Work

Joint Layer-wise Caching and Request Routing for Serverless Inference Acceleration at the Edge