Hugging Face's logo

qualcomm
/
Swin-Small

Image Classification
PyTorch
backbone
bu_auto
android
Model card Files
xet
 Community
Swin-Small / README.md
qaihm-bot's picture
qaihm-bot
v0.31.0
2ea22c7 verified
preview code
|
raw
history blame
17 kB
metadata 
library_name: pytorch
license: other
tags:
  - backbone
  - android
pipeline_tag: image-classification

Swin-Small: Optimized for Mobile Deployment

Imagenet classifier and general purpose backbone

SwinSmall 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 model is an implementation of Swin-Small found here.

This repository provides scripts to run Swin-Small on Qualcomm® devices. More details on model performance across various devices, can be found here.

Model Details

  • Model Type: Model_use_case.image_classification
  • Model Stats:
    • Model checkpoint: Imagenet
    • Input resolution: 224x224
    • Number of parameters: 50.4M
    • Model size (float): 193 MB
Model Precision Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Primary Compute Unit Target Model
Swin-Small float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 44.82 ms 0 - 277 MB NPU Swin-Small.tflite
Swin-Small float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 39.173 ms 1 - 248 MB NPU Swin-Small.dlc
Swin-Small float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 23.849 ms 0 - 271 MB NPU Swin-Small.tflite
Swin-Small float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 24.465 ms 1 - 245 MB NPU Swin-Small.dlc
Swin-Small float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 18.904 ms 0 - 30 MB NPU Swin-Small.tflite
Swin-Small float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 16.534 ms 0 - 55 MB NPU Swin-Small.dlc
Swin-Small float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 21.075 ms 0 - 278 MB NPU Swin-Small.tflite
Swin-Small float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 18.383 ms 0 - 521 MB NPU Swin-Small.dlc
Swin-Small float SA7255P ADP Qualcomm® SA7255P TFLITE 44.82 ms 0 - 277 MB NPU Swin-Small.tflite
Swin-Small float SA7255P ADP Qualcomm® SA7255P QNN_DLC 39.173 ms 1 - 248 MB NPU Swin-Small.dlc
Swin-Small float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 19.118 ms 0 - 31 MB NPU Swin-Small.tflite
Swin-Small float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 16.504 ms 0 - 56 MB NPU Swin-Small.dlc
Swin-Small float SA8295P ADP Qualcomm® SA8295P TFLITE 26.658 ms 0 - 272 MB NPU Swin-Small.tflite
Swin-Small float SA8295P ADP Qualcomm® SA8295P QNN_DLC 23.439 ms 1 - 245 MB NPU Swin-Small.dlc
Swin-Small float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 19.146 ms 0 - 32 MB NPU Swin-Small.tflite
Swin-Small float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 16.52 ms 0 - 58 MB NPU Swin-Small.dlc
Swin-Small float SA8775P ADP Qualcomm® SA8775P TFLITE 21.075 ms 0 - 278 MB NPU Swin-Small.tflite
Swin-Small float SA8775P ADP Qualcomm® SA8775P QNN_DLC 18.383 ms 0 - 521 MB NPU Swin-Small.dlc
Swin-Small float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile TFLITE 19.016 ms 0 - 28 MB NPU Swin-Small.tflite
Swin-Small float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile QNN_DLC 16.566 ms 0 - 59 MB NPU Swin-Small.dlc
Swin-Small float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile ONNX 16.206 ms 0 - 240 MB NPU Swin-Small.onnx
Swin-Small float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 12.585 ms 0 - 281 MB NPU Swin-Small.tflite
Swin-Small float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 10.714 ms 1 - 768 MB NPU Swin-Small.dlc
Swin-Small float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 10.665 ms 1 - 746 MB NPU Swin-Small.onnx
Swin-Small float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile TFLITE 12.478 ms 0 - 270 MB NPU Swin-Small.tflite
Swin-Small float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile QNN_DLC 9.809 ms 1 - 254 MB NPU Swin-Small.dlc
Swin-Small float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile ONNX 9.878 ms 1 - 517 MB NPU Swin-Small.onnx
Swin-Small float Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 19.435 ms 634 - 634 MB NPU Swin-Small.dlc
Swin-Small float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 18.542 ms 101 - 101 MB NPU Swin-Small.onnx
Swin-Small w8a16 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 31.44 ms 0 - 222 MB NPU Swin-Small.dlc
Swin-Small w8a16 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 21.392 ms 0 - 288 MB NPU Swin-Small.dlc
Swin-Small w8a16 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 17.284 ms 0 - 75 MB NPU Swin-Small.dlc
Swin-Small w8a16 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 18.09 ms 0 - 219 MB NPU Swin-Small.dlc
Swin-Small w8a16 RB3 Gen 2 (Proxy) Qualcomm® QCS6490 (Proxy) QNN_DLC 56.287 ms 0 - 936 MB NPU Swin-Small.dlc
Swin-Small w8a16 SA7255P ADP Qualcomm® SA7255P QNN_DLC 31.44 ms 0 - 222 MB NPU Swin-Small.dlc
Swin-Small w8a16 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 17.534 ms 0 - 65 MB NPU Swin-Small.dlc
Swin-Small w8a16 SA8295P ADP Qualcomm® SA8295P QNN_DLC 21.132 ms 0 - 230 MB NPU Swin-Small.dlc
Swin-Small w8a16 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 17.896 ms 0 - 86 MB NPU Swin-Small.dlc
Swin-Small w8a16 SA8775P ADP Qualcomm® SA8775P QNN_DLC 18.09 ms 0 - 219 MB NPU Swin-Small.dlc
Swin-Small w8a16 Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile QNN_DLC 17.845 ms 0 - 79 MB NPU Swin-Small.dlc
Swin-Small w8a16 Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile ONNX 94.038 ms 247 - 418 MB NPU Swin-Small.onnx
Swin-Small w8a16 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 11.705 ms 0 - 237 MB NPU Swin-Small.dlc
Swin-Small w8a16 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 79.813 ms 285 - 554 MB NPU Swin-Small.onnx
Swin-Small w8a16 Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile QNN_DLC 8.723 ms 0 - 214 MB NPU Swin-Small.dlc
Swin-Small w8a16 Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile ONNX 58.359 ms 286 - 520 MB NPU Swin-Small.onnx
Swin-Small w8a16 Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 20.456 ms 157 - 157 MB NPU Swin-Small.dlc
Swin-Small w8a16 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 94.998 ms 455 - 455 MB NPU Swin-Small.onnx

