Abstract: An inference system applies a machine-learning transformer model to a batch of requests with variable input length or variable target length or variable internal sate length by selectively batching a subset of operations in the transformer model but processing requests in the batch individually for a subset of operations in the transformer model. In one embodiment, the operation to be processed individually is an attention operation of an encoder or a decoder of the transformer model. By selective batching, the inference system can allow batching operations to be performed for a batch of requests with variable input or target length or internal state length to utilize the parallel computation capabilities of hardware accelerators while preventing unnecessary computations that occur for workarounds that restrain the data of a batch of requests to a same length.

Abstract: An inference system applies a machine-learning transformer model to a batch of requests with variable input length or variable target length or variable internal state length by selectively batching a subset of operations in the transformer model but processing requests in the batch individually for a subset of operations in the transformer model. In one embodiment, the operation to be processed individually is an attention operation of an encoder or a decoder of the transformer model. By selective batching, the inference system can allow batching operations to be performed for a batch of requests with variable input or target length or internal state length to utilize the parallel computation capabilities of hardware accelerators while preventing unnecessary computations that occur for workarounds that restrain the data of a batch of requests to a same length.

Abstract: Disclosed are a device and method for managing credentials. In a multi-decentralized identity management service environment, a user can safely manage credentials that have been once issued, and can use the credentials in various user terminals without being redundantly issued credentials.

Abstract: An inference system applies a machine-learning transformer model to a batch of requests with variable input length or variable target length or variable internal sate length by selectively batching a subset of operations in the transformer model but processing requests in the batch individually for a subset of operations in the transformer model. In one embodiment, the operation to be processed individually is an attention operation of an encoder or a decoder of the transformer model. By selective batching, the inference system can allow batching operations to be performed for a batch of requests with variable input or target length or internal state length to utilize the parallel computation capabilities of hardware accelerators while preventing unnecessary computations that occur for workarounds that restrain the data of a batch of requests to a same length.

Abstract: Disclosed is a method of executing deep learning programs. The method includes generating a symbolic graph corresponding to an imperative deep learning program, dividing the imperative deep learning program into a first portion related to a deep learning computation and a second portion not related to the deep learning computation, and performing a computation on the first portion using a graph runner and simultaneously performing a computation on the second portion using a language runner.

Abstract: Disclosed herein are an apparatus and method for issuing delegated credentials between digital wallets possessed by multiple users connected to a Decentralized Identifier (DID) registry through wired/wireless communication. The method includes receiving, by a digital wallet of a first user, a delegated credential issuance request message including a newly created DID document from a digital wallet of a second user, generating, by the digital wallet of the first user, a delegated credential using both the delegated credential issuance request message, received from the digital wallet of the second user, and an original credential, previously issued by the digital wallet of the first user, and transmitting, by the digital wallet of the first user, the generated delegated credential and the original credential, together with a delegated credential registration request message, to the digital wallet of the second user.

Abstract: Provided are a method and apparatus for providing a counseling service. A method of operating a counseling center server interworking with an instant messaging service includes acquiring, based on an input of a first user for requesting counseling through a channel of a second user registered in the instant messaging service, a counseling type of the requested counseling, creating a first type of counseling ticket corresponding to unmanned counseling or a second type of counseling ticket corresponding to manned counseling based on the counseling type, monitoring a state change event that changes a state of the counseling ticket, and controlling an interfacing object included in a chat interface for counseling provided to a terminal of the first user to correspond to the channel to indicate the counseling type and the changed state of the counseling ticket.

Abstract: An inference system applies a machine-learning transformer model to a batch of requests with variable input length or variable target length or variable internal state length by selectively batching a subset of operations in the transformer model but processing requests in the batch individually for a subset of operations in the transformer model. In one embodiment, the operation to be processed individually is an attention operation of an encoder or a decoder of the transformer model. By selective batching, the inference system can allow batching operations to be performed for a batch of requests with variable input or target length or internal state length to utilize the parallel computation capabilities of hardware accelerators while preventing unnecessary computations that occur for workarounds that restrain the data of a batch of requests to a same length.

Abstract: A tray assembly for loading a battery cap includes at least one first tray having a top surface, in which first grooves are provided, and a bottom surface, in which second grooves are provided, and at least one second tray having a top surface, in which third grooves are defined, and a bottom surface, in which fourth grooves are defined and disposed above or below the first tray. When the first tray is disposed above the second tray, each of the second grooves and each of the third grooves communicate with each other to provide a communication space. When the first tray is disposed below the second tray, each of the first grooves and each of the fourth grooves communicate with each other to provide a communication space. The battery cap is mounted in an erect state in the communication space provided by the first tray and the second tray.

