Task Parallel Processing for Neural Network Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional computer equipment faces inefficiencies in processing heavyweight artificial neural networks due to direct data processing, which affects speed and efficiency, and these networks may not be compatible with operating systems that only process lightweight data, limiting their application scope.
Innovation Solution
A computer system with first and second processors, where the first processor obtains and manages offline models of neural networks, including model parameters and instructions, and controls the second processor to run these models, reducing the need for direct computation node compilation and enabling efficient processing of heavyweight neural networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional computer equipment processes heavyweight neural network data directly, then the neural network can be executed, but the data processing speed and efficiency are affected
Solution Approach 1:
The patent segments the heavyweight neural network data into multiple lightweight data packets that can be processed separately. The neural network is divided into computation nodes that are distributed across multiple processors, transforming a single large data structure into multiple smaller, manageable units that improve processing efficiency.
Solution Approach 2:
The patent introduces a new dimension of processing by implementing a distributed multi-processor architecture. Instead of processing the neural network in a single sequential stream on one processor, the system distributes computation across multiple processors simultaneously, adding parallelism as a new dimension to the processing workflow.
2Adaptability or versatility
If the artificial neural network is processed on operating systems that only process lightweight data, then compatibility is achieved, but the application scope is limited
Solution Approach 1:
The patent segments the neural network into computation nodes that can be independently processed as lightweight data units. This segmentation allows the neural network to be executed on operating systems designed for lightweight data processing while maintaining the full functionality of the original network through coordinated processing across multiple nodes.
Solution Approach 2:
The patent introduces a runtime system as an intermediary layer between the heavyweight neural network and the lightweight operating system. This runtime system manages the distribution and coordination of computation nodes, enabling compatibility between systems with different data weight capabilities while preserving the neural network's application scope.
Data Source
AI summary
Computer systems, data processing methods, and computer-readable media are provided to run original networks. An exemplary computer system includes first and second processors a memory storing offline models and corresponding input data of a plurality of original networks, and a runtime system configured to run on the first processor. The runtime system, when runs on the first processor, causes the first processor to implement a plurality of virtual devices comprising a data processing device configured to obtain an offline model and corresponding input data of an original network from the memory, an equipment management device configured to control turning on or off of the second processor, and a task execution device configured to control the second processor to run the offline model of the original network.