DEEP LEARNING PLATFORM CAPABLE OF RESOURCE MANAGEMENT OF GRAPHIC PROCESSING UNIT AND METHOD FOR MANAGING RESOURCE OF GRAPHIC PROCESSING UNIT USING THE SAME

摘要

The present invention relates to a deep learning platform capable of resource management of a graphic processing unit (GPU) and a resource management method of a graphic processing unit using the same. More specifically, the deep learning platform capable of resource management of a graphic processing unit and the resource management method of a graphic processing unit using the same intercept a corresponding function in accordance with a process generation request for using a graphic processing unit (GPU) to return the corresponding function as a normal function after a multi-instance memory manager allocates an available memory of the GPU in accordance with a preset policy, change name and property information in accordance with an allocated resource policy, and then return the changed property information to allocate only a required amount of memory resources of the GPU instead of 100% of the memory resources of the GPU when a process of a specific user requests GPU resources by control for a function to transfer resource information and a function for the property information.