Hugging Face's logo

qualcomm
/
ResNet50Quantized

Image Classification
PyTorch
LiteRT
ONNX
backbone
quantized
android
Model card Files Community
qaihm-bot commited on
Commit
8628470
·
verified ·
1 Parent(s): 5ee4b02

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +42 -24
README.md CHANGED
@@ -36,28 +36,30 @@ 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 | 0.787 ms | 0 - 16 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
40
- | ResNet50Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 0.999 ms | 0 - 261 MB | INT8 | NPU | [ResNet50Quantized.so](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.so) |
41
- | ResNet50Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 0.59 ms | 0 - 63 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
42
- | ResNet50Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 0.779 ms | 0 - 18 MB | INT8 | NPU | [ResNet50Quantized.so](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.so) |
43
- | ResNet50Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 0.506 ms | 0 - 23 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
44
- | ResNet50Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 0.722 ms | 0 - 15 MB | INT8 | NPU | Use Export Script |
45
- | ResNet50Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | TFLITE | 2.693 ms | 0 - 25 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
46
- | ResNet50Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN | 4.071 ms | 0 - 8 MB | INT8 | NPU | Use Export Script |
47
- | ResNet50Quantized | RB5 (Proxy) | QCS8250 Proxy | TFLITE | 11.509 ms | 0 - 2 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
48
- | ResNet50Quantized | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 0.775 ms | 0 - 1 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
49
- | ResNet50Quantized | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 0.944 ms | 0 - 1 MB | INT8 | NPU | Use Export Script |
50
- | ResNet50Quantized | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 0.79 ms | 0 - 2 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
51
- | ResNet50Quantized | SA8255 (Proxy) | SA8255P Proxy | QNN | 0.951 ms | 0 - 1 MB | INT8 | NPU | Use Export Script |
52
- | ResNet50Quantized | SA8775 (Proxy) | SA8775P Proxy | TFLITE | 0.785 ms | 0 - 1 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
53
- | ResNet50Quantized | SA8775 (Proxy) | SA8775P Proxy | QNN | 0.949 ms | 0 - 2 MB | INT8 | NPU | Use Export Script |
54
- | ResNet50Quantized | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 0.788 ms | 0 - 1 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
55
- | ResNet50Quantized | SA8650 (Proxy) | SA8650P Proxy | QNN | 0.949 ms | 0 - 1 MB | INT8 | NPU | Use Export Script |
56
- | ResNet50Quantized | SA8295P ADP | SA8295P | TFLITE | 1.254 ms | 0 - 22 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
57
- | ResNet50Quantized | SA8295P ADP | SA8295P | QNN | 1.753 ms | 0 - 6 MB | INT8 | NPU | Use Export Script |
58
- | ResNet50Quantized | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 0.906 ms | 0 - 64 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
59
- | ResNet50Quantized | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 1.139 ms | 0 - 19 MB | INT8 | NPU | Use Export Script |
60
- | ResNet50Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 0.997 ms | 0 - 0 MB | INT8 | NPU | Use Export Script |
 
 
61
 
62
 
63
 
@@ -122,7 +124,7 @@ ResNet50Quantized
122
  Device : Samsung Galaxy S23 (13)
123
  Runtime : TFLITE
124
  Estimated inference time (ms) : 0.8
125
- Estimated peak memory usage (MB): [0, 16]
126
  Total # Ops : 82
127
  Compute Unit(s) : NPU (82 ops)
128
  ```
@@ -143,13 +145,29 @@ in memory using the `jit.trace` and then call the `submit_compile_job` API.
143
  import torch
144
 
145
  import qai_hub as hub
146
- from qai_hub_models.models.resnet50_quantized import
147
 
148
  # Load the model
 
149
 
150
  # Device
151
  device = hub.Device("Samsung Galaxy S23")
152
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
153
 
154
  ```
155
 
