ML Model Transport Packages for Cross-Platform Deployment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current MLOPs solutions are inflexible and require significant time and effort for cross-platform integration of machine learning models, leading to inefficiencies and software bugs due to the need for special configurations in different software environments.
Innovation Solution
A platform-agnostic system for managing, transporting, and storing machine learning models using model, station, and route objects, which are generated and stored in a decentralized network, allowing seamless deployment and sharing across various computing devices without specific configuration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If current MLOps solutions use specific machine learning frameworks for model deployment, then model operation reliability is improved, but cross-platform adaptability deteriorates due to requiring special configurations in different software environments
Solution Approach 1:
The patent introduces a standardized model package format as an intermediary layer between different machine learning frameworks and deployment environments. This package acts as a mediator that encapsulates model data in a unified structure, allowing models trained in one framework to be deployed in another without requiring framework-specific configuration modifications, thus resolving the contradiction between reliability and cross-platform adaptability
Solution Approach 2:
The invention creates a universal model package format that can be used across multiple different machine learning frameworks and software environments. This single standardized format serves multiple functions: it stores model data, provides metadata about the model, and enables interoperability between different platforms, eliminating the need for separate configuration approaches for each framework while maintaining operational reliability
2Reliability
If machine learning models are configured for specific software environments, then operational reliability is improved, but deployment time and effort increase significantly
Solution Approach 1:
The patent applies preliminary action by pre-packaging model data into a standardized format during the model development phase. This packaging process includes preparing all necessary model artifacts, metadata, and configuration information in advance within a unified structure. When deployment time arrives, the pre-packaged model can be directly transferred and deployed without requiring time-consuming configuration adjustments, thus reducing deployment time while maintaining operational reliability through the standardized format
3Adaptability or versatility
If platform-specific configurations are used for model deployment, then framework compatibility is improved, but system complexity increases due to multiple configuration requirements
Solution Approach 1:
The invention extracts the configuration information and model data into a separate, self-contained standardized package. By taking out the configuration requirements from the deployment process and encapsulating them within the unified model package format, the system eliminates the need for multiple separate configuration steps and reduces overall system complexity while maintaining framework compatibility across different platforms
Data Source
AI summary
Provided are systems, methods, and computer program products for delivery of machine learning models within a computing network including memory including object storage locations configured to store model objects, route objects, and station objects, a model storage device configured for storing machine learning models, and a processor configured with a class interface module and a class data aggregator module. The processor is configured to execute program code that, when executed, will cause the processor to execute the class interface module and the class data aggregator module to generate a model object based on a model class, generate a station object based on a station class, generate a route object based on a route class, extract the model object including parameters of a machine learning model from a first storage location, generate a model data package, and transmit the model data package to a computing node.