Knowing what resources your jobs actually used is key to requesting the optimal resources for future jobs.
Find the memory used by one job with ID <job ID>
sacct -j <job ID> --format JobID,start,elapsed,state,alloccpus,ReqMem,MaxRSS,AveRSS --units=g
Find out which nodes your jobs ran on
sacct --format=User,JobID,partition,state,time,elapsed,nodelist