Hugging Face's logo

qualcomm
/
GPUNet

Image Classification
PyTorch
LiteRT
backbone
android
Model card Files
xet
Community

GPUNet: Optimized for Mobile Deployment

Imagenet classifier and general purpose backbone

GPUNet 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 GPUNet found here.

This repository provides scripts to run GPUNet 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: 10.49M
    • Model size (float): 45.28MB
    • Model size (w8a8): 21.3MB
Model Precision Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Primary Compute Unit Target Model
GPUNet float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 4.724 ms 0 - 50 MB NPU GPUNet.tflite
GPUNet float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 4.619 ms 0 - 22 MB NPU GPUNet.dlc
GPUNet float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 1.777 ms 0 - 60 MB NPU GPUNet.tflite
GPUNet float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 2.248 ms 1 - 31 MB NPU GPUNet.dlc
GPUNet float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 1.221 ms 0 - 179 MB NPU GPUNet.tflite
GPUNet float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 1.242 ms 0 - 79 MB NPU GPUNet.dlc
GPUNet float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) ONNX 1.24 ms 0 - 89 MB NPU GPUNet.onnx.zip
GPUNet float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 1.665 ms 0 - 50 MB NPU GPUNet.tflite
GPUNet float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 1.693 ms 0 - 22 MB NPU GPUNet.dlc
GPUNet float SA7255P ADP Qualcomm® SA7255P TFLITE 4.724 ms 0 - 50 MB NPU GPUNet.tflite
GPUNet float SA7255P ADP Qualcomm® SA7255P QNN_DLC 4.619 ms 0 - 22 MB NPU GPUNet.dlc
GPUNet float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 1.227 ms 0 - 178 MB NPU GPUNet.tflite
GPUNet float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 1.243 ms 0 - 87 MB NPU GPUNet.dlc
GPUNet float SA8295P ADP Qualcomm® SA8295P TFLITE 2.223 ms 0 - 56 MB NPU GPUNet.tflite
GPUNet float SA8295P ADP Qualcomm® SA8295P QNN_DLC 2.209 ms 1 - 29 MB NPU GPUNet.dlc
GPUNet float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 1.232 ms 0 - 181 MB NPU GPUNet.tflite
GPUNet float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 1.246 ms 2 - 61 MB NPU GPUNet.dlc
GPUNet float SA8775P ADP Qualcomm® SA8775P TFLITE 1.665 ms 0 - 50 MB NPU GPUNet.tflite
GPUNet float SA8775P ADP Qualcomm® SA8775P QNN_DLC 1.693 ms 0 - 22 MB NPU GPUNet.dlc
GPUNet float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 0.886 ms 0 - 63 MB NPU GPUNet.tflite
GPUNet float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 0.902 ms 1 - 35 MB NPU GPUNet.dlc
GPUNet float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 0.88 ms 0 - 34 MB NPU GPUNet.onnx.zip
GPUNet float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile TFLITE 0.69 ms 0 - 57 MB NPU GPUNet.tflite
GPUNet float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 0.692 ms 1 - 29 MB NPU GPUNet.dlc
GPUNet float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile ONNX 0.714 ms 0 - 27 MB NPU GPUNet.onnx.zip
GPUNet float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile TFLITE 0.584 ms 0 - 56 MB NPU GPUNet.tflite
GPUNet float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile QNN_DLC 0.567 ms 0 - 30 MB NPU GPUNet.dlc
GPUNet float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile ONNX 0.638 ms 1 - 27 MB NPU GPUNet.onnx.zip
GPUNet float Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 1.334 ms 113 - 113 MB NPU GPUNet.dlc
GPUNet float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 1.108 ms 24 - 24 MB NPU GPUNet.onnx.zip
GPUNet w8a16 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 2.461 ms 0 - 29 MB NPU GPUNet.dlc
GPUNet w8a16 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 1.491 ms 0 - 45 MB NPU GPUNet.dlc
GPUNet w8a16 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 1.059 ms 0 - 57 MB NPU GPUNet.dlc
GPUNet w8a16 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) ONNX 1.118 ms 0 - 67 MB NPU GPUNet.onnx.zip
GPUNet w8a16 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 1.3 ms 0 - 29 MB NPU GPUNet.dlc
GPUNet w8a16 RB3 Gen 2 (Proxy) Qualcomm® QCS6490 (Proxy) QNN_DLC 4.008 ms 0 - 42 MB NPU GPUNet.dlc
GPUNet w8a16 RB3 Gen 2 (Proxy) Qualcomm® QCS6490 (Proxy) ONNX 49.526 ms 28 - 40 MB CPU GPUNet.onnx.zip
GPUNet w8a16 RB5 (Proxy) Qualcomm® QCS8250 (Proxy) ONNX 43.987 ms 24 - 41 MB CPU GPUNet.onnx.zip
GPUNet w8a16 SA7255P ADP Qualcomm® SA7255P QNN_DLC 2.461 ms 0 - 29 MB NPU GPUNet.dlc
GPUNet w8a16 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 1.064 ms 0 - 56 MB NPU GPUNet.dlc
GPUNet w8a16 SA8295P ADP Qualcomm® SA8295P QNN_DLC 1.649 ms 0 - 36 MB NPU GPUNet.dlc
GPUNet w8a16 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 1.067 ms 0 - 57 MB NPU GPUNet.dlc
GPUNet w8a16 SA8775P ADP Qualcomm® SA8775P QNN_DLC 1.3 ms 0 - 29 MB NPU GPUNet.dlc
GPUNet w8a16 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 0.763 ms 0 - 38 MB NPU GPUNet.dlc
