CPC G06F 9/4881 (2013.01) [G06F 9/5016 (2013.01)] | 20 Claims |
1. A method of scheduling an accelerator, the method comprising:
receiving at least one execution request for a first model and a second model that are executed independently from each other in the accelerator; and
performing layer-unit scheduling on the first model and the second model based on workload characteristics of the first model and the second model, such that layers of the first model and layers of the second model are alternately executed by the accelerator,
receiving at least one execution request for a first model and a second model that are executed independently from each other in the accelerator; and
determining an optimal scheduling result to find a path indicating an execution order from an input layer included in each of the first model and the second model to an output layer included in each of the first model and the second model, the determining being based on. for a current step. employing a current simulation of adding a layer of the input layer in a direction from a previous simulation from a previous step to the current step.
|