v0.30.5
Browse filesSee https://github.com/quic/ai-hub-models/releases/v0.30.5 for changelog.
- MobileNet-v3-Large_w8a16.onnx +2 -2
- README.md +51 -51
MobileNet-v3-Large_w8a16.onnx
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:fe4b69168c5795cf03e99a1a0cca0295ef4f648acb945c844d3a3661502858c8
|
| 3 |
+
size 22123072
|
README.md
CHANGED
|
@@ -37,53 +37,53 @@ More details on model performance across various devices, can be found
|
|
| 37 |
|
| 38 |
| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
|
| 39 |
|---|---|---|---|---|---|---|---|---|
|
| 40 |
-
| MobileNet-v3-Large | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE |
|
| 41 |
-
| MobileNet-v3-Large | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN |
|
| 42 |
-
| MobileNet-v3-Large | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 1.
|
| 43 |
-
| MobileNet-v3-Large | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN | 1.
|
| 44 |
-
| MobileNet-v3-Large | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 0.
|
| 45 |
-
| MobileNet-v3-Large | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN | 0.
|
| 46 |
-
| MobileNet-v3-Large | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 1.
|
| 47 |
-
| MobileNet-v3-Large | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN | 1.
|
| 48 |
-
| MobileNet-v3-Large | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE |
|
| 49 |
-
| MobileNet-v3-Large | float | SA7255P ADP | Qualcomm® SA7255P | QNN |
|
| 50 |
-
| MobileNet-v3-Large | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 0.
|
| 51 |
-
| MobileNet-v3-Large | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN | 0.
|
| 52 |
-
| MobileNet-v3-Large | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 1.
|
| 53 |
-
| MobileNet-v3-Large | float | SA8295P ADP | Qualcomm® SA8295P | QNN | 1.
|
| 54 |
-
| MobileNet-v3-Large | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 0.
|
| 55 |
-
| MobileNet-v3-Large | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN | 0.
|
| 56 |
-
| MobileNet-v3-Large | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 1.
|
| 57 |
-
| MobileNet-v3-Large | float | SA8775P ADP | Qualcomm® SA8775P | QNN | 1.
|
| 58 |
-
| MobileNet-v3-Large | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 0.
|
| 59 |
-
| MobileNet-v3-Large | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN | 0.
|
| 60 |
-
| MobileNet-v3-Large | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 0.
|
| 61 |
-
| MobileNet-v3-Large | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 0.
|
| 62 |
-
| MobileNet-v3-Large | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN | 0.
|
| 63 |
-
| MobileNet-v3-Large | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 0.
|
| 64 |
-
| MobileNet-v3-Large | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 0.
|
| 65 |
-
| MobileNet-v3-Large | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN | 0.
|
| 66 |
-
| MobileNet-v3-Large | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 0.565 ms |
|
| 67 |
-
| MobileNet-v3-Large | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 1.
|
| 68 |
-
| MobileNet-v3-Large | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 0.
|
| 69 |
-
| MobileNet-v3-Large | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN |
|
| 70 |
-
| MobileNet-v3-Large | w8a16 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN | 1.
|
| 71 |
-
| MobileNet-v3-Large | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN | 0.
|
| 72 |
-
| MobileNet-v3-Large | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN | 1.
|
| 73 |
-
| MobileNet-v3-Large | w8a16 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | QNN | 3.
|
| 74 |
-
| MobileNet-v3-Large | w8a16 | SA7255P ADP | Qualcomm® SA7255P | QNN |
|
| 75 |
-
| MobileNet-v3-Large | w8a16 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN | 0.
|
| 76 |
-
| MobileNet-v3-Large | w8a16 | SA8295P ADP | Qualcomm® SA8295P | QNN | 1.
|
| 77 |
-
| MobileNet-v3-Large | w8a16 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN | 0.
|
| 78 |
-
| MobileNet-v3-Large | w8a16 | SA8775P ADP | Qualcomm® SA8775P | QNN | 1.
|
| 79 |
-
| MobileNet-v3-Large | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN | 0.
|
| 80 |
-
| MobileNet-v3-Large | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 0.
|
| 81 |
-
| MobileNet-v3-Large | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN | 0.
|
| 82 |
-
| MobileNet-v3-Large | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 0.
|
| 83 |
-
| MobileNet-v3-Large | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN | 0.
|
| 84 |
-
| MobileNet-v3-Large | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 0.
|
| 85 |
-
| MobileNet-v3-Large | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 1.
|
| 86 |
-
| MobileNet-v3-Large | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 0.
|
| 87 |
|
| 88 |
|
| 89 |
|
|
@@ -147,8 +147,8 @@ Profiling Results
|
|
| 147 |
MobileNet-v3-Large
|
| 148 |
Device : cs_8275 (ANDROID 14)
|
| 149 |
Runtime : TFLITE
|
| 150 |
-
Estimated inference time (ms) :
|
| 151 |
-
Estimated peak memory usage (MB): [0,
|
| 152 |
Total # Ops : 128
|
| 153 |
Compute Unit(s) : npu (128 ops) gpu (0 ops) cpu (0 ops)
|
| 154 |
```
|
|
@@ -237,13 +237,13 @@ AI Hub. [Sign up for access](https://myaccount.qualcomm.com/signup).
