Tag: linear attention

A knowledge-distilled pyramid transformer with reduced attention achieved F1-scores of 90.92%, 97.76% and 94.59% on three 32 × 32-pixel image-classification ...