Custom ML Models accepted in ESG/EPBCS applications

Dear Experts,

I'm trying to create a custom ML model using Python and/or R for a ESG demo application. I have no problem implementing models such as "linear regression", but when I try to import PMML models that use Time Series (ARIMA or Exponential Smoothing) or Neural Networks these are not accepted as an error message appears indicating that "the library used is not supported".
Are there any restrictions on algorithms/models that can be imported in an ESG/EPBCS application? Is this information reported in any document?

Verifying the PMML standard ( ) both Time Series and Neural Networks are well accepted. Is there any restriction in application?