Chi Xie
Chi Xie
and could you please tell me if you can successfully reproduce the result? I tried it on cifar100 but FRN performs way worse than BN.
> So have you reproduce FRN on cifar100?@Charles-Xie @yukkyo I am still trying FRN on cifar100 and imagenet. I find it performs slightly worse than bn on imagenet using batch...
> > > So have you reproduce FRN on cifar100?@Charles-Xie @yukkyo > > > > > > I am still trying FRN on cifar100 and imagenet. I find it performs...
@T1anZhenYu Yes, linear multistep schedule. I will try cosine learning rate and warmup for FRN later. Can you tell me the results of FRN (under best condition) and BN in...
@T1anZhenYu also it is mentioned in paper that > While training InceptionV3 and VGG-A, it was crucial to use learning rate rampup (refer Section 4.1) and learned epsilon (refer Section...
> > @T1anZhenYu Yes, linear multistep schedule. > > I will try cosine learning rate and warmup for FRN later. > > Can you tell me the results of FRN...
@T1anZhenYu sure
I also came across such problem (the zero-shot performance of the official checkpoint drops if we load and save it (without training)). I try to compare the hyper-parameters in the...
I guess this is the expected behavior because I have followed every step in the doc and also get 500G+ data on my disk
+1 looking forward to the release of the training code!