OverflowError: (34, 'Result too large')

[norheim]

The code to reproduce the problem is a little long, so posting a link to it:
https://github.com/norheim/smallsat/blob/master/notebooks/minimalspace.ipynb

I'm getting a float overflow problem from a Signomial localsolve():

C:\Users\johan\Miniconda2\envs\gpsat\lib\site-packages\gpkit\nomials\math.pyc in mono_approximation(self, x0)
    123             e = val*diff/c0
    124             exp[vk] = e
--> 125             c /= val**e
    126         hmap = NomialMap({exp: c})
    127         hmap.units = self.units

OverflowError: (34, 'Result too large')

[bqpd]

Solves for me with all solvers, if you remove the middle GP solve.

The error you're seeing is because variables with specified values now give those values up to the first model they're a part of; I'd recommend creating these variables inside a Model object if you want to keep the code-block organization you have going here, so that you create new copies for the new model.

[norheim]

Got it thanks a lot Ned!

[bqpd]

Sure thing, sorry for the confusion!

…

On Mon, Apr 23, 2018 at 12:05 PM, norheim ***@***.***> wrote: Got it thanks a lot Ned! — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#1301 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABagGAPnMkGePqB7Q4mMNiiL0CyHaIIXks5triXggaJpZM4TfHv0> .

[norheim]

Ned, wanted to reopen the issue, as the Error code is the same, its just that the error comes from the formulation. I introduced a new variable(called it dummy to make it clear which one it is, but it's not a dummy variable). Actually the original constraint P comes from a linearization of this relationship, but I wanted to see if I could include the raw relationship(constraint with dummy and P2). During the signomial solve it definitly seems to hit the float overflow problem this time, even when I do Bounded(countraints). Let me know if you understand what's going on(from a modelling side, or a solver side)!

Just FYI, SingleSignomialEquality(dummy, R_earth+h), which is the actual constraint, still results in the same problem.

https://github.com/norheim/smallsat/blob/master/notebooks/minimalspace.ipynb

[bqpd]

The problem was that earth radius was in meters and gigantic, and the SP heuristic (since it assumes everything is close to 1) overflowed trying to raise it to a large power. Put the earth radius in kilometers and it works, and I'll add an error message to make this clearer next time!

[norheim]

Man, good insight!! :) and thank you so much for being so quick, its all working on my end too!

[bqpd]

Pull the new master and tell me what you think of the error message!

[norheim]

Nice, makes it very clear how to fix it if the problem ever shows up. Thanks again for the quick trigger, keeps the motivation up :) 2018-04-23 19:58 GMT-04:00 Ned Burnell <[email protected]>:

…

Pull the new master and tell me what you think of the error message! — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1301 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADTJymLc2jnpiBDglybdLhSmtwe_WmsUks5trmqtgaJpZM4TfHv0> .

-- Johannes J. Norheim MS Candidate '18 MIT Dept of Aeronautics and Astronautics +1 (857) 253-9901