Can you do the mean deviation for a grouped data?

Yes, but it rather pointless. The mean deviation for any data set will, by definition, be 0. Grouping may make it slightly different from 0 but this statistic has little, if any, useful properties.