Great work, thanks for sharing!

I see https://github.com/google/cld3, but how does this compare with https://github.com/CLD2Owners/cld2 which is used by the large https://commoncrawl.org project to classify billions of samples from the whole internet?