| --- |
| library_name: pytorch |
| license: other |
| tags: |
| - backbone |
| - android |
| pipeline_tag: image-classification |
|
|
| --- |
| |
|  |
|
|
| # VIT: Optimized for Qualcomm Devices |
|
|
| VIT is a machine learning model that can classify images from the Imagenet dataset. It can also be used as a backbone in building more complex models for specific use cases. |
|
|
| This is based on the implementation of VIT found [here](https://github.com/pytorch/vision/blob/main/torchvision/models/vision_transformer.py). |
| 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/v0.60.0/src/qai_hub_models/models/vit) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary). |
|
|
| 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. |
|
|
| ## Getting Started |
| There are two ways to deploy this model on your device: |
|
|
| ### Option 1: Download Pre-Exported Models |
|
|
| Below are pre-exported model assets ready for deployment. |
|
|
| | Runtime | Precision | Chipset | SDK Versions | Download | |
| |---|---|---|---|---| |
| | ONNX | float | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/vit/releases/v0.60.0/vit-onnx-float.zip) |
| | ONNX | w8a16 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/vit/releases/v0.60.0/vit-onnx-w8a16.zip) |
| | ONNX | w8a8 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/vit/releases/v0.60.0/vit-onnx-w8a8.zip) |
| | ONNX | w8a8_mixed_int16 | Universal | QAIRT 2.45, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/vit/releases/v0.60.0/vit-onnx-w8a8_mixed_int16.zip) |
| | QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/vit/releases/v0.60.0/vit-qnn_dlc-float.zip) |
| | QNN_DLC | w8a16 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/vit/releases/v0.60.0/vit-qnn_dlc-w8a16.zip) |
| | QNN_DLC | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/vit/releases/v0.60.0/vit-qnn_dlc-w8a8.zip) |
| | TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/vit/releases/v0.60.0/vit-tflite-float.zip) |
| | TFLITE | w8a8 | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/vit/releases/v0.60.0/vit-tflite-w8a8.zip) |
| |
| For more device-specific assets and performance metrics, visit **[VIT on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/vit)**. |
| |
| |
| ### Option 2: Export with Custom Configurations |
| |
| Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.60.0/src/qai_hub_models/models/vit) Python library to compile and export the model with your own: |
| - Custom weights (e.g., fine-tuned checkpoints) |
| - Custom input shapes |
| - Target device and runtime configurations |
| |
| This option is ideal if you need to customize the model beyond the default configuration provided here. |
| |
| See our repository for [VIT on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.60.0/src/qai_hub_models/models/vit) for usage instructions. |
| |
| ## Model Details |
| |
| **Model Type:** Model_use_case.image_classification |
|
|
| **Model Stats:** |
| - Input resolution: 224x224 |
| - Model checkpoint: Imagenet |
| - Model size (float): 330 MB |
| - Model size (w8a16): 86.2 MB |
| - Model size (w8a8): 83.2 MB |
| - Number of parameters: 86.6M |
|
|
| ## Performance Summary |
| | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
| |---|---|---|---|---|---|--- |
| | VIT | ONNX | float | Snapdragon® X2 Elite | 3.879 ms | 2 - 2 MB | NPU |
| | VIT | ONNX | float | Snapdragon® X Elite | 9.689 ms | 170 - 170 MB | NPU |
| | VIT | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 6.47 ms | 1 - 302 MB | NPU |
| | VIT | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 14.173 ms | 1 - 299 MB | NPU |
| | VIT | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 12.961 ms | 0 - 4 MB | NPU |
| | VIT | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 9.235 ms | 0 - 194 MB | NPU |
| | VIT | ONNX | float | Qualcomm® QCS8450 | 14.173 ms | 1 - 299 MB | NPU |
| | VIT | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 12.177 ms | 1 - 4 MB | NPU |
| | VIT | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 9.689 ms | 170 - 170 MB | NPU |
