MACs or FLOPS in ViG Paper

huawei-noah / Efficient-AI-Backbones

Efficient AI Backbones including GhostNet, TNT and MLP, developed by Huawei Noah's Ark Lab.

4.07k stars 708 forks source link

MACs or FLOPS in ViG Paper #141

Closed asif-hanif closed 2 years ago

asif-hanif commented 2 years ago

Hi Could you please clarify the following ambiguity.

Line 360 of train.py computes MACs. However, this number is reported as FLOPs in the paper e.g. Table-4 and Table-5. As I understand that 1 FLOPs = 2 x MACs. If this is the case, all numbers of 'FLOPs (B)' column in the tables, corresponding to ViG, should get multiplied by 2.

iamhankai commented 2 years ago

In our paper, FLOPs==MACs as in EfficientNet and Swin Transformer