2.5.0 Update (16th May 2025)
OwLite 2.5.0 Update (16 May 2025)
Hello from the OwLite team! 🎉
Today we’re thrilled to announce QNN (Qualcomm® Neural Network) output support, bringing hardware‑accelerated inference and deep profiling to Snapdragon™ & Qualcomm IoT devices — all straight from your familiar OwLite workflow.
Export Directly to QNN
QNN is Qualcomm’s cross-platform neural-network runtime and graph format that converts trained models into hardware-optimized executables and dynamically schedules them across Snapdragon CPU, Adreno GPU, Hexagon DSP, and NPU cores for ultra-low-latency, battery-efficient on-device AI.
QNN automatically routes each layer to the fastest engine available (CPU, Adreno GPU, Hexagon DSP, NPU) for lower latency and longer battery life.
Keep working with OwLite’s model visualization and detailed compression setting for the best accuracy‑vs‑speed trade‑off.
Getting Started (Cloud Plan)
Upgrade to OwLite Cloud Plan (
Settings → Subscription).Install QNN Runner and issue a Device Key — be sure to pick Target Framework = QNN during issuance.
With
owlite device connect, connect with OwLite Runner synced with Qualcomm AI Hub .Join Qualcomm AI Hub, copy your API Token, and paste it into Account → API Tokens.
https://aihub.qualcomm.com/
In code, call
owlite.init(device="SA8775P ADP")(or another device string from the list below).Open your OwLite project(framework = QNN) and configure your compression configuration.
Once you are done, build and benchmark your experiment with OwLite.
Supported Devices (Initial List)
Try it out and let us know what you think! Your feedback keeps OwLite evolving. 🚀
Last updated