 ms | 493 - 781 MB | NPU | [Swin-Base.onnx](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.onnx) |
|
| 84 |
-
| Swin-Base | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 22.
|
| 85 |
| Swin-Base | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 147.017 ms | 397 - 397 MB | NPU | [Swin-Base.onnx](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.onnx) |
|
| 86 |
|
| 87 |
|
|
@@ -140,17 +141,7 @@ device. This script does the following:
|
|
| 140 |
```bash
|
| 141 |
python -m qai_hub_models.models.swin_base.export
|
| 142 |
```
|
| 143 |
-
|
| 144 |
-
Profiling Results
|
| 145 |
-
------------------------------------------------------------
|
| 146 |
-
Swin-Base
|
| 147 |
-
Device : cs_8275 (ANDROID 14)
|
| 148 |
-
Runtime : TFLITE
|
| 149 |
-
Estimated inference time (ms) : 61.8
|
| 150 |
-
Estimated peak memory usage (MB): [116, 520]
|
| 151 |
-
Total # Ops : 1568
|
| 152 |
-
Compute Unit(s) : npu (1568 ops) gpu (0 ops) cpu (0 ops)
|
| 153 |
-
```
|
| 154 |
|
| 155 |
|
| 156 |
## How does this work?
|
|
|
|
| 24 |
[here](https://aihub.qualcomm.com/models/swin_base).
|
| 25 |
|
| 26 |
|
| 27 |
+
|
| 28 |
### Model Details
|
| 29 |
|
| 30 |
- **Model Type:** Model_use_case.image_classification
|
|
|
|
| 37 |
|
| 38 |
| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
|
| 39 |
|---|---|---|---|---|---|---|---|---|
|
| 40 |
+
| Swin-Base | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 61.815 ms | 0 - 387 MB | NPU | [Swin-Base.tflite](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.tflite) |
|
| 41 |
| Swin-Base | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 55.545 ms | 24 - 381 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 42 |
+
| Swin-Base | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 28.792 ms | 0 - 398 MB | NPU | [Swin-Base.tflite](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.tflite) |
|
| 43 |
| Swin-Base | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 28.781 ms | 1 - 379 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 44 |
+
| Swin-Base | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 22.056 ms | 0 - 40 MB | NPU | [Swin-Base.tflite](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.tflite) |
|
| 45 |
| Swin-Base | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 19.954 ms | 0 - 70 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 46 |
+
| Swin-Base | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 25.712 ms | 0 - 404 MB | NPU | [Swin-Base.tflite](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.tflite) |
|
| 47 |
| Swin-Base | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 22.962 ms | 0 - 357 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 48 |
+
| Swin-Base | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 61.815 ms | 0 - 387 MB | NPU | [Swin-Base.tflite](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.tflite) |
|
| 49 |
| Swin-Base | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 55.545 ms | 24 - 381 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 50 |
+
| Swin-Base | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 22.116 ms | 0 - 39 MB | NPU | [Swin-Base.tflite](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.tflite) |
|
| 51 |
| Swin-Base | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 19.98 ms | 0 - 71 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 52 |
+
| Swin-Base | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 31.954 ms | 0 - 393 MB | NPU | [Swin-Base.tflite](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.tflite) |
|
| 53 |
| Swin-Base | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 28.705 ms | 1 - 349 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 54 |
+
| Swin-Base | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 22.028 ms | 0 - 43 MB | NPU | [Swin-Base.tflite](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.tflite) |
|
| 55 |
| Swin-Base | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 20.04 ms | 0 - 66 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 56 |
+
| Swin-Base | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 25.712 ms | 0 - 404 MB | NPU | [Swin-Base.tflite](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.tflite) |
|
| 57 |
| Swin-Base | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 22.962 ms | 0 - 357 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 58 |
+
| Swin-Base | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 22.15 ms | 0 - 36 MB | NPU | [Swin-Base.tflite](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.tflite) |
|
| 59 |
| Swin-Base | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 20.111 ms | 0 - 59 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 60 |
| Swin-Base | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 19.716 ms | 0 - 46 MB | NPU | [Swin-Base.onnx](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.onnx) |
|
| 61 |
+
| Swin-Base | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 15.392 ms | 5 - 413 MB | NPU | [Swin-Base.tflite](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.tflite) |
|
| 62 |
| Swin-Base | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 13.609 ms | 1 - 388 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 63 |
| Swin-Base | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 13.294 ms | 1 - 387 MB | NPU | [Swin-Base.onnx](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.onnx) |
