README.md

---
library_name: pytorch
license: other
tags:
- bu_auto
- android
pipeline_tag: image-classification

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/convnext_base/web-assets/model_demo.png)

# ConvNext-Base: Optimized for Qualcomm Devices

ConvNextBase 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 ConvNext-Base found [here](https://github.com/pytorch/vision/blob/main/torchvision/models/convnext.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/v0.59.0/src/qai_hub_models/models/convnext_base) 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/convnext_base/releases/v0.59.0/convnext_base-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/convnext_base/releases/v0.59.0/convnext_base-onnx-w8a16.zip)
| QNN_DLC | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/convnext_base/releases/v0.59.0/convnext_base-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/convnext_base/releases/v0.59.0/convnext_base-qnn_dlc-w8a16.zip)
| TFLITE | float | Universal | QAIRT 2.45 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/convnext_base/releases/v0.59.0/convnext_base-tflite-float.zip)

For more device-specific assets and performance metrics, visit **[ConvNext-Base on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/convnext_base)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/convnext_base) 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 [ConvNext-Base on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.59.0/src/qai_hub_models/models/convnext_base) for usage instructions.

## Model Details

**Model Type:** Model_use_case.image_classification

**Model Stats:**
- Model checkpoint: Imagenet
- Input resolution: 224x224
- Number of parameters: 88.6M
- Model size (float): 338 MB
- Model size (w8a16): 88.7 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| ConvNext-Base | ONNX | float | Snapdragon® X2 Elite | 3.491 ms | 2 - 2 MB | NPU
| ConvNext-Base | ONNX | float | Snapdragon® X Elite | 7.229 ms | 176 - 176 MB | NPU
| ConvNext-Base | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 5.308 ms | 1 - 313 MB | NPU
| ConvNext-Base | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 19.187 ms | 1 - 301 MB | NPU
| ConvNext-Base | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 7.145 ms | 0 - 194 MB | NPU
| ConvNext-Base | ONNX | float | Qualcomm® QCS8450 | 19.187 ms | 1 - 301 MB | NPU
| ConvNext-Base | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 10.748 ms | 0 - 4 MB | NPU
| ConvNext-Base | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 7.229 ms | 176 - 176 MB | NPU
| ConvNext-Base | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 4.122 ms | 0 - 184 MB | NPU
| ConvNext-Base | ONNX | float | Snapdragon® 8 Elite Mobile | 4.122 ms | 0 - 184 MB | NPU
| ConvNext-Base | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.201 ms | 0 - 185 MB | NPU
| ConvNext-Base | ONNX | w8a16 | Snapdragon® X2 Elite | 2.383 ms | 1 - 1 MB | NPU
| ConvNext-Base | ONNX | w8a16 | Snapdragon® X Elite | 4.973 ms | 90 - 90 MB | NPU
| ConvNext-Base | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 3.455 ms | 0 - 263 MB | NPU
| ConvNext-Base | ONNX | w8a16 | Snapdragon® 8 Gen 1 Mobile | 8.154 ms | 0 - 262 MB | NPU
| ConvNext-Base | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 32.943 ms | 0 - 3 MB | NPU
| ConvNext-Base | ONNX | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 4.865 ms | 0 - 100 MB | NPU
| ConvNext-Base | ONNX | w8a16 | Qualcomm® QCS8450 | 8.154 ms | 0 - 262 MB | NPU
| ConvNext-Base | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 4.866 ms | 0 - 3 MB | NPU
| ConvNext-Base | ONNX | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 4.973 ms | 90 - 90 MB | NPU
| ConvNext-Base | ONNX | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 2.773 ms | 0 - 206 MB | NPU
| ConvNext-Base | ONNX | w8a16 | Snapdragon® 8 Elite Mobile | 2.773 ms | 0 - 206 MB | NPU
| ConvNext-Base | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 2.171 ms | 0 - 226 MB | NPU
| ConvNext-Base | QNN_DLC | float | Snapdragon® X2 Elite | 4.309 ms | 1 - 1 MB | NPU
| ConvNext-Base | QNN_DLC | float | Snapdragon® X Elite | 8.387 ms | 1 - 1 MB | NPU
| ConvNext-Base | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 5.842 ms | 1 - 306 MB | NPU
| ConvNext-Base | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 20.385 ms | 0 - 295 MB | NPU
| ConvNext-Base | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8275 | 41.888 ms | 1 - 180 MB | NPU
| ConvNext-Base | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 7.947 ms | 1 - 361 MB | NPU
