OpusMT-Zh-En: Optimized for Qualcomm Devices
OpusMT Chinese to English translation model is a state-of-the-art neural machine translation system designed for translating Chinese text into English. This model is based on the Marian transformer architecture and has been optimized for edge inference by splitting into encoder and decoder components with modified attention mechanisms. It exhibits robust performance for real-world translation tasks, making it highly reliable for practical applications. The model supports input sequences up to 256 tokens and can generate English translations with high accuracy.
This is based on the implementation of OpusMT-Zh-En found here. This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the Qualcomm® AI Hub Models library to export with custom configurations. More details on model performance across various devices, can be found here.
Qualcomm AI Hub Models uses Qualcomm AI Hub Workbench to compile, profile, and evaluate this model. Sign up 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 |
|---|---|---|---|---|
| PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | QAIRT 2.42, ONNX Runtime 1.24.1 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | QAIRT 2.42, ONNX Runtime 1.24.1 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | QAIRT 2.42, ONNX Runtime 1.24.1 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | QAIRT 2.42, ONNX Runtime 1.24.1 | Download |
| PRECOMPILED_QNN_ONNX | float | Qualcomm® QCS8550 (Proxy) | QAIRT 2.42, ONNX Runtime 1.24.1 | Download |
| PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | QAIRT 2.42, ONNX Runtime 1.24.1 | Download |
| PRECOMPILED_QNN_ONNX | float | Qualcomm® QCS9075 | QAIRT 2.42, ONNX Runtime 1.24.1 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 Mobile | QAIRT 2.43 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | QAIRT 2.43 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | QAIRT 2.43 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | QAIRT 2.43 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® QCS8550 (Proxy) | QAIRT 2.43 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® SA8775P | QAIRT 2.43 | Download |
| QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite For Galaxy Mobile | QAIRT 2.43 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® SA7255P | QAIRT 2.43 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | QAIRT 2.43 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® QCS9075 | QAIRT 2.43 | Download |
| QNN_CONTEXT_BINARY | float | Qualcomm® QCS8450 (Proxy) | QAIRT 2.43 | Download |
For more device-specific assets and performance metrics, visit OpusMT-Zh-En on Qualcomm® AI Hub.
Option 2: Export with Custom Configurations
Use the Qualcomm® AI Hub Models 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 OpusMT-Zh-En on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.text_generation
Model Stats:
- Model checkpoint: Helsinki-NLP/opus-mt-zh-en
- Input resolution: 256 tokens (Chinese text)
- Max input sequence length: 256 tokens
- Max output sequence length: 256 tokens
- Number of parameters (OpusMTEncoder): ~74M
- Model size (OpusMTEncoder) (float): ~280 MB
- Number of parameters (OpusMTDecoder): ~74M
- Model size (OpusMTDecoder) (float): ~280 MB
- Number of encoder layers: 6
- Number of decoder layers: 6
- Attention heads: 8
- Hidden dimension: 512
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| OpusMTDecoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.199 ms | 1 - 11 MB | NPU |
| OpusMTDecoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | 1.83 ms | 160 - 160 MB | NPU |
| OpusMTDecoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | 3.191 ms | 160 - 160 MB | NPU |
| OpusMTDecoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | 2.736 ms | 0 - 10 MB | NPU |
| OpusMTDecoder | PRECOMPILED_QNN_ONNX | float | Qualcomm® QCS8550 (Proxy) | 3.436 ms | 12 - 14 MB | NPU |
| OpusMTDecoder | PRECOMPILED_QNN_ONNX | float | Qualcomm® QCS9075 | 3.962 ms | 12 - 27 MB | NPU |
| OpusMTDecoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.405 ms | 0 - 12 MB | NPU |
| OpusMTDecoder | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.167 ms | 1 - 11 MB | NPU |
| OpusMTDecoder | QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | 2.223 ms | 12 - 12 MB | NPU |
| OpusMTDecoder | QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | 3.097 ms | 12 - 12 MB | NPU |
| OpusMTDecoder | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | 2.926 ms | 12 - 20 MB | NPU |
| OpusMTDecoder | QNN_CONTEXT_BINARY | float | Qualcomm® QCS8275 (Proxy) | 6.697 ms | 12 - 20 MB | NPU |
| OpusMTDecoder | QNN_CONTEXT_BINARY | float | Qualcomm® QCS8550 (Proxy) | 3.392 ms | 3 - 4 MB | NPU |
| OpusMTDecoder | QNN_CONTEXT_BINARY | float | Qualcomm® SA8775P | 4.237 ms | 12 - 20 MB | NPU |
| OpusMTDecoder | QNN_CONTEXT_BINARY | float | Qualcomm® QCS9075 | 3.945 ms | 14 - 28 MB | NPU |
| OpusMTDecoder | QNN_CONTEXT_BINARY | float | Qualcomm® QCS8450 (Proxy) | 4.72 ms | 12 - 21 MB | NPU |
| OpusMTDecoder | QNN_CONTEXT_BINARY | float | Qualcomm® SA7255P | 6.697 ms | 12 - 20 MB | NPU |
| OpusMTDecoder | QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | 4.895 ms | 10 - 16 MB | NPU |
| OpusMTDecoder | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.55 ms | 0 - 9 MB | NPU |
| OpusMTEncoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.756 ms | 0 - 10 MB | NPU |
| OpusMTEncoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® X2 Elite | 1.874 ms | 107 - 107 MB | NPU |
| OpusMTEncoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® X Elite | 4.034 ms | 107 - 107 MB | NPU |
| OpusMTEncoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Gen 3 Mobile | 2.657 ms | 0 - 7 MB | NPU |
| OpusMTEncoder | PRECOMPILED_QNN_ONNX | float | Qualcomm® QCS8550 (Proxy) | 3.772 ms | 0 - 114 MB | NPU |
| OpusMTEncoder | PRECOMPILED_QNN_ONNX | float | Qualcomm® QCS9075 | 4.708 ms | 16 - 18 MB | NPU |
| OpusMTEncoder | PRECOMPILED_QNN_ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.324 ms | 0 - 12 MB | NPU |
| OpusMTEncoder | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.738 ms | 0 - 10 MB | NPU |
| OpusMTEncoder | QNN_CONTEXT_BINARY | float | Snapdragon® X2 Elite | 2.125 ms | 0 - 0 MB | NPU |
| OpusMTEncoder | QNN_CONTEXT_BINARY | float | Snapdragon® X Elite | 3.975 ms | 0 - 0 MB | NPU |
| OpusMTEncoder | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Gen 3 Mobile | 2.643 ms | 0 - 8 MB | NPU |
| OpusMTEncoder | QNN_CONTEXT_BINARY | float | Qualcomm® QCS8275 (Proxy) | 12.725 ms | 0 - 8 MB | NPU |
| OpusMTEncoder | QNN_CONTEXT_BINARY | float | Qualcomm® QCS8550 (Proxy) | 3.609 ms | 0 - 2 MB | NPU |
| OpusMTEncoder | QNN_CONTEXT_BINARY | float | Qualcomm® SA8775P | 4.728 ms | 0 - 9 MB | NPU |
| OpusMTEncoder | QNN_CONTEXT_BINARY | float | Qualcomm® QCS9075 | 4.569 ms | 0 - 8 MB | NPU |
| OpusMTEncoder | QNN_CONTEXT_BINARY | float | Qualcomm® QCS8450 (Proxy) | 5.035 ms | 0 - 9 MB | NPU |
| OpusMTEncoder | QNN_CONTEXT_BINARY | float | Qualcomm® SA7255P | 12.725 ms | 0 - 8 MB | NPU |
| OpusMTEncoder | QNN_CONTEXT_BINARY | float | Qualcomm® SA8295P | 5.524 ms | 0 - 5 MB | NPU |
| OpusMTEncoder | QNN_CONTEXT_BINARY | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.232 ms | 0 - 9 MB | NPU |
License
- The license for the original implementation of OpusMT-Zh-En can be found here.
References
Community
- Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
- For questions or feedback please reach out to us.