SLIME expects its input data to be calibrated and to have weights that properly reflect the thermal noise. If you have not already done so, apply any a priori amplitude calibrations and fringe-fit the data. Run FIXWT on the data to make sure that the weights are realistic.
Although it is not essential to do so, it is a good idea to average you data in time before running SLIME. This will both reduce the amount of memory that SLIME needs and improve its responsiveness. SLIME will eventually be capable of time-averaging its own data but the current version does not support this. You will, therefore, have to run UVAVG to average your data.
When you have calibrated and averaged your data you are ready to start SLIME.