ResNet18 / README.md

v0.36.0

73888d2 verified 8 months ago

18.5 kB

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

	---

	![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/resnet18/web-assets/model_demo.png)

	# ResNet18: Optimized for Mobile Deployment
	## Imagenet classifier and general purpose backbone


	ResNet18 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 ResNet18 found [here](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py).


	This repository provides scripts to run ResNet18 on Qualcomm® devices.
	More details on model performance across various devices, can be found
	[here](https://aihub.qualcomm.com/models/resnet18).



	### Model Details

	- Model Type: Model_use_case.image_classification
	- Model Stats:
	- Model checkpoint: Imagenet
	- Input resolution: 224x224
	- Number of parameters: 11.7M
	- Model size (float): 44.6 MB
	- Model size (w8a8): 11.3 MB

	\| Model \| Precision \| Device \| Chipset \| Target Runtime \| Inference Time (ms) \| Peak Memory Range (MB) \| Primary Compute Unit \| Target Model
	\|---\|---\|---\|---\|---\|---\|---\|---\|---\|
	\| ResNet18 \| float \| QCS8275 (Proxy) \| Qualcomm® QCS8275 (Proxy) \| TFLITE \| 6.098 ms \| 0 - 20 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.tflite) \|
	\| ResNet18 \| float \| QCS8275 (Proxy) \| Qualcomm® QCS8275 (Proxy) \| QNN_DLC \| 5.88 ms \| 0 - 17 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.dlc) \|
	\| ResNet18 \| float \| QCS8450 (Proxy) \| Qualcomm® QCS8450 (Proxy) \| TFLITE \| 1.971 ms \| 0 - 53 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.tflite) \|
	\| ResNet18 \| float \| QCS8450 (Proxy) \| Qualcomm® QCS8450 (Proxy) \| QNN_DLC \| 2.339 ms \| 1 - 29 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.dlc) \|
	\| ResNet18 \| float \| QCS8550 (Proxy) \| Qualcomm® QCS8550 (Proxy) \| TFLITE \| 1.362 ms \| 0 - 222 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.tflite) \|
	\| ResNet18 \| float \| QCS8550 (Proxy) \| Qualcomm® QCS8550 (Proxy) \| QNN_DLC \| 1.284 ms \| 0 - 58 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.dlc) \|
	\| ResNet18 \| float \| QCS9075 (Proxy) \| Qualcomm® QCS9075 (Proxy) \| TFLITE \| 2.123 ms \| 0 - 20 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.tflite) \|
	\| ResNet18 \| float \| QCS9075 (Proxy) \| Qualcomm® QCS9075 (Proxy) \| QNN_DLC \| 1.918 ms \| 1 - 18 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.dlc) \|
	\| ResNet18 \| float \| SA7255P ADP \| Qualcomm® SA7255P \| TFLITE \| 6.098 ms \| 0 - 20 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.tflite) \|
	\| ResNet18 \| float \| SA7255P ADP \| Qualcomm® SA7255P \| QNN_DLC \| 5.88 ms \| 0 - 17 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.dlc) \|
	\| ResNet18 \| float \| SA8255 (Proxy) \| Qualcomm® SA8255P (Proxy) \| TFLITE \| 1.36 ms \| 0 - 222 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.tflite) \|
	\| ResNet18 \| float \| SA8255 (Proxy) \| Qualcomm® SA8255P (Proxy) \| QNN_DLC \| 1.264 ms \| 0 - 83 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.dlc) \|
	\| ResNet18 \| float \| SA8295P ADP \| Qualcomm® SA8295P \| TFLITE \| 2.506 ms \| 0 - 24 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.tflite) \|
	\| ResNet18 \| float \| SA8295P ADP \| Qualcomm® SA8295P \| QNN_DLC \| 2.324 ms \| 1 - 22 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.dlc) \|
	\| ResNet18 \| float \| SA8650 (Proxy) \| Qualcomm® SA8650P (Proxy) \| TFLITE \| 1.358 ms \| 0 - 220 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.tflite) \|
	\| ResNet18 \| float \| SA8650 (Proxy) \| Qualcomm® SA8650P (Proxy) \| QNN_DLC \| 1.266 ms \| 0 - 83 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.dlc) \|