| | VIT | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 4.585 ms | 0 - 170 MB | NPU |
| | VIT | ONNX | float | Snapdragon® 8 Elite Mobile | 4.585 ms | 0 - 170 MB | NPU |
| | VIT | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.555 ms | 0 - 158 MB | NPU |
| | VIT | ONNX | w8a16 | Snapdragon® X2 Elite | 3.575 ms | 1 - 1 MB | NPU |
| | VIT | ONNX | w8a16 | Snapdragon® X Elite | 8.191 ms | 85 - 85 MB | NPU |
| | VIT | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 5.329 ms | 0 - 321 MB | NPU |
| | VIT | ONNX | w8a16 | Snapdragon® 8 Gen 1 Mobile | 9.323 ms | 0 - 322 MB | NPU |
| | VIT | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 38.302 ms | 0 - 3 MB | NPU |
| | VIT | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 7.455 ms | 0 - 4 MB | NPU |
| | VIT | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 7.812 ms | 0 - 98 MB | NPU |
| | VIT | ONNX | w8a16 | Qualcomm® QCS8450 | 9.323 ms | 0 - 322 MB | NPU |
| | VIT | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 7.945 ms | 0 - 3 MB | NPU |
| | VIT | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 8.191 ms | 85 - 85 MB | NPU |
| | VIT | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 109.442 ms | 1 - 466 MB | NPU |
| | VIT | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 9.687 ms | 0 - 396 MB | NPU |
| | VIT | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 4.067 ms | 0 - 264 MB | NPU |
| | VIT | ONNX | w8a16 | Snapdragon® 8 Elite Mobile | 4.067 ms | 0 - 264 MB | NPU |
| | VIT | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 3.091 ms | 0 - 268 MB | NPU |
| | VIT | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 9.687 ms | 0 - 396 MB | NPU |
| | VIT | ONNX | w8a8 | Snapdragon® X2 Elite | 4.94 ms | 1 - 1 MB | NPU |
| | VIT | ONNX | w8a8 | Snapdragon® X Elite | 10.92 ms | 86 - 86 MB | NPU |
| | VIT | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 7.063 ms | 0 - 334 MB | NPU |
| | VIT | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 15.238 ms | 0 - 342 MB | NPU |
| | VIT | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 73.486 ms | 0 - 3 MB | NPU |
| | VIT | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 10.235 ms | 0 - 4 MB | NPU |
| | VIT | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 10.489 ms | 0 - 99 MB | NPU |
| | VIT | ONNX | w8a8 | Qualcomm® QCS8450 | 15.238 ms | 0 - 342 MB | NPU |
| | VIT | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 9.894 ms | 0 - 3 MB | NPU |
| | VIT | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 10.92 ms | 86 - 86 MB | NPU |
| | VIT | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 250.726 ms | 0 - 504 MB | NPU |
| | VIT | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 18.89 ms | 0 - 338 MB | NPU |
| | VIT | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 5.975 ms | 0 - 213 MB | NPU |
| | VIT | ONNX | w8a8 | Snapdragon® 8 Elite Mobile | 5.975 ms | 0 - 213 MB | NPU |
| | VIT | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 4.298 ms | 0 - 221 MB | NPU |
| | VIT | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 18.89 ms | 0 - 338 MB | NPU |
| | VIT | ONNX | w8a8_mixed_int16 | Snapdragon® 8 Gen 3 Mobile | 15.939 ms | 19 - 501 MB | NPU |
| | VIT | ONNX | w8a8_mixed_int16 | Snapdragon® 8 Gen 1 Mobile | 31.211 ms | 16 - 487 MB | NPU |
| | VIT | ONNX | w8a8_mixed_int16 | Qualcomm® Dragonwing™ IQ-8275 | 24.165 ms | 16 - 19 MB | NPU |
| | VIT | ONNX | w8a8_mixed_int16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 21.795 ms | 16 - 20 MB | NPU |
| | VIT | ONNX | w8a8_mixed_int16 | Qualcomm® QCS8450 | 31.211 ms | 16 - 487 MB | NPU |
| | VIT | ONNX | w8a8_mixed_int16 | Qualcomm® Dragonwing™ IQ-9075 | 25.143 ms | 17 - 19 MB | NPU |
| | VIT | ONNX | w8a8_mixed_int16 | Qualcomm® Dragonwing™ Q-8750 | 13.521 ms | 17 - 371 MB | NPU |
| | VIT | ONNX | w8a8_mixed_int16 | Snapdragon® 8 Elite Mobile | 13.521 ms | 17 - 371 MB | NPU |
| | VIT | ONNX | w8a8_mixed_int16 | Snapdragon® 8 Elite Gen 5 Mobile | 10.058 ms | 19 - 384 MB | NPU |
| | VIT | QNN_DLC | float | Snapdragon® X2 Elite | 5.755 ms | 1 - 1 MB | NPU |
