Using jsonpickle instead of pickle has benefits.

pickle-jsonPickle.py

# jsonpickle is human readable for inspection (security, etc.) before unpickling.

# It is smaller, when native.
#
#  246099 cd33.jpkl
# 1432509 cd33.pkl
#
#    241K cd33.jpkl
#    1.4M cd33.pkl


# Though may be larger when compressed.
#
# 183K cd33.jpkl.xz
#  99K cd33.pkl.xz

import pandas

import jsonpickle
import jsonpickle.ext.pandas as jsonpickle_pandas

jsonpickle_pandas.register_handlers()

pfile = 'cd33.pkl'

# On Python 3, data pickled with Python 2.7 may not be unpickled as usual. However, it may unpickled with pandas itself.
# https://stackoverflow.com/a/46389140/420947
#
#   with open(pfile, 'rb') as file:
#     d = pickle.load(file)
# 
#   ==> ModuleNotFoundError: No module named 'pandas.indexes'
#
d = pandas.read_pickle(pfile)

json = jsonpickle.encode(d)

with open('cd33.jpkl', 'w') as file:
  file.write(json)