|
| 64 |
+
| Swin-Base | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 14.594 ms | 0 - 395 MB | NPU | [Swin-Base.tflite](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.tflite) |
|
| 65 |
| Swin-Base | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 10.108 ms | 1 - 357 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 66 |
| Swin-Base | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 10.027 ms | 1 - 354 MB | NPU | [Swin-Base.onnx](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.onnx) |
|
| 67 |
| Swin-Base | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 23.11 ms | 1155 - 1155 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.dlc) |
|
| 68 |
| Swin-Base | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 22.311 ms | 175 - 175 MB | NPU | [Swin-Base.onnx](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base.onnx) |
|
| 69 |
+
| Swin-Base | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 35.912 ms | 0 - 261 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.dlc) |
|
| 70 |
+
| Swin-Base | w8a16 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 24.675 ms | 0 - 263 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.dlc) |
|
| 71 |
+
| Swin-Base | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 19.04 ms | 0 - 79 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.dlc) |
|
| 72 |
+
| Swin-Base | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 19.172 ms | 0 - 261 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.dlc) |
|
| 73 |
+
| Swin-Base | w8a16 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | QNN_DLC | 76.505 ms | 0 - 997 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.dlc) |
|
| 74 |
+
| Swin-Base | w8a16 | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 35.912 ms | 0 - 261 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.dlc) |
|
| 75 |
+
| Swin-Base | w8a16 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 18.947 ms | 0 - 75 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.dlc) |
|
| 76 |
+
| Swin-Base | w8a16 | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 22.771 ms | 0 - 263 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.dlc) |
|
| 77 |
+
| Swin-Base | w8a16 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 18.987 ms | 0 - 81 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.dlc) |
|
| 78 |
+
| Swin-Base | w8a16 | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 19.172 ms | 0 - 261 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.dlc) |
|
| 79 |
+
| Swin-Base | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN_DLC | 18.907 ms | 0 - 79 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.dlc) |
|
| 80 |
| Swin-Base | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 149.0 ms | 452 - 587 MB | NPU | [Swin-Base.onnx](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.onnx) |
|
| 81 |
+
| Swin-Base | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 12.525 ms | 0 - 269 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.dlc) |
|
| 82 |
| Swin-Base | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 118.606 ms | 471 - 756 MB | NPU | [Swin-Base.onnx](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.onnx) |
|
| 83 |
+
| Swin-Base | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN_DLC | 11.756 ms | 0 - 258 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.dlc) |
|
| 84 |
| Swin-Base | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 103.835 ms | 493 - 781 MB | NPU | [Swin-Base.onnx](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.onnx) |
|
| 85 |
+
| Swin-Base | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 22.269 ms | 302 - 302 MB | NPU | [Swin-Base.dlc](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.dlc) |
|
| 86 |
| Swin-Base | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 147.017 ms | 397 - 397 MB | NPU | [Swin-Base.onnx](https://huggingface.co/qualcomm/Swin-Base/blob/main/Swin-Base_w8a16.onnx) |
|
| 87 |
|
| 88 |
|
|
|
|
| 141 |
```bash
|
| 142 |
python -m qai_hub_models.models.swin_base.export
|
| 143 |
```
|
| 144 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 145 |
|
| 146 |
|
| 147 |
## How does this work?
|
Swin-Base_w8a16.dlc
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 94617349
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:062f41918d1cdb3b7793202efdf382ce4304421ae8a6da9dff94bb50c977cef2
|
| 3 |
size 94617349
|
precompiled/qualcomm-snapdragon-x-elite/Swin-Base.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 182820000
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a02c0783dd8ef237cd284ec9c6d0c883339d2578655605ba7c88d948a30f639d
|
| 3 |
size 182820000
|
precompiled/qualcomm-snapdragon-x-elite/Swin-Base.onnx.zip
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1184b55307bd3a6163df4aa93696f67b6c3e7e27d7a3cee2b5abbbac888a0684
|
| 3 |
+
size 164660642
|
precompiled/qualcomm-snapdragon-x-elite/Swin-Base_w8a16.onnx.zip
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 82732598
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fca3b61bf2a6a08c3499268ada365a2a18ea39eac1eb93acbe1a612126b2027b
|
| 3 |
size 82732598
|
precompiled/qualcomm-snapdragon-x-elite/sdk_versions.yml
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
sdk_versions:
|
| 2 |
+
qnn_context_binary:
|
| 3 |
+
qairt: 2.34.2.250528164111_119506
|
| 4 |
+
precompiled_qnn_onnx:
|
| 5 |
+
qairt: 2.33.2.250410134701_117956
|