Fix typo in unsqueeze

adding the ifd conversion as a method on the model

adding some debug logging for timings

changing interface to accept words instead of a sentence string. Documenting the truncation flag

more vectorization

more vectorization of logits to labels

add vectorized and more stable way to aggregate subwords to words

improve comments

fix bad readme and remove warnings about uninitialized tensors

add support for IFD tags and some refactoring

Fix batched word masking

Fix inconsistencies with the old model - now works equally

Add readme with directions on how to run the model

Delete __init__.py

Delete __pycache__

Upload folder using huggingface_hub

initial commit