# Scaling out custom functions

**URL:** <https://discuss.ray.io/t/scaling-out-custom-functions/15930>\
**Category:** Ray Data\
**Created:** [October 2, 2024, 10:06am UTC](https://discuss.ray.io/t/scaling-out-custom-functions/15930 "2024-10-02T10:06:00Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![aeroaks](https://sea2.discourse-cdn.com/flex020/user_avatar/discuss.ray.io/aeroaks/32/1948_2.png) [@aeroaks](https://discuss.ray.io/u/aeroaks)\
**Post date:** [October 2, 2024, 10:06am UTC](https://discuss.ray.io/t/scaling-out-custom-functions/15930/1 "2024-10-02T10:06:01Z")

</div>

I have function(s) that expects input dataframe and config (dict). The function can return multiple dataframes of different sizes depending on the function.  
I was trying to use Ray Data with map\_groups but the return in map\_groups expects dataframes of same size, which will not be the case in my functions.  
Can use the above approach or Do I have to combine Ray Data with Ray Task to scale the computation?