HyperAI超神经

首页算力平台文档资讯论文教程数据集百科 SOTA LLM 模型天梯 GPU 天梯顶会

中文

HyperAI超神经

Efficient Vits On Imagenet 1K With Lv Vit S

评估指标

GFLOPs

Top 1 Accuracy

评测结果

各个模型在此基准测试上的表现结果

			Paper Title	Repository
Base (LV-ViT-S)	6.6	83.3	All Tokens Matter: Token Labeling for Training Better Vision Transformers
DynamicViT (90%)	5.8	83.3	DynamicViT: Efficient Vision Transformers with Dynamic Token Sparsification
DynamicViT (80%)	5.1	83.2	DynamicViT: Efficient Vision Transformers with Dynamic Token Sparsification
MCTF ($r=8$)	4.9	83.5	Multi-criteria Token Fusion with One-step-ahead Attention for Efficient Vision Transformers
PS-LV-ViT-S	4.7	82.4	Patch Slimming for Efficient Vision Transformers	-
EViT (70%)	4.7	83.0	Not All Patches are What You Need: Expediting Vision Transformers via Token Reorganizations
BAT	4.7	83.1	Beyond Attentive Tokens: Incorporating Token Importance and Diversity for Efficient Vision Transformers
DynamicViT (70%)	4.6	83.0	DynamicViT: Efficient Vision Transformers with Dynamic Token Sparsification
AS-LV-S (70%)	4.6	83.1	Adaptive Sparse ViT: Towards Learnable Adaptive Token Pruning by Fully Exploiting Self-Attention
PPT	4.6	83.1	PPT: Token Pruning and Pooling for Efficient Vision Transformers
DPS-LV-ViT-S	4.5	82.9	Patch Slimming for Efficient Vision Transformers	-
SPViT	4.3	83.1	SPViT: Enabling Faster Vision Transformers via Soft Token Pruning
MCTF ($r=12$)	4.2	83.4	Multi-criteria Token Fusion with One-step-ahead Attention for Efficient Vision Transformers
EViT (50%)	3.9	82.5	Not All Patches are What You Need: Expediting Vision Transformers via Token Reorganizations
DiffRate	3.9	82.6	DiffRate : Differentiable Compression Rate for Efficient Vision Transformers
AS-LV-S (60%)	3.9	82.6	Adaptive Sparse ViT: Towards Learnable Adaptive Token Pruning by Fully Exploiting Self-Attention
eTPS	3.8	82.5	Joint Token Pruning and Squeezing Towards More Aggressive Compression of Vision Transformers
dTPS	3.8	82.6	Joint Token Pruning and Squeezing Towards More Aggressive Compression of Vision Transformers
MCTF ($r=16$)	3.6	82.3	Multi-criteria Token Fusion with One-step-ahead Attention for Efficient Vision Transformers

0 of 19 row(s) selected.

Efficient Vits On Imagenet 1K With Lv Vit S | SOTA | HyperAI超神经