Robert Parker

> I did a quick check of setting the default tolerance for near-parallel components to 1e-6 which brought the number of failures down to around 14. Tightening the tolerance to...

In this framework, is there a difference between extensions in `supported` and "top-level" extensions such as DAE, GDP, and Network? Or are you proposing to combine these somehow?

Test failures are in a GDP baseline file comparison, likely due to changes in variable order generated by `get_vars_from_components`.

I had to update some tests to use `ComponentSet` to get `test_visitor.py` to work, so this doesn't give the same variable order that `identify_variables` used to. For that reason, I...

With this patch, the motivating example from https://github.com/Pyomo/pyomo/pull/3186 looks pretty good: ```console Identifier ncalls cumtime percall % -------------------------------------------------------------------------------- root 1 2.348 2.348 100.0 --------------------------------------------------------------------------- full model post-solve 1 0.460 0.460...

I believe this is because Ipopt's internal data structures and AMPL interface use different multiplier conventions. I believe the rationale was for the AMPL interface to be consistent with CPLEX's...

Assuming #2830 is localized to MindtPy, it will not fix this issue.

@ZedongPeng Here is the `load_state_into_pyomo` function of `PyomoNLPwithGreyBoxBlocks`, which is called by our CyIpopt interface: ```python def load_state_into_pyomo(self, bound_multipliers=None): # load the values of the primals into the pyomo primals...

Would it make sense to migrate just the Hungarian algorithm from GraphsMatching into Graphs.jl? I would like to use this implementation, but am hesitant to depend on GraphsMatching due to...

I would be willing to give this a try. However, I just realized that GraphsMatching's Hungarian algorithm depends on [Hungarian.jl](https://github.com/Gnimuc/Hungarian.jl). I am assuming that this is the kind of one-off...