/bertopic

Getting started with BERTopic

Primary LanguageHTML

BERTopic

BERTopic is a modeling technique that creates interpretable topics. The cuML package in RAPIDS accelerates the UMAP and DBSCAN steps of the BERTopic algorithm, resulting in significant performance gains:

image

Notebooks

This repos contains notebooks for getting started with BERTopic with RAPIDS. To reproduce, we recommend using the PyTorch container from NGC.

Benchmarks

Contains benchmarks for BERTopic with RAPIDS compared to CPU.

References