We have several resources that may be helpful to you. In particular, this blog post from 2017 motivates the use of sampling weights and links to a variety of IPUMS project-specific resources on applying sampling weights. In IPUMS USA data, the sampling weight represents how many individuals in the full population a specific individual (or household) in the sample represents. Specifically for your case, it sounds like you should be using the person-level weights. In IPUMS USA this is the PERWT variable. The key reason for this is that the statistic you are aiming to calculate is a person-level characteristic (e.g., marriages). If, on the other hand, you intend to characterize households by the marital status of the reference person (or something similar), then you would want to use the household-level weights (e.g., HHWT).