Hugging Face's logo

qualcomm
/
NASNet

Image Classification
PyTorch
bu_auto
android
Model card Files
xet
 Community
qaihm-bot commited on
Commit
a4598d5
·
verified ·
1 Parent(s): eb547a4

v0.47.0

Browse files

See https://github.com/qualcomm/ai-hub-models/releases/v0.47.0 for changelog.

Files changed (1) hide show
  1. README.md +47 -11
README.md CHANGED
@@ -15,7 +15,7 @@ pipeline_tag: image-classification
15
  NASNet is a vision transformer model that can classify images from the Imagenet dataset.
16
 
17
  This is based on the implementation of NASNet found [here](https://github.com/huggingface/pytorch-image-models/tree/main).
18
- This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/nasnet) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).
19
 
20
  Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device.
21
 
@@ -28,21 +28,25 @@ Below are pre-exported model assets ready for deployment.
28
 
29
  | Runtime | Precision | Chipset | SDK Versions | Download |
30
  |---|---|---|---|---|
31
- | TFLITE | float | Universal | QAIRT 2.42, TFLite 2.17.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/nasnet/releases/v0.46.0/nasnet-tflite-float.zip)
 
 
 
 
32
 
33
  For more device-specific assets and performance metrics, visit **[NASNet on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/nasnet)**.
34
 
35
 
36
  ### Option 2: Export with Custom Configurations
37
 
38
- Use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/nasnet) Python library to compile and export the model with your own:
39
  - Custom weights (e.g., fine-tuned checkpoints)
40
  - Custom input shapes
41
  - Target device and runtime configurations
42
 
43
  This option is ideal if you need to customize the model beyond the default configuration provided here.
44
 
45
- See our repository for [NASNet on GitHub](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/nasnet) for usage instructions.
46
 
47
  ## Model Details
48
 
@@ -59,13 +63,45 @@ See our repository for [NASNet on GitHub](https://github.com/quic/ai-hub-models/
59
  ## Performance Summary
60
  | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
61
  |---|---|---|---|---|---|---
62
- | NASNet | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 5.589 ms | 0 - 671 MB | NPU
63
- | NASNet | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 36.767 ms | 0 - 526 MB | NPU
64
- | NASNet | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 7.663 ms | 0 - 3 MB | NPU
65
- | NASNet | TFLITE | float | Qualcomm® QCS9075 | 10.262 ms | 0 - 191 MB | NPU
66
- | NASNet | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 23.477 ms | 0 - 648 MB | NPU
67
- | NASNet | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 4.596 ms | 0 - 524 MB | NPU
68
- | NASNet | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 4.11 ms | 0 - 529 MB | NPU
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
69
 
70
  ## License
71
  * The license for the original implementation of NASNet can be found
 
15
  NASNet is a vision transformer model that can classify images from the Imagenet dataset.
16
 
17
  This is based on the implementation of NASNet found [here](https://github.com/huggingface/pytorch-image-models/tree/main).
18
+ This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/qai_hub_models/models/nasnet) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).
19
 
20
  Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device.
21
 
 
28
 
29
  | Runtime | Precision | Chipset | SDK Versions | Download |
30
  |---|---|---|---|---|
31
+ | ONNX | float | Universal | QAIRT 2.42, ONNX Runtime 1.24.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/nasnet/releases/v0.47.0/nasnet-onnx-float.zip)
32
+ | ONNX | w8a8_mixed_fp16 | Universal | QAIRT 2.42, ONNX Runtime 1.24.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/nasnet/releases/v0.47.0/nasnet-onnx-w8a8_mixed_fp16.zip)
33
+ | QNN_DLC | float | Universal | QAIRT 2.43 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/nasnet/releases/v0.47.0/nasnet-qnn_dlc-float.zip)
34
+ | QNN_DLC | w8a8_mixed_fp16 | Universal | QAIRT 2.43 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/nasnet/releases/v0.47.0/nasnet-qnn_dlc-w8a8_mixed_fp16.zip)
35
+ | TFLITE | float | Universal | QAIRT 2.43, TFLite 2.17.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/nasnet/releases/v0.47.0/nasnet-tflite-float.zip)
36
 
37
  For more device-specific assets and performance metrics, visit **[NASNet on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/nasnet)**.
38
 
39
 
40
  ### Option 2: Export with Custom Configurations
41
 
42
+ Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/qai_hub_models/models/nasnet) Python library to compile and export the model with your own:
43
  - Custom weights (e.g., fine-tuned checkpoints)
44
  - Custom input shapes
45
  - Target device and runtime configurations
46
 
47
  This option is ideal if you need to customize the model beyond the default configuration provided here.
48
 
49
+ See our repository for [NASNet on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/qai_hub_models/models/nasnet) for usage instructions.
50
 