GPUNet w8a16 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 0.74 ms 0 - 47 MB NPU GPUNet.onnx.zip
GPUNet w8a16 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 0.514 ms 0 - 38 MB NPU GPUNet.dlc
GPUNet w8a16 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile ONNX 0.577 ms 14 - 54 MB NPU GPUNet.onnx.zip
GPUNet w8a16 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile QNN_DLC 0.433 ms 0 - 37 MB NPU GPUNet.dlc
GPUNet w8a16 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile ONNX 0.525 ms 0 - 45 MB NPU GPUNet.onnx.zip
GPUNet w8a16 Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 1.214 ms 55 - 55 MB NPU GPUNet.dlc
GPUNet w8a16 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 1.039 ms 14 - 14 MB NPU GPUNet.onnx.zip
GPUNet w8a8 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 1.118 ms 0 - 29 MB NPU GPUNet.tflite
GPUNet w8a8 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 1.407 ms 0 - 28 MB NPU GPUNet.dlc
GPUNet w8a8 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 0.573 ms 0 - 43 MB NPU GPUNet.tflite
GPUNet w8a8 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 0.858 ms 0 - 41 MB NPU GPUNet.dlc
GPUNet w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 0.436 ms 0 - 62 MB NPU GPUNet.tflite
GPUNet w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 0.615 ms 0 - 62 MB NPU GPUNet.dlc
GPUNet w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) ONNX 0.842 ms 0 - 39 MB NPU GPUNet.onnx.zip
GPUNet w8a8 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 0.628 ms 0 - 29 MB NPU GPUNet.tflite
GPUNet w8a8 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 0.815 ms 0 - 29 MB NPU GPUNet.dlc
GPUNet w8a8 RB3 Gen 2 (Proxy) Qualcomm® QCS6490 (Proxy) TFLITE 1.577 ms 0 - 37 MB NPU GPUNet.tflite
GPUNet w8a8 RB3 Gen 2 (Proxy) Qualcomm® QCS6490 (Proxy) QNN_DLC 2.213 ms 0 - 36 MB NPU GPUNet.dlc
GPUNet w8a8 RB3 Gen 2 (Proxy) Qualcomm® QCS6490 (Proxy) ONNX 12.425 ms 12 - 26 MB CPU GPUNet.onnx.zip
GPUNet w8a8 RB5 (Proxy) Qualcomm® QCS8250 (Proxy) TFLITE 8.038 ms 0 - 3 MB NPU GPUNet.tflite
GPUNet w8a8 RB5 (Proxy) Qualcomm® QCS8250 (Proxy) ONNX 10.502 ms 5 - 23 MB CPU GPUNet.onnx.zip
GPUNet w8a8 SA7255P ADP Qualcomm® SA7255P TFLITE 1.118 ms 0 - 29 MB NPU GPUNet.tflite
GPUNet w8a8 SA7255P ADP Qualcomm® SA7255P QNN_DLC 1.407 ms 0 - 28 MB NPU GPUNet.dlc
GPUNet w8a8 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 0.442 ms 6 - 67 MB NPU GPUNet.tflite
GPUNet w8a8 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 0.617 ms 0 - 62 MB NPU GPUNet.dlc
GPUNet w8a8 SA8295P ADP Qualcomm® SA8295P TFLITE 0.874 ms 0 - 35 MB NPU GPUNet.tflite
GPUNet w8a8 SA8295P ADP Qualcomm® SA8295P QNN_DLC 1.067 ms 0 - 35 MB NPU GPUNet.dlc
GPUNet w8a8 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 0.435 ms 0 - 62 MB NPU GPUNet.tflite
GPUNet w8a8 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 0.614 ms 0 - 62 MB NPU GPUNet.dlc
GPUNet w8a8 SA8775P ADP Qualcomm® SA8775P TFLITE 0.628 ms 0 - 29 MB NPU GPUNet.tflite
GPUNet w8a8 SA8775P ADP Qualcomm® SA8775P QNN_DLC 0.815 ms 0 - 29 MB NPU GPUNet.dlc
GPUNet w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 0.34 ms 0 - 40 MB NPU GPUNet.tflite
GPUNet w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 0.469 ms 0 - 40 MB NPU GPUNet.dlc
GPUNet w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 0.626 ms 0 - 42 MB NPU GPUNet.onnx.zip
GPUNet w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile TFLITE 0.268 ms 0 - 32 MB NPU GPUNet.tflite
GPUNet w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 0.34 ms 0 - 32 MB NPU GPUNet.dlc
GPUNet w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile ONNX 0.525 ms 0 - 41 MB NPU GPUNet.onnx.zip
GPUNet w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile TFLITE 0.26 ms 0 - 37 MB NPU GPUNet.tflite
GPUNet w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile QNN_DLC 0.292 ms 0 - 37 MB NPU GPUNet.dlc
GPUNet w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile ONNX 0.529 ms 0 - 42 MB NPU GPUNet.onnx.zip
GPUNet w8a8 Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 0.721 ms 68 - 68 MB NPU GPUNet.dlc
GPUNet w8a8 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 0.761 ms 14 - 14 MB NPU GPUNet.onnx.zip

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.gpunet.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.gpunet.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.gpunet.export

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.gpunet import Model

# Load the model
torch_model = Model.from_pretrained()

# Device
device = hub.Device("Samsung Galaxy S25")

# 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.

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 GPUNet's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

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

References

Community

Downloads last month
39
Inference Providers NEW
Image Classification
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support