mpi4py how to scatter large pandas dataframe

Question

I am trying to scatter a pandas dataframe to spawned children processes.

Here is what I have tried.

master.py

from mpi4py import MPI
import numpy as np
import sys
import pandas as pd

comm = MPI.COMM_SELF.Spawn(sys.executable,
                           args=['child.py'],
                           maxprocs=5)

class Model:
    def __init__(self, num_draw):
        self.num_draw = num_draw

    def test(self):
        id = np.arange(14)
        id_repeat = np.repeat(id, self.num_draw)
        draws = np.arange(self.num_draw*14)
        data = pd.DataFrame({"id": id_repeat, "draws": draws})
        chunks = np.array_split(data,5)
        comm.scatter(chunks, root=MPI.ROOT)
        sum_gather = None
        sum_gather = comm.gather(sum_gather, root=MPI.ROOT)
        print('parent', sum_gather)

model=Model(500)
model.test()

comm.Disconnect()

child.py

from mpi4py import MPI


comm = MPI.Comm.Get_parent()
size = comm.Get_size()
rank = comm.Get_rank()

data = None
data = comm.scatter(data, root=0)

def li(data):
    data_sum = data.sum()
    return data_sum

data_sum=li(data)
print("child rank", rank, "after scatter", data_sum)
comm.gather(data_sum, root=0)
comm.Disconnect()

The program hangs without error message. If I do model=Model(5), it works. So perhaps this is related to the size of dataframe? How can I scatter a large dataframe?