ToFe freezes unimportant ViT tokens temporarily and reuses them later, achieving about 50% FLOPs reduction with less than 2% top-1 accuracy drop on ImageNet.
Width & depth pruning for vision transformers,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
ToFe: Lagged Token Freezing and Reusing for Efficient Vision Transformer Inference
ToFe freezes unimportant ViT tokens temporarily and reuses them later, achieving about 50% FLOPs reduction with less than 2% top-1 accuracy drop on ImageNet.