Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
@@ -36,29 +36,31 @@ 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 |
-
| Person-Foot-Detection | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 4.
|
40 |
-
| Person-Foot-Detection | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 5.
|
41 |
-
| Person-Foot-Detection | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 5.
|
42 |
-
| Person-Foot-Detection | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 3.
|
43 |
-
| Person-Foot-Detection | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 3.
|
44 |
-
| Person-Foot-Detection | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX |
|
45 |
-
| Person-Foot-Detection | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 3.
|
46 |
-
| Person-Foot-Detection | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 3.
|
47 |
-
| Person-Foot-Detection | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 3.
|
48 |
-
| Person-Foot-Detection | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 4.
|
49 |
-
| Person-Foot-Detection | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 4.
|
50 |
-
| Person-Foot-Detection |
|
51 |
-
| Person-Foot-Detection |
|
52 |
-
| Person-Foot-Detection |
|
53 |
-
| Person-Foot-Detection |
|
54 |
-
| Person-Foot-Detection |
|
55 |
-
| Person-Foot-Detection |
|
56 |
-
| Person-Foot-Detection |
|
57 |
-
| Person-Foot-Detection |
|
58 |
-
| Person-Foot-Detection |
|
59 |
-
| Person-Foot-Detection |
|
60 |
-
| Person-Foot-Detection |
|
61 |
-
| Person-Foot-Detection |
|
|
|
|
|
62 |
|
63 |
|
64 |
|
@@ -122,8 +124,8 @@ Profiling Results
|
|
122 |
Person-Foot-Detection
|
123 |
Device : Samsung Galaxy S23 (13)
|
124 |
Runtime : TFLITE
|
125 |
-
Estimated inference time (ms) : 4.
|
126 |
-
Estimated peak memory usage (MB): [5,
|
127 |
Total # Ops : 135
|
128 |
Compute Unit(s) : NPU (135 ops)
|
129 |
```
|
@@ -144,13 +146,29 @@ in memory using the `jit.trace` and then call the `submit_compile_job` API.
|
|
144 |
import torch
|
145 |
|
146 |
import qai_hub as hub
|
147 |
-
from qai_hub_models.models.foot_track_net import
|
148 |
|
149 |
# Load the model
|
|
|
150 |
|
151 |
# Device
|
152 |
device = hub.Device("Samsung Galaxy S23")
|
153 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
154 |
|
155 |
```
|
156 |
|
|
|
36 |
|
37 |
| Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
|
38 |
|---|---|---|---|---|---|---|---|---|
|
39 |
+
| Person-Foot-Detection | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 4.804 ms | 5 - 13 MB | FP16 | NPU | [Person-Foot-Detection.tflite](https://huggingface.co/qualcomm/Person-Foot-Detection/blob/main/Person-Foot-Detection.tflite) |
|
40 |
+
| Person-Foot-Detection | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 5.05 ms | 4 - 9 MB | FP16 | NPU | [Person-Foot-Detection.so](https://huggingface.co/qualcomm/Person-Foot-Detection/blob/main/Person-Foot-Detection.so) |
|
41 |
+
| Person-Foot-Detection | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 5.12 ms | 16 - 19 MB | FP16 | NPU | [Person-Foot-Detection.onnx](https://huggingface.co/qualcomm/Person-Foot-Detection/blob/main/Person-Foot-Detection.onnx) |
|
42 |
+
| Person-Foot-Detection | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 3.333 ms | 0 - 20 MB | FP16 | NPU | [Person-Foot-Detection.tflite](https://huggingface.co/qualcomm/Person-Foot-Detection/blob/main/Person-Foot-Detection.tflite) |
|
43 |
+
| Person-Foot-Detection | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 3.511 ms | 0 - 21 MB | FP16 | NPU | [Person-Foot-Detection.so](https://huggingface.co/qualcomm/Person-Foot-Detection/blob/main/Person-Foot-Detection.so) |
|
44 |
+
| Person-Foot-Detection | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 3.621 ms | 2 - 71 MB | FP16 | NPU | [Person-Foot-Detection.onnx](https://huggingface.co/qualcomm/Person-Foot-Detection/blob/main/Person-Foot-Detection.onnx) |
|
45 |
+
| Person-Foot-Detection | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 3.407 ms | 0 - 18 MB | FP16 | NPU | [Person-Foot-Detection.tflite](https://huggingface.co/qualcomm/Person-Foot-Detection/blob/main/Person-Foot-Detection.tflite) |
|
46 |
+
| Person-Foot-Detection | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 3.585 ms | 0 - 18 MB | FP16 | NPU | Use Export Script |
|
47 |
+
