Device-Cloud Collaboration
An architectural model that dynamically divides computing tasks between end devices and cloud servers. Privacy-sensitive and latency-sensitive tasks are executed locally, while compute-intensive tasks or those requiring the latest model capabilities are uploaded to the cloud.
Timeline (last 90 days)
Device-cloud collaboration refers to an architecture model that divides computing tasks between end devices and cloud servers, dynamically allocating compute power according to task characteristics.
The difficulty of device-cloud collaboration lies in making users unaware of the boundary of task switching, and in keeping the capabilities at both ends consistent when model versions are updated.
At the hardware level, the compute power, power consumption and heat dissipation of home devices all need to be balanced within consumer-grade costs; at the software level, the scheduling of device-cloud collaboration, model version management and data consistency have not yet been fully solved.
On smartphones, voice wake-word recognition is done locally while complex semantic understanding is handed to the cloud, which is the prototype of a device-cloud collaboration architecture.
AI photo-retouching products supporting millions of DAUs need to control per-inference costs through model distillation, quantization compression, and edge-cloud collaboration.
All Facts (5)
Device-cloud collaboration refers to an architecture model that divides computing tasks between end devices and cloud servers, dynamically allocating compute power according to task characteristics.
50%UnverifiedThe difficulty of device-cloud collaboration lies in making users unaware of the boundary of task switching, and in keeping the capabilities at both ends consistent when model versions are updated.
50%UnverifiedAt the hardware level, the compute power, power consumption and heat dissipation of home devices all need to be balanced within consumer-grade costs; at the software level, the scheduling of device-cloud collaboration, model version management and data consistency have not yet been fully solved.
50%UnverifiedOn smartphones, voice wake-word recognition is done locally while complex semantic understanding is handed to the cloud, which is the prototype of a device-cloud collaboration architecture.
50%UnverifiedAI photo-retouching products supporting millions of DAUs need to control per-inference costs through model distillation, quantization compression, and edge-cloud collaboration.
50%