Experimental Work in progress!
A chain veld repo for publishing self-trained MARA models to huggingface.
These spacy models were trained with spacy 2.2.4 . But only from 3.* onwards, does spacy offer a native huggingface integration which converts model metadata into harmonized huggingface metadata. Spacy 2.* does not offer this and it also does not package into whl formats which are required by huggingface.
Further work must be invested into bridging this gap, potentially by manually building a whl file
with build setuptools wheel and a pyproject.toml file.