Skip to content

Comment on Faster and smaller Hugging Face BERT on CPU via compound sparsification

Comments

combining multiple sparsification methods to improve the Hugging Face BERT base model (uncased) performance on CPUs. We combine distillation with both unstructured pruning and structured layer dropping. This “compound sparsification” technique enables up to 14x faster and 4.1x smaller BERT on CPUs depending on accuracy constraints.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.