Abstract: This application relates to an internet of things technology-based data processing method. In the method, the business procedure of an edge node is decoupled from an application, allowing a user, via a cloud management platform, to dynamically configure the business procedure of the edge node as well as the application that executes the business procedure, and to send a business model to the edge node indicating the edge node to execute a corresponding business based on the business model. In the method, both the business procedure of the edge node and the execution manner of the business procedure can be dynamically adjusted based on the requirements of the use case. This allows adaptation to requirements of diversified use cases, better coping with rapid changes in user-side business, and ensures the business execution effect of the application of the edge node in various use cases.

Abstract: Methods and systems are disclosed. The method comprises receiving the input text sequence; generating, a first likelihood score for each of a plurality of candidate output tokens using an expert model on the input text sequence, the expert model being a LoRA adapter trained to generate text based on a task; generating a respective second likelihood score for each of the plurality of candidate output tokens using an amateur model on the input text sequence, the amateur model having been previously used to train the expert model; generating a contrastive score for each of the plurality of candidate output tokens by maximizing a positive difference in their respective first likelihood score and second likelihood score; selecting a given candidate output text token as the output token based on the respective contrastive score.

Abstract: A domain name query method includes: an HTTPDNS server receives an HTTP request packet from a terminal device; the HTTPDNS server constructs a DNS query packet based on the HTTP request packet; the HTTPDNS server sends the DNS query packet to an authoritative DNS server; the HTTPDNS server receives a DNS reply packet from the authoritative domain name server; the HTTPDNS server constructs an HTTP reply packet based on the DNS reply packet; and the HTTPDNS server sends the HTTP reply packet to the terminal device. The HTTP reply packet carries the IP address of the to-be-accessed site.

Abstract: A method includes that obtaining a to-be-inferred node in an abstract syntax tree during development of program code, where the to-be-inferred node indicates a to-be-inferred field in the program code, and the to-be-inferred field includes one of a variable, a class, a function, or a parameter; obtaining a constraint related to the to-be-inferred node, where the constraint indicates a relationship related to the to-be-inferred node in the abstract syntax tree; and determining a type of the to-be-inferred node based on the constraint, where the type of the to-be-inferred node is one of a number type, a sequence type, a mapping type, a set type, an iterator type, a context manager type, a module type, a metaclass type, a none type, or a callable type. The solutions of this disclosure help reduce memory occupied for type inference.

Abstract: A cross-realm access control method based on a cloud management platform including: For a plurality of realms included in an organization of the tenant, when the tenant needs to perform access control on a first cloud service included in a first realm among the plurality of realms, the tenant inputs, via an access control interface provided by the cloud management platform, an access control policy. The cloud management platform receives the access control policy via the access control interface. If the access control policy indicates that an access authorization scope of the first cloud service included in the first realm is a second cloud service included in the organization of the tenant, the cloud management platform allows a second cloud service included in any realm in the organization of the tenant to access the first cloud service included in the first realm.

Abstract: A data management method, apparatus, and system, and a storage medium are provided which belong to the communication field. The method includes: obtaining first description information, wherein the first description information describes a first row group, the first row group is a row group stored by a first storage system within a specified time period, and the first row group includes at least one row of data; and storing the first row group in a second storage system based on the first description information. According to the application, operation costs and maintenance difficulty can be reduced.

Abstract: A content delivery network (CDN) system includes an optical network terminal configured to communicate with a CDN edge node and an optical network control center through a dedicated WAN interface, and enjoy bandwidth resources of the dedicated WAN interface, without occupying home bandwidth resources of a user. In the system, after determining that a first CDN edge node does not cache target content requested by a first client, the first CDN edge node sends a first query request to the optical network control center. Then, the optical network control center controls, based on the first query request, the first CDN edge node to request the target content from a CDN central node.

Abstract: This is a method and a system for integrating a cloud service engine framework with an open source service mesh (e.g. ISTIO) to bridge communication barriers between the two systems. The disclosure may enable users to adopt iterative cloud modernization model by embracing hybrid microservice architecture. Adoption of a multi-cloud architecture may also help to cut expenses due to small resources footprint of proxy less feature.

Abstract: A plurality of pieces of to-be-processed data are obtained, and a plurality of groups corresponding to the plurality of pieces of to-be-processed data are further obtained. Watermarks are added to to-be-processed data in the target group by using the target group as a granularity. Because watermarks are added to the plurality of pieces of to-be-processed data by using the group as the granularity, although the to-be-processed data in the group is modified due to watermark adding, the to-be-processed data changes within a range of a small amount of to-be-processed data included in the target group, and does not change within a range of the to-be-processed data, thereby alleviating a problem that the data is severely distorted after watermark adding. When to-be-processed data with similar semantics is placed in a same group, semantics of the to-be-processed data in the target group has a relatively small change after watermark adding.

Abstract: An access control method based on a cloud management platform, including: the first cloud instance sends a credential acquisition request to a cloud management platform, where the credential acquisition request carries an access control policy for the plurality of cloud instances, the access control policy is provided by the tenant for the first cloud instance, and the access control policy is used to indicate at least one cloud instance in the plurality of cloud instances that is allowed to obtain a credential. Then, the cloud management platform detect, based on the credential acquisition request, whether the first cloud instance is in the at least one cloud instance indicated by the access control policy. If the first cloud instance is in the at least one cloud instance indicated by the access control policy, the cloud management platform sends the credential of the tenant to the first cloud instance.

