how to scale learning rate?

[fcakyon]

The original YOLOv5 repo calculates the learning rate and schedule automatically based on the given epoch number, batch size, and GPU count.

In MMYOLO, default lr is 0.01, but I assume it is calculated for 8 batch size x 8 GPU.

How should I scale it in my fine-tuning task when using N GPUs with M batch size per GPU?

[fcakyon]

Do you have any suggestions? @hhaAndroid

[RangeKing]

Thank you for your attention. Related PR #486.