Closed thchang closed 1 year ago
Ooo, @Deathn0t , one additional question about this that I forgot to ask in the PR message:
When running examples/bench_dtlz.py
, I immediately get a message that tensorflow is being loaded.
Why is this?
I'm not using tensorflow for anything and my understanding of deephyper is that it shouldn't need tensorflow unless it is specifically doing a tensorflow model's hyperparameter optimization, but I could be mistaken
Hey @thchang , yeah good point and good question. So, when running from deephyper.problem import HpProblem
the deephyper/problem/__init__.py
(link) file is being loaded which if NAS requirements are installed will load deephyper/problem/_neuralarchitecture.py
to have the NaProblem
and tensorflow is loaded there...
It would be better to have a lazy loader, trying to import NaProblem
only when from deephyper.problem import NaProblem
but I did not implement it for now (also I am not sure how the documentation will work if lazy loading is used...);
Modified the test problems in ParMOO's sample simulation library, to fit with DeepHyper and provided an example script importing and solving DTLZ2.
Addresses Issue #5