|
|
| 237 |
You can also run the demo on-device.
|
| 238 |
|
| 239 |
```bash
|
| 240 |
-
python -m qai_hub_models.models.mobilenet_v3_large.demo --on-device
|
| 241 |
```
|
| 242 |
|
| 243 |
**NOTE**: If you want running in a Jupyter Notebook or Google Colab like
|
| 244 |
environment, please add the following to your cell (instead of the above).
|
| 245 |
```
|
| 246 |
-
%run -m qai_hub_models.models.mobilenet_v3_large.demo -- --on-device
|
| 247 |
```
|
| 248 |
|
| 249 |
|
|
|
|
| 37 |
|
| 38 |
| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
|
| 39 |
|---|---|---|---|---|---|---|---|---|
|
| 40 |
+
| MobileNet-v3-Large | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 14.978 ms | 0 - 26 MB | NPU | [MobileNet-v3-Large.tflite](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large.tflite) |
|
| 41 |
+
| MobileNet-v3-Large | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN | 14.34 ms | 1 - 10 MB | NPU | Use Export Script |
|
| 42 |
+
| MobileNet-v3-Large | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 1.373 ms | 0 - 39 MB | NPU | [MobileNet-v3-Large.tflite](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large.tflite) |
|
| 43 |
+
| MobileNet-v3-Large | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN | 1.761 ms | 0 - 33 MB | NPU | Use Export Script |
|
| 44 |
+
| MobileNet-v3-Large | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 0.962 ms | 0 - 94 MB | NPU | [MobileNet-v3-Large.tflite](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large.tflite) |
|
| 45 |
+
| MobileNet-v3-Large | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN | 0.937 ms | 1 - 4 MB | NPU | Use Export Script |
|
| 46 |
+
| MobileNet-v3-Large | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 1.352 ms | 0 - 27 MB | NPU | [MobileNet-v3-Large.tflite](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large.tflite) |
|
| 47 |
+
| MobileNet-v3-Large | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN | 1.272 ms | 1 - 13 MB | NPU | Use Export Script |
|
| 48 |
+
| MobileNet-v3-Large | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 14.978 ms | 0 - 26 MB | NPU | [MobileNet-v3-Large.tflite](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large.tflite) |
|
| 49 |
+
| MobileNet-v3-Large | float | SA7255P ADP | Qualcomm® SA7255P | QNN | 14.34 ms | 1 - 10 MB | NPU | Use Export Script |
|
| 50 |
+
| MobileNet-v3-Large | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 0.964 ms | 0 - 94 MB | NPU | [MobileNet-v3-Large.tflite](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large.tflite) |
|
| 51 |
+
| MobileNet-v3-Large | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN | 0.934 ms | 1 - 3 MB | NPU | Use Export Script |
|
| 52 |
+
| MobileNet-v3-Large | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 1.801 ms | 0 - 28 MB | NPU | [MobileNet-v3-Large.tflite](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large.tflite) |
|
| 53 |
+
| MobileNet-v3-Large | float | SA8295P ADP | Qualcomm® SA8295P | QNN | 1.726 ms | 0 - 17 MB | NPU | Use Export Script |
|
| 54 |
+
| MobileNet-v3-Large | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 0.963 ms | 0 - 94 MB | NPU | [MobileNet-v3-Large.tflite](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large.tflite) |
|
| 55 |
+
| MobileNet-v3-Large | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN | 0.93 ms | 1 - 4 MB | NPU | Use Export Script |
|
| 56 |
+
| MobileNet-v3-Large | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 1.352 ms | 0 - 27 MB | NPU | [MobileNet-v3-Large.tflite](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large.tflite) |
|
| 57 |
+
| MobileNet-v3-Large | float | SA8775P ADP | Qualcomm® SA8775P | QNN | 1.272 ms | 1 - 13 MB | NPU | Use Export Script |
|
| 58 |
+
| MobileNet-v3-Large | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 0.966 ms | 0 - 93 MB | NPU | [MobileNet-v3-Large.tflite](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large.tflite) |
|
| 59 |
+
| MobileNet-v3-Large | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN | 0.925 ms | 0 - 72 MB | NPU | Use Export Script |
|
| 60 |
+
| MobileNet-v3-Large | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 0.808 ms | 0 - 51 MB | NPU | [MobileNet-v3-Large.onnx](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large.onnx) |
|
| 61 |
+
| MobileNet-v3-Large | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 0.647 ms | 0 - 36 MB | NPU | [MobileNet-v3-Large.tflite](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large.tflite) |
|
| 62 |
+
| MobileNet-v3-Large | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN | 0.638 ms | 0 - 31 MB | NPU | Use Export Script |
|
| 63 |
+
