UDapter is a multilingual dependency parser that uses "contextual" adapters together with language-typology features for language-specific adaptation. This repository includes the code for "UDapter: Language Adaptation for Truly Universal Dependency Parsing"
Would it be possible to host or share the trained model weights from the paper?
This would be very helpful for others to build on your work and test in your settings without having to retrain the 13-lang and monolingual models from scratch. Especially since retraining may also introduce variance in final due to gpu nondeterminism / different training compute.
I am running this scripts/concat_ud_data.sh, after installing the dependencies mentioned in the requirements file in a conda virtual environment. First, I got an error from package six, which was ModuleNotFoundError: No module named 'six.moves'; 'six' is not a package. I resolved this following this link. I also had to install tqdm. Then I got another missing package error, namely import lang2vec.lang2vec as l2v ModuleNotFoundError: No module named 'lang2vec'. I installed it and now it is running!
Add these in the instructions or the requirements file, if possible. Just thought I would send this because other people will probably face the same issues.