sisrformer / GRFormer

the code of GRFormer: Grouped Residual Self-Attention for Lightweight Single Image Super-Resolution
9 stars 0 forks source link

Macs #2

Open HilbertLucus opened 1 month ago

HilbertLucus commented 1 month ago

Your calculation is wrong, you probably used thop library to calculate, this library calculation is inaccurate, it is recommended to use fvcore library to calculate, the calculation of ×4 model is about 75G, please refer to the