Accenture / AmpliGraph

Python library for Representation Learning on Knowledge Graphs https://docs.ampligraph.org
Apache License 2.0
2.14k stars 251 forks source link

Script to clean up validation and test set - to remove unseen entities/relations during train #62

Closed sumitpai closed 5 years ago

sumitpai commented 5 years ago

Background and Context If the test/validation sets have entities that were not present during training, the ampligraph crashes with an error.

Description

  1. We need to have a graceful exit(raise an exception with meaningful message)

  2. We need to have a script to help the user clean up test/valid set of entities/relations unseen during train. (Move clean_data from predictive_performance.py to a meaningful location and update documentation. )

  3. Remove unseen entities while loading fb15k-237 and wnrr18

sumitpai commented 5 years ago

@chanlevan Can you look at points 2(documentation) and 3