Efficient Vits On Imagenet 1K With Deit T

Metrics

GFLOPs

Top 1 Accuracy

Results

Performance results of various models on this benchmark

Model Name	GFLOPs	Top 1 Accuracy	Paper Title	Repository
SPViT (1.0G)	1.0	72.2	SPViT: Enabling Faster Vision Transformers via Soft Token Pruning
LTMP (60%)	0.8	71.5	Learned Thresholds Token Merging and Pruning for Vision Transformers
MCTF ($r=20$)	0.6	71.4	Multi-criteria Token Fusion with One-step-ahead Attention for Efficient Vision Transformers
PS-ViT	0.7	72.0	Patch Slimming for Efficient Vision Transformers	-
LTMP (45%)	0.7	69.8	Learned Thresholds Token Merging and Pruning for Vision Transformers
ToMe ($r=16$)	0.6	70.7	Token Merging: Your ViT But Faster
DPS-ViT	0.6	72.1	Patch Slimming for Efficient Vision Transformers	-
MCTF ($r=8$)	1.0	72.9	Multi-criteria Token Fusion with One-step-ahead Attention for Efficient Vision Transformers
BAT	0.8	72.3	Beyond Attentive Tokens: Incorporating Token Importance and Diversity for Efficient Vision Transformers
ToMe ($r=12$)	0.8	71.4	Token Merging: Your ViT But Faster
SPViT (0.9G)	0.9	72.1	SPViT: Enabling Faster Vision Transformers via Soft Token Pruning
S$^2$ViTE	0.9	70.1	Chasing Sparsity in Vision Transformers: An End-to-End Exploration
EvoViT	0.8	72.0	Evo-ViT: Slow-Fast Token Evolution for Dynamic Vision Transformer
HVT-Ti-1	0.6	69.6	Scalable Vision Transformers with Hierarchical Pooling
SPViT	1.0	70.7	Pruning Self-attentions into Convolutional Layers in Single Path
Base (DeiT-T)	1.2	72.2	Training data-efficient image transformers & distillation through attention	-
eTPS	0.8	72.3	Joint Token Pruning and Squeezing Towards More Aggressive Compression of Vision Transformers
MCTF ($r=16$)	0.7	72.7	Multi-criteria Token Fusion with One-step-ahead Attention for Efficient Vision Transformers
LTMP (80%)	1.0	72.0	Learned Thresholds Token Merging and Pruning for Vision Transformers
PPT	0.8	72.1	PPT: Token Pruning and Pooling for Efficient Vision Transformers

0 of 22 row(s) selected.