• search
    search
  • reviewers
    reviewers
  • feeds
    feeds
  • assignments
    assignments
  • settings
  • logout

Weight subcloning: direct initialization of transformers using larger pretrained ones.

Mohammad SamraghMehrdad FarajtabarSachin MehtaRaviteja VemulapalliFartash FaghriDevang NaikOncel TuzelMohammad Rastegari
Published in: CoRR (2023)
Keyphrases
  • real time
  • decision making
  • website
  • k means
  • database
  • real world
  • computer vision
  • multimedia
  • objective function
  • weighting scheme
  • small size
  • initial conditions
  • weight function