FastBERT is a BERT variant that uses only 0.3% of its neurons during inference, achieving similar performance to other BERT models. It selectively engages 12 out of 4095 neurons per layer using fast feedforward networks, resulting in significant speedup.

https://arxiv.org/abs//2311.10770

YouTube: https://www.youtube.com/@ArxivPapers

TikTok: https://www.tiktok.com/@arxiv_papers

Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016

Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers