List your company

Block-wise quantization complexity analysis on OPT Models

0.24GFLOPSaespa
-1.39049.614820.6231.6252Feb 14, 2024

Evaluation results

MethodLinks
aespaModel size=125M2024.020.24
aespaModel size=350M2024.020.42
aespaModel size=1.3B2024.021.6
aespaModel size=2.7B2024.023.2
Conventional block-wise quantizationModel size=125M, Batch size (B)=42024.026.7
Conventional block-wise quantizationModel size=350M, Batch size (B)=42024.027.5
Conventional block-wise quantizationModel size=1.3B, Batch size (B)=42024.0211
aespaModel size=6.7B2024.0213
Conventional block-wise quantizationModel size=2.7B, Batch size (B)=42024.0215
aespaModel size=13B2024.0220
Conventional block-wise quantizationModel size=6.7B, Batch size (B)=42024.0234
Conventional block-wise quantizationModel size=13B, Batch size (B)=42024.0241