AI Model Partitioning for Edge Computing Scheduling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current technologies face challenges in efficiently deploying and executing large-scale artificial intelligence (AI) models due to limitations in computing resources and communication bandwidth, particularly in edge computing environments where heterogeneous devices with restricted capabilities are used.
Innovation Solution
The proposed solution involves generating an intermediate representation of an AI model, partitioning it into subsets, and scheduling these partitions for execution across multiple edge computing devices, leveraging model parallelism to facilitate parallel computation and resource-efficient deployment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single computing device is used to execute AI models, then device complexity is reduced, but computing capability and resource efficiency are insufficient for large-scale models
Solution Approach 1:
The patent divides a large-scale AI model into multiple partitions and distributes them across multiple computing devices in the edge computing environment. This segmentation allows the system to handle models that exceed the capacity of a single device while maintaining manageable complexity through modular architecture.
Solution Approach 2:
The patent transitions from a single-device execution model to a multi-device distributed execution model, adding the dimension of spatial distribution across the edge computing network. This enables parallel computation and scales the system's computing capability beyond individual device limits.
2Productivity
If AI models are partitioned and distributed across multiple edge computing devices, then resource efficiency and parallel computation are improved, but scheduling and coordination complexity increases
Solution Approach 1:
The patent performs model partitioning and scheduling decisions in advance before execution. The system analyzes the AI model structure and edge device capabilities beforehand to create an optimal partitioning scheme, which simplifies runtime coordination and reduces scheduling complexity during actual execution.
Solution Approach 2:
The patent introduces a model management system that acts as an intermediary between the AI model and edge computing devices. This intermediary handles the complex tasks of partitioning, scheduling, and coordination, isolating the complexity from both the model and the individual devices while enabling efficient parallel execution.
3Adaptability or versatility
If AI models are executed on edge computing devices with heterogeneous capabilities, then adaptability to different devices is improved, but resource usage efficiency deteriorates due to capability mismatches
Solution Approach 1:
The patent assigns different model partitions to edge computing devices based on their specific capabilities and resource characteristics. Each device executes partitions that match its local quality or capability profile, optimizing resource usage efficiency while maintaining adaptability across heterogeneous device types through capability-aware partitioning.
Data Source
AI summary
Techniques are disclosed for managing artificial intelligence model partitions for execution in an information processing system with edge computing resources. For example, a method comprises the following steps. An intermediate representation of an artificial intelligence model is obtained. A computation graph is generated based on the intermediate representation. The computation graph is partitioned into a set of partitions. The method then schedules the set of partitions for respective execution on a set of computing devices in an edge computing environment, and causes deployment of the set of partitions respectively to the set of computing devices for execution in the edge computing environment.