Fetching the paper…

HET: Scaling out Huge Embedding Model Training via Cache-enabled Distributed Framework · Around