help with timedelta error, trying to sum

[alkaline3]

There is quite a bit of info around summing up time formats in Python, but I'm not able to solve my use case.

I would like to sum up times based on certain criteria in my data frame. My data looks like this, where time is timedelta type:

User: 1 Flag: 0 Time: 04:00:03
User: 1 Flag: 0 Time: 00:25:00

User: 2 Flag: 1 Time: 04:20:00
User: 2 Flag: 1 Time: 01:26:00

User: 3 Flag: 2 Time: 1:00:01
User: 3 Flag: 2 Time: 14:00:02

The criteria is: sum Time when flag is 0, otherwise it just takes the max value.

I want it to look like this:

User: 1 Flag: 0 Time: 04:25:03
User: 1 Flag: 0 Time: 04:25:03

User: 2 Flag: 1 Time: 04:20:00
User: 2 Flag: 1 Time: 04:20:00

User: 3 Flag: 2 Time: 14:00:02
User: 3 Flag: 2 Time: 14:00:02

However, I get an error when I run this code:

groups = sample.groupby('user')['time']
flag = sample.groupby('user')['flag'].transform('max')
sample['time_new'] = np.select([flag.eq(0), flag.isin([1,2])], [groups.transform('sum'), groups.transform('max')]) 

TypeError: Cannot cast scalar from dtype('<m8[ns]') to dtype('<m8') according to the rule 'same_kind'

from using this code

sample.loc[:,'time'] = pd.to_timedelta(sample['time'])

Is there another way to sum time? What am I doing wrong? Thank you very much for any help.