loss nan!

[jigongbao]

when i train in visdrone, i just modify the batchsize and the max iter ,but there is a promblem in 768iter

[Frank-jinchuan]

maybe you can try to reduce your lr

[ChenhongyiYang]

Hi, we have recently updated a new VisDrone training config (gradient clip added), which can avoid NaN now.

[simon1624]

Hi, I set the Batchsize to 16 and augment the lr to 0.02 respectively, and I got the same loss nan error

[minigemzhen]

Hi, we have recently updated a new VisDrone training config (gradient clip added), which can avoid NaN now.

Hello, when I set lr to 0.01 or 0.001 when training private datasets, I have loss to nan, and when I set lr to 0.0001, training is normal, but the final result AP is nan.

[shaoyandea]

#97 when I set lr to 0.01 or 0.001 when training private datasets,training is normal, and the final result ap is lower than paper eight percent

[lebron-2016]

#97 when I set lr to 0.01 or 0.001 when training private datasets,training is normal, and the final result ap is lower than paper eight percent

Hi, have you solved it?