MATH
·
IN
·
MODELS
The map
Structures
Hypotheses
Classification
Matrix
Methods bubble
Genealogy
Families
Papers
Methods
Why
BEiT (BERT Pre-Training of Image Transformers)
Microsoft
Structures found in this family (1)
Linear Subspace
By model (1)
BEiT-Large (ImageNet-22k pretrained)
Linear Subspace
Observations (1)
One unconstrained linear map injects vision embeddings into frozen GPT-J
Papers
Linearly Mapping from Image to Text Space (2023)