Abstract: The present disclosure relates to an assembly-type chair and provides an assembly-type chair that has a specific structure, thereby conveniently storing and transporting the chair with a minimized volume, and has a structure in which a chair body part including a seat plate and a backrest is rotatable, thereby effectively storing a plurality of chairs by stacking the chairs in the horizontal direction so as to be in close contact with each other.

Abstract: An inference system applies a machine-learning transformer model to a batch of requests with variable input length or variable target length or variable internal state length by selectively batching a subset of operations in the transformer model but processing requests in the batch individually for a subset of operations in the transformer model. In one embodiment, the operation to be processed individually is an attention operation of an encoder or a decoder of the transformer model. By selective batching, the inference system can allow batching operations to be performed for a batch of requests with variable input or target length or internal state length to utilize the parallel computation capabilities of hardware accelerators while preventing unnecessary computations that occur for workarounds that restrain the data of a batch of requests to a same length.

Abstract: Disclosed herein are an apparatus and a method for generating an audio-embedded image. The apparatus for generating an audio-embedded image includes one or more processors, and execution memory for storing at least one program that is executed by the one or more processors, wherein the at least one program is configured to receive audio and an image, convert the audio into audio information having a preset image format and generate an audio-embedded image in which the audio information is embedded in the image, and discriminate the audio-embedded image using the audio information and discrimination audio information extracted from the audio-embedded image.

Abstract: Disclosed herein are an apparatus and method for managing a history of object owners. The method includes generating, by an object owner history management apparatus, a product registration Verifiable Credential (VC) in a form of a Merkle tree, and registering the product registration VC in a blockchain in order to manage a history of owners of an object, acquiring, by a first user terminal device, the product registration VC from the blockchain and registering a first owner VC, generated based on the product registration VC, in the blockchain to register an ownership of the object, and acquiring, by a second user terminal device, a product transfer Verifiable Presentation (VP) from the first user terminal device and registering a second owner VC, generated based on the product transfer VP, in the blockchain to receive transfer of the ownership of the object.

Abstract: Disclosed herein are an apparatus and method for managing personal information.

Abstract: Disclosed herein are a method and apparatus for detecting a disaster based on images. The apparatus includes an image capture unit for capturing video using at least one camera and controlling the camera based on a camera control signal received from the outside; a disaster detection unit for generating a disaster log based on the video captured using the camera; a disaster analysis unit for calculating a disaster occurrence probability value based on the disaster log and determining whether to enter a camera control mode based on the disaster occurrence probability value; and a disaster alert unit for warning of a disaster based on a disaster alert request signal.

Abstract: The present disclosure relates to an assembly-type chair and provides an assembly-type chair that has a specific structure, thereby conveniently storing and transporting the chair with a minimized volume, and has a structure in which a chair body part including a seat plate and a backrest is rotatable, thereby effectively storing a plurality of chairs by stacking the chairs in the horizontal direction so as to be in close contact with each other.

Abstract: A method for compressing time series data includes: reading original data including time series data; measuring a unit of the original data; determining a threshold for determining a range allowing for a difference between compressed data and the original data; performing longest distance downsampling to preserve an abnormal point of the original data; storing a start point and an end point of the original data; performing drop-out on the longest distance downsampling result, wherein the drop-out is not performed for the start point and the end point; and if the number of values of the original data is smaller than ? times the original data length, storing at least one of the values of the original data as a reference point, wherein ? is a real number having a value between 0 and 1.

Abstract: A method for compressing time series data includes: reading original data including time series data; measuring a unit of the original data; determining a threshold for determining a range allowing for a difference between compressed data and the original data; performing longest distance downsampling to preserve an abnormal point of the original data; storing a start point and an end point of the original data; performing drop-out on the longest distance downsampling result, wherein the drop-out is not performed for the start point and the end point; and if the number of values of the original data is smaller than a times the original data length, storing at least one of the values of the original data as a reference point, wherein ? is a real number having a value between 0 and 1.

Abstract: An accessory apparatus and an information display method are provided. The accessory apparatus include a housing, a display exposed through a first portion of the housing, a conductive pattern, a wireless communication circuit configured to support a short-range wireless communication protocol, at least one processor, and a memory, wherein the memory stores instructions that allow the at least one processor to control the wireless communication circuit to operate in the first mode, operate in the second mode in which the second type of external electronic device is allowed to read the second data for the second time period following the first time period, detect the existence of the second type of external electronic device at one or more points of time in the first time period, and switch into the second mode when the existence of the second type of external electronic device is detected.

Abstract: An electronic device of the present disclosure is provided. The electronic device includes a housing having a first plate facing a first direction, a second plate facing a second direction, and at least one side part, between the first plate and the second plate, facing a third direction, a first conductive pattern comprising a first coil having a first axis substantially extending in the first direction or the second direction, a second conductive pattern comprising a second coil having a second axis substantially extending in the third direction, and a communication circuit connected with the first conductive pattern and the second conductive pattern.