Question re: 'gsample' procedure.

I am currently analyzing the Healthcare Utilization Project (HCUP) National Inpatient Sample (NIS) 2016 dataset which is a stratified systematic sample of discharges--estimated 7 million observations with patient ICD-10 diagnostic and procedure codes. The NIS data documentation recommend utilizing the 'svyset' command to account for the complex survey design and for weights

svyset [pweight= discwt], strata(nis_stratum) psu(hosp_nis)

I'd like to work with a smaller sample of the data given limited computer resources to run such a large 10 gigabyte dataset, but removing observations would significantly alter the nationally estimated SE's because it is weighted data. This is where I'd like to consider 'gsample'--but as per my understanding of the documentation the 'gsample' procedure is utilized for frequency weights, and NOT probability weights, which is utilized in my data. However, I've been informed that I should still be able to use the 'gsample' procedure--I'm still not convinced. I wanted to know if anyone who has utilized a similar dataset or the 'gsample' procedure in a similar case and what suggestions they might have. I thought I had found a similar post here, but I'm not sure if this is answerign my exact question regarding frequency vs. probability weights for the gsample procedure.

Cheers!