Issue with return types of evaluation functions

[ronald-jaepel]

This issue is to collect my notes and thoughts to look over tomorrow.

ax.NEHVI is broken with

[INFO 03-25 16:44:05] ax.core.experiment: No trials are in a state expecting data.
C:\Users\ronal\mambaforge\envs\cadet-process-debug\Lib\site-packages\CADETProcess\optimization\axAdapater.py:395: UserWarning: Initial population smaller than popsize. Creating missing entries.
  warnings.warn(
Traceback (most recent call last):
  File "C:\Users\ronal\mambaforge\envs\cadet-process-debug\Lib\site-packages\IPython\core\interactiveshell.py", line 3577, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "<ipython-input-2-de9f0bd67ea9>", line 1, in <module>
    runfile('C:\\Users\\ronal\\PycharmProjects\\pamenter\\code\\MSSMA_3comp_3grad_UCL.py', wdir='C:\\Users\\ronal\\PycharmProjects\\pamenter\\code')
  File "C:\Program Files\JetBrains\PyCharm Community Edition 2023.2.3\plugins\python-ce\helpers\pydev\_pydev_bundle\pydev_umd.py", line 197, in runfile
    pydev_imports.execfile(filename, global_vars, local_vars)  # execute the script
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Program Files\JetBrains\PyCharm Community Edition 2023.2.3\plugins\python-ce\helpers\pydev\_pydev_imps\_pydev_execfile.py", line 18, in execfile
    exec(compile(contents+"\n", file, 'exec'), glob, loc)
  File "C:\Users\ronal\PycharmProjects\pamenter\code\MSSMA_3comp_3grad_UCL.py", line 293, in <module>
    optimization_results = optimizer.optimize(
                           ^^^^^^^^^^^^^^^^^^^
  File "C:\Users\ronal\mambaforge\envs\cadet-process-debug\Lib\site-packages\CADETProcess\optimization\optimizer.py", line 254, in optimize
    self.run(self.optimization_problem, x0, *args, **kwargs)
  File "C:\Users\ronal\mambaforge\envs\cadet-process-debug\Lib\site-packages\CADETProcess\optimization\axAdapater.py", line 418, in run
    self._create_manual_trial(x0_init_transformed)
  File "C:\Users\ronal\mambaforge\envs\cadet-process-debug\Lib\site-packages\CADETProcess\optimization\axAdapater.py", line 274, in _create_manual_trial
    trial.run()
  File "C:\Users\ronal\mambaforge\envs\cadet-process-debug\Lib\site-packages\ax\core\base_trial.py", line 408, in run
    self._run_metadata = not_none(self._runner).run(self)
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\ronal\mambaforge\envs\cadet-process-debug\Lib\site-packages\CADETProcess\optimization\axAdapater.py", line 129, in run
    trial_metadata = self.get_metadata(
                     ^^^^^^^^^^^^^^^^^^
  File "C:\Users\ronal\mambaforge\envs\cadet-process-debug\Lib\site-packages\CADETProcess\optimization\axAdapater.py", line 143, in get_metadata
    for metric, f_metric in zip(objective_labels, F.T)
                                                  ^^^
AttributeError: 'list' object has no attribute 'T'

Git bisect shows commit f765f3f to be the first commit with that issue.

I think, axAdapter.get_metadata() expects F (the objectives) to be a numpy array and have a .T transformed attribute.
The new transform_maximization returns a list

s = np.multiply(factors, s).tolist()

return s

@schmoelder let's discuss tomorrow.

[ronald-jaepel]

Plan is to always return np.ndarrays as F.