BESS-KGE API Reference

besskge.pipeline	High-level APIs for training/inference with BESS.
besskge.dataset	Utilities for building and storing knowledge graph datasets as collections of (h,r,t) triples.
besskge.sharding	Sharding of embedding tables and triple sets for distributed execution.
besskge.batch_sampler	Classes for sampling batches of positive and negative triples for each processing device, according to the BESS distribution scheme.
besskge.negative_sampler	Classes for sampling entities to use as corrupted heads/tails when constructing negative samples.
besskge.bess	PyTorch modules implementing the BESS distribution scheme [CJM+22] for KGE training and inference on multiple IPUs.
besskge.scoring	Functions for scoring positive and negative triples, specific to each KGE model.
besskge.loss	Functions for computing the batch loss based on the scores of positive and negative samples.
besskge.metric	Utilities for computing metrics to evaluate the predictions of KGE models.
besskge.embedding	Utilities for initializing and managing entity/relation embedding tables.
besskge.utils	General purpose utilities.