Upload README.md with huggingface_hub

1c37514 verified 9 months ago

13.2 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/swin_small/web-assets/model_demo.png)

	# 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](https://github.com/pytorch/vision/blob/main/torchvision/models/swin_transformer.py).


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


	### 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 \| 44.432 ms \| 0 - 258 MB \| NPU \| [Swin-Small.tflite](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small.tflite) \|
	\| Swin-Small \| float \| QCS8275 (Proxy) \| Qualcomm® QCS8275 (Proxy) \| QNN \| 41.766 ms \| 1 - 10 MB \| NPU \| Use Export Script \|
	\| Swin-Small \| float \| QCS8450 (Proxy) \| Qualcomm® QCS8450 (Proxy) \| TFLITE \| 23.171 ms \| 0 - 258 MB \| NPU \| [Swin-Small.tflite](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small.tflite) \|
	\| Swin-Small \| float \| QCS8450 (Proxy) \| Qualcomm® QCS8450 (Proxy) \| QNN \| 26.233 ms \| 1 - 248 MB \| NPU \| Use Export Script \|
	\| Swin-Small \| float \| QCS8550 (Proxy) \| Qualcomm® QCS8550 (Proxy) \| TFLITE \| 18.44 ms \| 0 - 27 MB \| NPU \| [Swin-Small.tflite](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small.tflite) \|
	\| Swin-Small \| float \| QCS8550 (Proxy) \| Qualcomm® QCS8550 (Proxy) \| QNN \| 17.358 ms \| 1 - 4 MB \| NPU \| Use Export Script \|
	\| Swin-Small \| float \| QCS9075 (Proxy) \| Qualcomm® QCS9075 (Proxy) \| TFLITE \| 20.688 ms \| 0 - 258 MB \| NPU \| [Swin-Small.tflite](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small.tflite) \|
	\| Swin-Small \| float \| QCS9075 (Proxy) \| Qualcomm® QCS9075 (Proxy) \| QNN \| 19.601 ms \| 1 - 11 MB \| NPU \| Use Export Script \|
	\| Swin-Small \| float \| SA7255P ADP \| Qualcomm® SA7255P \| TFLITE \| 44.432 ms \| 0 - 258 MB \| NPU \| [Swin-Small.tflite](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small.tflite) \|
	\| Swin-Small \| float \| SA7255P ADP \| Qualcomm® SA7255P \| QNN \| 41.766 ms \| 1 - 10 MB \| NPU \| Use Export Script \|
	\| Swin-Small \| float \| SA8255 (Proxy) \| Qualcomm® SA8255P (Proxy) \| TFLITE \| 18.394 ms \| 0 - 30 MB \| NPU \| [Swin-Small.tflite](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small.tflite) \|
	\| Swin-Small \| float \| SA8255 (Proxy) \| Qualcomm® SA8255P (Proxy) \| QNN \| 17.639 ms \| 0 - 2 MB \| NPU \| Use Export Script \|
	\| Swin-Small \| float \| SA8295P ADP \| Qualcomm® SA8295P \| TFLITE \| 25.415 ms \| 0 - 251 MB \| NPU \| [Swin-Small.tflite](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small.tflite) \|
	\| Swin-Small \| float \| SA8295P ADP \| Qualcomm® SA8295P \| QNN \| 24.516 ms \| 1 - 18 MB \| NPU \| Use Export Script \|
	\| Swin-Small \| float \| SA8650 (Proxy) \| Qualcomm® SA8650P (Proxy) \| TFLITE \| 18.519 ms \| 0 - 25 MB \| NPU \| [Swin-Small.tflite](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small.tflite) \|
	\| Swin-Small \| float \| SA8650 (Proxy) \| Qualcomm® SA8650P (Proxy) \| QNN \| 17.621 ms \| 1 - 3 MB \| NPU \| Use Export Script \|
	\| Swin-Small \| float \| SA8775P ADP \| Qualcomm® SA8775P \| TFLITE \| 20.688 ms \| 0 - 258 MB \| NPU \| [Swin-Small.tflite](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small.tflite) \|
	\| Swin-Small \| float \| SA8775P ADP \| Qualcomm® SA8775P \| QNN \| 19.601 ms \| 1 - 11 MB \| NPU \| Use Export Script \|
	\| Swin-Small \| float \| Samsung Galaxy S23 \| Snapdragon® 8 Gen 2 Mobile \| TFLITE \| 18.385 ms \| 0 - 28 MB \| NPU \| [Swin-Small.tflite](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small.tflite) \|
	\| Swin-Small \| float \| Samsung Galaxy S23 \| Snapdragon® 8 Gen 2 Mobile \| QNN \| 17.545 ms \| 0 - 59 MB \| NPU \| Use Export Script \|
	\| Swin-Small \| float \| Samsung Galaxy S23 \| Snapdragon® 8 Gen 2 Mobile \| ONNX \| 16.352 ms \| 0 - 281 MB \| NPU \| [Swin-Small.onnx](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small.onnx) \|