| | VIT | QNN_DLC | float | Snapdragon® X Elite | 12.644 ms | 1 - 1 MB | NPU |
| | VIT | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 8.263 ms | 0 - 305 MB | NPU |
| | VIT | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 15.437 ms | 0 - 303 MB | NPU |
| | VIT | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 14.903 ms | 1 - 4 MB | NPU |
| | VIT | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 42.281 ms | 1 - 168 MB | NPU |
| | VIT | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 11.86 ms | 1 - 2 MB | NPU |
| | VIT | QNN_DLC | float | Qualcomm® SA8775P | 14.623 ms | 1 - 169 MB | NPU |
| | VIT | QNN_DLC | float | Qualcomm® SA8650P | 14.623 ms | 1 - 169 MB | NPU |
| | VIT | QNN_DLC | float | Qualcomm® SA8255P | 14.623 ms | 1 - 169 MB | NPU |
| | VIT | QNN_DLC | float | Qualcomm® QCS8450 | 15.437 ms | 0 - 303 MB | NPU |
| | VIT | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 17.975 ms | 1 - 3 MB | NPU |
| | VIT | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 12.644 ms | 1 - 1 MB | NPU |
| | VIT | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 6.18 ms | 1 - 173 MB | NPU |
| | VIT | QNN_DLC | float | Qualcomm® SA7255P | 42.281 ms | 1 - 168 MB | NPU |
| | VIT | QNN_DLC | float | Qualcomm® SA8295P | 16.193 ms | 1 - 165 MB | NPU |
| | VIT | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 6.18 ms | 1 - 173 MB | NPU |
| | VIT | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 5.184 ms | 1 - 158 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 5.813 ms | 0 - 0 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Snapdragon® X Elite | 13.177 ms | 0 - 0 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 8.736 ms | 0 - 337 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 11.837 ms | 0 - 3 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-8275 | 28.674 ms | 0 - 276 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 12.39 ms | 0 - 2 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Qualcomm® SA8775P | 12.786 ms | 0 - 278 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Qualcomm® SA8650P | 12.786 ms | 0 - 278 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Qualcomm® SA8255P | 12.786 ms | 0 - 278 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 12.805 ms | 0 - 2 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 13.177 ms | 0 - 0 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 140.385 ms | 0 - 477 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 18.295 ms | 0 - 427 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 6.322 ms | 0 - 277 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Qualcomm® SA7255P | 28.674 ms | 0 - 276 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Snapdragon® 8 Elite Mobile | 6.322 ms | 0 - 277 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 4.777 ms | 0 - 280 MB | NPU |
| | VIT | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 18.295 ms | 0 - 427 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 5.373 ms | 0 - 0 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Snapdragon® X Elite | 11.376 ms | 1 - 1 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 7.045 ms | 0 - 322 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 14.712 ms | 0 - 335 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 73.62 ms | 0 - 2 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 10.152 ms | 0 - 3 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 31.163 ms | 0 - 202 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 10.635 ms | 0 - 11 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Qualcomm® SA8775P | 10.359 ms | 0 - 202 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Qualcomm® SA8650P | 10.359 ms | 0 - 202 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Qualcomm® SA8255P | 10.359 ms | 0 - 202 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 14.712 ms | 0 - 335 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 10.045 ms | 2 - 4 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 11.376 ms | 1 - 1 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 251.718 ms | 0 - 508 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 18.693 ms | 0 - 334 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 6.011 ms | 0 - 203 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Qualcomm® SA7255P | 31.163 ms | 0 - 202 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Qualcomm® SA8295P | 17.47 ms | 0 - 211 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Snapdragon® 8 Elite Mobile | 6.011 ms | 0 - 203 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 4.339 ms | 0 - 211 MB | NPU |
