Dfine nano version learning rate config error

[qwuzer]

DEIM/configs/deim_dfine/deim_hgnetv2_n_coco.yml

Lines 29 to 32 in 8f28fe6

	## Our LR-Scheduler
	flat_epoch: 7800 # 4 + epoch // 2, e.g., 40 = 4 + 72 / 2
	no_aug_epoch: 12
	lr_gamma: 1.0

In this nano version config, the flat epoch seems to be set wrong to 7800, which should be 78 instead. Also, why is lr_gamma set as 1.0, that is strange since wouldn't that cause the learning rate to be the same? Should I modify that to 0.5 so it match the other versions' base lr in

DEIM/configs/base/deim.yml

Line 39 in 8f28fe6

lr_gamma: 0.5

with 0.5?

Have been trying to train a DEIM model that works better on long tail objects for my machine learning course, if anyone could help answer my question here would be a great help, thanks!