Hugging Face's logo

qualcomm
/
DenseNet-121

Image Classification
PyTorch
TF Lite
ONNX
backbone
android
Model card Files Community
bhushans commited on
Commit
efbd8d1
·
verified ·
1 Parent(s): b76809d

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +43 -25
README.md CHANGED
@@ -35,29 +35,31 @@ More details on model performance across various devices, can be found
35
 
36
  | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
37
  |---|---|---|---|---|---|---|---|---|
38
- | DenseNet-121 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 1.927 ms | 0 - 230 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
39
- | DenseNet-121 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 1.99 ms | 0 - 29 MB | FP16 | NPU | [DenseNet-121.so](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.so) |
40
- | DenseNet-121 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 1.879 ms | 0 - 18 MB | FP16 | NPU | [DenseNet-121.onnx](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.onnx) |
41
- | DenseNet-121 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 1.287 ms | 0 - 100 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
42
- | DenseNet-121 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 1.325 ms | 1 - 20 MB | FP16 | NPU | [DenseNet-121.so](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.so) |
43
- | DenseNet-121 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 1.321 ms | 0 - 106 MB | FP16 | NPU | [DenseNet-121.onnx](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.onnx) |
44
- | DenseNet-121 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 1.224 ms | 0 - 26 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
45
- | DenseNet-121 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 1.292 ms | 0 - 18 MB | FP16 | NPU | Use Export Script |
46
- | DenseNet-121 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 1.297 ms | 0 - 31 MB | FP16 | NPU | [DenseNet-121.onnx](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.onnx) |
47
- | DenseNet-121 | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 1.921 ms | 0 - 1 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
48
- | DenseNet-121 | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 1.786 ms | 1 - 2 MB | FP16 | NPU | Use Export Script |
49
- | DenseNet-121 | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 1.925 ms | 0 - 1 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
50
- | DenseNet-121 | SA8255 (Proxy) | SA8255P Proxy | QNN | 1.806 ms | 0 - 1 MB | FP16 | NPU | Use Export Script |
51
- | DenseNet-121 | SA8775 (Proxy) | SA8775P Proxy | TFLITE | 1.92 ms | 0 - 2 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
52
- | DenseNet-121 | SA8775 (Proxy) | SA8775P Proxy | QNN | 1.795 ms | 1 - 2 MB | FP16 | NPU | Use Export Script |
53
- | DenseNet-121 | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 1.93 ms | 0 - 2 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
54
- | DenseNet-121 | SA8650 (Proxy) | SA8650P Proxy | QNN | 1.802 ms | 1 - 2 MB | FP16 | NPU | Use Export Script |
55
- | DenseNet-121 | SA8295P ADP | SA8295P | TFLITE | 3.339 ms | 0 - 26 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
56
- | DenseNet-121 | SA8295P ADP | SA8295P | QNN | 3.289 ms | 1 - 6 MB | FP16 | NPU | Use Export Script |
57
- | DenseNet-121 | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 2.607 ms | 0 - 102 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
58
- | DenseNet-121 | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 2.695 ms | 1 - 22 MB | FP16 | NPU | Use Export Script |
59
- | DenseNet-121 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 2.072 ms | 1 - 1 MB | FP16 | NPU | Use Export Script |
60
- | DenseNet-121 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 2.051 ms | 16 - 16 MB | FP16 | NPU | [DenseNet-121.onnx](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.onnx) |
 
 
61
 
62
 
63
 
@@ -122,7 +124,7 @@ DenseNet-121
122
  Device : Samsung Galaxy S23 (13)
123
  Runtime : TFLITE
124
  Estimated inference time (ms) : 1.9
125
- Estimated peak memory usage (MB): [0, 230]
126
  Total # Ops : 312
127
  Compute Unit(s) : NPU (312 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.densenet121 import
147
 
148
  # Load the model
 
149
 
150
  # Device
151
  device = hub.Device("Samsung Galaxy S23")
152
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
153
 
