Good morning, recently I have investigated how to calculate the space needed to store a model in memory, I thought that that is enough, however now I realize that it is not only necessary to have space available in memory, it is also necessary to know the power of compute from my cpu or gpu to run the model efficiently. There must be some way to know if an 8 million parameter model needs such a compute level requirement for a cpu or gpu, be it the cores, clock frequency, instruction per cycle, etc. Is there a formula to estimate the computing power needed to run models just by knowing the number of parameters?