Customize learning rates for modules in architecture

[aditya9710]

I am running the PointPillars architecture on KITTI dataset. This architecture has mainly three modules, a Feature Extractor, a 2D CNN Backbone, and a Detection Head. I want to use different learning rates for these modules. How can I do so with the mmdetection3d framework? Any examples would be greatly appreciated.

[Xiangxu-0103]

You can refer to this tutorial for usage.