[ERROR] ActiveHyperparameterNotSetError

[jwpark0921]

OS: Ubuntu 24.04(x86_64)
Python version : 3.8.19
Install method : pip install autoPyTorch[forecasting]

Code (In a nutshell example)

from autoPyTorch.api.tabular_classification import TabularClassificationTask

data and metric imports

import sklearn.model_selection
import sklearn.datasets
import sklearn.metrics
X, y = sklearn.datasets.load_digits(return_X_y=True)
X_train, X_test, y_train, y_test =
sklearn.model_selection.train_test_split(X, y, random_state=1)

initialise Auto-PyTorch api

api = TabularClassificationTask()

Search for an ensemble of machine learning algorithms

api.search(
X_train=X_train,
y_train=y_train,
X_test=X_test,
y_test=y_test,
optimize_metric='accuracy',
total_walltime_limit=300,
func_eval_time_limit_secs=50
)

Calculate test accuracy

y_pred = api.predict(X_test)
score = api.score(y_pred, y_test)
print("Accuracy score", score)

Error message

ActiveHyperparameterNotSetError Traceback (most recent call last)
Cell In[5], line 1
----> 1 api.search(
2 X_train=X_train,
3 y_train=y_train,
4 X_test=X_test,
5 y_test=y_test,
6 dataset_name='digits',
7 optimize_metric='accuracy',
8 total_walltime_limit=300,
9 func_eval_time_limit_secs=50
10 )

File ~/miniconda3/envs/auto-pytorch/lib/python3.8/site-packages/autoPyTorch/api/tabular_classification.py:450, in TabularClassificationTask.search(self, optimize_metric, X_train, y_train, X_test, y_test, dataset_name, feat_types, budget_type, min_budget, max_budget, total_walltime_limit, func_eval_time_limit_secs, enable_traditional_pipeline, memory_limit, smac_scenario_args, get_smac_object_callback, all_supported_metrics, precision, disable_file_output, load_models, portfolio_selection, dataset_compression)
437 self._dataset_compression = get_dataset_compression_mapping(memory_limit, dataset_compression)
439 self.dataset, self.input_validator = self._get_dataset_input_validator(
440 X_train=X_train,
441 y_train=y_train,
(...)
447 dataset_compression=self._dataset_compression,
448 feat_types=feat_types)
--> 450 return self._search(
451 dataset=self.dataset,
452 optimize_metric=optimize_metric,
453 budget_type=budget_type,
454 min_budget=min_budget,
455 max_budget=max_budget,
456 total_walltime_limit=total_walltime_limit,
457 func_eval_time_limit_secs=func_eval_time_limit_secs,
458 enable_traditional_pipeline=enable_traditional_pipeline,
459 memory_limit=memory_limit,
460 smac_scenario_args=smac_scenario_args,
461 get_smac_object_callback=get_smac_object_callback,
462 all_supported_metrics=all_supported_metrics,
463 precision=precision,
464 disable_file_output=disable_file_output,
465 load_models=load_models,
466 portfolio_selection=portfolio_selection,
467 )

File ~/miniconda3/envs/auto-pytorch/lib/python3.8/site-packages/autoPyTorch/api/base_task.py:1011, in BaseTask._search(self, optimize_metric, dataset, budget_type, min_budget, max_budget, total_walltime_limit, func_eval_time_limit_secs, enable_traditional_pipeline, memory_limit, smac_scenario_args, get_smac_object_callback, tae_func, all_supported_metrics, precision, disable_file_output, load_models, portfolio_selection, dask_client, **kwargs)
1009 # Initialise information needed for the experiment
1010 experiment_task_name: str = 'runSearch'
-> 1011 dataset_requirements = get_dataset_requirements(
1012 info=dataset.get_required_dataset_info(),
1013 include=self.include_components,
1014 exclude=self.exclude_components,
1015 search_space_updates=self.search_space_updates)
1016 self._dataset_requirements = dataset_requirements
1017 dataset_properties = dataset.get_dataset_properties(dataset_requirements)

File ~/miniconda3/envs/auto-pytorch/lib/python3.8/site-packages/autoPyTorch/utils/pipeline.py:75, in get_dataset_requirements(info, include, exclude, search_space_updates)
69 return _get_regression_dataset_requirements(info,
70 include if include is not None else {},
71 exclude if exclude is not None else {},
72 search_space_updates=search_space_updates
73 )
74 elif task_type in CLASSIFICATION_TASKS:
---> 75 return _get_classification_dataset_requirements(info,
76 include if include is not None else {},
77 exclude if exclude is not None else {},
78 search_space_updates=search_space_updates
79 )
80 else:
81 if not forecasting_dependencies_installed:

File ~/miniconda3/envs/auto-pytorch/lib/python3.8/site-packages/autoPyTorch/utils/pipeline.py:116, in _get_classification_dataset_requirements(info, include, exclude, search_space_updates)
113 task_type = STRING_TO_TASK_TYPES[info['task_type']]
115 if task_type in TABULAR_TASKS:
--> 116 return TabularClassificationPipeline(
117 dataset_properties=info,
118 include=include, exclude=exclude,
119 search_space_updates=search_space_updates
120 ).get_dataset_requirements()
121 elif task_type in IMAGE_TASKS:
122 return ImageClassificationPipeline(
123 dataset_properties=info,
124 include=include, exclude=exclude,
125 search_space_updates=search_space_updates
126 ).get_dataset_requirements()

