I got unexpected behaviors of BiT-M-ResNet152x4 on PyTorch. When I use BiT-M-R152x4-ILSVRC2012.npz on ImageNet-1k dataset, most of model outputs are nan. Hence, the resultant performance is very low. Given that other models work well, I think that there might be some bugs on the pretrained weights of BiT-M-R152x4.
Hello,
I got unexpected behaviors of
BiT-M-ResNet152x4
on PyTorch. When I useBiT-M-R152x4-ILSVRC2012.npz
on ImageNet-1k dataset, most of model outputs arenan
. Hence, the resultant performance is very low. Given that other models work well, I think that there might be some bugs on the pretrained weights ofBiT-M-R152x4
.