lumen-org / modelbase

A SQL-like interface for python and the web to query probabilistic machine learning models and its data.
GNU Lesser General Public License v3.0
4 stars 1 forks source link

Implementation of shared variables #71

Open jong42 opened 5 years ago

jong42 commented 5 years ago

Currently, when a model is copied, each instance has access to the same shared variables object. That means, when the shared variables for one model are changed, the shared variables of each derived model are also changed. This behaviour is impractical and should be changed so that each copied model has its own shared variables. I already tried to fix that a number of times and I did not find a way to do it once the model was specified. A solution could be to create a full new model when copying, so that the new model has other shared variables than the original one. Maybe do it by giving a function that creates the model?