Factoranalyzer calculates different scores using Python 3.5 versus Python 3.7

I did a factor analysis using Factor_Analyzer in Python 3.5 and saved the picke file.

I built a predict function to obtain the factor scores with new customerdata on a monthly basis:

This worked in Python 3.5, but since we upgraded to Python 3.7 this doesn't work anymore: 1/ either he refuses to calculate the factor scores giving an error that ha can't compute the inverse matrix of the correlations: 'raise LinAlgError("Singular matrix") 2/ or he gives me different scores for the same dataset compared to Python 3.5:

mean scores of factor scores in Python 3.7:

0   -2.498980e+08
1   -2.556053e+09
2    2.969148e+08
3    2.569302e+07

mean scores of factor scores in Python 3.5:

0    0.011848
1    0.005893
2    0.016926
3    0.003448

What is the explanation for this difference in Python 3.5 and 3.7 and how can we solve it in 3.7 ?

Thanks for you help !