[MRG] Explicitly set dtype=object when generating permutations (#534)

* Explicitly set dtype=object

* Try uninstall/install instead of upgrade

* Upgrade scipy and numpy

.github/workflows/build_and_deploy.yml

@@ -80,7 +80,7 @@ jobs:
           # Runs tests and checks for a numpy regression by upgrading numpy and running tests again
           CIBW_TEST_COMMAND: >
             pytest --showlocals --durations=20 --pyargs pmdarima &&
-            pip install --upgrade numpy &&
+            pip install --upgrade scipy numpy &&
             pytest --showlocals --durations=20 --pyargs pmdarima
           # Avoid testing on emulated architectures
           CIBW_TEST_SKIP: "*-*linux_{aarch64,ppc64le,s390x} *-macosx_arm64"

pmdarima/arima/_auto_solvers.py

@@ -99,8 +99,10 @@ def __init__(self, y, X, fit_partial, d, D, m, max_order,
         if random:
             random_state = check_random_state(random_state)
 
-            # make a list to scramble...
-            gen = random_state.permutation(list(gen))[:n_fits]
+            # make a list to scramble... `gen` may have a ragged nested
+            # sequence, so we have to explicitly use dtype='object', otherwise
+            # it will raise a ValueError on numpy >= 1.24
+            gen = random_state.permutation(np.array(list(gen), dtype='object'))[:n_fits]  # noqa: E501
 
         self.gen = gen
         self.n_jobs = n_jobs