Data filtering method, device and medium based on hardware processing capability

By dividing the data into multiple groups and filtering candidate data based on hardware processing capabilities, the processing length limitation of hardware acceleration devices is solved, achieving efficient and accurate data filtering that adapts to the processing capabilities of hardware computing units.

CN121835758BActive Publication Date: 2026-06-09MOXIN ARTIFICIAL INTELLIGENCE TECH (SHENZHEN) CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
MOXIN ARTIFICIAL INTELLIGENCE TECH (SHENZHEN) CO LTD
Filing Date
2026-03-13
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Current hardware acceleration devices (such as GPUs and dedicated AI chips) have a limited single-processing length in their built-in Top-K computing modules, which creates a bottleneck between efficiency and accuracy when processing massive amounts of data, and existing solutions cannot effectively solve this problem.

Method used

The data to be processed is divided into multiple data groups. Candidate data is selected based on the grouping ratio of hardware processing capabilities and merged into a candidate data set. The final selection is then performed by the hardware computing unit, achieving a balance between hardware adaptability, processing efficiency, and result accuracy.

Benefits of technology

It improves the processing efficiency of the hardware computing unit and appropriately ensures the accuracy of the processing results, achieving a triple balance between hardware adaptability, processing efficiency and result accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121835758B_ABST
    Figure CN121835758B_ABST
Patent Text Reader

Abstract

The present disclosure relates to a hardware processing capability-based data screening method, device and medium. A hardware processing capability-based data screening method is provided, comprising: obtaining N data to be processed and a target number K from a large language model; dividing the N data into G data groups based on the hardware processing capability, wherein the hardware processing capability comprises a maximum data quantity M that can be processed in parallel by a hardware computing unit, and; using the hardware computing unit to screen candidate data from each data group based on a preset grouping screening ratio P, and merging the candidate data of all data groups into a candidate data set; and using the hardware computing unit to screen the target number K target data from the candidate data set.
Need to check novelty before this filing date? Find Prior Art