Hugging Face's logo

qualcomm
/
EfficientNet-B4

Image Classification
PyTorch
LiteRT
ONNX
backbone
android
Model card Files Community
bhushans commited on
Commit
d70f219
·
verified ·
1 Parent(s): 45316a0

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +33 -17
README.md CHANGED
@@ -35,21 +35,21 @@ 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
- | EfficientNet-B4 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 3.624 ms | 0 - 3 MB | FP16 | NPU | [EfficientNet-B4.tflite](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.tflite) |
39
- | EfficientNet-B4 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 3.726 ms | 0 - 230 MB | FP16 | NPU | [EfficientNet-B4.so](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.so) |
40
- | EfficientNet-B4 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 3.571 ms | 0 - 50 MB | FP16 | NPU | [EfficientNet-B4.onnx](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.onnx) |
41
- | EfficientNet-B4 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 2.629 ms | 0 - 159 MB | FP16 | NPU | [EfficientNet-B4.tflite](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.tflite) |
42
- | EfficientNet-B4 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 2.694 ms | 0 - 27 MB | FP16 | NPU | [EfficientNet-B4.so](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.so) |
43
- | EfficientNet-B4 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 2.589 ms | 0 - 164 MB | FP16 | NPU | [EfficientNet-B4.onnx](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.onnx) |
44
- | EfficientNet-B4 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 2.106 ms | 0 - 63 MB | FP16 | NPU | [EfficientNet-B4.tflite](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.tflite) |
45
- | EfficientNet-B4 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 2.548 ms | 0 - 25 MB | FP16 | NPU | Use Export Script |
46
- | EfficientNet-B4 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 2.505 ms | 0 - 68 MB | FP16 | NPU | [EfficientNet-B4.onnx](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.onnx) |
47
- | EfficientNet-B4 | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 3.61 ms | 0 - 2 MB | FP16 | NPU | [EfficientNet-B4.tflite](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.tflite) |
48
- | EfficientNet-B4 | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 3.321 ms | 1 - 2 MB | FP16 | NPU | Use Export Script |
49
- | EfficientNet-B4 | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 7.289 ms | 0 - 174 MB | FP16 | NPU | [EfficientNet-B4.tflite](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.tflite) |
50
- | EfficientNet-B4 | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 7.403 ms | 0 - 34 MB | FP16 | NPU | Use Export Script |
51
- | EfficientNet-B4 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 3.659 ms | 1 - 1 MB | FP16 | NPU | Use Export Script |
52
- | EfficientNet-B4 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 3.728 ms | 47 - 47 MB | FP16 | NPU | [EfficientNet-B4.onnx](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.onnx) |
53
 
54
 
55
 
@@ -114,7 +114,7 @@ EfficientNet-B4
114
  Device : Samsung Galaxy S23 (13)
115
  Runtime : TFLITE
116
  Estimated inference time (ms) : 3.6
117
- Estimated peak memory usage (MB): [0, 3]
118
  Total # Ops : 482
119
  Compute Unit(s) : NPU (482 ops)
120
  ```
@@ -135,13 +135,29 @@ in memory using the `jit.trace` and then call the `submit_compile_job` API.
135
  import torch
136
 
137
  import qai_hub as hub
138
- from qai_hub_models.models.efficientnet_b4 import
139
 
140
  # Load the model
 
141
 
142
  # Device
143
  device = hub.Device("Samsung Galaxy S23")
144
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
145
 
146
  ```
147
 
 
35
 
36
  | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
37
  |---|---|---|---|---|---|---|---|---|
38
+ | EfficientNet-B4 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 3.619 ms | 0 - 280 MB | FP16 | NPU | [EfficientNet-B4.tflite](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.tflite) |
39
+ | EfficientNet-B4 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 3.731 ms | 0 - 239 MB | FP16 | NPU | [EfficientNet-B4.so](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.so) |
40
+ | EfficientNet-B4 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 3.562 ms | 0 - 50 MB | FP16 | NPU | [EfficientNet-B4.onnx](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.onnx) |
41
+ | EfficientNet-B4 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 2.635 ms | 0 - 26 MB | FP16 | NPU | [EfficientNet-B4.tflite](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.tflite) |
42
+ | EfficientNet-B4 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 2.691 ms | 1 - 25 MB | FP16 | NPU | [EfficientNet-B4.so](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.so) |
43
+ | EfficientNet-B4 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 2.586 ms | 0 - 163 MB | FP16 | NPU | [EfficientNet-B4.onnx](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.onnx) |
44
+ | EfficientNet-B4 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 2.123 ms | 0 - 26 MB | FP16 | NPU | [EfficientNet-B4.tflite](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.tflite) |
45
+ | EfficientNet-B4 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 2.547 ms | 0 - 25 MB | FP16 | NPU | Use Export Script |
46
+ | EfficientNet-B4 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 2.504 ms | 1 - 69 MB | FP16 | NPU | [EfficientNet-B4.onnx](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.onnx) |
47
+ | EfficientNet-B4 | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 3.62 ms | 0 - 281 MB | FP16 | NPU | [EfficientNet-B4.tflite](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.tflite) |
48
+ | EfficientNet-B4 | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 3.337 ms | 1 - 2 MB | FP16 | NPU | Use Export Script |
49
+ | EfficientNet-B4 | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 7.24 ms | 0 - 36 MB | FP16 | NPU | [EfficientNet-B4.tflite](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.tflite) |
50
+ | EfficientNet-B4 | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 7.454 ms | 0 - 36 MB | FP16 | NPU | Use Export Script |
51
+ | EfficientNet-B4 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 3.627 ms | 1 - 1 MB | FP16 | NPU | Use Export Script |
52
+ | EfficientNet-B4 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 3.746 ms | 46 - 46 MB | FP16 | NPU | [EfficientNet-B4.onnx](https://huggingface.co/qualcomm/EfficientNet-B4/blob/main/EfficientNet-B4.onnx) |
53
 
54
 
55
 
 
114
  Device : Samsung Galaxy S23 (13)
115
  Runtime : TFLITE
116
  Estimated inference time (ms) : 3.6
117
+ Estimated peak memory usage (MB): [0, 280]
118
  Total # Ops : 482
119
  Compute Unit(s) : NPU (482 ops)
120
  ```
 
135
  import torch
136
 
137
  import qai_hub as hub
138
+ from qai_hub_models.models.efficientnet_b4 import Model
139
 
140
  # Load the model
141
+ torch_model = Model.from_pretrained()
142
 
143
  # Device
144
  device = hub.Device("Samsung Galaxy S23")
145
 
146
+ # Trace model
147
+ input_shape = torch_model.get_input_spec()
148
+ sample_inputs = torch_model.sample_inputs()
149
+
150
+ pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])
151
+
152
+ # Compile model on a specific device
153
+ compile_job = hub.submit_compile_job(
154
+ model=pt_model,
155
+ device=device,
156
+ input_specs=torch_model.get_input_spec(),
157
+ )
158
+
159
+ # Get target model to run on-device
160
+ target_model = compile_job.get_target_model()
161
 
162
  ```
163