Tensor Operator Exchange for Data Model Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data models suffer from reduced operational efficiency due to the interlaced operation orders of tensor-transpose and tensor-reshape operators, which complicates data distribution adjustments.
Innovation Solution
A data model adjustment method that identifies and exchanges the operation orders of target tensor-reshape and tensor-transpose operators within the data model, ensuring that these operators are adjacent, thereby simplifying calculations and improving operational efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If tensor-reshape and tensor-transpose operators are used alternatively in data models, then data distribution adjustment flexibility is improved, but operational efficiency deteriorates
Solution Approach 1:
The patent merges adjacent tensor-reshape operators or tensor-transpose operators into single unified operators. When multiple tensor-reshape operators are adjacent, they are combined into one operator that performs the cumulative reshape operation. Similarly, adjacent tensor-transpose operators are merged. This reduces the total number of operator executions while maintaining the same data transformation effect, thereby improving operational efficiency without sacrificing data distribution adjustment flexibility.
2Adaptability or versatility
If tensor-reshape and tensor-transpose operators are used alternatively in data models, then data distribution adjustment capability is improved, but calculation complexity increases
Solution Approach 1:
The patent combines multiple adjacent tensor-reshape operators into a single operator and merges adjacent tensor-transpose operators. This consolidation reduces calculation complexity by eliminating redundant operations while preserving the full data distribution adjustment capability through the unified operators that encapsulate the cumulative transformation logic.
Solution Approach 2:
The patent extracts and removes redundant tensor-reshape and tensor-transpose operators from the data model by identifying and eliminating duplicate or adjacent operations. This extraction process simplifies the calculation pipeline while maintaining the essential data distribution adjustment functions through the remaining optimized operators.
Data Source
AI summary
The present disclosure relates to the field of data processing, and specifically, it relates to a data model adjustment method and device, model construction method and device, electronic device, and computer-readable storage medium. The data model adjustment method includes obtaining the target tensor-reshape operator and target tensor-transpose operator in the target data model; determining whether the target tensor-reshape operator and the target tensor-transpose operator satisfy exchange condition; exchanging, if the target tensor-reshape operator and the target tensor-transpose operator satisfy the exchange condition, the operation order of the target tensor-reshape operator and the target tensor-transpose operator in the target data model; and adjusting the operator expression of the tensor-transpose operator.