PR #13 Improvements and fixes for the LUKE integration

add model_size argument to the conversion script

ikuyamada committed 5 years ago

add support binary cross entropy loss in all head models

ikuyamada committed 5 years ago

rename [ENT] to <ent>

ikuyamada committed 5 years ago

improve LukeTokenizer (work in progress)

ikuyamada committed 5 years ago

bug fixes

ikuyamada committed 5 years ago

update the conversion script to use the tokenizer's encode function

ikuyamada committed 5 years ago

remove tokenization_luke_fast.py

ikuyamada committed 5 years ago

add docstring

ikuyamada committed 5 years ago

fix issues in tokenizer

ikuyamada committed 5 years ago

fix model names and urls

ikuyamada committed 5 years ago

transformers Improvements and fixes for the LUKE integration #13 Merged