如何蒸馏和量化

[DWCTOD]

大佬，感谢您的分享，效果实测非常好。
但是显存占用和速度优化，有什么方案吗？

[Mountchicken]

Hi @DWCTOD
Sorry for the late reply. MAERec is a pure transformer architecture and there might be some off-the-shelf optimization methods. However I am not an expert of this field😂