Hugging Face's logo

qualcomm
/
EfficientFormer

Image Classification
PyTorch
bu_auto
android
Model card Files
xet
 Community
qaihm-bot commited on
Commit
a8cb4a4
·
verified ·
1 Parent(s): 3753942

v0.49.1

Browse files

See https://github.com/qualcomm/ai-hub-models/releases/v0.49.1 for changelog.

Files changed (1) hide show
  1. README.md +49 -49
README.md CHANGED
@@ -15,7 +15,7 @@ pipeline_tag: image-classification
15
  EfficientFormer is a vision transformer model that can classify images from the Imagenet dataset.
16
 
17
  This is based on the implementation of EfficientFormer found [here](https://github.com/snap-research/EfficientFormer).
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/efficientformer) 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
 
@@ -28,25 +28,25 @@ Below are pre-exported model assets ready for deployment.
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/efficientformer/releases/v0.48.0/efficientformer-onnx-float.zip)
32
- | ONNX | w8a16 | Universal | QAIRT 2.42, ONNX Runtime 1.24.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientformer/releases/v0.48.0/efficientformer-onnx-w8a16.zip)
33
- | QNN_DLC | float | Universal | QAIRT 2.43 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientformer/releases/v0.48.0/efficientformer-qnn_dlc-float.zip)
34
- | QNN_DLC | w8a16 | Universal | QAIRT 2.43 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientformer/releases/v0.48.0/efficientformer-qnn_dlc-w8a16.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/efficientformer/releases/v0.48.0/efficientformer-tflite-float.zip)
36
 
37
  For more device-specific assets and performance metrics, visit **[EfficientFormer on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/efficientformer)**.
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/efficientformer) 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 [EfficientFormer on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/qai_hub_models/models/efficientformer) for usage instructions.
50
 
51
  ## Model Details
52
 
@@ -62,49 +62,49 @@ See our repository for [EfficientFormer on GitHub](https://github.com/qualcomm/a
62
  ## Performance Summary
63
  | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
64
  |---|---|---|---|---|---|---
65
- | EfficientFormer | ONNX | float | Snapdragon® X2 Elite | 0.647 ms | 25 - 25 MB | NPU
66
- | EfficientFormer | ONNX | float | Snapdragon® X Elite | 1.558 ms | 24 - 24 MB | NPU
67
- | EfficientFormer | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 0.957 ms | 0 - 88 MB | NPU
68
- | EfficientFormer | ONNX | float | Qualcomm® QCS8550 (Proxy) | 1.332 ms | 1 - 3 MB | NPU
69
- | EfficientFormer | ONNX | float | Qualcomm® QCS9075 | 1.882 ms | 1 - 3 MB | NPU
70
- | EfficientFormer | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.712 ms | 0 - 53 MB | NPU
71
- | EfficientFormer | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.607 ms | 0 - 53 MB | NPU
72
- | EfficientFormer | ONNX | w8a16 | Snapdragon® X2 Elite | 0.556 ms | 12 - 12 MB | NPU
73
- | EfficientFormer | ONNX | w8a16 | Snapdragon® X Elite | 1.657 ms | 12 - 12 MB | NPU
74
- | EfficientFormer | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 0.941 ms | 0 - 91 MB | NPU
75
- | EfficientFormer | ONNX | w8a16 | Qualcomm® QCS6490 | 140.473 ms | 20 - 26 MB | CPU
76
- | EfficientFormer | ONNX | w8a16 | Qualcomm® QCS8550 (Proxy) | 1.436 ms | 0 - 18 MB | NPU
77
- | EfficientFormer | ONNX | w8a16 | Qualcomm® QCS9075 | 1.627 ms | 0 - 3 MB | NPU
78
- | EfficientFormer | ONNX | w8a16 | Qualcomm® QCM6690 | 66.333 ms | 12 - 21 MB | CPU
79
- | EfficientFormer | ONNX | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 0.644 ms | 0 - 58 MB | NPU
80
- | EfficientFormer | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 62.664 ms | 21 - 30 MB | CPU
81
  | EfficientFormer | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.525 ms | 0 - 67 MB | NPU
82
- | EfficientFormer | QNN_DLC | float | Snapdragon® X2 Elite | 0.899 ms | 1 - 1 MB | NPU
83
- | EfficientFormer | QNN_DLC | float | Snapdragon® X Elite | 1.757 ms | 1 - 1 MB | NPU
84
- | EfficientFormer | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 1.055 ms | 0 - 80 MB | NPU
85
- | EfficientFormer | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 4.931 ms | 1 - 46 MB | NPU
86
- | EfficientFormer | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 1.526 ms | 1 - 2 MB | NPU
87
- | EfficientFormer | QNN_DLC | float | Qualcomm® QCS9075 | 1.978 ms | 1 - 3 MB | NPU
88
- | EfficientFormer | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 5.53 ms | 0 - 81 MB | NPU
89
- | EfficientFormer | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.776 ms | 1 - 49 MB | NPU
 
