attr_standardizer shouldnt load from huggingface for each call

[nleroy917]

I was following the READNE's instructions:

from attribute_standardizer.attribute_standardizer import attr_standardizer

attr_standardizer(pep=/path/to/pep, schema="ENCODE")

It looks like this function is calling load_from_huggingface for each PEP you want to standardize. While huggingface does cache the model files on disk.. it's very unnecessary to load the model from disk for each PEP you want to standardize.

I would really recommend creating some class that holds the model in memory and can be called many times over for many PEPs:

from bedmess import AttrStandardizer

model = AttrStandardizer("databio/encode-bm")

model.standardize(path_to_pep)

[nleroy917]

This will rear its head for the first user to use the standardizer on a public server since it downloads the model not when the server starts up but when the user requests the standardization pipeline

[saanikat]

Solved with new PR #15