# Fisher's exact test, contingency tables

Are there any other methods for an $m\times n$ contingency table with $m$ or $n$ greater than 2 for use with small samples ($np<5$) other than Fisher's exact test?