90
  | EfficientFormer | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.652 ms | 1 - 49 MB | NPU
91
- | EfficientFormer | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 0.837 ms | 0 - 0 MB | NPU
92
- | EfficientFormer | QNN_DLC | w8a16 | Snapdragon® X Elite | 1.845 ms | 0 - 0 MB | NPU
93
- | EfficientFormer | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 1.091 ms | 0 - 80 MB | NPU
94
- | EfficientFormer | QNN_DLC | w8a16 | Qualcomm® QCS8275 (Proxy) | 3.278 ms | 0 - 57 MB | NPU
95
- | EfficientFormer | QNN_DLC | w8a16 | Qualcomm® QCS8550 (Proxy) | 1.616 ms | 0 - 96 MB | NPU
96
- | EfficientFormer | QNN_DLC | w8a16 | Qualcomm® QCS9075 | 1.79 ms | 0 - 2 MB | NPU
97
- | EfficientFormer | QNN_DLC | w8a16 | Qualcomm® QCM6690 | 6.926 ms | 0 - 176 MB | NPU
98
- | EfficientFormer | QNN_DLC | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 0.726 ms | 0 - 50 MB | NPU
99
- | EfficientFormer | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 1.732 ms | 0 - 60 MB | NPU
100
- | EfficientFormer | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.584 ms | 0 - 58 MB | NPU
101
- | EfficientFormer | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 1.037 ms | 0 - 105 MB | NPU
102
- | EfficientFormer | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 4.9 ms | 0 - 65 MB | NPU
103
- | EfficientFormer | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 1.521 ms | 0 - 4 MB | NPU
104
- | EfficientFormer | TFLITE | float | Qualcomm® QCS9075 | 1.974 ms | 0 - 27 MB | NPU
105
- | EfficientFormer | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 5.505 ms | 0 - 101 MB | NPU
106
- | EfficientFormer | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.776 ms | 0 - 66 MB | NPU
107
- | EfficientFormer | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.651 ms | 0 - 69 MB | NPU
 
 
 
 
 
 
 
 
108
 
109
  ## License
110
  * The license for the original implementation of EfficientFormer can be found
 
15
  EfficientFormer is a vision transformer model that can classify images from the Imagenet dataset.
16
 
17
  This is based on the implementation of EfficientFormer found [here](https://github.com/snap-research/EfficientFormer).
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/tree/v0.49.1/qai_hub_models/models/efficientformer) 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/efficientformer/releases/v0.49.1/efficientformer-onnx-float.zip)
32
+ | ONNX | w8a16 | Universal | QAIRT 2.42, ONNX Runtime 1.24.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientformer/releases/v0.49.1/efficientformer-onnx-w8a16.zip)
33
+ | QNN_DLC | float | Universal | QAIRT 2.43 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientformer/releases/v0.49.1/efficientformer-qnn_dlc-float.zip)
34
+ | QNN_DLC | w8a16 | Universal | QAIRT 2.43 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/efficientformer/releases/v0.49.1/efficientformer-qnn_dlc-w8a16.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/efficientformer/releases/v0.49.1/efficientformer-tflite-float.zip)
36
 
37
  For more device-specific assets and performance metrics, visit **[EfficientFormer on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/efficientformer)**.
38
 
39
 
40
  ### Option 2: Export with Custom Configurations
41
 
42
+ Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/tree/v0.49.1/qai_hub_models/models/efficientformer) 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 [EfficientFormer on GitHub](https://github.com/qualcomm/ai-hub-models/tree/v0.49.1/qai_hub_models/models/efficientformer) for usage instructions.
50
 
51
  ## Model Details
52
 
 
62
  ## Performance Summary
63
  | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
64
  |---|---|---|---|---|---|---
65
+ | EfficientFormer | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.609 ms | 1 - 53 MB | NPU
66
+ | EfficientFormer | ONNX | float | Snapdragon® X2 Elite | 0.646 ms | 25 - 25 MB | NPU
67
+ | EfficientFormer | ONNX | float | Snapdragon® X Elite | 1.547 ms | 24 - 24 MB | NPU
68
+ | EfficientFormer | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 0.945 ms | 0 - 89 MB | NPU
69
+ | EfficientFormer | ONNX | float | Qualcomm® QCS8550 (Proxy) | 1.322 ms | 0 - 3 MB | NPU
70
+ | EfficientFormer | ONNX | float | Qualcomm® QCS9075 | 1.905 ms | 1 - 3 MB | NPU
71
+ | EfficientFormer | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.705 ms | 0 - 46 MB | NPU
 
