josephsdavid / N2D

Library implementation of https://arxiv.org/abs/1908.05968v5
https://n2d.readthedocs.io/en/latest/
MIT License
23 stars 6 forks source link

Make model saving more convenient/useful #5

Open josephsdavid opened 4 years ago

josephsdavid commented 4 years ago

Currently, I am implementing a two-part saving strategy where the keras part of the model is saved to h5, and the sklearn/umap part is saved as a pickle (or joblib maybe?) file, however I am sure this is not that convenient. If anybody would like to propose a more elegant way to save N2D objects for easy deployment, I am open to suggestions!