Defining search problems

The --problem and --run arguments to DeepHyper search methods must be loadable Python objects, or paths to files containing the objects. To ensure that imports work correctly when DeepHyper loads your code, it is best to place your modules inside a problem directory installed in a virtual environment.

This is easily accomplished with:

bash
$ deephyper start-project demo
$ cd demo/demo/
$ deephyper new-problem hps my_problem

This will create a my_problem package that is globally importable within your virtual environment. Now you can write modules with your Problem instance and run function inside in the my_problem subdirectory. You are free to define and import helper modules with the from my_problem ... syntax.

You can then pass --problem my_problem.problem.Problem to DeepHyper to indicate that the Problem instance is located in the problem.py module of the my_problem package.