AI Model Format Alignment for Cross-Device Communication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing communication systems face challenges in ensuring compatibility of AI models across diverse devices due to differing deep learning frameworks, leading to inefficiencies in model deployment and increased computational and air interface overhead.
Innovation Solution
A communication method and device that utilize model format configuration information to facilitate adaptive conversion of AI models between devices, ensuring compatibility and reducing unnecessary training and transmission overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If AI models are deployed across diverse devices with different deep learning frameworks, then device diversity and flexibility are improved, but model compatibility deteriorates
Solution Approach 1:
The patent introduces a model format conversion mechanism that acts as an intermediary between different deep learning frameworks. The source device converts AI models into a target format compatible with the target device's framework before transmission, ensuring compatibility across diverse devices without requiring both devices to use the same framework.
Solution Approach 2:
The patent changes the parameter of model format from framework-specific to framework-agnostic. By converting models between different formats (e.g., TensorFlow to PyTorch), the system adapts the model's structural parameters to match the target device's requirements, enabling cross-framework compatibility.
2Reliability
If AI models are converted between different formats, then model compatibility is improved, but computational overhead increases
Solution Approach 1:
The patent performs model format conversion in advance at the source device before transmission. This preliminary action ensures that the model is already in the correct format when received, eliminating the need for conversion at the target device and reducing overall computational overhead in the system.
Solution Approach 2:
The source device performs the conversion operation itself using its own computational resources, rather than requiring the target device to perform conversion. This self-service approach distributes the computational burden to the device that initiates the model transfer.
3Adaptability or versatility
If AI models are transmitted between devices, then model deployment flexibility is improved, but air interface overhead increases
Solution Approach 1:
The model format conversion mechanism serves as an intermediary that prepares models for efficient transmission. By converting models to the appropriate format before transmission, the system optimizes the data structure for the target device, reducing unnecessary data overhead during air interface transmission.
4Reliability
If model format conversion is performed, then compatibility among diverse devices is improved, but device complexity increases
Solution Approach 1:
The patent implements a standardized conversion mechanism that acts as an intermediary layer between different frameworks. This intermediary approach abstracts the complexity of format conversion, providing a unified interface that simplifies the overall system architecture despite the underlying conversion complexity.
Data Source
Figure 1~3
Figure 4a~5
Figure 6
AI summary
The present disclosure relates to a communication method and a device. The method includes a first communication device receiving model format configuration information from a second communication device. The method further includes the first communication device performing a model format-related operation according to the model format configuration information. In the embodiment of the present disclosure, a model format of a communication device can be indicated through the model format configuration information, thereby improving the compatibility of the model among different devices.