[MODEL] TODOs for Deep Biaffine Attention Dependency Parser

hankcs commented 5 years ago

This issue is used to track further work for #408.

[ ] Vocabulary should be switched to gluonnlp.Vocab once #393 is addressed.
[ ] The rest of the data pipeline should be switched to GluonNLP classes. This includes:
- ConllWord/ConllSentence -> Transform functions. We also have conll datasets, the code of which can be extracted for this use case.
- DataLoader -> Gluon's DataLoader with Bucketing functions. See http://gluon-nlp.mxnet.io/master/api/notes/data_api.html#bucketing-and-dataloader
- Hybridize
[ ] Refine CoNLL's eval script (though preferably a python/cython version)
[ ] Maybe move ExponentialScheduler/KMeans/Tarjan/SaveableConfig to API
[x] Remove reshape
[ ] Remove unnecessary third party code

eric-haibin-lin commented 5 years ago

Any update on this?

hankcs commented 5 years ago

Sorry, I've been busy with a paper till yesterday. I'll get on to this now.

dmlc / gluon-nlp