Abstract: A method for batch generation of test cases includes that obtaining a to-be-tested function of a target software product; obtaining an action factor and a data factor that correspond to the to-be-tested function, where the action factor is used to describe at least one of the following information in a test case: a prerequisite, a test step, and an expected result, and the data factor includes at least one test input parameter corresponding to the to-be-tested function and a plurality of values of each test input parameter; combining a plurality of values of the at least one test input parameter to obtain a plurality of data combination results; generating a plurality of corresponding test cases in batches based on the action factor and the plurality of data combination results; and displaying the plurality of generated test cases to a user.

Abstract: A code generation method includes that a code development platform receives a code generation task, and extracts a code context based on the code generation task, where the code context includes a code fragment related to to-be-generated code in a project; the code development platform determines, based on the code context using a language model, whether to call an application programming interface (API) of a language service; when it is determined to call the API of the language service, the code development platform sends an API call request to the language service and obtains an API call response of the language service; and the code development platform performs inference based on the code context and the API call response of the language service using the language model to obtain generated code.

Abstract: A natural language query statement processing method includes obtaining a natural language query statement; inputting the natural language query statement and the to-be-queried data table into a child table recall model to obtain at least one candidate data table; determining a target data table from the at least one candidate data table; and inputting the natural language query statement and the target data table into a natural language to structured query statement model to obtain a structured query statement.

Abstract: A traffic analysis method is applied to a cloud platform that includes a management node and a plurality of traffic analysis engines. The management node obtains program code of a domain-specific language corresponding to a target traffic analysis task, and the management node verifies the program code. When the verification of the program code succeeds, the management node allocates the program code to a target engine. The target engine determines a traffic analysis result based on the program code. After the management node in the cloud platform obtains the program code corresponding to the target traffic analysis task, the traffic analysis engine may directly analyze traffic in the cloud platform.

Abstract: In a database access method, if an application needs to access a database instance, the application sends an access request of the application to the database access initiation module. The access request carries an identifier of the database instance. When receiving the access request, the database access initiation module determines a database access service engine used to access the database instance, and sends the access request to the determined database access service engine. The database access service engine is deployed independently of the application. The database access service engine is configured to establish a connection used to access the database instance. In this way, the database access initiation module may forward the access request to the database instance using the database access engine such that the database instance may process the access request.

Abstract: In an application remodeling method, an application management system obtains an environment variable of an application in a runtime environment and a code package of the application. The system then obtains a remodeling result of the application based on the environment variable and the code package. The remodeling result includes an image set of the application, the image set includes a plurality of images, and a container created based on each image is configured to execute modules of the application. Thus, automatic application remodeling is implemented based on the environment variable and the code package of the application.

Abstract: One example virtual avatar control method includes obtaining live streaming interaction data of a live streaming room, where the live streaming interaction data includes live streaming bullet comment data and live streamer speech data, and the live streaming room includes a virtual avatar and a live streamer, determining a target to-be-answered question based on the live streaming bullet comment data, where the live streaming bullet comment data is sent by a viewer in the live streaming room, determining target interaction content based on the target to-be-answered question, where the target interaction content includes reply content of the target to-be-answered question, determining target interaction data based on the target interaction content, and adding the target interaction data to live streaming data of the live streaming room.

Abstract: A multimedia processing system communicates with a plurality of terminal devices and includes a forwarding apparatus and an adaptation apparatus. The forwarding apparatus is configured to receive a first video stream with a first resolution from a first terminal device. The adaptation apparatus is configured to generate a second video stream with a second resolution based on the first video stream. The second resolution is greater than the first resolution. The forwarding apparatus is further configured to send the second video stream to a second terminal device. Both the first terminal device and the second terminal device access a same video application. In this way, even if computing power of the first terminal device is low, the second terminal device can still obtain the second video stream with the higher resolution through resolution conversion performed by the multimedia processing system on the video stream.

Abstract: Systems, methods, and processor-readable media for elastic allocation of resources for deep learning jobs are described. A machine-learning-as-a-service (MLaaS) of a cloud computing system includes an elastic training module which includes resource allocator for allocating resources to training jobs that optimizes overall estimated time to completion (ETC) for all training jobs received by the system and uses node-based resource allocation. The elastic training module may realize a combination of high resource utilization, short training times, and low queueing delay relative to existing approaches, thereby potentially enabling the realization of higher profits for a cloud computing system which provides MLaaS to users (i.e. customers). An improved user interface is described, enabling users to specify a range of resources to elastically allocate to the user's training job, and/or informing users of training time saved through the use of elastic resource allocation.

Abstract: A method and an apparatus for transferring knowledge from a Large Language Model (LLM) to a Small Language Model (SLM) are provided. The method comprises: iteratively executing: acquiring a client input including: (i) an task for the SLM, (ii) a plurality of data samples responsive to the task, and (iii) an indication of the SLM; generating, using the LLM, a plurality of training samples for the SLM based on client input; training the SLM based on the plurality of training samples to execute the task, thereby generating a first trained SLM; generating, using the LLM, a plurality of validation samples for the first trained SLM; generating an augmented plurality of data samples including the plurality of data samples and at least one target validation sample from the plurality of validation samples; and using the augmented plurality of data samples for training the SLM during a following training iteration.