Hugging Face's logo

qualcomm
/
ConvNext-Base

Image Classification
PyTorch
LiteRT
bu_auto
android
Model card Files
xet
Community
qaihm-bot commited on
Commit
e6cb09f
·
verified ·
1 Parent(s): bf7e6f0

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +42 -29
README.md CHANGED
@@ -1,6 +1,6 @@
1
  ---
2
  library_name: pytorch
3
- license: bsd-3-clause
4
  tags:
5
  - android
6
  pipeline_tag: image-classification
@@ -25,34 +25,47 @@ More details on model performance across various devices, can be found
25
 
26
  ### Model Details
27
 
28
- - **Model Type:** Image classification
29
  - **Model Stats:**
30
  - Model checkpoint: Imagenet
31
  - Input resolution: 224x224
32
  - Number of parameters: 28.6M
33
- - Model size: 109 MB
34
 
35
- | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
36
  |---|---|---|---|---|---|---|---|---|
37
- | ConvNext-Base | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 7.479 ms | 0 - 25 MB | FP16 | NPU | [ConvNext-Base.tflite](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.tflite) |
38
- | ConvNext-Base | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 8.131 ms | 0 - 25 MB | FP16 | NPU | [ConvNext-Base.so](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.so) |
39
- | ConvNext-Base | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 7.787 ms | 0 - 449 MB | FP16 | NPU | [ConvNext-Base.onnx](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.onnx) |
40
- | ConvNext-Base | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 5.558 ms | 0 - 272 MB | FP16 | NPU | [ConvNext-Base.tflite](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.tflite) |
41
- | ConvNext-Base | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 5.994 ms | 1 - 272 MB | FP16 | NPU | [ConvNext-Base.so](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.so) |
42
- | ConvNext-Base | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 6.05 ms | 1 - 282 MB | FP16 | NPU | [ConvNext-Base.onnx](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.onnx) |
43
- | ConvNext-Base | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 5.048 ms | 0 - 261 MB | FP16 | NPU | [ConvNext-Base.tflite](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.tflite) |
44
- | ConvNext-Base | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 5.359 ms | 1 - 266 MB | FP16 | NPU | Use Export Script |
45
- | ConvNext-Base | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 4.55 ms | 1 - 267 MB | FP16 | NPU | [ConvNext-Base.onnx](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.onnx) |
46
- | ConvNext-Base | QCS8275 (Proxy) | QCS8275 Proxy | TFLITE | 303.019 ms | 0 - 258 MB | FP16 | NPU | [ConvNext-Base.tflite](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.tflite) |
47
- | ConvNext-Base | QCS8275 (Proxy) | QCS8275 Proxy | QNN | 302.898 ms | 1 - 10 MB | FP16 | NPU | Use Export Script |
48
- | ConvNext-Base | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 7.555 ms | 0 - 12 MB | FP16 | NPU | [ConvNext-Base.tflite](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.tflite) |
49
- | ConvNext-Base | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 8.061 ms | 1 - 3 MB | FP16 | NPU | Use Export Script |
50
- | ConvNext-Base | QCS9075 (Proxy) | QCS9075 Proxy | TFLITE | 13.943 ms | 0 - 257 MB | FP16 | NPU | [ConvNext-Base.tflite](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.tflite) |
51
- | ConvNext-Base | QCS9075 (Proxy) | QCS9075 Proxy | QNN | 14.251 ms | 0 - 10 MB | FP16 | NPU | Use Export Script |
52
- | ConvNext-Base | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 19.147 ms | 0 - 275 MB | FP16 | NPU | [ConvNext-Base.tflite](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.tflite) |
53
- | ConvNext-Base | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 19.635 ms | 1 - 271 MB | FP16 | NPU | Use Export Script |
54
- | ConvNext-Base | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 8.14 ms | 1 - 1 MB | FP16 | NPU | Use Export Script |
55
- | ConvNext-Base | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 7.91 ms | 176 - 176 MB | FP16 | NPU | [ConvNext-Base.onnx](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.onnx) |
 
 
 
 
 
 
 
 
 
 
 
 
 
