AI Model Transformation via Device-Aware Operator Compatibility
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge lies in efficiently transforming an artificial intelligence-based model to be executable on a target device, considering the device's hardware capabilities and limitations, while ensuring optimal performance and compatibility.
Innovation Solution
A method is proposed that involves obtaining model information and target device information, determining model operators and target operators, comparing these lists, and then modifying the artificial intelligence-based model to include only operators that are compatible with the target device, using specific change algorithms based on operator types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If an AI model with excellent performance in a specific domain is used, then the model performance is improved, but the compatibility with specific hardware may deteriorate
Solution Approach 1:
The system transforms the AI model by changing its operational parameters to match hardware capabilities. It analyzes the target hardware's operator support list and modifies the model's operator list accordingly, replacing unsupported operators with supported alternatives while maintaining model functionality and performance
Solution Approach 2:
The system segments the model transformation process into distinct steps: obtaining model information, determining operator lists, comparing operator compatibility, and selectively transforming operators. This segmentation allows systematic handling of compatibility issues without compromising overall model performance
2Adaptability or versatility
If manual transformation of AI models for each hardware is performed, then hardware compatibility is improved, but the time and resource consumption deteriorates
Solution Approach 1:
The system implements self-service automation where the transformation process autonomously obtains model information, analyzes hardware compatibility, determines necessary operator transformations, and executes the model conversion without requiring manual intervention for each hardware platform
Solution Approach 2:
The system creates a universal transformation framework that can handle multiple hardware platforms through a single automated process. The method maintains a database of hardware operator lists and automatically adapts the model to any target hardware by comparing operator compatibility and applying appropriate transformations
Data Source
AI summary
Disclosed is a method for changing an artificial intelligence-based model to suit a target device based on a device awareness, performed by a computing device. The method comprises: obtaining model information corresponding to the artificial intelligence-based model, and target device information indicating a characteristic of the target device on which the artificial intelligence-based model is executed; obtaining a model operator list comprising model operators by determining the model operators included in the artificial intelligence-based model based on the model information, and obtaining a target operator list comprising target operators by determining the target operators which are criteria for changing the model operators, based on the target device information; comparing the model operator list and the target operator list; and changing the artificial intelligence-based model to a target model which is executable at the target device, based on a result of the comparison.