I would suggest using pickle when you prefer a compact file format.
# import packages
import pandas as pd
import numpy as np
import pickle
import os
# create dictionary of dataframes
nrows, ncols, ndataframes = 1_000, 50, 100
my_dict = {k:v for (k,v) in [[f'df_{n}', pd.DataFrame(np.random.rand(nrows, ncols))] for n in range(ndataframes)]}
# save dictionary as pickle file
pickle_out = open('my_dict.pickle', 'wb')
pickle.dump(my_dict, pickle_out)
pickle_out.close()
# create new dictionary from pickle file
pickle_in = open('my_dict.pickle', 'rb')
new_dict = pickle.load(pickle_in)
# print file size
print('File size pickle file is', round(os.path.getsize('my_dict.pickle') / (1024**2), 1), 'MB')
# sample
new_dict['df_10'].iloc[:5, :5]
Result:
File size pickle file is 38.2 MB
0 1 2 3 4
0 0.338838 0.501158 0.406240 0.693233 0.567305
1 0.092142 0.569312 0.952694 0.083705 0.006950
2 0.684314 0.373091 0.550300 0.391419 0.877889
3 0.117929 0.597653 0.726894 0.763094 0.466603
4 0.530755 0.472033 0.553457 0.863435 0.906389