| | VIT | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 18.693 ms | 0 - 334 MB | NPU |
| | VIT | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 6.798 ms | 0 - 302 MB | NPU |
| | VIT | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 15.253 ms | 0 - 300 MB | NPU |
| | VIT | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 13.013 ms | 0 - 173 MB | NPU |
| | VIT | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 38.633 ms | 0 - 168 MB | NPU |
| | VIT | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 9.349 ms | 0 - 3 MB | NPU |
| | VIT | TFLITE | float | Qualcomm® SA8775P | 12.52 ms | 0 - 170 MB | NPU |
| | VIT | TFLITE | float | Qualcomm® SA8650P | 12.52 ms | 0 - 170 MB | NPU |
| | VIT | TFLITE | float | Qualcomm® SA8255P | 12.52 ms | 0 - 170 MB | NPU |
| | VIT | TFLITE | float | Qualcomm® QCS8450 | 15.253 ms | 0 - 300 MB | NPU |
| | VIT | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 12.582 ms | 0 - 173 MB | NPU |
| | VIT | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 4.822 ms | 0 - 176 MB | NPU |
| | VIT | TFLITE | float | Qualcomm® SA7255P | 38.633 ms | 0 - 168 MB | NPU |
| | VIT | TFLITE | float | Qualcomm® SA8295P | 15.651 ms | 0 - 162 MB | NPU |
| | VIT | TFLITE | float | Snapdragon® 8 Elite Mobile | 4.822 ms | 0 - 176 MB | NPU |
| | VIT | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.737 ms | 0 - 166 MB | NPU |
| | VIT | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 9.505 ms | 0 - 474 MB | NPU |
| | VIT | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 21.726 ms | 0 - 466 MB | NPU |
| | VIT | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 147.572 ms | 1 - 100 MB | NPU |
| | VIT | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 13.645 ms | 0 - 87 MB | NPU |
| | VIT | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 37.336 ms | 0 - 341 MB | NPU |
| | VIT | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 13.797 ms | 0 - 2 MB | NPU |
| | VIT | TFLITE | w8a8 | Qualcomm® SA8775P | 13.795 ms | 0 - 342 MB | NPU |
| | VIT | TFLITE | w8a8 | Qualcomm® SA8650P | 13.795 ms | 0 - 342 MB | NPU |
| | VIT | TFLITE | w8a8 | Qualcomm® SA8255P | 13.795 ms | 0 - 342 MB | NPU |
| | VIT | TFLITE | w8a8 | Qualcomm® QCS8450 | 21.726 ms | 0 - 466 MB | NPU |
| | VIT | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 14.268 ms | 0 - 87 MB | NPU |
| | VIT | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 203.327 ms | 1 - 259 MB | NPU |
| | VIT | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 30.996 ms | 1 - 187 MB | NPU |
| | VIT | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 7.791 ms | 0 - 344 MB | NPU |
| | VIT | TFLITE | w8a8 | Qualcomm® SA7255P | 37.336 ms | 0 - 341 MB | NPU |
| | VIT | TFLITE | w8a8 | Qualcomm® SA8295P | 21.403 ms | 0 - 335 MB | NPU |
| | VIT | TFLITE | w8a8 | Snapdragon® 8 Elite Mobile | 7.791 ms | 0 - 344 MB | NPU |
| | VIT | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 5.523 ms | 0 - 350 MB | NPU |
| | VIT | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 30.996 ms | 1 - 187 MB | NPU |
|
|
| ## License |
| * The license for the original implementation of VIT can be found |
| [here](https://github.com/pytorch/vision/blob/main/LICENSE). |
|
|
| ## References |
| * [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) |
| * [Source Model Implementation](https://github.com/pytorch/vision/blob/main/torchvision/models/vision_transformer.py) |
|
|
| ## Community |
| * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI. |
| * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com). |
|
|