Contribute Your Valuable Data to NPU.top (NPU.tech)¶
Thank you very much for taking the time out of your busy schedule to contribute valuable data to our project.
NPU.top is a startup project and still has many imperfections, especially in terms of features and hardware data. Therefore, we urgently need actual hardware runtime data. You can refer to the following websites for data testing. For NPU testing, we recommend using Windows 11 and the latest version of Edge Canary.
1. Microsoft WebNN Image Classification¶
It can be used to test first-run latency, average inference latency, and throughput, and supports MobileNet, ResNet, and EfficientNet.
2. WebNN Samples Image Classification¶
It can be used to analyze model execution, loading, and building. You can switch between CPU/GPU/NPU, making it suitable for checking whether the NPU can be invoked normally.
3. Microsoft WebNN Demo Collection¶
It can be used to test speech recognition, image segmentation, text-to-image generation, and more. Each project indicates GPU/NPU support, allowing you to test real AI tasks.
Submit Your Data¶
After you complete the tests, you can submit the data to GitHub Issues. Thank you again for your support and participation.