--- library_name: pytorch license: other tags: - real_time - android pipeline_tag: audio-classification --- ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/yamnet/web-assets/model_demo.png) # YamNet: Optimized for Mobile Deployment ## Audio Event classification Model An audio event classifier trained on the AudioSet dataset to predict audio events from the AudioSet ontology employing the Mobilenet_v1 depthwise-separable convolution architecture. This model is an implementation of YamNet found [here](https://github.com/w-hc/torch_audioset). This repository provides scripts to run YamNet on Qualcomm® devices. More details on model performance across various devices, can be found [here](https://aihub.qualcomm.com/models/yamnet). ### Model Details - **Model Type:** Model_use_case.audio_classification - **Model Stats:** - Model checkpoint: yamnet.pth - Input resolution: 1x1x96x64 - Number of parameters: 3.73M - Model size (float): 14.2 MB | Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model |---|---|---|---|---|---|---|---|---| | YamNet | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 0.67 ms | 0 - 122 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.tflite) | | YamNet | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 0.652 ms | 0 - 115 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.dlc) | | YamNet | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 0.355 ms | 0 - 150 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.tflite) | | YamNet | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 0.359 ms | 0 - 137 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.dlc) | | YamNet | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 0.216 ms | 0 - 2 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.tflite) | | YamNet | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 0.208 ms | 0 - 2 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.dlc) | | YamNet | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 0.419 ms | 0 - 10 MB | NPU | [YamNet.onnx.zip](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.onnx.zip) | | YamNet | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 0.379 ms | 0 - 122 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.tflite) | | YamNet | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 1.1 ms | 0 - 115 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.dlc) | | YamNet | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 0.67 ms | 0 - 122 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.tflite) | | YamNet | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 0.652 ms | 0 - 115 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.dlc) | | YamNet | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 0.209 ms | 0 - 3 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.tflite) | | YamNet | float | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 0.223 ms | 0 - 3 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.dlc) | | YamNet | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 0.548 ms | 0 - 129 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.tflite) | | YamNet | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 0.543 ms | 0 - 122 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.dlc) | | YamNet | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 0.22 ms | 0 - 4 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.tflite) | | YamNet | float | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 0.224 ms | 0 - 2 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.dlc) | | YamNet | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 0.379 ms | 0 - 122 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.tflite) | | YamNet | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 1.1 ms | 0 - 115 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.dlc) | | YamNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 0.175 ms | 0 - 147 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.tflite) | | YamNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 0.175 ms | 0 - 139 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.dlc) | | YamNet | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 0.266 ms | 0 - 112 MB | NPU | [YamNet.onnx.zip](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.onnx.zip) | | YamNet | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 0.153 ms | 0 - 126 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.tflite) | | YamNet | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 0.163 ms | 0 - 119 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.dlc) | | YamNet | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 0.245 ms | 0 - 94 MB | NPU | [YamNet.onnx.zip](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.onnx.zip) | | YamNet | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | TFLITE | 0.158 ms | 0 - 124 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.tflite) | | YamNet | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 0.157 ms | 0 - 117 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.dlc) | | YamNet | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 0.259 ms | 0 - 92 MB | NPU | [YamNet.onnx.zip](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.onnx.zip) | | YamNet | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 0.276 ms | 0 - 0 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.dlc) | | YamNet | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 0.272 ms | 8 - 8 MB | NPU | [YamNet.onnx.zip](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet.onnx.zip) | | YamNet | w8a8 | Dragonwing Q-6690 MTP | Qualcomm® Qcm6690 | TFLITE | 0.422 ms | 0 - 121 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.tflite) | | YamNet | w8a8 | Dragonwing Q-6690 MTP | Qualcomm® Qcm6690 | QNN_DLC | 0.423 ms | 0 - 122 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.dlc) | | YamNet | w8a8 | Dragonwing Q-6690 MTP | Qualcomm® Qcm6690 | ONNX | 0.882 ms | 0 - 14 MB | CPU | [YamNet.onnx.zip](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.onnx.zip) | | YamNet | w8a8 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | TFLITE | 0.522 ms | 0 - 6 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.tflite) | | YamNet | w8a8 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | QNN_DLC | 0.635 ms | 0 - 2 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.dlc) | | YamNet | w8a8 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | ONNX | 1.632 ms | 0 - 7 MB | CPU | [YamNet.onnx.zip](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.onnx.zip) | | YamNet | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 0.398 ms | 0 - 117 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.tflite) | | YamNet | w8a8 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 0.375 ms | 0 - 118 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.dlc) | | YamNet | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 0.22 ms | 0 - 140 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.tflite) | | YamNet | w8a8 | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 0.224 ms | 0 - 136 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.dlc) | | YamNet | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 0.138 ms | 0 - 3 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.tflite) | | YamNet | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 0.136 ms | 0 - 2 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.dlc) | | YamNet | w8a8 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 0.333 ms | 0 - 7 MB | NPU | [YamNet.onnx.zip](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.onnx.zip) | | YamNet | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 0.295 ms | 0 - 117 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.tflite) | | YamNet | w8a8 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 0.274 ms | 0 - 118 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.dlc) | | YamNet | w8a8 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | TFLITE | 1.673 ms | 0 - 25 MB | GPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.tflite) | | YamNet | w8a8 | RB5 (Proxy) | Qualcomm® QCS8250 (Proxy) | ONNX | 0.772 ms | 0 - 10 MB | CPU | [YamNet.onnx.zip](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.onnx.zip) | | YamNet | w8a8 | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 0.398 ms | 0 - 117 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.tflite) | | YamNet | w8a8 | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 0.375 ms | 0 - 118 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.dlc) | | YamNet | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | TFLITE | 0.136 ms | 0 - 4 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.tflite) | | YamNet | w8a8 | SA8255 (Proxy) | Qualcomm® SA8255P (Proxy) | QNN_DLC | 0.139 ms | 0 - 2 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.dlc) | | YamNet | w8a8 | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 0.442 ms | 0 - 123 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.tflite) | | YamNet | w8a8 | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 0.406 ms | 0 - 124 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.dlc) | | YamNet | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | TFLITE | 0.124 ms | 0 - 4 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.tflite) | | YamNet | w8a8 | SA8650 (Proxy) | Qualcomm® SA8650P (Proxy) | QNN_DLC | 0.139 ms | 0 - 2 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.dlc) | | YamNet | w8a8 | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 0.295 ms | 0 - 117 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.tflite) | | YamNet | w8a8 | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 0.274 ms | 0 - 118 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.dlc) | | YamNet | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 0.11 ms | 0 - 140 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.tflite) | | YamNet | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 0.112 ms | 0 - 138 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.dlc) | | YamNet | w8a8 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 0.206 ms | 0 - 110 MB | NPU | [YamNet.onnx.zip](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.onnx.zip) | | YamNet | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 0.097 ms | 0 - 121 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.tflite) | | YamNet | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 0.091 ms | 0 - 122 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.dlc) | | YamNet | w8a8 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 0.197 ms | 0 - 98 MB | NPU | [YamNet.onnx.zip](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.onnx.zip) | | YamNet | w8a8 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | TFLITE | 0.172 ms | 0 - 122 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.tflite) | | YamNet | w8a8 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | QNN_DLC | 0.165 ms | 0 - 122 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.dlc) | | YamNet | w8a8 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | ONNX | 0.714 ms | 0 - 15 MB | CPU | [YamNet.onnx.zip](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.onnx.zip) | | YamNet | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | TFLITE | 0.088 ms | 0 - 119 MB | NPU | [YamNet.tflite](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.tflite) | | YamNet | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | QNN_DLC | 0.094 ms | 0 - 120 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.dlc) | | YamNet | w8a8 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen5 Mobile | ONNX | 0.184 ms | 0 - 96 MB | NPU | [YamNet.onnx.zip](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.onnx.zip) | | YamNet | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 0.208 ms | 0 - 0 MB | NPU | [YamNet.dlc](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.dlc) | | YamNet | w8a8 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 0.224 ms | 4 - 4 MB | NPU | [YamNet.onnx.zip](https://huggingface.co/qualcomm/YamNet/blob/main/YamNet_w8a8.onnx.zip) | ## Installation Install the package via pip: ```bash # NOTE: 3.10 <= PYTHON_VERSION < 3.14 is supported. pip install "qai-hub-models[yamnet]" git+https://github.com/w-hc/torch_audioset.git ``` ## Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device Sign-in to [Qualcomm® AI Hub Workbench](https://workbench.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://workbench.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.yamnet.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.yamnet.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.yamnet.export ``` ## How does this work? This [export script](https://aihub.qualcomm.com/models/yamnet/qai_hub_models/models/YamNet/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.yamnet 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. ```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 Workbench. [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.yamnet.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.yamnet.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 YamNet's performance across various devices [here](https://aihub.qualcomm.com/models/yamnet). Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/) ## License * The license for the original implementation of YamNet can be found [here](https://github.com/w-hc/torch_audioset/blob/master/LICENSE). ## References * [MobileNets Efficient Convolutional Neural Networks for Mobile Vision Applications](https://arxiv.org/abs/1704.04861) * [Source Model Implementation](https://github.com/w-hc/torch_audioset) ## 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).