--- library_name: pytorch license: mit tags: - real_time - android pipeline_tag: automatic-speech-recognition --- ![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/funasr_conformer_en/web-assets/model_demo.png) # FunASR-Conformer-EN: Optimized for Qualcomm Devices FunASR Conformer-EN is a large-scale English ASR model from Alibaba DAMO Academy, trained on ~50,000 hours of English speech. It uses a 32-block Conformer encoder(512 hidden dim, 16 attention heads) with a CTC head for decoding. The model accepts raw 16kHz audio and outputs transcribed text via CTC greedy decoding. This is based on the implementation of FunASR-Conformer-EN found [here](https://huggingface.co/funasr/conformer-en). This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.63.0/src/qai_hub_models/models/funasr_conformer_en) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary). Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device. ## Getting Started There are two ways to deploy this model on your device: ### Option 1: Download Pre-Exported Models Below are pre-exported model assets ready for deployment. | Runtime | Precision | Chipset | SDK Versions | Download | |---|---|---|---|---| | ONNX | float | Universal | QAIRT 2.50, ONNX Runtime 1.27.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/funasr_conformer_en/releases/v0.63.0/funasr_conformer_en-onnx-float.zip) | QNN_DLC | float | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/funasr_conformer_en/releases/v0.63.0/funasr_conformer_en-qnn_dlc-float.zip) | TFLITE | float | Universal | QAIRT 2.50 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/funasr_conformer_en/releases/v0.63.0/funasr_conformer_en-tflite-float.zip) For more device-specific assets and performance metrics, visit **[FunASR-Conformer-EN on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/funasr_conformer_en)**. ### Option 2: Export with Custom Configurations Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.63.0/src/qai_hub_models/models/funasr_conformer_en) Python library to compile and export the model with your own: - Custom weights (e.g., fine-tuned checkpoints) - Custom input shapes - Target device and runtime configurations This option is ideal if you need to customize the model beyond the default configuration provided here. See our repository for [FunASR-Conformer-EN on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.63.0/src/qai_hub_models/models/funasr_conformer_en) for usage instructions. ## Model Details **Model Type:** Model_use_case.speech_recognition **Model Stats:** - Input resolution: 1x160000 (10s at 16kHz) - Model checkpoint: funasr/conformer-en - Model size (float): ~840 MB - Number of parameters: 220M ## Performance Summary | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |---|---|---|---|---|---|--- | FunASR-Conformer-EN | ONNX | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 17.576 ms | 6 - 659 MB | NPU | FunASR-Conformer-EN | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 22.504 ms | 6 - 637 MB | NPU | FunASR-Conformer-EN | ONNX | float | Snapdragon® X2 Elite | 23.161 ms | 9 - 9 MB | NPU | FunASR-Conformer-EN | ONNX | float | Snapdragon® X Elite | 44.434 ms | 325 - 325 MB | NPU | FunASR-Conformer-EN | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 30.544 ms | 8 - 980 MB | NPU | FunASR-Conformer-EN | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 55.817 ms | 6 - 990 MB | NPU | FunASR-Conformer-EN | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 48.118 ms | 5 - 10 MB | NPU | FunASR-Conformer-EN | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 40.864 ms | 0 - 387 MB | NPU | FunASR-Conformer-EN | ONNX | float | Qualcomm® QCS8450 | 55.817 ms | 6 - 990 MB | NPU | FunASR-Conformer-EN | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 49.29 ms | 6 - 10 MB | NPU | FunASR-Conformer-EN | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 44.434 ms | 325 - 325 MB | NPU | FunASR-Conformer-EN | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 22.504 ms | 6 - 637 MB | NPU | FunASR-Conformer-EN | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 18.035 ms | 0 - 596 MB | NPU | FunASR-Conformer-EN | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 22.912 ms | 0 - 592 MB | NPU | FunASR-Conformer-EN | QNN_DLC | float | Snapdragon® X2 Elite | 23.844 ms | 0 - 0 MB | NPU | FunASR-Conformer-EN | QNN_DLC | float | Snapdragon® X Elite | 44.805 ms | 0 - 0 MB | NPU | FunASR-Conformer-EN | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 30.33 ms | 1 - 896 MB | NPU | FunASR-Conformer-EN | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 57.557 ms | 0 - 905 MB | NPU | FunASR-Conformer-EN | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 48.282 ms | 0 - 7 MB | NPU | FunASR-Conformer-EN | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 40.752 ms | 1 - 3 MB | NPU | FunASR-Conformer-EN | QNN_DLC | float | Qualcomm® QCS8450 | 57.557 ms | 0 - 905 MB | NPU | FunASR-Conformer-EN | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 49.067 ms | 2 - 8 MB | NPU | FunASR-Conformer-EN | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 44.805 ms | 0 - 0 MB | NPU | FunASR-Conformer-EN | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 22.912 ms | 0 - 592 MB | NPU | FunASR-Conformer-EN | QNN_DLC | float | Qualcomm® SA8295P | 54.397 ms | 1 - 595 MB | NPU | FunASR-Conformer-EN | TFLITE | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 18.027 ms | 3 - 626 MB | NPU | FunASR-Conformer-EN | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 22.862 ms | 3 - 622 MB | NPU | FunASR-Conformer-EN | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 30.43 ms | 3 - 963 MB | NPU | FunASR-Conformer-EN | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 56.922 ms | 3 - 959 MB | NPU | FunASR-Conformer-EN | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 48.473 ms | 3 - 335 MB | NPU | FunASR-Conformer-EN | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 41.152 ms | 0 - 4 MB | NPU | FunASR-Conformer-EN | TFLITE | float | Qualcomm® SA8775P | 48.186 ms | 3 - 621 MB | NPU | FunASR-Conformer-EN | TFLITE | float | Qualcomm® SA8650P | 48.186 ms | 3 - 621 MB | NPU | FunASR-Conformer-EN | TFLITE | float | Qualcomm® SA8255P | 48.186 ms | 3 - 621 MB | NPU | FunASR-Conformer-EN | TFLITE | float | Qualcomm® QCS8450 | 56.922 ms | 3 - 959 MB | NPU | FunASR-Conformer-EN | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 48.86 ms | 3 - 335 MB | NPU | FunASR-Conformer-EN | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 22.862 ms | 3 - 622 MB | NPU | FunASR-Conformer-EN | TFLITE | float | Qualcomm® SA7255P | 106.305 ms | 3 - 622 MB | NPU | FunASR-Conformer-EN | TFLITE | float | Qualcomm® SA8295P | 54.877 ms | 3 - 615 MB | NPU ## License * The license for the original implementation of FunASR-Conformer-EN can be found [here](https://github.com/modelscope/FunASR?tab=MIT-1-ov-file). ## References * [Conformer: Convolution-augmented Transformer for Speech Recognition](https://arxiv.org/abs/2005.08100) * [Source Model Implementation](https://huggingface.co/funasr/conformer-en) ## 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).