154
  ```
155
 
 
35
 
36
  | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
37
  |---|---|---|---|---|---|---|---|---|
38
+ | DenseNet-121 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 1.926 ms | 0 - 47 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
39
+ | DenseNet-121 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 1.996 ms | 0 - 27 MB | FP16 | NPU | [DenseNet-121.so](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.so) |
40
+ | DenseNet-121 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 1.866 ms | 0 - 17 MB | FP16 | NPU | [DenseNet-121.onnx](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.onnx) |
41
+ | DenseNet-121 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 1.281 ms | 0 - 21 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
42
+ | DenseNet-121 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 1.334 ms | 1 - 20 MB | FP16 | NPU | [DenseNet-121.so](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.so) |
43
+ | DenseNet-121 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 1.327 ms | 0 - 106 MB | FP16 | NPU | [DenseNet-121.onnx](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.onnx) |
44
+ | DenseNet-121 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 1.248 ms | 0 - 20 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
45
+ | DenseNet-121 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 1.064 ms | 1 - 19 MB | FP16 | NPU | Use Export Script |
46
+ | DenseNet-121 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 1.299 ms | 0 - 31 MB | FP16 | NPU | [DenseNet-121.onnx](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.onnx) |
47
+ | DenseNet-121 | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 1.921 ms | 0 - 8 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
48
+ | DenseNet-121 | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 1.789 ms | 1 - 2 MB | FP16 | NPU | Use Export Script |
49
+ | DenseNet-121 | SA7255P ADP | SA7255P | TFLITE | 53.464 ms | 0 - 19 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
50
+ | DenseNet-121 | SA7255P ADP | SA7255P | QNN | 53.597 ms | 3 - 8 MB | FP16 | NPU | Use Export Script |
51
+ | DenseNet-121 | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 1.921 ms | 0 - 69 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
52
+ | DenseNet-121 | SA8255 (Proxy) | SA8255P Proxy | QNN | 1.799 ms | 1 - 2 MB | FP16 | NPU | Use Export Script |
53
+ | DenseNet-121 | SA8295P ADP | SA8295P | TFLITE | 3.319 ms | 0 - 17 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
54
+ | DenseNet-121 | SA8295P ADP | SA8295P | QNN | 3.286 ms | 0 - 6 MB | FP16 | NPU | Use Export Script |
55
+ | DenseNet-121 | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 1.932 ms | 0 - 6 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
56
+ | DenseNet-121 | SA8650 (Proxy) | SA8650P Proxy | QNN | 1.794 ms | 0 - 1 MB | FP16 | NPU | Use Export Script |
57
+ | DenseNet-121 | SA8775P ADP | SA8775P | TFLITE | 3.637 ms | 0 - 19 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
58
+ | DenseNet-121 | SA8775P ADP | SA8775P | QNN | 3.571 ms | 1 - 6 MB | FP16 | NPU | Use Export Script |
59
+ | DenseNet-121 | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 2.635 ms | 0 - 23 MB | FP16 | NPU | [DenseNet-121.tflite](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.tflite) |
60
+ | DenseNet-121 | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 2.696 ms | 0 - 22 MB | FP16 | NPU | Use Export Script |
61
+ | DenseNet-121 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 2.064 ms | 1 - 1 MB | FP16 | NPU | Use Export Script |
62
+ | DenseNet-121 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 2.004 ms | 16 - 16 MB | FP16 | NPU | [DenseNet-121.onnx](https://huggingface.co/qualcomm/DenseNet-121/blob/main/DenseNet-121.onnx) |
63
 
64
 
65
 
 
124
  Device : Samsung Galaxy S23 (13)
125
  Runtime : TFLITE
126
  Estimated inference time (ms) : 1.9
127
+ Estimated peak memory usage (MB): [0, 47]
128
  Total # Ops : 312
129
  Compute Unit(s) : NPU (312 ops)
130
  ```
 
145
  import torch
146
 
147
  import qai_hub as hub
148
+ from qai_hub_models.models.densenet121 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