How should I refer when the two training stages have converged？

[Hryxyhe]

Dear Author : I really like your great contribution and try to utilize this framework to my task. But when I respectively train the local and global adapters, I do not know how to refer when the two training stages have converged? Since my sub-task needs pixel level performance, it is timely and have poor performance at two stages. So I would like to know what is the respective final performance of this two adapters before the joint inference?

[hhh388]

Dear Author : 亲爱的作者： I really like your great contribution and try to utilize this framework to my task. But when I respectively train the local and global adapters, I do not know how to refer when the two training stages have converged? Since my sub-task needs pixel level performance, it is timely and have poor performance at two stages. So I would like to know what is the respective final performance of this two adapters before the joint inference?我真的很喜欢你的巨大贡献，并尝试利用这个框架来完成我的任务。 但是，当我分别训练本地和全局适配器时，我不知道如何引用两个训练阶段何时收敛？由于我的子任务需要像素级的性能，所以它是及时的，并且在两个阶段的性能都很差。所以我想知道这两个适配器在联合推理之前各自的最终性能如何？

Hello, did you solve this problem? I would also like to know what the respective final performance of this two adapters looks like，my training loss is not decreasing.

[Hryxyhe]

Dear Author : 亲爱的作者： I really like your great contribution and try to utilize this framework to my task. But when I respectively train the local and global adapters, I do not know how to refer when the two training stages have converged? Since my sub-task needs pixel level performance, it is timely and have poor performance at two stages. So I would like to know what is the respective final performance of this two adapters before the joint inference?我真的很喜欢你的巨大贡献，并尝试利用这个框架来完成我的任务。 但是，当我分别训练本地和全局适配器时，我不知道如何引用两个训练阶段何时收敛？由于我的子任务需要像素级的性能，所以它是及时的，并且在两个阶段的性能都很差。所以我想知道这两个适配器在联合推理之前各自的最终性能如何？

Hello, did you solve this problem? I would also like to know what the respective final performance of this two adapters looks like，my training loss is not decreasing.

Not yet. I have moved back to original ControlNet. Maybe you could try some other new open-source works