Ask HN: Estimating AI/ML Compute Resources

Putting aside the knowledge, code, confirms, hardware,datasets, etc. to do a full build, are there any common ways to estimate the compute resources required to do ML learning for a given ML method over a data set of X size?