Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
@@ -36,38 +36,38 @@ More details on model performance across various devices, can be found
|
|
36 |
|
37 |
| Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
|
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 |
|
@@ -128,11 +128,11 @@ python -m qai_hub_models.models.resnet50_quantized.export
|
|
128 |
```
|
129 |
Profiling Results
|
130 |
------------------------------------------------------------
|
131 |
-
|
132 |
Device : Samsung Galaxy S23 (13)
|
133 |
Runtime : TFLITE
|
134 |
Estimated inference time (ms) : 0.8
|
135 |
-
Estimated peak memory usage (MB): [0,
|
136 |
Total # Ops : 82
|
137 |
Compute Unit(s) : NPU (82 ops)
|
138 |
```
|
@@ -216,20 +216,6 @@ AI Hub. [Sign up for access](https://myaccount.qualcomm.com/signup).
|
|
216 |
|
217 |
|
218 |
|
219 |
-
## Run demo on a cloud-hosted device
|
220 |
-
|
221 |
-
You can also run the demo on-device.
|
222 |
-
|
223 |
-
```bash
|
224 |
-
python -m qai_hub_models.models.resnet50_quantized.demo --on-device
|
225 |
-
```
|
226 |
-
|
227 |
-
**NOTE**: If you want running in a Jupyter Notebook or Google Colab like
|
228 |
-
environment, please add the following to your cell (instead of the above).
|
229 |
-
```
|
230 |
-
%run -m qai_hub_models.models.resnet50_quantized.demo -- --on-device
|
231 |
-
```
|
232 |
-
|
233 |
|
234 |
## Deploying compiled model to Android
|
235 |
|
|
|
36 |
|
37 |
| Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
|
38 |
|---|---|---|---|---|---|---|---|---|
|
39 |
+
| ResNet50 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 0.772 ms | 0 - 95 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.tflite) |
|
40 |
+
| ResNet50 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 0.89 ms | 0 - 3 MB | INT8 | NPU | [ResNet50Quantized.so](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.so) |
|
41 |
+
| ResNet50 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 1.059 ms | 0 - 133 MB | INT8 | NPU | [ResNet50Quantized.onnx](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.onnx) |
|
42 |
+
| ResNet50 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 0.596 ms | 0 - 58 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.tflite) |
|
43 |
+
| ResNet50 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 0.699 ms | 0 - 19 MB | INT8 | NPU | [ResNet50Quantized.so](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.so) |
|
44 |
+
| ResNet50 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 0.784 ms | 0 - 57 MB | INT8 | NPU | [ResNet50Quantized.onnx](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.onnx) |
|
45 |
+
| ResNet50 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 0.557 ms | 0 - 29 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.tflite) |
|
46 |
+
| ResNet50 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 0.635 ms | 0 - 32 MB | INT8 | NPU | Use Export Script |
|
47 |
+
| ResNet50 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 0.83 ms | 0 - 40 MB | INT8 | NPU | [ResNet50Quantized.onnx](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.onnx) |
|
48 |
+
| ResNet50 | SA7255P ADP | SA7255P | TFLITE | 8.702 ms | 0 - 26 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.tflite) |
|
49 |
+
| ResNet50 | SA7255P ADP | SA7255P | QNN | 9.153 ms | 0 - 10 MB | INT8 | NPU | Use Export Script |
|
50 |
+
| ResNet50 | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 0.785 ms | 0 - 94 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.tflite) |
|
51 |
+
| ResNet50 | SA8255 (Proxy) | SA8255P Proxy | QNN | 0.891 ms | 0 - 3 MB | INT8 | NPU | Use Export Script |
|
52 |
+
| ResNet50 | SA8295P ADP | SA8295P | TFLITE | 1.263 ms | 0 - 29 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.tflite) |
|
53 |
+
| ResNet50 | SA8295P ADP | SA8295P | QNN | 1.408 ms | 0 - 18 MB | INT8 | NPU | Use Export Script |
|
54 |
+
| ResNet50 | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 0.781 ms | 0 - 6 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.tflite) |
|
55 |
+
| ResNet50 | SA8650 (Proxy) | SA8650P Proxy | QNN | 0.895 ms | 0 - 3 MB | INT8 | NPU | Use Export Script |
|
56 |
+
| ResNet50 | SA8775P ADP | SA8775P | TFLITE | 1.059 ms | 0 - 27 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.tflite) |
|
57 |
+
| ResNet50 | SA8775P ADP | SA8775P | QNN | 1.193 ms | 0 - 10 MB | INT8 | NPU | Use Export Script |
|
58 |
+
| ResNet50 | RB3 Gen 2 (Proxy) | QCS6490 Proxy | TFLITE | 2.758 ms | 0 - 47 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.tflite) |
|
59 |
+
| ResNet50 | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN | 3.869 ms | 0 - 14 MB | INT8 | NPU | Use Export Script |
|
60 |
+
| ResNet50 | RB5 (Proxy) | QCS8250 Proxy | TFLITE | 11.782 ms | 0 - 3 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.tflite) |
|
61 |
+
| ResNet50 | QCS8275 (Proxy) | QCS8275 Proxy | TFLITE | 8.702 ms | 0 - 26 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.tflite) |
|
62 |
+
| ResNet50 | QCS8275 (Proxy) | QCS8275 Proxy | QNN | 9.153 ms | 0 - 10 MB | INT8 | NPU | Use Export Script |
|
63 |
+
| ResNet50 | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 0.778 ms | 0 - 92 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.tflite) |
|
64 |
+
| ResNet50 | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 0.893 ms | 0 - 3 MB | INT8 | NPU | Use Export Script |
|
65 |
+
| ResNet50 | QCS9075 (Proxy) | QCS9075 Proxy | TFLITE | 1.059 ms | 0 - 27 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.tflite) |
|
66 |
+
| ResNet50 | QCS9075 (Proxy) | QCS9075 Proxy | QNN | 1.193 ms | 0 - 10 MB | INT8 | NPU | Use Export Script |
|
67 |
+
| ResNet50 | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 1.006 ms | 0 - 55 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.tflite) |
|
68 |
+
| ResNet50 | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 1.171 ms | 0 - 54 MB | INT8 | NPU | Use Export Script |
|
69 |
+
| ResNet50 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 0.951 ms | 0 - 0 MB | INT8 | NPU | Use Export Script |
|
70 |
+
| ResNet50 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 1.005 ms | 27 - 27 MB | INT8 | NPU | [ResNet50Quantized.onnx](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50.onnx) |
|
71 |
|
72 |
|
73 |
|
|
|
128 |
```
|
129 |
Profiling Results
|
130 |
------------------------------------------------------------
|
131 |
+
ResNet50
|
132 |
Device : Samsung Galaxy S23 (13)
|
133 |
Runtime : TFLITE
|
134 |
Estimated inference time (ms) : 0.8
|
135 |
+
Estimated peak memory usage (MB): [0, 95]
|
136 |
Total # Ops : 82
|
137 |
Compute Unit(s) : NPU (82 ops)
|
138 |
```
|
|
|
216 |
|
217 |
|
218 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
219 |
|
220 |
## Deploying compiled model to Android
|
221 |
|