BayesSearchCV: Not possible to provide list of tuples as search space parameter

[CodingKaiser]

I have a list of tuples that I would like to cross-validate over, e.g. [(1,), (2,), (1,2)]. Providing this as the value for a key in the search_space parameter dictionary that I pass to BayesSearchCV causes an error however when I call fit():

Traceback (most recent call last):
  File "/path/to/conda/envlib/python3.7/runpy.py", line 193, in _run_module_as_main
    "__main__", mod_spec)
  File "/path/to/conda/envlib/python3.7/runpy.py", line 85, in _run_code
    exec(code, run_globals)
  File "/path/to/src/main.py", line 256, in <module>
    main()
  File "/path/to/src/main.py", line 229, in main
    opt.fit(X_train, y=y_train)
  File "/path/to/conda/envlib/python3.7/site-packages/skopt/searchcv.py", line 680, in fit
    groups=groups, n_points=n_points_adjusted
  File "/path/to/conda/envlib/python3.7/site-packages/skopt/searchcv.py", line 555, in _step
    params = [[np.array(v).item() for v in p] for p in params]
  File "/path/to/conda/envlib/python3.7/site-packages/skopt/searchcv.py", line 555, in <listcomp>
    params = [[np.array(v).item() for v in p] for p in params]
  File "/path/to/conda/envlib/python3.7/site-packages/skopt/searchcv.py", line 555, in <listcomp>
    params = [[np.array(v).item() for v in p] for p in params]
ValueError: can only convert an array of size 1 to a Python scalar

I am using the following packages:
scikit-learn 0.22.2.post1
scikit-optimize 0.7.4

This is the dictionary I'm passing to BayesSearchCV:

    params = {
        'module__param1': [(1, ), (2, ), (1, 2)],
        'module__param2': [(1, )],
    }

It is interesting to note that this problem does not occur if I use GridSearchCV instead of BayesSearchCV. I've managed to find a work-around by simply wrapping the tuples inside my own class, but it would be nice to get it to work as usual.

[ellengiacometti]

Another interesting thing is that it works with Random Forest but not with ANN

[fjatelnicki]

Hey @CodingKaiser!
Could you show how did you wrap tuples to solve this issue?