Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -36,42 +36,42 @@ 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 |
-
| ResNet50Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 0.
|
| 40 |
-
| ResNet50Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 1.
|
| 41 |
-
| ResNet50Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 1.
|
| 42 |
-
| ResNet50Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 0.
|
| 43 |
-
| ResNet50Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 0.
|
| 44 |
-
| ResNet50Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 1.
|
| 45 |
-
| ResNet50Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 0.
|
| 46 |
-
| ResNet50Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 0.
|
| 47 |
-
| ResNet50Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX |
|
| 48 |
-
| ResNet50Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | TFLITE | 2.
|
| 49 |
-
| ResNet50Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN |
|
| 50 |
-
| ResNet50Quantized | RB5 (Proxy) | QCS8250 Proxy | TFLITE |
|
| 51 |
-
| ResNet50Quantized | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 0.
|
| 52 |
-
| ResNet50Quantized | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 0.
|
| 53 |
-
| ResNet50Quantized | SA7255P ADP | SA7255P | TFLITE | 8.
|
| 54 |
-
| ResNet50Quantized | SA7255P ADP | SA7255P | QNN | 9.
|
| 55 |
-
| ResNet50Quantized | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 0.
|
| 56 |
-
| ResNet50Quantized | SA8255 (Proxy) | SA8255P Proxy | QNN | 0.
|
| 57 |
-
| ResNet50Quantized | SA8295P ADP | SA8295P | TFLITE | 1.
|
| 58 |
-
| ResNet50Quantized | SA8295P ADP | SA8295P | QNN | 1.
|
| 59 |
-
| ResNet50Quantized | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 0.789 ms | 0 -
|
| 60 |
| ResNet50Quantized | SA8650 (Proxy) | SA8650P Proxy | QNN | 0.949 ms | 0 - 2 MB | INT8 | NPU | Use Export Script |
|
| 61 |
-
| ResNet50Quantized | SA8775P ADP | SA8775P | TFLITE | 1.
|
| 62 |
-
| ResNet50Quantized | SA8775P ADP | SA8775P | QNN | 1.
|
| 63 |
-
| ResNet50Quantized | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 0.
|
| 64 |
-
| ResNet50Quantized | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 1.
|
| 65 |
-
| ResNet50Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN |
|
| 66 |
-
| ResNet50Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 1.
|
| 67 |
|
| 68 |
|
| 69 |
|
| 70 |
|
| 71 |
## Installation
|
| 72 |
|
| 73 |
-
This model can be installed as a Python package via pip.
|
| 74 |
|
|
|
|
| 75 |
```bash
|
| 76 |
pip install qai-hub-models
|
| 77 |
```
|
|
@@ -128,7 +128,7 @@ ResNet50Quantized
|
|
| 128 |
Device : Samsung Galaxy S23 (13)
|
| 129 |
Runtime : TFLITE
|
| 130 |
Estimated inference time (ms) : 0.8
|
| 131 |
-
Estimated peak memory usage (MB): [0,
|
| 132 |
Total # Ops : 82
|
| 133 |
Compute Unit(s) : NPU (82 ops)
|
| 134 |
```
|
|
@@ -155,7 +155,7 @@ from qai_hub_models.models.resnet50_quantized import Model
|
|
| 155 |
torch_model = Model.from_pretrained()
|
| 156 |
|
| 157 |
# Device
|
| 158 |
-
device = hub.Device("Samsung Galaxy
|
| 159 |
|
| 160 |
# Trace model
|
| 161 |
input_shape = torch_model.get_input_spec()
|
|
@@ -247,7 +247,8 @@ Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
|
|
| 247 |
|
| 248 |
|
| 249 |
## License
|
| 250 |
-
* The license for the original implementation of ResNet50Quantized can be found
|
|
|
|
| 251 |
* The license for the compiled assets for on-device deployment can be found [here](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/Qualcomm+AI+Hub+Proprietary+License.pdf)
|
| 252 |
|
| 253 |
|
|
|
|
| 36 |
|
| 37 |
| Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
|
| 38 |
|---|---|---|---|---|---|---|---|---|
|
| 39 |
+
| ResNet50Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 0.787 ms | 0 - 44 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
|
| 40 |
+
| ResNet50Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 1.004 ms | 0 - 42 MB | INT8 | NPU | [ResNet50Quantized.so](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.so) |
|
| 41 |
+
| ResNet50Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 1.437 ms | 0 - 62 MB | INT8 | NPU | [ResNet50Quantized.onnx](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.onnx) |
|
| 42 |
+
| ResNet50Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 0.592 ms | 0 - 34 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
|
| 43 |
+
| ResNet50Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 0.754 ms | 0 - 31 MB | INT8 | NPU | [ResNet50Quantized.so](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.so) |
|
| 44 |
+
| ResNet50Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 1.046 ms | 0 - 41 MB | INT8 | NPU | [ResNet50Quantized.onnx](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.onnx) |
|
| 45 |
+
| ResNet50Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 0.577 ms | 0 - 22 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
|
| 46 |
+
| ResNet50Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 0.628 ms | 0 - 24 MB | INT8 | NPU | Use Export Script |
|
| 47 |
+
| ResNet50Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 1.059 ms | 0 - 33 MB | INT8 | NPU | [ResNet50Quantized.onnx](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.onnx) |
|
| 48 |
+
| ResNet50Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | TFLITE | 2.762 ms | 0 - 23 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
|
| 49 |
+
| ResNet50Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN | 4.009 ms | 0 - 12 MB | INT8 | NPU | Use Export Script |
|
| 50 |
+
| ResNet50Quantized | RB5 (Proxy) | QCS8250 Proxy | TFLITE | 11.738 ms | 0 - 2 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
|
| 51 |
+
| ResNet50Quantized | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 0.789 ms | 0 - 44 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
|
| 52 |
+
| ResNet50Quantized | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 0.954 ms | 0 - 3 MB | INT8 | NPU | Use Export Script |
|
| 53 |
+
| ResNet50Quantized | SA7255P ADP | SA7255P | TFLITE | 8.685 ms | 0 - 17 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
|
| 54 |
+
| ResNet50Quantized | SA7255P ADP | SA7255P | QNN | 9.474 ms | 0 - 9 MB | INT8 | NPU | Use Export Script |
|
| 55 |
+
| ResNet50Quantized | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 0.786 ms | 0 - 44 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
|
| 56 |
+
| ResNet50Quantized | SA8255 (Proxy) | SA8255P Proxy | QNN | 0.952 ms | 0 - 2 MB | INT8 | NPU | Use Export Script |
|
| 57 |
+
| ResNet50Quantized | SA8295P ADP | SA8295P | TFLITE | 1.248 ms | 0 - 22 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
|
| 58 |
+
| ResNet50Quantized | SA8295P ADP | SA8295P | QNN | 1.727 ms | 0 - 14 MB | INT8 | NPU | Use Export Script |
|
| 59 |
+
| ResNet50Quantized | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 0.789 ms | 0 - 7 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
|
| 60 |
| ResNet50Quantized | SA8650 (Proxy) | SA8650P Proxy | QNN | 0.949 ms | 0 - 2 MB | INT8 | NPU | Use Export Script |
|
| 61 |
+
| ResNet50Quantized | SA8775P ADP | SA8775P | TFLITE | 1.07 ms | 0 - 17 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
|
| 62 |
+
| ResNet50Quantized | SA8775P ADP | SA8775P | QNN | 1.434 ms | 0 - 10 MB | INT8 | NPU | Use Export Script |
|
| 63 |
+
| ResNet50Quantized | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 0.925 ms | 0 - 31 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
|
| 64 |
+
| ResNet50Quantized | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 1.174 ms | 0 - 19 MB | INT8 | NPU | Use Export Script |
|
| 65 |
+
| ResNet50Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 1.006 ms | 0 - 0 MB | INT8 | NPU | Use Export Script |
|
| 66 |
+
| ResNet50Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 1.477 ms | 29 - 29 MB | INT8 | NPU | [ResNet50Quantized.onnx](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.onnx) |
|
| 67 |
|
| 68 |
|
| 69 |
|
| 70 |
|
| 71 |
## Installation
|
| 72 |
|
|
|
|
| 73 |
|
| 74 |
+
Install the package via pip:
|
| 75 |
```bash
|
| 76 |
pip install qai-hub-models
|
| 77 |
```
|
|
|
|
| 128 |
Device : Samsung Galaxy S23 (13)
|
| 129 |
Runtime : TFLITE
|
| 130 |
Estimated inference time (ms) : 0.8
|
| 131 |
+
Estimated peak memory usage (MB): [0, 44]
|
| 132 |
Total # Ops : 82
|
| 133 |
Compute Unit(s) : NPU (82 ops)
|
| 134 |
```
|
|
|
|
| 155 |
torch_model = Model.from_pretrained()
|
| 156 |
|
| 157 |
# Device
|
| 158 |
+
device = hub.Device("Samsung Galaxy S24")
|
| 159 |
|
| 160 |
# Trace model
|
| 161 |
input_shape = torch_model.get_input_spec()
|
|
|
|
| 247 |
|
| 248 |
|
| 249 |
## License
|
| 250 |
+
* The license for the original implementation of ResNet50Quantized can be found
|
| 251 |
+
[here](https://github.com/pytorch/vision/blob/main/LICENSE).
|
| 252 |
* The license for the compiled assets for on-device deployment can be found [here](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/Qualcomm+AI+Hub+Proprietary+License.pdf)
|
| 253 |
|
| 254 |
|