never used it, but my guess is:

absolute tolerance has to do with the absolute error: if the actual value is 0.001, then absolute tolerance says how close you have to be to 0.001

relative tolerance has to do with a ratio of the error to the actual value: if the total error is 0.0001, this is very close if the actual value is 1, but inaccurate if the value is 0.00001