WORKSHOP 2: AI FOR EVERYONE
summary
Huawei demonstrates moving models onto the HiAI Foundation with a drag‑and‑drop converter (offline via Docker or an upcoming online service) and an Android Studio plugin called DevEco IDE that auto‑generates JNI, Java wrappers and Gradle entries. On‑device NPU (MPU) inference claims 25× CPU speed and 50× power savings (ResNet‑50: 2,005 vs iPhone X 880 img/min). Prisma reports a style‑transfer app running ~300 ms on NPU vs ~1 s on CPU after tuning. Tools report unsupported operators and can fall back to CPU for those layers.
AI-written summary · about
chapters (7)
- 0:00Introduction and poll
- 2:52On-device AI with MPU
- 14:03Prisma integration experience
- 22:13Tooling support for AI models
- 31:55Demo setup and remote devices
- 37:07Live coding demo
- 59:07Q&A session
related talks
36mHuaweu HiAI Unlock The Future
39mRunning On-Device AI With Qualcomm AI Engine Using LiteRT & ExecuTorch
59mKeynote: AI without the BS, for humans
32mBuilding Once, Running Everywhere: The Future of Portable AI Applications
26mAI computing power in Edge Industry use cases
15mSimplifying & Scaling Edge AI
18mThe Cutting Edge of AI and IoT
39mServing AI at Massive Scale: The Cloud Native Inference Plane Behind Huawei...