| ConvNext-Base | QNN_DLC | float | Qualcomm® SA8775P | 11.826 ms | 1 - 181 MB | NPU
| ConvNext-Base | QNN_DLC | float | Qualcomm® SA8650P | 11.826 ms | 1 - 181 MB | NPU
| ConvNext-Base | QNN_DLC | float | Qualcomm® SA8255P | 11.826 ms | 1 - 181 MB | NPU
| ConvNext-Base | QNN_DLC | float | Qualcomm® QCS8450 | 20.385 ms | 0 - 295 MB | NPU
| ConvNext-Base | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 11.529 ms | 1 - 3 MB | NPU
| ConvNext-Base | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 8.387 ms | 1 - 1 MB | NPU
| ConvNext-Base | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 4.542 ms | 0 - 180 MB | NPU
| ConvNext-Base | QNN_DLC | float | Qualcomm® SA7255P | 41.888 ms | 1 - 180 MB | NPU
| ConvNext-Base | QNN_DLC | float | Qualcomm® SA8295P | 19.68 ms | 1 - 171 MB | NPU
| ConvNext-Base | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 4.542 ms | 0 - 180 MB | NPU
| ConvNext-Base | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.493 ms | 1 - 187 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 3.074 ms | 0 - 0 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Snapdragon® X Elite | 6.244 ms | 0 - 0 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 4.081 ms | 0 - 251 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Snapdragon® 8 Gen 1 Mobile | 9.257 ms | 0 - 250 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS6490 | 29.847 ms | 0 - 2 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS8275 | 14.615 ms | 0 - 203 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 5.871 ms | 0 - 2 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Qualcomm® SA8775P | 6.194 ms | 0 - 205 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Qualcomm® SA8650P | 6.194 ms | 0 - 205 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Qualcomm® SA8255P | 6.194 ms | 0 - 205 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Qualcomm® QCS8450 | 9.257 ms | 0 - 250 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-9075 | 5.916 ms | 0 - 2 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ IQ-X7181 | 6.244 ms | 0 - 0 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-6690 | 71.178 ms | 0 - 397 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-7790 | 7.723 ms | 0 - 253 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Qualcomm® Dragonwing™ Q-8750 | 3.267 ms | 0 - 195 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Qualcomm® SA7255P | 14.615 ms | 0 - 203 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Qualcomm® SA8295P | 9.393 ms | 0 - 204 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Snapdragon® 8 Elite Mobile | 3.267 ms | 0 - 195 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 2.568 ms | 0 - 217 MB | NPU
| ConvNext-Base | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 7.723 ms | 0 - 253 MB | NPU
| ConvNext-Base | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 5.456 ms | 0 - 304 MB | NPU
| ConvNext-Base | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 19.709 ms | 0 - 290 MB | NPU
| ConvNext-Base | TFLITE | float | Qualcomm® Dragonwing™ QCS8275 | 40.977 ms | 0 - 175 MB | NPU
| ConvNext-Base | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 7.255 ms | 0 - 2 MB | NPU
| ConvNext-Base | TFLITE | float | Qualcomm® SA8775P | 11.062 ms | 0 - 176 MB | NPU
| ConvNext-Base | TFLITE | float | Qualcomm® SA8650P | 11.062 ms | 0 - 176 MB | NPU
| ConvNext-Base | TFLITE | float | Qualcomm® SA8255P | 11.062 ms | 0 - 176 MB | NPU
| ConvNext-Base | TFLITE | float | Qualcomm® QCS8450 | 19.709 ms | 0 - 290 MB | NPU
| ConvNext-Base | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 11.72 ms | 0 - 177 MB | NPU
| ConvNext-Base | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 4.096 ms | 0 - 178 MB | NPU
| ConvNext-Base | TFLITE | float | Qualcomm® SA7255P | 40.977 ms | 0 - 175 MB | NPU
| ConvNext-Base | TFLITE | float | Qualcomm® SA8295P | 18.786 ms | 0 - 161 MB | NPU
| ConvNext-Base | TFLITE | float | Snapdragon® 8 Elite Mobile | 4.096 ms | 0 - 178 MB | NPU
| ConvNext-Base | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.216 ms | 0 - 180 MB | NPU

## License
* The license for the original implementation of ConvNext-Base can be found
  [here](https://github.com/pytorch/vision/blob/main/LICENSE).

## References
* [A ConvNet for the 2020s](https://arxiv.org/abs/2201.03545)
* [Source Model Implementation](https://github.com/pytorch/vision/blob/main/torchvision/models/convnext.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).