ChenccLOL
Results
1
comments of
ChenccLOL
> > Thanks for your great work！ > > When I try to train maptrv2 with vovnet99 as backbone,the grad_norm is always nan. Reducing the learning rate and applying gradient...