Is that possible to convert a caffe layer to pytorch layer?

the newest version of pytorch already have c++ interfaces. So is that possible to convert a caffe layer to pytorch layer?

Some layers might be portable, but as far as I know, Caffe2 layers make some assumptions about the input, e.g. contiguity, which might make these layers unportable. Take this information with a grain of salt as I’ve just scratched Caffe2 on the surface so far.

1 Like