ITU-T F.748.20 PDF

This Recommendation aims to specify the technical framework of deep neural network (DNN) model partition and collaborative execution, including the establishment of a latency prediction model, the formulation of a DNN model partition strategy, resource allocation optimization and the collaborative execution of the model, so as to meet the demand of DNN model inference and realize the efficient resource utilization between the end devices and the edge device.

The scope of this Recommendation includes:

– Overview of DNN model partition and collaborative execution;

– Latency prediction model establishment of a DNN model partition;

– Partition strategy formulation of a DNN model;

– Collaborative execution of a DNN model after DNN model partition.