Knowing the number of characters in your dataset helps us determine the most efficient way to run raxml. The number of patterns is the number of unique columns in the multiple sequence alignment matrix. You can get this number from the output of the intermediate results once a job begins. Entering the number of characters per taxon in your matirx, or 1000 as the number of patterns is an ok start. Look at the intermediate results, and see if that is reasonably close. If it is not, kill the job, and adjust the number.