ViTDet icon indicating copy to clipboard operation
ViTDet copied to clipboard

Published 20 hours ago • verified publisher icon ViTAE-Transformer

Why loading pre-trained models does not match？

Open fadaishaitaiyang opened this issue 2 years ago • 2 comments

e68847d37da998b0bda6d0fd406fc0b

fadaishaitaiyang avatar Jun 15 '22 00:06 fadaishaitaiyang

Hi,

It seems that you have modified the backbone structures with convolution and batch normalization in the FFN part.

Annbless avatar Jun 15 '22 00:06 Annbless

I just modified the backbone structures with convolution in vit part.

fadaishaitaiyang avatar Jun 15 '22 00:06 fadaishaitaiyang