Hugging Face's logo

qualcomm
/
SqueezeNet-1.1

Image Classification
PyTorch
LiteRT
backbone
bu_auto
android
Model card Files
xet
Community

SqueezeNet-1.1: Optimized for Mobile Deployment

Imagenet classifier and general purpose backbone

SqueezeNet 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 SqueezeNet-1.1 found here.

This repository provides scripts to run SqueezeNet-1.1 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: 1.24M
    • Model size (float): 4.73 MB
    • Model size (w8a8): 1.30 MB
Model Precision Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Primary Compute Unit Target Model
SqueezeNet-1.1 float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 1.98 ms 0 - 117 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 1.933 ms 1 - 116 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 1.1 ms 0 - 136 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 1.097 ms 1 - 135 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 0.597 ms 0 - 3 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 0.596 ms 1 - 3 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) ONNX 0.602 ms 0 - 4 MB NPU SqueezeNet-1.1.onnx.zip
SqueezeNet-1.1 float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 0.919 ms 0 - 118 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 0.925 ms 1 - 116 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 float SA7255P ADP Qualcomm® SA7255P TFLITE 1.98 ms 0 - 117 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 float SA7255P ADP Qualcomm® SA7255P QNN_DLC 1.933 ms 1 - 116 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 0.594 ms 0 - 3 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 0.599 ms 1 - 3 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 float SA8295P ADP Qualcomm® SA8295P TFLITE 1.088 ms 0 - 124 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 float SA8295P ADP Qualcomm® SA8295P QNN_DLC 1.074 ms 0 - 122 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 0.6 ms 0 - 4 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 0.597 ms 1 - 3 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 float SA8775P ADP Qualcomm® SA8775P TFLITE 0.919 ms 0 - 118 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 float SA8775P ADP Qualcomm® SA8775P QNN_DLC 0.925 ms 1 - 116 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 0.409 ms 0 - 137 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 0.415 ms 0 - 134 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 0.373 ms 0 - 105 MB NPU SqueezeNet-1.1.onnx.zip
SqueezeNet-1.1 float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile TFLITE 0.315 ms 0 - 121 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 0.313 ms 0 - 120 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 float Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile ONNX 0.308 ms 0 - 93 MB NPU SqueezeNet-1.1.onnx.zip
SqueezeNet-1.1 float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile TFLITE 0.244 ms 0 - 121 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile QNN_DLC 0.238 ms 0 - 119 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 float Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile ONNX 0.282 ms 1 - 92 MB NPU SqueezeNet-1.1.onnx.zip
SqueezeNet-1.1 float Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 0.738 ms 1 - 1 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 0.496 ms 2 - 2 MB NPU SqueezeNet-1.1.onnx.zip
SqueezeNet-1.1 w8a8 Dragonwing Q-6690 MTP Qualcomm® Qcm6690 TFLITE 0.936 ms 0 - 118 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 w8a8 Dragonwing Q-6690 MTP Qualcomm® Qcm6690 QNN_DLC 1.471 ms 0 - 120 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 w8a8 Dragonwing Q-6690 MTP Qualcomm® Qcm6690 ONNX 2.948 ms 0 - 14 MB CPU SqueezeNet-1.1.onnx.zip
SqueezeNet-1.1 w8a8 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 TFLITE 0.635 ms 0 - 3 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 w8a8 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 QNN_DLC 1.054 ms 0 - 2 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 w8a8 Dragonwing RB3 Gen 2 Vision Kit Qualcomm® QCS6490 ONNX 5.539 ms 5 - 9 MB CPU SqueezeNet-1.1.onnx.zip
SqueezeNet-1.1 w8a8 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 0.623 ms 0 - 115 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 w8a8 QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN_DLC 0.934 ms 0 - 115 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 w8a8 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 0.269 ms 0 - 136 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 w8a8 QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN_DLC 0.476 ms 0 - 129 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 0.199 ms 0 - 2 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN_DLC 0.383 ms 0 - 3 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 w8a8 QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) ONNX 0.529 ms 0 - 3 MB NPU SqueezeNet-1.1.onnx.zip
