Hugging Face's logo

qualcomm
/
SqueezeNet-1_1Quantized

Image Classification
PyTorch
LiteRT
ONNX
backbone
quantized
android
Model card Files Community
bhushans commited on
Commit
b20e0df
·
verified ·
1 Parent(s): 7fc5243

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +44 -23
README.md CHANGED
@@ -36,29 +36,34 @@ 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
- | SqueezeNet-1_1Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 0.208 ms | 0 - 1 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
40
  | SqueezeNet-1_1Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 0.466 ms | 0 - 3 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.so](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.so) |
41
- | SqueezeNet-1_1Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 0.151 ms | 0 - 26 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
42
- | SqueezeNet-1_1Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 0.34 ms | 0 - 13 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.so](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.so) |
43
- | SqueezeNet-1_1Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 0.152 ms | 0 - 16 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
44
- | SqueezeNet-1_1Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 0.284 ms | 0 - 9 MB | INT8 | NPU | Use Export Script |
45
- | SqueezeNet-1_1Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | TFLITE | 0.482 ms | 0 - 17 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
46
- | SqueezeNet-1_1Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN | 0.979 ms | 0 - 8 MB | INT8 | NPU | Use Export Script |
47
- | SqueezeNet-1_1Quantized | RB5 (Proxy) | QCS8250 Proxy | TFLITE | 4.151 ms | 0 - 5 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
48
- | SqueezeNet-1_1Quantized | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 0.206 ms | 0 - 1 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
 
 
 
49
  | SqueezeNet-1_1Quantized | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 0.429 ms | 0 - 1 MB | INT8 | NPU | Use Export Script |
50
- | SqueezeNet-1_1Quantized | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 0.203 ms | 0 - 69 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
51
- | SqueezeNet-1_1Quantized | SA8255 (Proxy) | SA8255P Proxy | QNN | 0.428 ms | 0 - 2 MB | INT8 | NPU | Use Export Script |
52
- | SqueezeNet-1_1Quantized | SA8775 (Proxy) | SA8775P Proxy | TFLITE | 0.203 ms | 0 - 1 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
53
- | SqueezeNet-1_1Quantized | SA8775 (Proxy) | SA8775P Proxy | QNN | 0.427 ms | 0 - 1 MB | INT8 | NPU | Use Export Script |
54
- | SqueezeNet-1_1Quantized | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 0.204 ms | 0 - 1 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
55
- | SqueezeNet-1_1Quantized | SA8650 (Proxy) | SA8650P Proxy | QNN | 0.423 ms | 0 - 1 MB | INT8 | NPU | Use Export Script |
56
- | SqueezeNet-1_1Quantized | SA8295P ADP | SA8295P | TFLITE | 0.533 ms | 0 - 15 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
57
- | SqueezeNet-1_1Quantized | SA8295P ADP | SA8295P | QNN | 0.893 ms | 0 - 6 MB | INT8 | NPU | Use Export Script |
58
- | SqueezeNet-1_1Quantized | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 0.239 ms | 0 - 25 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
59
- | SqueezeNet-1_1Quantized | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 0.512 ms | 0 - 13 MB | INT8 | NPU | Use Export Script |
60
- | SqueezeNet-1_1Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 0.57 ms | 1 - 1 MB | INT8 | NPU | Use Export Script |
61
- | SqueezeNet-1_1Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 43.615 ms | 17 - 17 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.onnx](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.onnx) |
 
