How to call shared memory to accelerate model training when video memory is insufficient in WSL2 environment?

When deploying and training a model using Windows Subsystem for Linux (WSL2), I need to use the GPU to speed up the training process. However, I encountered a dilemma: the video memory requirements of the model exceeded the capacity limit of the independent video card. In Windows systems, when the independent video memory is insufficient, the system will automatically call the shared video memory to meet the demand. Is there any way to implement a similar function in the WSL2 environment, that is, to call the shared video memory when the independent video memory is insufficient to continue training the model?

The issue names no source files, tests, or entry points. Start by checking the WSL2 GPU and memory documentation and existing issue discussion for support of shared video memory; done would be a confirmed answer about whether this capability is supported and how it can be used, if at all.

Written by the indexing model from the issue text.