	\| ResNet18 \| float \| SA8775P ADP \| Qualcomm® SA8775P \| TFLITE \| 2.123 ms \| 0 - 20 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.tflite) \|
	\| ResNet18 \| float \| SA8775P ADP \| Qualcomm® SA8775P \| QNN_DLC \| 1.918 ms \| 1 - 18 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.dlc) \|
	\| ResNet18 \| float \| Samsung Galaxy S23 \| Snapdragon® 8 Gen 2 Mobile \| TFLITE \| 1.356 ms \| 0 - 226 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.tflite) \|
	\| ResNet18 \| float \| Samsung Galaxy S23 \| Snapdragon® 8 Gen 2 Mobile \| QNN_DLC \| 1.264 ms \| 0 - 60 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.dlc) \|
	\| ResNet18 \| float \| Samsung Galaxy S23 \| Snapdragon® 8 Gen 2 Mobile \| ONNX \| 1.149 ms \| 0 - 64 MB \| NPU \| [ResNet18.onnx.zip](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.onnx.zip) \|
	\| ResNet18 \| float \| Samsung Galaxy S24 \| Snapdragon® 8 Gen 3 Mobile \| TFLITE \| 0.947 ms \| 0 - 50 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.tflite) \|
	\| ResNet18 \| float \| Samsung Galaxy S24 \| Snapdragon® 8 Gen 3 Mobile \| QNN_DLC \| 0.877 ms \| 1 - 25 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.dlc) \|
	\| ResNet18 \| float \| Samsung Galaxy S24 \| Snapdragon® 8 Gen 3 Mobile \| ONNX \| 0.84 ms \| 0 - 24 MB \| NPU \| [ResNet18.onnx.zip](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.onnx.zip) \|
	\| ResNet18 \| float \| Snapdragon 8 Elite QRD \| Snapdragon® 8 Elite Mobile \| TFLITE \| 0.872 ms \| 0 - 27 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.tflite) \|
	\| ResNet18 \| float \| Snapdragon 8 Elite QRD \| Snapdragon® 8 Elite Mobile \| QNN_DLC \| 0.826 ms \| 1 - 23 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.dlc) \|
	\| ResNet18 \| float \| Snapdragon 8 Elite QRD \| Snapdragon® 8 Elite Mobile \| ONNX \| 0.849 ms \| 1 - 18 MB \| NPU \| [ResNet18.onnx.zip](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.onnx.zip) \|
	\| ResNet18 \| float \| Snapdragon X Elite CRD \| Snapdragon® X Elite \| QNN_DLC \| 1.398 ms \| 88 - 88 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.dlc) \|
	\| ResNet18 \| float \| Snapdragon X Elite CRD \| Snapdragon® X Elite \| ONNX \| 1.298 ms \| 22 - 22 MB \| NPU \| [ResNet18.onnx.zip](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18.onnx.zip) \|
	\| ResNet18 \| w8a8 \| QCS8275 (Proxy) \| Qualcomm® QCS8275 (Proxy) \| TFLITE \| 1.034 ms \| 0 - 18 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.tflite) \|
	\| ResNet18 \| w8a8 \| QCS8275 (Proxy) \| Qualcomm® QCS8275 (Proxy) \| QNN_DLC \| 1.281 ms \| 0 - 18 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.dlc) \|
	\| ResNet18 \| w8a8 \| QCS8450 (Proxy) \| Qualcomm® QCS8450 (Proxy) \| TFLITE \| 0.471 ms \| 0 - 43 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.tflite) \|
	\| ResNet18 \| w8a8 \| QCS8450 (Proxy) \| Qualcomm® QCS8450 (Proxy) \| QNN_DLC \| 0.743 ms \| 0 - 44 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.dlc) \|
	\| ResNet18 \| w8a8 \| QCS8550 (Proxy) \| Qualcomm® QCS8550 (Proxy) \| TFLITE \| 0.401 ms \| 0 - 89 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.tflite) \|
	\| ResNet18 \| w8a8 \| QCS8550 (Proxy) \| Qualcomm® QCS8550 (Proxy) \| QNN_DLC \| 0.548 ms \| 0 - 89 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.dlc) \|
	\| ResNet18 \| w8a8 \| QCS9075 (Proxy) \| Qualcomm® QCS9075 (Proxy) \| TFLITE \| 0.57 ms \| 0 - 18 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.tflite) \|
	\| ResNet18 \| w8a8 \| QCS9075 (Proxy) \| Qualcomm® QCS9075 (Proxy) \| QNN_DLC \| 0.699 ms \| 0 - 18 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.dlc) \|
	\| ResNet18 \| w8a8 \| RB3 Gen 2 (Proxy) \| Qualcomm® QCS6490 (Proxy) \| TFLITE \| 1.392 ms \| 0 - 33 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.tflite) \|
