RISE: Relay Inference and Online Scheduling for Efficient Edge-Device Collaborative Diffusion Model Services
For edge AI services deploying text-to-image diffusion models, RISE addresses the trade-off between quality and latency by enabling efficient collaboration between edge and device models.
RISE proposes a training-free relay inference mechanism for edge-device diffusion model services, where a large edge model handles early denoising steps and a small device model refines details, achieving up to 2.1× speedup while preserving full-model quality. A contextual bandit scheduler selects relay configurations based on prompt complexity, user preferences, network quality, and node loads.
Text-to-image diffusion models are increasingly deployed at the network edge to serve heterogeneous workloads with diverse quality and latency requirements. However, existing deployment strategies choose either large edge-side models with high fidelity but high latency or lightweight device-side models that offer speed at the cost of semantic coherence. Moreover, these approaches rarely split the denoising workload between models of different sizes across edge servers and user devices. To bridge this gap, we propose RISE, a method for edge-device diffusion model services that combines relay inference with online scheduling. Driven by the finding that the latent intensity exhibits minimal deviation after a model handoff, RISE uses a training-free relay mechanism that exploits the shared latent space within a model family: the large model on the edge handles the early denoising steps that shape semantic structure, then passes the intermediate latent to a small device-side model for detail refinement. To deploy this mechanism as a practical service, a contextual bandit scheduler selects the best relay configuration based on prompt complexity, user preferences, network quality and real-time node loads. Experiments on two benchmarks show that RISE's relay mechanism achieves up to 2.1$\times$ speedup while preserving full-model quality, and its context-aware scheduler effectively balances quality and latency under mixed workloads.