List your company

Topic Classification on AG News BERT-base width (test)

91.52AccuracyT4-linear
91.46891.481591.49591.5085Mar 5, 2026

Evaluation results

MethodLinks
T4-linearModel width (d)=768, p=4, Encoder parameters=7.1 M, Compression ratio=0.25×, Peak memory=1.66 GB2026.0391.52
StdModel width (d)=768, p=4, Encoder parameters=28.4 M, Compression ratio=1.00×, Peak memory=1.95 GB2026.0391.47