	\| Swin-Small \| float \| Samsung Galaxy S24 \| Snapdragon® 8 Gen 3 Mobile \| TFLITE \| 12.448 ms \| 64 - 328 MB \| NPU \| [Swin-Small.tflite](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small.tflite) \|
	\| Swin-Small \| float \| Samsung Galaxy S24 \| Snapdragon® 8 Gen 3 Mobile \| QNN \| 11.803 ms \| 1 - 252 MB \| NPU \| Use Export Script \|
	\| Swin-Small \| float \| Samsung Galaxy S24 \| Snapdragon® 8 Gen 3 Mobile \| ONNX \| 10.903 ms \| 0 - 245 MB \| NPU \| [Swin-Small.onnx](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small.onnx) \|
	\| Swin-Small \| float \| Snapdragon 8 Elite QRD \| Snapdragon® 8 Elite Mobile \| TFLITE \| 11.832 ms \| 0 - 258 MB \| NPU \| [Swin-Small.tflite](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small.tflite) \|
	\| Swin-Small \| float \| Snapdragon 8 Elite QRD \| Snapdragon® 8 Elite Mobile \| QNN \| 9.316 ms \| 1 - 242 MB \| NPU \| Use Export Script \|
	\| Swin-Small \| float \| Snapdragon 8 Elite QRD \| Snapdragon® 8 Elite Mobile \| ONNX \| 8.439 ms \| 1 - 235 MB \| NPU \| [Swin-Small.onnx](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small.onnx) \|
	\| Swin-Small \| float \| Snapdragon X Elite CRD \| Snapdragon® X Elite \| QNN \| 18.354 ms \| 1 - 1 MB \| NPU \| Use Export Script \|
	\| Swin-Small \| float \| Snapdragon X Elite CRD \| Snapdragon® X Elite \| ONNX \| 19.156 ms \| 100 - 100 MB \| NPU \| [Swin-Small.onnx](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small.onnx) \|
	\| Swin-Small \| w8a16 \| Samsung Galaxy S23 \| Snapdragon® 8 Gen 2 Mobile \| ONNX \| 100.576 ms \| 238 - 418 MB \| NPU \| [Swin-Small.onnx](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small_w8a16.onnx) \|
	\| Swin-Small \| w8a16 \| Samsung Galaxy S24 \| Snapdragon® 8 Gen 3 Mobile \| ONNX \| 85.053 ms \| 276 - 521 MB \| NPU \| [Swin-Small.onnx](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small_w8a16.onnx) \|
	\| Swin-Small \| w8a16 \| Snapdragon 8 Elite QRD \| Snapdragon® 8 Elite Mobile \| ONNX \| 70.937 ms \| 284 - 500 MB \| NPU \| [Swin-Small.onnx](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small_w8a16.onnx) \|
	\| Swin-Small \| w8a16 \| Snapdragon X Elite CRD \| Snapdragon® X Elite \| ONNX \| 107.725 ms \| 458 - 458 MB \| NPU \| [Swin-Small.onnx](https://huggingface.co/qualcomm/Swin-Small/blob/main/Swin-Small_w8a16.onnx) \|




	## 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.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.

	```bash
	python -m qai_hub_models.models.swin_small.export
	```
	```
	Profiling Results
	------------------------------------------------------------
	Swin-Small
	Device : cs_8275 (ANDROID 14)
	Runtime : TFLITE
	Estimated inference time (ms) : 44.4
	Estimated peak memory usage (MB): [0, 258]
	Total # Ops : 1563
	Compute Unit(s) : npu (1563 ops) gpu (0 ops) cpu (0 ops)
	```


	## How does this work?

	This [export script](https://aihub.qualcomm.com/models/swin_small/qai_hub_models/models/Swin-Small/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.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.
	```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.swin_small.demo --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 -- --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 Swin-Small's performance across various devices [here](https://aihub.qualcomm.com/models/swin_small).
	Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)


	## License
	* The license for the original implementation of Swin-Small 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
	* [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://arxiv.org/abs/2103.14030)
	* [Source Model Implementation](https://github.com/pytorch/vision/blob/main/torchvision/models/swin_transformer.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).