56
 
57
 
58
 
@@ -114,12 +127,12 @@ python -m qai_hub_models.models.convnext_base.export
114
  Profiling Results
115
  ------------------------------------------------------------
116
  ConvNext-Base
117
- Device : Samsung Galaxy S23 (13)
118
- Runtime : TFLITE
119
- Estimated inference time (ms) : 7.5
120
- Estimated peak memory usage (MB): [0, 25]
121
- Total # Ops : 598
122
- Compute Unit(s) : NPU (598 ops)
123
  ```
124
 
125
 
 
1
  ---
2
  library_name: pytorch
3
+ license: other
4
  tags:
5
  - android
6
  pipeline_tag: image-classification
 
25
 
26
  ### Model Details
27
 
28
+ - **Model Type:** Model_use_case.image_classification
29
  - **Model Stats:**
30
  - Model checkpoint: Imagenet
31
  - Input resolution: 224x224
32
  - Number of parameters: 28.6M
33
+ - Model size (float): 109 MB
34
 
35
+ | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
36
  |---|---|---|---|---|---|---|---|---|
37
+ | ConvNext-Base | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 41.83 ms | 0 - 258 MB | NPU | [ConvNext-Base.tflite](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.tflite) |
38
+ | ConvNext-Base | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN | 41.657 ms | 1 - 10 MB | NPU | Use Export Script |
39
+ | ConvNext-Base | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 19.154 ms | 0 - 276 MB | NPU | [ConvNext-Base.tflite](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.tflite) |
40
+ | ConvNext-Base | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN | 22.814 ms | 0 - 279 MB | NPU | Use Export Script |
41
+ | ConvNext-Base | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 7.543 ms | 0 - 30 MB | NPU | [ConvNext-Base.tflite](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.tflite) |
42
+ | ConvNext-Base | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN | 8.033 ms | 1 - 4 MB | NPU | Use Export Script |
43
+ | ConvNext-Base | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 11.565 ms | 0 - 257 MB | NPU | [ConvNext-Base.tflite](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.tflite) |
44
+ | ConvNext-Base | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN | 11.962 ms | 1 - 10 MB | NPU | Use Export Script |
45
+ | ConvNext-Base | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | TFLITE | 7.472 ms | 0 - 20 MB | NPU | [ConvNext-Base.tflite](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.tflite) |
46
+ | ConvNext-Base | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN | 8.079 ms | 0 - 19 MB | NPU | Use Export Script |
47
+ | ConvNext-Base | float | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 7.773 ms | 0 - 418 MB | NPU | [ConvNext-Base.onnx](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.onnx) |
48
+ | ConvNext-Base | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 5.574 ms | 0 - 273 MB | NPU | [ConvNext-Base.tflite](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.tflite) |
49
+ | ConvNext-Base | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN | 6.151 ms | 1 - 276 MB | NPU | Use Export Script |
50
+ | ConvNext-Base | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 6.031 ms | 0 - 280 MB | NPU | [ConvNext-Base.onnx](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.onnx) |
51
+ | ConvNext-Base | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | TFLITE | 5.049 ms | 0 - 261 MB | NPU | [ConvNext-Base.tflite](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.tflite) |
52
+ | ConvNext-Base | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN | 5.166 ms | 1 - 268 MB | NPU | Use Export Script |
53
+ | ConvNext-Base | float | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 5.261 ms | 1 - 267 MB | NPU | [ConvNext-Base.onnx](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.onnx) |
54
+ | ConvNext-Base | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 8.189 ms | 1 - 1 MB | NPU | Use Export Script |
55
+ | ConvNext-Base | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 7.891 ms | 176 - 176 MB | NPU | [ConvNext-Base.onnx](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base.onnx) |
56
+ | ConvNext-Base | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN | 14.553 ms | 0 - 10 MB | NPU | Use Export Script |
57
+ | ConvNext-Base | w8a16 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN | 10.511 ms | 0 - 131 MB | NPU | Use Export Script |
58
+ | ConvNext-Base | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN | 5.943 ms | 0 - 3 MB | NPU | Use Export Script |
59
+ | ConvNext-Base | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN | 6.121 ms | 0 - 11 MB | NPU | Use Export Script |
60
+ | ConvNext-Base | w8a16 | RB3 Gen 2 (Proxy) | Qualcomm® QCS6490 (Proxy) | QNN | 39.533 ms | 0 - 13 MB | NPU | Use Export Script |
61
+ | ConvNext-Base | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | QNN | 5.903 ms | 0 - 31 MB | NPU | Use Export Script |
62
+ | ConvNext-Base | w8a16 | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 Mobile | ONNX | 232.599 ms | 666 - 1049 MB | NPU | [ConvNext-Base.onnx](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base_w8a16.onnx) |
63
+ | ConvNext-Base | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN | 4.135 ms | 0 - 127 MB | NPU | Use Export Script |
64
+ | ConvNext-Base | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 212.919 ms | 685 - 1284 MB | NPU | [ConvNext-Base.onnx](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base_w8a16.onnx) |
65
+ | ConvNext-Base | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | QNN | 3.613 ms | 0 - 122 MB | NPU | Use Export Script |
66
+ | ConvNext-Base | w8a16 | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite Mobile | ONNX | 169.574 ms | 687 - 1253 MB | NPU | [ConvNext-Base.onnx](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base_w8a16.onnx) |
67
+ | ConvNext-Base | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 6.274 ms | 0 - 0 MB | NPU | Use Export Script |
68
+ | ConvNext-Base | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 264.165 ms | 926 - 926 MB | NPU | [ConvNext-Base.onnx](https://huggingface.co/qualcomm/ConvNext-Base/blob/main/ConvNext-Base_w8a16.onnx) |
69
 
70
 
71
 
 
127
  Profiling Results
128
  ------------------------------------------------------------
129
  ConvNext-Base
130
+ Device : cs_8275 (ANDROID 14)
131
+ Runtime : TFLITE
132
+ Estimated inference time (ms) : 41.8
133
+ Estimated peak memory usage (MB): [0, 258]
134
+ Total # Ops : 598
135
+ Compute Unit(s) : npu (598 ops) gpu (0 ops) cpu (0 ops)
136
  ```
137
 
138