mamba
mamba copied to clipboard
Issue about the FLOPs of selective scan
Hi author! Thanks for your brilliant work first. I try to calculate the flops of mamba through calculate_flops from calflops library. I am wondering if the efficiency of selective scan can change when the sequence has more 0 or in other word more sparse. Will more sparse sequence may lead Flops smaller?I cannot test if from the calflops. Thank you!