4results about How to "Reduce verification times" patented technology

The present disclosure provides a storage chip test method and device, computer equipment and storage medium, and relates to the technical field of semiconductors. The method comprises the following steps: performing M times of burn-in cycle tests on a storage chip, M being a positive integer greater than 0, each burn-in cycle test comprising the following steps: performing N times of write operations on the storage chip, wherein N is a positive integer greater than 1; verifying the storage chip after the N times of write operations are performed to obtain a verification result; and obtaining a burn-in test result based on M verification results of the storage chip after the burn-in cycle test. The present disclosure obtains the burn-in test result based on the M verification results after performing the M times of burn-in cycle tests, thereby reducing the probability of misjudging the storage chip as a failed product due to slight poor contact. In addition, the storage chip is verified after the N times of write operations are performed, thereby reducing the number of verifications in the burn-in test process, and further reducing the burn-in test duration and test cost.

The application provides an acceleration method, system, device and storage medium for quantization speculation sampling, which comprises the following steps: obtaining context content input in a natural language generation task; outputting a candidate sequence in a drafting model of quantization speculation sampling according to the context content; extracting entropy fluctuation features and instantaneous entropy difference features of the candidate sequence, judging the difficulty of task generation of the candidate sequence based on the entropy fluctuation features and the instantaneous entropy difference features, routing the candidate sequence as a text sequence to be verified to a target model with different precisions according to the difficulty, and enabling the target model with a selected precision to perform parallel verification on a predetermined number of text sequences determined. The application realizes a substantial optimization of inference efficiency and memory occupation without significantly sacrificing generation quality.

The application relates to the technical field of intelligent manufacturing knowledge management, and discloses a semantic enhancement knowledge graph dynamic integrity protection method, which comprises the following steps: semantic enhancement modeling, constructing a unified data model in the discrete manufacturing field, and the unified data model at least comprising an entity set, a relation set, an attribute set, a static structure constraint set and a dynamic operation constraint set; formulating the static structure constraint set and the dynamic operation constraint set into an SHACL rule base; performing a structure constraint shape in the SHACL rule base on full-amount data before knowledge graph deployment, generating a structured verification report, intercepting a real-time data operation request, extracting a semantic subgraph related to the current operation, and only performing local verification on the semantic subgraph by loading corresponding SPARQL operation constraint shapes, and if the verification fails, the operation request is rejected and a correction prompt is returned. The application realizes the elimination of semantic fragmentation by internalizing manufacturing business rules into dynamic operation constraints.

The application relates to the technical field of data analysis, in particular to a production variable verification method and device, equipment and a computer readable storage medium, the method comprising the following steps: packaging processing logic codes of various production variables to obtain variable code packages; obtaining verification data, calling the variable code packages by taking the verification data as input parameters, and obtaining processing results of the various production variables; generating check results of the various production variables according to the processing logic codes of the various production variables and the verification data; and if the processing result of a target production variable in the various production variables is consistent with the check result of the target production variable, it is determined that the target production variable passes the verification. The application improves the online timeliness effect of the production variable.