I tried the simple CNN network provided in this repo, however, model.meta_named_parameters() gives only the parameters of MetaBatchNorm2d() and fully ignore MetaConv2d() and MetaLinear(). I checked model.named_parameters() and it is able to output all the parameters of the model.
My torchmeta version is 1.8.0 and pytorch version is 1.8.1.
Could you please give me a hand about this? Thanks
Sorry, I figure out the reason, since I defined these metaconv2d and metalinear by myself in some places and I accidently call these functions as well, so these my own functions mess with yours of the same name.
Hi,
I tried the simple CNN network provided in this repo, however, model.meta_named_parameters() gives only the parameters of MetaBatchNorm2d() and fully ignore MetaConv2d() and MetaLinear(). I checked model.named_parameters() and it is able to output all the parameters of the model.
My torchmeta version is 1.8.0 and pytorch version is 1.8.1.
Could you please give me a hand about this? Thanks