multiply by 2 when calculating floating-point operations?

[lxy5513]

from thop import profile 
input = torch.rand((1, 6, 256, 256))
gflops,params = profile(model,inputs=(input,))
**gflops = gflops*2 / 10**9**
params = params / 10**6
print(gflops,'==============')
print(params,'==============')

 Hello, why do we need to multiply by 2 when calculating floating-point operations? @ZZZHANG-jx 

[ZZZHANG-jx]

The profile function in the thop library only considers the number of multiplications. If you also account for addition operations, the final FLOPs are approximately twice the result. For more details, you can refer to this link.

[lxy5513]

@ZZZHANG-jx I get it, thanks for u kind replay