SqueezeNet-1.1 w8a8 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 1.296 ms 0 - 114 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 w8a8 QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN_DLC 0.553 ms 0 - 115 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 w8a8 RB5 (Proxy) Qualcomm® QCS8250 (Proxy) TFLITE 3.876 ms 0 - 11 MB GPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 w8a8 RB5 (Proxy) Qualcomm® QCS8250 (Proxy) ONNX 2.579 ms 0 - 5 MB CPU SqueezeNet-1.1.onnx.zip
SqueezeNet-1.1 w8a8 SA7255P ADP Qualcomm® SA7255P TFLITE 0.623 ms 0 - 115 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 w8a8 SA7255P ADP Qualcomm® SA7255P QNN_DLC 0.934 ms 0 - 115 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 w8a8 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 0.204 ms 0 - 3 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 w8a8 SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN_DLC 0.383 ms 0 - 2 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 w8a8 SA8295P ADP Qualcomm® SA8295P TFLITE 0.496 ms 0 - 121 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 w8a8 SA8295P ADP Qualcomm® SA8295P QNN_DLC 0.722 ms 0 - 121 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 w8a8 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 0.201 ms 0 - 3 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 w8a8 SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN_DLC 0.382 ms 0 - 3 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 w8a8 SA8775P ADP Qualcomm® SA8775P TFLITE 1.296 ms 0 - 114 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 w8a8 SA8775P ADP Qualcomm® SA8775P QNN_DLC 0.553 ms 0 - 115 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 0.147 ms 0 - 133 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN_DLC 0.27 ms 0 - 133 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 w8a8 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 0.37 ms 0 - 105 MB NPU SqueezeNet-1.1.onnx.zip
SqueezeNet-1.1 w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile TFLITE 0.112 ms 0 - 119 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile QNN_DLC 0.179 ms 0 - 119 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 w8a8 Samsung Galaxy S25 Snapdragon® 8 Elite For Galaxy Mobile ONNX 0.337 ms 0 - 94 MB NPU SqueezeNet-1.1.onnx.zip
SqueezeNet-1.1 w8a8 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile TFLITE 0.207 ms 0 - 119 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 w8a8 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile QNN_DLC 0.378 ms 0 - 120 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 w8a8 Snapdragon 7 Gen 4 QRD Snapdragon® 7 Gen 4 Mobile ONNX 2.646 ms 0 - 15 MB CPU SqueezeNet-1.1.onnx.zip
SqueezeNet-1.1 w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile TFLITE 0.095 ms 0 - 117 MB NPU SqueezeNet-1.1.tflite
SqueezeNet-1.1 w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile QNN_DLC 0.148 ms 0 - 118 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 w8a8 Snapdragon 8 Elite Gen 5 QRD Snapdragon® 8 Elite Gen5 Mobile ONNX 0.32 ms 0 - 94 MB NPU SqueezeNet-1.1.onnx.zip
SqueezeNet-1.1 w8a8 Snapdragon X Elite CRD Snapdragon® X Elite QNN_DLC 0.502 ms 0 - 0 MB NPU SqueezeNet-1.1.dlc
SqueezeNet-1.1 w8a8 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 0.477 ms 0 - 0 MB NPU SqueezeNet-1.1.onnx.zip

Installation

Install the package via pip:

pip install qai-hub-models

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

Sign-in to Qualcomm® AI Hub Workbench 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.squeezenet1_1.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.squeezenet1_1.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.squeezenet1_1.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.squeezenet1_1 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 Workbench. 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.squeezenet1_1.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.squeezenet1_1.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 SqueezeNet-1.1's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

  • The license for the original implementation of SqueezeNet-1.1 can be found here.

References

Community

Downloads last month
54
Inference Providers NEW
Image Classification
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Paper for qualcomm/SqueezeNet-1.1