 
62
 
63
 
64
 
@@ -123,7 +128,7 @@ SqueezeNet-1_1Quantized
123
  Device : Samsung Galaxy S23 (13)
124
  Runtime : TFLITE
125
  Estimated inference time (ms) : 0.2
126
- Estimated peak memory usage (MB): [0, 1]
127
  Total # Ops : 43
128
  Compute Unit(s) : NPU (43 ops)
129
  ```
@@ -144,13 +149,29 @@ 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.squeezenet1_1_quantized 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
+ | SqueezeNet-1_1Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 0.21 ms | 0 - 9 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
40
  | SqueezeNet-1_1Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 0.466 ms | 0 - 3 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.so](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.so) |
41
+ | SqueezeNet-1_1Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 36.352 ms | 10 - 14 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.onnx](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.onnx) |
42
+ | SqueezeNet-1_1Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 0.151 ms | 0 - 14 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
43
+ | SqueezeNet-1_1Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 0.346 ms | 0 - 10 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.so](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.so) |
44
+ | SqueezeNet-1_1Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 29.424 ms | 10 - 993 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.onnx](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.onnx) |
45
+ | SqueezeNet-1_1Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 0.153 ms | 0 - 9 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
46
+ | SqueezeNet-1_1Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 0.348 ms | 0 - 9 MB | INT8 | NPU | Use Export Script |
47
+ | SqueezeNet-1_1Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 27.57 ms | 11 - 934 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.onnx](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.onnx) |
48
+ | SqueezeNet-1_1Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | TFLITE | 0.478 ms | 0 - 12 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
49
+ | SqueezeNet-1_1Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN | 0.994 ms | 0 - 7 MB | INT8 | NPU | Use Export Script |
50
+ | SqueezeNet-1_1Quantized | RB5 (Proxy) | QCS8250 Proxy | TFLITE | 4.133 ms | 0 - 7 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
51
+ | SqueezeNet-1_1Quantized | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 0.203 ms | 0 - 3 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
52
  | SqueezeNet-1_1Quantized | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 0.429 ms | 0 - 1 MB | INT8 | NPU | Use Export Script |
53
+ | SqueezeNet-1_1Quantized | SA7255P ADP | SA7255P | TFLITE | 1.378 ms | 0 - 9 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
54
+ | SqueezeNet-1_1Quantized | SA7255P ADP | SA7255P | QNN | 1.956 ms | 0 - 6 MB | INT8 | NPU | Use Export Script |
55
+ | SqueezeNet-1_1Quantized | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 0.202 ms | 0 - 28 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
56
+ | SqueezeNet-1_1Quantized | SA8255 (Proxy) | SA8255P Proxy | QNN | 0.432 ms | 0 - 2 MB | INT8 | NPU | Use Export Script |
57
+ | SqueezeNet-1_1Quantized | SA8295P ADP | SA8295P | TFLITE | 0.523 ms | 0 - 9 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
58
+ | SqueezeNet-1_1Quantized | SA8295P ADP | SA8295P | QNN | 1.084 ms | 0 - 6 MB | INT8 | NPU | Use Export Script |
59
+ | SqueezeNet-1_1Quantized | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 0.204 ms | 0 - 77 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
60
+ | SqueezeNet-1_1Quantized | SA8650 (Proxy) | SA8650P Proxy | QNN | 0.431 ms | 0 - 1 MB | INT8 | NPU | Use Export Script |
61
+ | SqueezeNet-1_1Quantized | SA8775P ADP | SA8775P | TFLITE | 0.459 ms | 0 - 8 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
62
+ | SqueezeNet-1_1Quantized | SA8775P ADP | SA8775P | QNN | 0.913 ms | 0 - 6 MB | INT8 | NPU | Use Export Script |
63
+ | SqueezeNet-1_1Quantized | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 0.244 ms | 0 - 14 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.tflite](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.tflite) |
64
+ | SqueezeNet-1_1Quantized | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 0.52 ms | 0 - 14 MB | INT8 | NPU | Use Export Script |
65
+ | SqueezeNet-1_1Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 0.565 ms | 1 - 1 MB | INT8 | NPU | Use Export Script |
66
+ | SqueezeNet-1_1Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 44.19 ms | 17 - 17 MB | INT8 | NPU | [SqueezeNet-1_1Quantized.onnx](https://huggingface.co/qualcomm/SqueezeNet-1_1Quantized/blob/main/SqueezeNet-1_1Quantized.onnx) |
67
 
68
 
69
 
 
128
  Device : Samsung Galaxy S23 (13)
129
  Runtime : TFLITE
130
  Estimated inference time (ms) : 0.2
131
+ Estimated peak memory usage (MB): [0, 9]
132
  Total # Ops : 43
133
  Compute Unit(s) : NPU (43 ops)
134
  ```
 
149
  import torch
150
 
151
  import qai_hub as hub
152
+ from qai_hub_models.models.squeezenet1_1_quantized import Model
153
 
154
  # Load the model
155
+ torch_model = Model.from_pretrained()
156
 
157
  # Device
158
  device = hub.Device("Samsung Galaxy S23")
159
 
160
+ # Trace model
161
+ input_shape = torch_model.get_input_spec()
162
+ sample_inputs = torch_model.sample_inputs()
163
+
164
+ pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])
165
+
166
+ # Compile model on a specific device
167
+ compile_job = hub.submit_compile_job(
168
+ model=pt_model,
169
+ device=device,
170
+ input_specs=torch_model.get_input_spec(),
171
+ )
172
+
173
+ # Get target model to run on-device
174
+ target_model = compile_job.get_target_model()
175
 
176
  ```
177