List your company
ResearchFormerly WizWand

Inference Efficiency on HAGRID

4.75#MATSAM-Decoding[E2]
1.32842.21673.1053.9933Nov 16, 2024
Updated 8mo ago

Evaluation results

MethodLinks
SAM-Decoding[E2]Backbone=Vicuna-7B-v1.3, Auxiliary Method=EAGLE-2, Data Type=float16, Batch size=12024.114.7596.62.81
EAGLE-2Backbone=Vicuna-7B-v1.3, Data Type=float16, Batch size=12024.114.1582.612.41
SAM-Decoding[T]Backbone=Vicuna-7B-v1.3, Auxiliary Method=Token Recycling, Data Type=float16, Batch size=12024.113.2387.932.57
Token RecyclingBackbone=Vicuna-7B-v1.3, Data Type=float16, Batch size=12024.112.8866.171.93
SAM-DecodingBackbone=Vicuna-7B-v1.3, Data Type=float16, Batch size=12024.112.4476.722.24
PIABackbone=Vicuna-7B-v1.3, Data Type=float16, Batch size=12024.112.4366.651.95
PLDBackbone=Vicuna-7B-v1.3, Data Type=float16, Batch size=12024.112.0344.111.29
REST*Backbone=Vicuna-7B-v1.3, Data Type=float16, Batch size=12024.111.5324.911.39
Lookahead*Backbone=Vicuna-7B-v1.3, Data Type=float16, Batch size=12024.111.4623.581.32