Question about matrix rank of floating points

[Wsine]

Hi, guys

I read the paper and code and notice that there is no additional description about quantized model, so I assume that the experiment is based on floating point model.

How did you compute the matrix rank of floating point filters? Will it always get the full rank or just those values lower than tol/eps were pruned? In most of the case in my test, they are full rank. Am I right?

Just for discussion.

Thanks.