- fixed dependency bug
- decoupled xlinear and other models to relief the library requirements when installing the wheel
- updated setup.py file
- excludes terms that have the attribute "Do not use"
- excludes terms that we manually want to exclude using
descriptors_not_to_use_manual.csv
- #180: Bug fix
Bug fixes:
- Fix download_model to download version specific mesh model
- #152 Fix grants_tagger package which was failing due to missing dependencies
Improvements:
- #76 Implements
grants_tagger downloadwhich downloads data from EPMC that contain mesh tags - #74 Implements
grants_tagger visualizethat uses streamlit to interact with the models - #84 Implements
grants_tagger explainthat produces local and global explanations using SHAP - #78 Extends
grants_tagger downloadto download models from Github releases - #90 Add
MeshXLinearclass, config and results for MeSH - #100 Add time to train and ec2_instance information when training
This is the first release that packages everything developed to this point. Grants tagger comes with a nice CLI, commands are
- preprocess
- train
- evaluate
- predict
- tune
- pretrain
- tag