 
 
 
 
 
 
 
 
72
  | EfficientFormer | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.525 ms | 0 - 67 MB | NPU
73
+ | EfficientFormer | ONNX | w8a16 | Snapdragon® X2 Elite | 0.554 ms | 12 - 12 MB | NPU
74
+ | EfficientFormer | ONNX | w8a16 | Snapdragon® X Elite | 1.66 ms | 12 - 12 MB | NPU
75
+ | EfficientFormer | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 0.959 ms | 0 - 92 MB | NPU
76
+ | EfficientFormer | ONNX | w8a16 | Qualcomm® QCS6490 | 138.705 ms | 20 - 26 MB | CPU
77
+ | EfficientFormer | ONNX | w8a16 | Qualcomm® QCS8550 (Proxy) | 1.402 ms | 0 - 17 MB | NPU
78
+ | EfficientFormer | ONNX | w8a16 | Qualcomm® QCS9075 | 1.619 ms | 0 - 3 MB | NPU
79
+ | EfficientFormer | ONNX | w8a16 | Qualcomm® QCM6690 | 65.758 ms | 23 - 31 MB | CPU
80
+ | EfficientFormer | ONNX | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 0.641 ms | 0 - 56 MB | NPU
81
+ | EfficientFormer | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 62.964 ms | 22 - 31 MB | CPU
82
  | EfficientFormer | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.652 ms | 1 - 49 MB | NPU
83
+ | EfficientFormer | QNN_DLC | float | Snapdragon® X2 Elite | 0.913 ms | 1 - 1 MB | NPU
84
+ | EfficientFormer | QNN_DLC | float | Snapdragon® X Elite | 1.767 ms | 1 - 1 MB | NPU
85
+ | EfficientFormer | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 1.062 ms | 0 - 84 MB | NPU
86
+ | EfficientFormer | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 4.914 ms | 1 - 46 MB | NPU
87
+ | EfficientFormer | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 1.528 ms | 1 - 3 MB | NPU
88
+ | EfficientFormer | QNN_DLC | float | Qualcomm® QCS9075 | 1.973 ms | 3 - 5 MB | NPU
89
+ | EfficientFormer | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 5.597 ms | 0 - 81 MB | NPU
90
+ | EfficientFormer | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.781 ms | 0 - 45 MB | NPU
91
+ | EfficientFormer | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 0.582 ms | 0 - 58 MB | NPU
92
+ | EfficientFormer | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 0.843 ms | 0 - 0 MB | NPU
93
+ | EfficientFormer | QNN_DLC | w8a16 | Snapdragon® X Elite | 1.854 ms | 0 - 0 MB | NPU
94
+ | EfficientFormer | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 1.088 ms | 0 - 78 MB | NPU
95
+ | EfficientFormer | QNN_DLC | w8a16 | Qualcomm® QCS8275 (Proxy) | 3.296 ms | 0 - 56 MB | NPU
96
+ | EfficientFormer | QNN_DLC | w8a16 | Qualcomm® QCS8550 (Proxy) | 1.616 ms | 0 - 2 MB | NPU
97
+ | EfficientFormer | QNN_DLC | w8a16 | Qualcomm® QCS9075 | 1.773 ms | 2 - 4 MB | NPU
98
+ | EfficientFormer | QNN_DLC | w8a16 | Qualcomm® QCM6690 | 6.968 ms | 0 - 176 MB | NPU
99
+ | EfficientFormer | QNN_DLC | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 0.729 ms | 0 - 55 MB | NPU
100
+ | EfficientFormer | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 1.739 ms | 0 - 60 MB | NPU
101
+ | EfficientFormer | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 0.652 ms | 0 - 69 MB | NPU
102
+ | EfficientFormer | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 1.048 ms | 0 - 106 MB | NPU
103
+ | EfficientFormer | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 4.89 ms | 0 - 65 MB | NPU
104
+ | EfficientFormer | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 1.507 ms | 0 - 2 MB | NPU
105
+ | EfficientFormer | TFLITE | float | Qualcomm® QCS9075 | 1.969 ms | 0 - 27 MB | NPU
106
+ | EfficientFormer | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 5.52 ms | 0 - 101 MB | NPU
107
+ | EfficientFormer | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 0.776 ms | 0 - 63 MB | NPU
108
 
109
  ## License
110
  * The license for the original implementation of EfficientFormer can be found