51
  ## Model Details
52
 
 
63
  ## Performance Summary
64
  | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
65
  |---|---|---|---|---|---|---
66
+ | NASNet | ONNX | float | Snapdragon® X2 Elite | 8.272 ms | 189 - 189 MB | NPU
67
+ | NASNet | ONNX | float | Snapdragon® X Elite | 17.835 ms | 188 - 188 MB | NPU
68
+ | NASNet | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 12.849 ms | 1 - 839 MB | NPU
69
+ | NASNet | ONNX | float | Qualcomm® QCS8550 (Proxy) | 17.435 ms | 0 - 196 MB | NPU
70
+ | NASNet | ONNX | float | Qualcomm® QCS9075 | 28.348 ms | 0 - 4 MB | NPU
71
+ | NASNet | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 10.583 ms | 1 - 676 MB | NPU
72
+ | NASNet | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 8.382 ms | 1 - 673 MB | NPU
73
+ | NASNet | ONNX | w8a8_mixed_fp16 | Snapdragon® X2 Elite | 4.504 ms | 100 - 100 MB | NPU
74
+ | NASNet | ONNX | w8a8_mixed_fp16 | Snapdragon® X Elite | 11.095 ms | 98 - 98 MB | NPU
75
+ | NASNet | ONNX | w8a8_mixed_fp16 | Snapdragon® 8 Gen 3 Mobile | 6.945 ms | 6 - 479 MB | NPU
76
+ | NASNet | ONNX | w8a8_mixed_fp16 | Qualcomm® QCS8550 (Proxy) | 9.919 ms | 5 - 10 MB | NPU
77
+ | NASNet | ONNX | w8a8_mixed_fp16 | Qualcomm® QCS9075 | 11.66 ms | 5 - 8 MB | NPU
78
+ | NASNet | ONNX | w8a8_mixed_fp16 | Snapdragon® 8 Elite For Galaxy Mobile | 5.651 ms | 6 - 362 MB | NPU
79
+ | NASNet | ONNX | w8a8_mixed_fp16 | Snapdragon® 8 Elite Gen 5 Mobile | 4.589 ms | 5 - 358 MB | NPU
80
+ | NASNet | QNN_DLC | float | Snapdragon® X2 Elite | 9.047 ms | 1 - 1 MB | NPU
81
+ | NASNet | QNN_DLC | float | Snapdragon® X Elite | 19.34 ms | 1 - 1 MB | NPU
82
+ | NASNet | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 12.455 ms | 0 - 814 MB | NPU
83
+ | NASNet | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 53.959 ms | 1 - 659 MB | NPU
84
+ | NASNet | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 19.142 ms | 1 - 2 MB | NPU
85
+ | NASNet | QNN_DLC | float | Qualcomm® QCS9075 | 29.033 ms | 1 - 3 MB | NPU
86
+ | NASNet | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 34.65 ms | 0 - 794 MB | NPU
87
+ | NASNet | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 11.042 ms | 0 - 655 MB | NPU
88
+ | NASNet | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 8.769 ms | 1 - 657 MB | NPU
89
+ | NASNet | QNN_DLC | w8a8_mixed_fp16 | Snapdragon® X2 Elite | 4.092 ms | 0 - 0 MB | NPU
90
+ | NASNet | QNN_DLC | w8a8_mixed_fp16 | Snapdragon® X Elite | 9.174 ms | 0 - 0 MB | NPU
91
+ | NASNet | QNN_DLC | w8a8_mixed_fp16 | Snapdragon® 8 Gen 3 Mobile | 6.034 ms | 0 - 494 MB | NPU
92
+ | NASNet | QNN_DLC | w8a8_mixed_fp16 | Qualcomm® QCS8275 (Proxy) | 16.411 ms | 0 - 378 MB | NPU
93
+ | NASNet | QNN_DLC | w8a8_mixed_fp16 | Qualcomm® QCS8550 (Proxy) | 8.89 ms | 0 - 4 MB | NPU
94
+ | NASNet | QNN_DLC | w8a8_mixed_fp16 | Qualcomm® QCS9075 | 9.427 ms | 0 - 2 MB | NPU
95
+ | NASNet | QNN_DLC | w8a8_mixed_fp16 | Qualcomm® QCS8450 (Proxy) | 11.055 ms | 0 - 506 MB | NPU
96
+ | NASNet | QNN_DLC | w8a8_mixed_fp16 | Snapdragon® 8 Elite For Galaxy Mobile | 5.004 ms | 0 - 378 MB | NPU
97
+ | NASNet | QNN_DLC | w8a8_mixed_fp16 | Snapdragon® 8 Elite Gen 5 Mobile | 3.867 ms | 0 - 379 MB | NPU
98
+ | NASNet | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 8.775 ms | 0 - 783 MB | NPU
99
+ | NASNet | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 44.328 ms | 0 - 629 MB | NPU
100
+ | NASNet | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 12.574 ms | 0 - 4 MB | NPU
101
+ | NASNet | TFLITE | float | Qualcomm® QCS9075 | 15.543 ms | 0 - 192 MB | NPU
102
+ | NASNet | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 28.997 ms | 0 - 766 MB | NPU
103
+ | NASNet | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 6.954 ms | 0 - 634 MB | NPU
104
+ | NASNet | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 5.666 ms | 0 - 616 MB | NPU
105
 
106
  ## License
107
  * The license for the original implementation of NASNet can be found