	\| ResNet18 \| w8a8 \| RB3 Gen 2 (Proxy) \| Qualcomm® QCS6490 (Proxy) \| QNN_DLC \| 2.011 ms \| 0 - 32 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.dlc) \|
	\| ResNet18 \| w8a8 \| RB5 (Proxy) \| Qualcomm® QCS8250 (Proxy) \| TFLITE \| 7.426 ms \| 0 - 3 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.tflite) \|
	\| ResNet18 \| w8a8 \| SA7255P ADP \| Qualcomm® SA7255P \| TFLITE \| 1.034 ms \| 0 - 18 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.tflite) \|
	\| ResNet18 \| w8a8 \| SA7255P ADP \| Qualcomm® SA7255P \| QNN_DLC \| 1.281 ms \| 0 - 18 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.dlc) \|
	\| ResNet18 \| w8a8 \| SA8255 (Proxy) \| Qualcomm® SA8255P (Proxy) \| TFLITE \| 0.404 ms \| 0 - 89 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.tflite) \|
	\| ResNet18 \| w8a8 \| SA8255 (Proxy) \| Qualcomm® SA8255P (Proxy) \| QNN_DLC \| 0.537 ms \| 0 - 87 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.dlc) \|
	\| ResNet18 \| w8a8 \| SA8295P ADP \| Qualcomm® SA8295P \| TFLITE \| 0.745 ms \| 0 - 23 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.tflite) \|
	\| ResNet18 \| w8a8 \| SA8295P ADP \| Qualcomm® SA8295P \| QNN_DLC \| 0.908 ms \| 0 - 24 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.dlc) \|
	\| ResNet18 \| w8a8 \| SA8650 (Proxy) \| Qualcomm® SA8650P (Proxy) \| TFLITE \| 0.402 ms \| 0 - 87 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.tflite) \|
	\| ResNet18 \| w8a8 \| SA8650 (Proxy) \| Qualcomm® SA8650P (Proxy) \| QNN_DLC \| 0.551 ms \| 0 - 88 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.dlc) \|
	\| ResNet18 \| w8a8 \| SA8775P ADP \| Qualcomm® SA8775P \| TFLITE \| 0.57 ms \| 0 - 18 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.tflite) \|
	\| ResNet18 \| w8a8 \| SA8775P ADP \| Qualcomm® SA8775P \| QNN_DLC \| 0.699 ms \| 0 - 18 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.dlc) \|
	\| ResNet18 \| w8a8 \| Samsung Galaxy S23 \| Snapdragon® 8 Gen 2 Mobile \| QNN_DLC \| 0.548 ms \| 0 - 87 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.dlc) \|
	\| ResNet18 \| w8a8 \| Samsung Galaxy S23 \| Snapdragon® 8 Gen 2 Mobile \| ONNX \| 28.145 ms \| 35 - 123 MB \| NPU \| [ResNet18.onnx.zip](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.onnx.zip) \|
	\| ResNet18 \| w8a8 \| Samsung Galaxy S24 \| Snapdragon® 8 Gen 3 Mobile \| TFLITE \| 0.312 ms \| 0 - 42 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.tflite) \|
	\| ResNet18 \| w8a8 \| Samsung Galaxy S24 \| Snapdragon® 8 Gen 3 Mobile \| QNN_DLC \| 0.414 ms \| 0 - 41 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.dlc) \|
	\| ResNet18 \| w8a8 \| Samsung Galaxy S24 \| Snapdragon® 8 Gen 3 Mobile \| ONNX \| 21.54 ms \| 28 - 531 MB \| NPU \| [ResNet18.onnx.zip](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.onnx.zip) \|
	\| ResNet18 \| w8a8 \| Snapdragon 8 Elite QRD \| Snapdragon® 8 Elite Mobile \| TFLITE \| 0.307 ms \| 0 - 22 MB \| NPU \| [ResNet18.tflite](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.tflite) \|
	\| ResNet18 \| w8a8 \| Snapdragon 8 Elite QRD \| Snapdragon® 8 Elite Mobile \| QNN_DLC \| 0.326 ms \| 0 - 24 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.dlc) \|
	\| ResNet18 \| w8a8 \| Snapdragon 8 Elite QRD \| Snapdragon® 8 Elite Mobile \| ONNX \| 21.18 ms \| 15 - 545 MB \| NPU \| [ResNet18.onnx.zip](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.onnx.zip) \|
	\| ResNet18 \| w8a8 \| Snapdragon X Elite CRD \| Snapdragon® X Elite \| QNN_DLC \| 0.614 ms \| 71 - 71 MB \| NPU \| [ResNet18.dlc](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.dlc) \|
	\| ResNet18 \| w8a8 \| Snapdragon X Elite CRD \| Snapdragon® X Elite \| ONNX \| 27.589 ms \| 58 - 58 MB \| NPU \| [ResNet18.onnx.zip](https://huggingface.co/qualcomm/ResNet18/blob/main/ResNet18_w8a8.onnx.zip) \|




