AutoML reload gives me typeError: '<' not supported between instances of 'float' and 'str'

[bardamchop]

My process crashes after 9 hours or so and rerunning
automl = AutoML(results_path=f"./model/{final_assets}_{shift_l}", mode="Optuna",
total_time_limit=48* 3600,
optuna_time_budget=3600)

I am getting the following error
TypeError: '<' not
supported between instances of 'float' and 'str'

File /anaconda/envs/azureml_py310_sdkv2/lib/python3.10/site-packages/supervised/base_automl.py:485, in BaseAutoML._perform_model_stacking(self)
482 if self._fairness_metric is not None:
483 # get only fair models if we train with sensitive features
484 ldb = ldb[ldb["is_fair"]]
--> 485 ldb = ldb.sort_values(by="metric_value", ascending=True)
486 models_map = {m.get_name(): m for m in self._models if not m._is_stacked}
487 self._stacked_models = []

File /anaconda/envs/azureml_py310_sdkv2/lib/python3.10/site-packages/pandas/core/frame.py:6955, in DataFrame.sort_values(self, by, axis, ascending, inplace, kind, na_position, ignore_index, key)
6952 if isinstance(ascending, (tuple, list)):
6953 ascending = ascending[0]
-> 6955 indexer = nargsort(
6956 k, kind=kind, ascending=ascending, na_position=na_position, key=key
6957 )
6958 else:
6959 if inplace:

File /anaconda/envs/azureml_py310_sdkv2/lib/python3.10/site-packages/pandas/core/sorting.py:483, in nargsort(items, kind, ascending, na_position, key, mask)
481 non_nans = non_nans[::-1]
482 non_nan_idx = non_nan_idx[::-1]
--> 483 indexer = non_nan_idx[non_nans.argsort(kind=kind)]
484 if not ascending:
485 indexer = indexer[::-1]

I had Pandas 2.1.3 but I also tried with 2.1.0 with the same error
!{sys.executable} -m pip install --upgrade xgboost==2.0.1
!{sys.executable} -m pip install pandas==2.1.0
!{sys.executable} -m pip install --upgrade mljar-supervised

Love this package so far by the way!

[pplonski]

Thank you @bardamchop for reporting the issue. Looks like a bug.