Abstract:
An accelerator resource management method and apparatus are disclosed. The accelerator resource management method includes receiving a task request for a neural network-related task and a resource scheduling policy for the neural network-related task, obtaining information on a current resource utilization status of an accelerator cluster comprising a plurality of accelerators, in response to the task request, and allocating an accelerator resource for performing the task based on a utility of a resource allocation that is based on the resource scheduling policy and the information.
Abstract:
A method and a device for preventing access to an administrative privilege are provided. The method includes acquiring a flash memory identifier from a flash memory, starting kernel loading, decoding an encrypted administrative privilege granting code using the flash memory identifier, and completing booting when the decoding succeeds. According to an embodiment of the present disclosure, the code for accessing the administrative privilege is encrypted using the identifier of the flash memory so that the unique password code can be loaded for each device, thereby enhancing the system security.
Abstract:
Disclosed are an audio processing apparatus and method including obtaining at least one substream and additional information by parsing a bitstream, obtaining at least one audio signal of at least one channel group (CG) by decompressing the at least one substream, and obtaining a multi-channel audio signal by de-mixing the at least one audio signal of the at least one CG, based on the additional information. The additional information includes a weight index offset identified based on an energy value of a height channel of the multi-channel audio signal and an energy value of a surround channel of the multi-channel audio signal.