Antoine Prouvost

Adding this would also be useful for tests - [ ] Automatically detect the stage needed to test the observation in `make_obs`, `unit_tests`.

Yes, because we do not currently support M1 Macs for Ecole. In the meantime, you can launch your development environment with Rosetta.

Basically CMake is telling you it cannot find Python. This should be set properly in `setup.py` by `scikit-build`. I do not recommend using `python setup.py install`, `pip` is more appropriate....

Two ways this could be handled: - 1. Throw a couple of `deepcopy` in the parsing/operators of the data functions - 2. Add a mechanism counting the transitions in the...

So it seems the best approach for us would be to use _Generalized Upper Bound_ (GUB) branching. For _integer_ variables `x_i` provided by the user, and their LP relaxation value...

One (open ?) question that remain is the value for the `estimate` and `priority` parameters of the [`SCIPcreateChild`](https://www.scipopt.org/doc-7.0.2/html/group__PublicBranchingMethods.php#ga72c19678b1199a8d73066e3a5078af91) function.

I removed `BranchingSum` and reopened #299 for the current implementation. @gasse do we have more insight on what was wrong with it?

I cannot not reproduce on MacOS, but I can reproduce on Linux (rossoblu). Running with SCIP in Debug, I could get a failing assertion: ``` scipoptsuite-7.0.1/scip/src/scip/var.c:937 boundchgReleaseData: Assertion `boundchg !=...

I've managed to reproduce it with `pyscipopt` only, taking Ecole out of the equation ```python import pyscipopt model = pyscipopt.Model() model.readProblem("test4.lp") model.setParams({ "branching/scorefunc" : "s", "branching/scorefac" : 0.027616771873704711037778025684019667096436023712158203125, "randomization/randomseedshift" :...

I have not reported. I'll wait for the conda-forge PR for SCIP to be merged before building 7.0.2.