FLOPs 和 Multi-adds是否搞反了?
zl11250422 opened this issue · comments
For example, in this article, IMDN shows FLOPs as 41, while in BSRN’s paper, IMDN’s Multi-adds are shown as 41. Similar conclusions appear not only in one paper, suggesting that many people confuse these two parameters. Additionally, the torchstat library also confuses these two parameters.