 
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 | 0.789 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 | 1.01 ms | 0 - 32 MB | INT8 | NPU | [ResNet50Quantized.so](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.so) |
41
+ | ResNet50Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 0.597 ms | 0 - 22 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
42
+ | ResNet50Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 0.755 ms | 0 - 17 MB | INT8 | NPU | [ResNet50Quantized.so](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.so) |
43
+ | ResNet50Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 0.544 ms | 0 - 17 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
44
+ | ResNet50Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 0.726 ms | 0 - 17 MB | INT8 | NPU | Use Export Script |
45
+ | ResNet50Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | TFLITE | 2.713 ms | 0 - 22 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
46
+ | ResNet50Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN | 3.965 ms | 0 - 8 MB | INT8 | NPU | Use Export Script |
47
+ | ResNet50Quantized | RB5 (Proxy) | QCS8250 Proxy | TFLITE | 11.577 ms | 0 - 2 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
48
+ | ResNet50Quantized | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 0.786 ms | 0 - 34 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
49
+ | ResNet50Quantized | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 0.945 ms | 0 - 2 MB | INT8 | NPU | Use Export Script |
50
+ | ResNet50Quantized | SA7255P ADP | SA7255P | TFLITE | 8.838 ms | 0 - 15 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
51
+ | ResNet50Quantized | SA7255P ADP | SA7255P | QNN | 9.593 ms | 0 - 5 MB | INT8 | NPU | Use Export Script |
52
+ | ResNet50Quantized | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 0.79 ms | 0 - 265 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
53
+ | ResNet50Quantized | SA8255 (Proxy) | SA8255P Proxy | QNN | 0.952 ms | 0 - 2 MB | INT8 | NPU | Use Export Script |
54
+ | ResNet50Quantized | SA8295P ADP | SA8295P | TFLITE | 1.241 ms | 0 - 14 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
55
+ | ResNet50Quantized | SA8295P ADP | SA8295P | QNN | 1.874 ms | 0 - 6 MB | INT8 | NPU | Use Export Script |
56
+ | ResNet50Quantized | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 0.781 ms | 0 - 43 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
57
+ | ResNet50Quantized | SA8650 (Proxy) | SA8650P Proxy | QNN | 0.951 ms | 0 - 1 MB | INT8 | NPU | Use Export Script |
58
+ | ResNet50Quantized | SA8775P ADP | SA8775P | TFLITE | 1.081 ms | 0 - 16 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
59
+ | ResNet50Quantized | SA8775P ADP | SA8775P | QNN | 1.462 ms | 0 - 6 MB | INT8 | NPU | Use Export Script |
60
+ | ResNet50Quantized | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 0.906 ms | 0 - 23 MB | INT8 | NPU | [ResNet50Quantized.tflite](https://huggingface.co/qualcomm/ResNet50Quantized/blob/main/ResNet50Quantized.tflite) |
61
+ | ResNet50Quantized | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 1.13 ms | 0 - 18 MB | INT8 | NPU | Use Export Script |
62
+ | ResNet50Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 1.419 ms | 0 - 0 MB | INT8 | NPU | Use Export Script |
63
 
64
 
65
 
 
124
  Device : Samsung Galaxy S23 (13)
125
  Runtime : TFLITE
126
  Estimated inference time (ms) : 0.8
127
+ Estimated peak memory usage (MB): [0, 44]
128
  Total # Ops : 82
129
  Compute Unit(s) : NPU (82 ops)
130
  ```
 
145
  import torch
146
 
147
  import qai_hub as hub
148
+ from qai_hub_models.models.resnet50_quantized import Model
149
 
150
  # Load the model
151
+ torch_model = Model.from_pretrained()
152
 
153
  # Device
154
  device = hub.Device("Samsung Galaxy S23")
155
 
156
+ # Trace model
157
+ input_shape = torch_model.get_input_spec()
158
+ sample_inputs = torch_model.sample_inputs()
159
+
160
+ pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])
161
+
162
+ # Compile model on a specific device
163
+ compile_job = hub.submit_compile_job(
164
+ model=pt_model,
165
+ device=device,
166
+ input_specs=torch_model.get_input_spec(),
167
+ )
168
+
169
+ # Get target model to run on-device
170
+ target_model = compile_job.get_target_model()
171
 
172
  ```
173