Data set failing using OR-Tools about draft

[ferryjul]

Thank you for your feedback @torressa ! I've been able to reproduce this error. It was due to an edge case where all training data belongs to the same class - and the resulting trees have a single leaf node (and no internal node). I fixed that in our code, the error should no longer appear! (however, note that in such a case the forest brings no valuable information for reconstruction)

Regarding the fact that we include the model-building time in the returned solve_duration_time, it is due to the fact that we aim at measuring the runtime of our entire method - and not the performances of the solvers themselves. Indeed, model creation would be way faster using the C++ APIs rather than the Python ones!
We could have done it the other way to actually quantify the solver's ability to solve the problem, and may modify that later.
In our experiments, we assign limited run-times using the timeout parameter, which applies to the solvers' runtimes only as we cannot bound the models' creation time (and so the comparisons presented in the Appendices of our paper only compare the solvers' runtimes once the model is built).

I'm closing this issue now as the problem is fixed, but don't hesitate to reach out for any additional comment/feedback!

[torressa]

Cool! Thanks for the clarification!
Very cool piece of work BTW!