| Person-Foot-Detection | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 3.539 ms | 0 - 36 MB | FP16 | NPU | [Person-Foot-Detection.onnx](https://huggingface.co/qualcomm/Person-Foot-Detection/blob/main/Person-Foot-Detection.onnx) |
|
48 |
+
| Person-Foot-Detection | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 4.772 ms | 5 - 11 MB | FP16 | NPU | [Person-Foot-Detection.tflite](https://huggingface.co/qualcomm/Person-Foot-Detection/blob/main/Person-Foot-Detection.tflite) |
|
49 |
+
| Person-Foot-Detection | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 4.873 ms | 4 - 5 MB | FP16 | NPU | Use Export Script |
|
50 |
+
| Person-Foot-Detection | SA7255P ADP | SA7255P | TFLITE | 156.75 ms | 5 - 21 MB | FP16 | NPU | [Person-Foot-Detection.tflite](https://huggingface.co/qualcomm/Person-Foot-Detection/blob/main/Person-Foot-Detection.tflite) |
|
51 |
+
| Person-Foot-Detection | SA7255P ADP | SA7255P | QNN | 156.903 ms | 3 - 9 MB | FP16 | NPU | Use Export Script |
|
52 |
+
| Person-Foot-Detection | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 4.87 ms | 5 - 13 MB | FP16 | NPU | [Person-Foot-Detection.tflite](https://huggingface.co/qualcomm/Person-Foot-Detection/blob/main/Person-Foot-Detection.tflite) |
|
53 |
+
| Person-Foot-Detection | SA8255 (Proxy) | SA8255P Proxy | QNN | 4.867 ms | 4 - 6 MB | FP16 | NPU | Use Export Script |
|
54 |
+
| Person-Foot-Detection | SA8295P ADP | SA8295P | TFLITE | 8.296 ms | 5 - 20 MB | FP16 | NPU | [Person-Foot-Detection.tflite](https://huggingface.co/qualcomm/Person-Foot-Detection/blob/main/Person-Foot-Detection.tflite) |
|
55 |
+
| Person-Foot-Detection | SA8295P ADP | SA8295P | QNN | 8.873 ms | 0 - 6 MB | FP16 | NPU | Use Export Script |
|
56 |
+
| Person-Foot-Detection | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 4.87 ms | 5 - 14 MB | FP16 | NPU | [Person-Foot-Detection.tflite](https://huggingface.co/qualcomm/Person-Foot-Detection/blob/main/Person-Foot-Detection.tflite) |
|
57 |
+
| Person-Foot-Detection | SA8650 (Proxy) | SA8650P Proxy | QNN | 4.939 ms | 4 - 6 MB | FP16 | NPU | Use Export Script |
|
58 |
+
| Person-Foot-Detection | SA8775P ADP | SA8775P | TFLITE | 9.24 ms | 5 - 22 MB | FP16 | NPU | [Person-Foot-Detection.tflite](https://huggingface.co/qualcomm/Person-Foot-Detection/blob/main/Person-Foot-Detection.tflite) |
|
59 |
+
| Person-Foot-Detection | SA8775P ADP | SA8775P | QNN | 9.613 ms | 3 - 8 MB | FP16 | NPU | Use Export Script |
|
60 |
+
| Person-Foot-Detection | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 7.029 ms | 5 - 27 MB | FP16 | NPU | [Person-Foot-Detection.tflite](https://huggingface.co/qualcomm/Person-Foot-Detection/blob/main/Person-Foot-Detection.tflite) |
|
61 |
+
| Person-Foot-Detection | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 7.509 ms | 4 - 25 MB | FP16 | NPU | Use Export Script |
|
62 |
+
| Person-Foot-Detection | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 5.46 ms | 4 - 4 MB | FP16 | NPU | Use Export Script |
|
63 |
+
| Person-Foot-Detection | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 5.765 ms | 17 - 17 MB | FP16 | NPU | [Person-Foot-Detection.onnx](https://huggingface.co/qualcomm/Person-Foot-Detection/blob/main/Person-Foot-Detection.onnx) |
|
64 |
|
65 |
|
66 |
|
|
|
124 |
Person-Foot-Detection
|
125 |
Device : Samsung Galaxy S23 (13)
|
126 |
Runtime : TFLITE
|
127 |
+
Estimated inference time (ms) : 4.8
|
128 |
+
Estimated peak memory usage (MB): [5, 13]
|
129 |
Total # Ops : 135
|
130 |
Compute Unit(s) : NPU (135 ops)
|
131 |
```
|
|
|
146 |
import torch
|
147 |
|
148 |
import qai_hub as hub
|
149 |
+
from qai_hub_models.models.foot_track_net import Model
|
150 |
|
151 |
# Load the model
|
152 |
+
torch_model = Model.from_pretrained()
|
153 |
|
154 |
# Device
|
155 |
device = hub.Device("Samsung Galaxy S23")
|
156 |
|
157 |
+
# Trace model
|
158 |
+
input_shape = torch_model.get_input_spec()
|
159 |
+
sample_inputs = torch_model.sample_inputs()
|
160 |
+
|
161 |
+
pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])
|
162 |
+
|
163 |
+
# Compile model on a specific device
|
164 |
+
compile_job = hub.submit_compile_job(
|
165 |
+
model=pt_model,
|
166 |
+
device=device,
|
167 |
+
input_specs=torch_model.get_input_spec(),
|
168 |
+
)
|
169 |
+
|
170 |
+
# Get target model to run on-device
|
171 |
+
target_model = compile_job.get_target_model()
|
172 |
|
173 |
```
|
174 |
|