| MobileNet-v3-Large | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 0.552 ms | 0 - 35 MB | NPU | [MobileNet-v3-Large.onnx](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large.onnx) |
|
| 64 |
+
| MobileNet-v3-Large | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 0.64 ms | 0 - 32 MB | NPU | [MobileNet-v3-Large.tflite](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large.tflite) |
|
| 65 |
+
| MobileNet-v3-Large | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN | 0.599 ms | 1 - 27 MB | NPU | Use Export Script |
|
| 66 |
+
| MobileNet-v3-Large | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 0.565 ms | 1 - 28 MB | NPU | [MobileNet-v3-Large.onnx](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large.onnx) |
|
| 67 |
+
| MobileNet-v3-Large | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 1.193 ms | 1 - 1 MB | NPU | Use Export Script |
|
| 68 |
+
| MobileNet-v3-Large | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 0.871 ms | 13 - 13 MB | NPU | [MobileNet-v3-Large.onnx](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large.onnx) |
|
| 69 |
+
| MobileNet-v3-Large | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN | 4.016 ms | 0 - 10 MB | NPU | Use Export Script |
|
| 70 |
+
| MobileNet-v3-Large | w8a16 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN | 1.218 ms | 0 - 29 MB | NPU | Use Export Script |
|
| 71 |
+
| MobileNet-v3-Large | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN | 0.943 ms | 0 - 4 MB | NPU | Use Export Script |
|
| 72 |
+
| MobileNet-v3-Large | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN | 1.163 ms | 0 - 13 MB | NPU | Use Export Script |
|
| 73 |
+
| MobileNet-v3-Large | w8a16 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | QNN | 3.115 ms | 0 - 12 MB | NPU | Use Export Script |
|
| 74 |
+
| MobileNet-v3-Large | w8a16 | SA7255P ADP | Qualcomm® SA7255P | QNN | 4.016 ms | 0 - 10 MB | NPU | Use Export Script |
|
| 75 |
+
| MobileNet-v3-Large | w8a16 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN | 0.939 ms | 0 - 2 MB | NPU | Use Export Script |
|
| 76 |
+
| MobileNet-v3-Large | w8a16 | SA8295P ADP | Qualcomm® SA8295P | QNN | 1.512 ms | 0 - 17 MB | NPU | Use Export Script |
|
| 77 |
+
| MobileNet-v3-Large | w8a16 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN | 0.949 ms | 0 - 3 MB | NPU | Use Export Script |
|
| 78 |
+
| MobileNet-v3-Large | w8a16 | SA8775P ADP | Qualcomm® SA8775P | QNN | 1.163 ms | 0 - 13 MB | NPU | Use Export Script |
|
| 79 |
+
| MobileNet-v3-Large | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN | 0.95 ms | 0 - 37 MB | NPU | Use Export Script |
|
| 80 |
+
| MobileNet-v3-Large | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 0.824 ms | 0 - 44 MB | NPU | [MobileNet-v3-Large.onnx](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large_w8a16.onnx) |
|
| 81 |
+
| MobileNet-v3-Large | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN | 0.65 ms | 0 - 37 MB | NPU | Use Export Script |
|
| 82 |
+
| MobileNet-v3-Large | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 0.567 ms | 0 - 40 MB | NPU | [MobileNet-v3-Large.onnx](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large_w8a16.onnx) |
|
| 83 |
+
| MobileNet-v3-Large | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN | 0.551 ms | 0 - 21 MB | NPU | Use Export Script |
|
| 84 |
+
| MobileNet-v3-Large | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 0.58 ms | 0 - 30 MB | NPU | [MobileNet-v3-Large.onnx](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large_w8a16.onnx) |
|
| 85 |
+
| MobileNet-v3-Large | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 1.076 ms | 0 - 0 MB | NPU | Use Export Script |
|
| 86 |
+
| MobileNet-v3-Large | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 0.887 ms | 8 - 8 MB | NPU | [MobileNet-v3-Large.onnx](https://huggingface.co/qualcomm/MobileNet-v3-Large/blob/main/MobileNet-v3-Large_w8a16.onnx) |
|
| 87 |
|
| 88 |
|
| 89 |
|
|
|
|
| 147 |
MobileNet-v3-Large
|
| 148 |
Device : cs_8275 (ANDROID 14)
|
| 149 |
Runtime : TFLITE
|
| 150 |
+
Estimated inference time (ms) : 15.0
|
| 151 |
+
Estimated peak memory usage (MB): [0, 26]
|
| 152 |
Total # Ops : 128
|
| 153 |
Compute Unit(s) : npu (128 ops) gpu (0 ops) cpu (0 ops)
|
| 154 |
```
|
|
|
|
| 237 |
You can also run the demo on-device.
|
| 238 |
|
| 239 |
```bash
|
| 240 |
+
python -m qai_hub_models.models.mobilenet_v3_large.demo --eval-mode on-device
|
| 241 |
```
|
| 242 |
|
| 243 |
**NOTE**: If you want running in a Jupyter Notebook or Google Colab like
|
| 244 |
environment, please add the following to your cell (instead of the above).
|
| 245 |
```
|
| 246 |
+
%run -m qai_hub_models.models.mobilenet_v3_large.demo -- --eval-mode on-device
|
| 247 |
```
|
| 248 |
|
| 249 |
|