	## Installation


	Install the package via pip:
	```bash
	pip install qai-hub-models
	```


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

	Sign-in to [Qualcomm® AI Hub](https://app.aihub.qualcomm.com/) 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.
	```bash
	qai-hub configure --api_token API_TOKEN
	```
	Navigate to [docs](https://app.aihub.qualcomm.com/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.

	```bash
	python -m qai_hub_models.models.resnet18.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.resnet18.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.

	```bash
	python -m qai_hub_models.models.resnet18.export
	```



	## How does this work?

	This [export script](https://aihub.qualcomm.com/models/resnet18/qai_hub_models/models/ResNet18/export.py)
	leverages [Qualcomm® AI Hub](https://aihub.qualcomm.com/) 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.

	```python
	import torch

	import qai_hub as hub
	from qai_hub_models.models.resnet18 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.
	```python
	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.
	```python
	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](https://myaccount.qualcomm.com/signup).



	## Run demo on a cloud-hosted device

	You can also run the demo on-device.

	```bash
	python -m qai_hub_models.models.resnet18.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.resnet18.demo -- --eval-mode on-device
	```


	## Deploying compiled model to Android


	The models can be deployed using multiple runtimes:
	- TensorFlow Lite (`.tflite` export): [This
	tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a
	guide to deploy the .tflite model in an Android application.


	- QNN (`.so` export ): This [sample
	app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html)
	provides instructions on how to use the `.so` shared library in an Android application.


	## View on Qualcomm® AI Hub
	Get more details on ResNet18's performance across various devices [here](https://aihub.qualcomm.com/models/resnet18).
	Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)


	## License
	* The license for the original implementation of ResNet18 can be found
	[here](https://github.com/pytorch/vision/blob/main/LICENSE).
	* The license for the compiled assets for on-device deployment can be found [here](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/Qualcomm+AI+Hub+Proprietary+License.pdf)



	## References
	* [Deep Residual Learning for Image Recognition](https://arxiv.org/abs/1512.03385)
	* [Source Model Implementation](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.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).