Hugging Face's logo

qualcomm
/
FFNet-78S

Image Segmentation
PyTorch
LiteRT
android
Model card Files Community
bhushans commited on
Commit
a9120b6
·
verified ·
1 Parent(s): 8fd2cfa

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +44 -26
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
- | FFNet-78S | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 23.43 ms | 2 - 5 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
39
- | FFNet-78S | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 23.858 ms | 24 - 47 MB | FP16 | NPU | [FFNet-78S.so](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.so) |
40
- | FFNet-78S | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 32.75 ms | 24 - 55 MB | FP16 | NPU | [FFNet-78S.onnx](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.onnx) |
41
- | FFNet-78S | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 17.145 ms | 1 - 130 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
42
- | FFNet-78S | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 17.555 ms | 168 - 209 MB | FP16 | NPU | [FFNet-78S.so](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.so) |
43
- | FFNet-78S | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 24.814 ms | 30 - 185 MB | FP16 | NPU | [FFNet-78S.onnx](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.onnx) |
44
- | FFNet-78S | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 14.037 ms | 2 - 55 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
45
- | FFNet-78S | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 16.989 ms | 24 - 63 MB | FP16 | NPU | Use Export Script |
46
- | FFNet-78S | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 20.887 ms | 28 - 87 MB | FP16 | NPU | [FFNet-78S.onnx](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.onnx) |
47
- | FFNet-78S | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 23.314 ms | 2 - 4 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
48
- | FFNet-78S | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 23.039 ms | 24 - 25 MB | FP16 | NPU | Use Export Script |
49
- | FFNet-78S | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 23.167 ms | 2 - 4 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
50
- | FFNet-78S | SA8255 (Proxy) | SA8255P Proxy | QNN | 23.598 ms | 24 - 25 MB | FP16 | NPU | Use Export Script |
51
- | FFNet-78S | SA8775 (Proxy) | SA8775P Proxy | TFLITE | 23.92 ms | 2 - 4 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
52
- | FFNet-78S | SA8775 (Proxy) | SA8775P Proxy | QNN | 23.417 ms | 24 - 28 MB | FP16 | NPU | Use Export Script |
53
- | FFNet-78S | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 23.66 ms | 2 - 5 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
54
- | FFNet-78S | SA8650 (Proxy) | SA8650P Proxy | QNN | 23.682 ms | 24 - 25 MB | FP16 | NPU | Use Export Script |
55
- | FFNet-78S | SA8295P ADP | SA8295P | TFLITE | 44.269 ms | 2 - 44 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
56
- | FFNet-78S | SA8295P ADP | SA8295P | QNN | 43.534 ms | 24 - 30 MB | FP16 | NPU | Use Export Script |
57
- | FFNet-78S | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 41.048 ms | 2 - 108 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
58
- | FFNet-78S | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 39.495 ms | 14 - 50 MB | FP16 | NPU | Use Export Script |
59
- | FFNet-78S | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 23.189 ms | 24 - 24 MB | FP16 | NPU | Use Export Script |
60
- | FFNet-78S | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 36.432 ms | 32 - 32 MB | FP16 | NPU | [FFNet-78S.onnx](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.onnx) |
 
 
61
 
62
 
63
 
@@ -122,8 +124,8 @@ Profiling Results
122
  FFNet-78S
123
  Device : Samsung Galaxy S23 (13)
124
  Runtime : TFLITE
125
- Estimated inference time (ms) : 23.4
126
- Estimated peak memory usage (MB): [2, 5]
127
  Total # Ops : 149
128
  Compute Unit(s) : NPU (149 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.ffnet_78s import
148
 
149
  # Load the model
 
150
 
151
  # Device
152
  device = hub.Device("Samsung Galaxy S23")
153
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
154
 
155
  ```
156
 
 
35
 
36
  | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
37
  |---|---|---|---|---|---|---|---|---|
38
+ | FFNet-78S | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 23.72 ms | 2 - 30 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
39
+ | FFNet-78S | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 24.001 ms | 24 - 52 MB | FP16 | NPU | [FFNet-78S.so](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.so) |
40
+ | FFNet-78S | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 37.609 ms | 24 - 54 MB | FP16 | NPU | [FFNet-78S.onnx](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.onnx) |
41
+ | FFNet-78S | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 17.32 ms | 1 - 44 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
42
+ | FFNet-78S | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 17.928 ms | 20 - 59 MB | FP16 | NPU | [FFNet-78S.so](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.so) |
43
+ | FFNet-78S | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 26.395 ms | 32 - 178 MB | FP16 | NPU | [FFNet-78S.onnx](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.onnx) |
44
+ | FFNet-78S | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 14.248 ms | 1 - 43 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
45
+ | FFNet-78S | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 16.848 ms | 20 - 60 MB | FP16 | NPU | Use Export Script |
46
+ | FFNet-78S | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 25.427 ms | 28 - 86 MB | FP16 | NPU | [FFNet-78S.onnx](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.onnx) |
47
+ | FFNet-78S | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 23.647 ms | 2 - 30 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
48
+ | FFNet-78S | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 23.086 ms | 24 - 25 MB | FP16 | NPU | Use Export Script |
49
+ | FFNet-78S | SA7255P ADP | SA7255P | TFLITE | 1164.068 ms | 0 - 40 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
50
+ | FFNet-78S | SA7255P ADP | SA7255P | QNN | 1162.892 ms | 24 - 30 MB | FP16 | NPU | Use Export Script |
51
+ | FFNet-78S | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 23.214 ms | 2 - 32 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
52
+ | FFNet-78S | SA8255 (Proxy) | SA8255P Proxy | QNN | 24.106 ms | 24 - 25 MB | FP16 | NPU | Use Export Script |
53
+ | FFNet-78S | SA8295P ADP | SA8295P | TFLITE | 44.303 ms | 2 - 34 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
54
+ | FFNet-78S | SA8295P ADP | SA8295P | QNN | 43.366 ms | 24 - 30 MB | FP16 | NPU | Use Export Script |
55
+ | FFNet-78S | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 23.216 ms | 2 - 32 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
56
+ | FFNet-78S | SA8650 (Proxy) | SA8650P Proxy | QNN | 23.393 ms | 24 - 25 MB | FP16 | NPU | Use Export Script |
57
+ | FFNet-78S | SA8775P ADP | SA8775P | TFLITE | 49.885 ms | 2 - 42 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
58
+ | FFNet-78S | SA8775P ADP | SA8775P | QNN | 49.835 ms | 24 - 29 MB | FP16 | NPU | Use Export Script |
59
+ | FFNet-78S | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 39.548 ms | 2 - 41 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
60
+ | FFNet-78S | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 39.617 ms | 24 - 58 MB | FP16 | NPU | Use Export Script |
61
+ | FFNet-78S | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 23.164 ms | 24 - 24 MB | FP16 | NPU | Use Export Script |
62
+ | FFNet-78S | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 36.863 ms | 31 - 31 MB | FP16 | NPU | [FFNet-78S.onnx](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.onnx) |
63
 
64
 
65
 
 
124
  FFNet-78S
125
  Device : Samsung Galaxy S23 (13)
126
  Runtime : TFLITE
127
+ Estimated inference time (ms) : 23.7
128
+ Estimated peak memory usage (MB): [2, 30]
129
  Total # Ops : 149
130
  Compute Unit(s) : NPU (149 ops)
131
  ```
 
146
  import torch
147
 
148
  import qai_hub as hub
149
+ from qai_hub_models.models.ffnet_78s 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