Method of managing data representation for deep learning, method of processing data for deep learning and deep learning system performing the same

A deep learning system comprising:

• a plurality of heterogeneous resources;
• a first processor configured to generate first path information indicating a first computing path for a first task when the first task includes at least one of a plurality of operations to be performed by at least one of the plurality of heterogeneous resources, the first computing path including a sequence of the operations included in the first task and a driving sequence of the resources for performing the operations included in the first task;
• and
• a second processor configured to set data representation formats of the resources for performing the operations included in the first task based on data representation information and the first path information, the data representation information indicating an optimized data representation format for each of the plurality of heterogeneous resources.

A deep learning system comprising

• a memory storing a computer program comprising a neural network configured to classify data into one of a plurality of classes, the neural network including a task comprising a plurality of sequential operations;
• a first processor of a first type initially configured to output data in a first data format;
• a second processor of a second type different from the first type initially configured to operate on data in a second data format different from the first data format;
• a third processor configured to determine a first operation among the operations that is to be performed by the first processor having an output that is to be provided to a second operation among the operations that is to be performed by the second processor, and set the first processor to output data in the second data format when the first processor is about to perform the first operation.