Summing within bouts in pandas dataframe

I am working with data generated by an accelerometer, and this particular model writes a new line of data every time there is a posture change (e.g., going from sitting to standing) or each time there is an individual step, and each row then has a duration. A few example rows (~30k rows per file):

Event type describes whether it's laying, sitting, stepping, etc, and stepping is coded as a 2.

For each day in the dataset I'd like to sum up the duration column for each set of continuous rows with an event type of 2 so that I have the duration of each stepping bouts within each day (I'll later do some groupby math on those bouts to get things like average bout duration). I'm having a hard time figuring out what to search for here, so I appreciate any help!

I'm having a hard time figuring out what to search for here, so I appreciate any help! Obviously just grouping by Event type won't work as I need to group only within specific uninterrupted stepping bouts.

You can create a custom grouper:

# identify events of type 2
m = df['Event Type'].eq(2)

# select only the type 2 rows
# group by ID and successive rows
# aggregate as sum
out = (df[m]
       .groupby(['ID', (~m).groupby(df['ID']).cumsum()],
                as_index=False)
       ['Duration'].sum()
      )

Output (using an alternative input):

     ID  Duration
0  1885      18.1
1  1885       6.0

Used input:

      ID        Date  Event Type  Duration
0   1885  2021-06-29         4.0   31208.5
1   1885  2021-07-07         2.0       1.1 #
2   1885  2021-07-07         2.0       3.2 # set 1
3   1885  2021-07-07         2.0       3.5 #
4   1885  2021-07-07         2.0       4.3 #
5   1885  2021-07-07         2.0       3.7 #
6   1885  2021-07-07         2.0       1.2 #
7   1885  2021-07-07         2.0       1.1 #
8   1885  2021-07-07         1.0       1.0
9   1885  2021-07-07         2.0       3.7 #
10  1885  2021-07-07         2.0       1.2 # set 2
11  1885  2021-07-07         2.0       1.1 #
12  1885  2021-07-07         1.0       1.0