AI Model Development Platform Operator Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The development of on-device AI models for mobile terminals is hindered by cumbersome and costly processes, high memory usage of cloud-based models, and the need for tailored optimizations for diverse mobile terminal environments, leading to high development costs and difficulties in debugging.
Innovation Solution
A development platform that obtains platform and service indication information, selects operators from a database, generates candidate policies, performs sample training, and simulates AI model execution to match performance requirements, automatically generating model code deployable on various running platforms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If cloud-based AI models are configured on mobile terminals, then AI processing capability is improved, but memory space occupation increases and computing capability requirements become higher
Solution Approach 1:
The patent segments the AI model into multiple operators and processes them in stages. The development platform divides the AI model into discrete operator components, allowing selective deployment and execution. This segmentation enables the system to process data in smaller chunks rather than loading the entire model into memory, thereby reducing memory space occupation while maintaining AI processing capability.
Solution Approach 2:
The patent introduces a new dimension of operator-level granularity for AI model execution. Instead of treating the AI model as a monolithic unit, the system operates at the operator level, enabling fine-grained control over memory usage and computation. This dimensional change allows the system to optimize the balance between processing capability and memory requirements by selecting and executing only necessary operators.
2Speed
If dedicated AI processing chips are carried in mobile terminals, then AI processing speed is improved, but development process complexity and costs increase
Solution Approach 1:
The patent creates a universal development platform that can generate AI model code applicable across different mobile terminal types and operating systems. The platform produces standardized operator code that can be executed on various devices without requiring device-specific customization, thereby reducing development process complexity while maintaining AI processing speed through efficient operator implementation.
Solution Approach 2:
The patent changes the fundamental parameter of AI model representation from high-level model descriptions to low-level operator implementations. By transforming AI models into sequences of executable operators with specific parameters, the system achieves faster processing speeds while simplifying the development process, as operators can be generated and optimized automatically without complex device-specific adaptations.
3Reliability
If AI models are optimized for different terminal environments, then model performance is improved, but development difficulty and costs increase
Solution Approach 1:
The patent implements self-service through automated code generation and optimization. The development platform automatically analyzes terminal environment characteristics and generates optimized AI model code without requiring manual intervention from developers. This self-service approach maintains high model performance across different terminals while significantly reducing development difficulty and costs by eliminating the need for developers to manually optimize models for each environment.
Solution Approach 2:
The patent systematically changes optimization parameters based on terminal environment characteristics. The development platform automatically adjusts operator parameters, computation graphs, and execution strategies according to the specific terminal's capabilities, ensuring optimal model performance. This parameter-based adaptation approach replaces complex manual optimization processes, reducing development difficulty while maintaining high performance across diverse terminal environments.
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
Embodiments of this application disclose an AI model development method and apparatus, and relate to the field of AI technologies, so that an AI model that satisfies a running environment and a performance requirement of a terminal can be developed with comparatively low costs. A specific solution is as follows: A development platform obtains platform indication information, service indication information, and parameter requirement information of a to-be-generated AI model; selects, from an operator database, a plurality of available operators that correspond to a running platform and that are used to execute a first service; selects an operator from the plurality of available operators, and sets a parameter of the selected operator, to generate a first candidate policy; performs sample training on the first candidate policy to obtain a first candidate AI model; invokes a simulator component of the running platform to run the first candidate AI model to execute processing on test data of the first service, so as to obtain a first running parameter; and if the first running parameter matches the parameter requirement information, determines that the first candidate AI model is a target AI model.