Installation

Install the package via pip:

pip install qai-hub-models

Configure Qualcomm® AI Hub to run this model on a cloud-hosted device

Sign-in to Qualcomm® AI Hub with your Qualcomm® ID. Once signed in navigate to Account -> Settings -> API Token.

With this API token, you can configure your client to run models on the cloud hosted devices.

qai-hub configure --api_token API_TOKEN

Navigate to docs for more information.

Demo off target

The package contains a simple end-to-end demo that downloads pre-trained weights and runs this model on a sample input.

python -m qai_hub_models.models.swin_small.demo

The above demo runs a reference implementation of pre-processing, model inference, and post processing.

NOTE: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above).

%run -m qai_hub_models.models.swin_small.demo

Run model on a cloud-hosted device

In addition to the demo, you can also run the model on a cloud-hosted Qualcomm® device. This script does the following:

  • Performance check on-device on a cloud-hosted device
  • Downloads compiled assets that can be deployed on-device for Android.
  • Accuracy check between PyTorch and on-device outputs.
python -m qai_hub_models.models.swin_small.export

Profiling Results
------------------------------------------------------------
Swin-Small
Device                          : cs_8275 (ANDROID 14)                  
Runtime                         : TFLITE                                
Estimated inference time (ms)   : 44.8                                  
Estimated peak memory usage (MB): [0, 277]                              
Total # Ops                     : 1563                                  
Compute Unit(s)                 : npu (1563 ops) gpu (0 ops) cpu (0 ops)

How does this work?

This export script leverages Qualcomm® AI Hub to optimize, validate, and deploy this model on-device. Lets go through each step below in detail:

Step 1: Compile model for on-device deployment

To compile a PyTorch model for on-device deployment, we first trace the model in memory using the jit.trace and then call the submit_compile_job API.

import torch

import qai_hub as hub
from qai_hub_models.models.swin_small import Model

# Load the model
torch_model = Model.from_pretrained()

# Device
device = hub.Device("Samsung Galaxy S24")

# Trace model
input_shape = torch_model.get_input_spec()
sample_inputs = torch_model.sample_inputs()

pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])

# Compile model on a specific device
compile_job = hub.submit_compile_job(
    model=pt_model,
    device=device,
    input_specs=torch_model.get_input_spec(),
)

# Get target model to run on-device
target_model = compile_job.get_target_model()

Step 2: Performance profiling on cloud-hosted device

After compiling models from step 1. Models can be profiled model on-device using the target_model. Note that this scripts runs the model on a device automatically provisioned in the cloud. Once the job is submitted, you can navigate to a provided job URL to view a variety of on-device performance metrics.

profile_job = hub.submit_profile_job(
    model=target_model,
    device=device,
)

Step 3: Verify on-device accuracy

To verify the accuracy of the model on-device, you can run on-device inference on sample input data on the same cloud hosted device.

input_data = torch_model.sample_inputs()
inference_job = hub.submit_inference_job(
    model=target_model,
    device=device,
    inputs=input_data,
)
    on_device_output = inference_job.download_output_data()

With the output of the model, you can compute like PSNR, relative errors or spot check the output with expected output.

Note: This on-device profiling and inference requires access to Qualcomm® AI Hub. Sign up for access.

Run demo on a cloud-hosted device

You can also run the demo on-device.

python -m qai_hub_models.models.swin_small.demo --eval-mode on-device

NOTE: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above).

%run -m qai_hub_models.models.swin_small.demo -- --eval-mode on-device

Deploying compiled model to Android

The models can be deployed using multiple runtimes:

  • TensorFlow Lite (.tflite export): This tutorial provides a guide to deploy the .tflite model in an Android application.

  • QNN (.so export ): This sample app provides instructions on how to use the .so shared library in an Android application.

View on Qualcomm® AI Hub

Get more details on Swin-Small's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

  • The license for the original implementation of Swin-Small can be found here.
  • The license for the compiled assets for on-device deployment can be found here

References

Community