File ~/miniconda3/envs/auto-pytorch/lib/python3.8/site-packages/autoPyTorch/pipeline/tabular_classification.py:128, in TabularClassificationPipeline.init(self, config, steps, dataset_properties, include, exclude, random_state, init_params, search_space_updates)
117 def init(
118 self,
119 config: Optional[Configuration] = None,
(...)
126 search_space_updates: Optional[HyperparameterSearchSpaceUpdates] = None
127 ):
--> 128 super().init(
129 config, steps, dataset_properties, include, exclude,
130 random_state, init_params, search_space_updates)
132 # Because a pipeline is passed to a worker, we need to honor the random seed
133 # in this context. A tabular classification pipeline will implement a torch
134 # model, so we comply with https://pytorch.org/docs/stable/notes/randomness.html
135 torch.manual_seed(self.random_state.get_state()[1][0])

File ~/miniconda3/envs/auto-pytorch/lib/python3.8/site-packages/autoPyTorch/pipeline/base_pipeline.py:104, in BasePipeline.init(self, config, steps, dataset_properties, include, exclude, random_state, init_params, search_space_updates)
101 else:
102 self.steps = steps
--> 104 self.config_space = self.get_hyperparameter_search_space()
106 if config is None:
107 self.config = self.config_space.get_default_configuration()

File ~/miniconda3/envs/auto-pytorch/lib/python3.8/site-packages/autoPyTorch/pipeline/base_pipeline.py:262, in BasePipeline.get_hyperparameter_search_space(self)
256 """Return the configuration space for the CASH problem.
257
258 Returns:
259 ConfigurationSpace: The configuration space describing the Pipeline.
260 """
261 if not hasattr(self, 'config_space') or self.config_space is None:
--> 262 self.config_space = self._get_hyperparameter_search_space(
263 dataset_properties=self.dataset_properties,
264 include=self.include,
265 exclude=self.exclude,
266 )
267 return self.config_space

File ~/miniconda3/envs/auto-pytorch/lib/python3.8/site-packages/autoPyTorch/pipeline/tabular_classification.py:260, in TabularClassificationPipeline._get_hyperparameter_search_space(self, dataset_properties, include, exclude)
256 dataset_properties['target_type'] = 'tabular_classification'
257 # get the base search space given this
258 # dataset properties. Then overwrite with custom
259 # classification requirements
--> 260 cs = self._get_base_search_space(
261 cs=cs, dataset_properties=dataset_properties,
262 exclude=exclude, include=include, pipeline=self.steps)
264 # Here we add custom code, that is used to ensure valid configurations, For example
265 # Learned Entity Embedding is only valid when encoder is one hot encoder
266 if 'network_embedding' in self.named_steps.keys() and 'encoder' in self.named_steps.keys():

File ~/miniconda3/envs/auto-pytorch/lib/python3.8/site-packages/autoPyTorch/pipeline/base_pipeline.py:376, in BasePipeline._get_base_search_space(self, cs, dataset_properties, include, exclude, pipeline)
367 choices_list = find_active_choices(
368 matches, node, node_idx,
369 dataset_properties,
370 include.get(node_name),
371 exclude.get(node_name)
372 )
374 # ignore type check here as mypy is not able to infer
375 # that isinstance(node, autoPyTorchChooice) = True
--> 376 sub_config_space = node.get_hyperparameter_search_space( # type: ignore[call-arg]
377 dataset_properties, include=choices_list)
378 cs.add_configuration_space(node_name, sub_config_space)
380 # If the node is a choice, we have to figure out which of its
381 # choices are actually legal choices
382 else:

File ~/miniconda3/envs/auto-pytorch/lib/python3.8/site-packages/autoPyTorch/pipeline/components/preprocessing/tabular_preprocessing/feature_preprocessing/init.py:213, in FeatureProprocessorChoice.get_hyperparameter_search_space(self, dataset_properties, default, include, exclude)
210 config_space = available_[name].get_hyperparameter_search_space(dataset_properties, # type:ignore
211 **updates)
212 parent_hyperparameter = {'parent': preprocessor, 'value': name}
--> 213 cs.add_configuration_space(name, config_space,
214 parent_hyperparameter=parent_hyperparameter)
216 self.configuration_space = cs
217 self.dataset_properties = dataset_properties

File ~/miniconda3/envs/auto-pytorch/lib/python3.8/site-packages/ConfigSpace/configuration_space.py:446, in ConfigurationSpace.add_configuration_space(self, prefix, configuration_space, delimiter, parent_hyperparameter)
443 conditions_to_add.append(EqualsCondition(param, parent, value))
445 if len(conditions_to_add) > 0:
--> 446 self.add(conditions_to_add)
448 return configuration_space

File ~/miniconda3/envs/auto-pytorch/lib/python3.8/site-packages/ConfigSpace/configuration_space.py:351, in ConfigurationSpace.add(self, *args)
348 self._dag.add_forbidden(forbidden)
350 self._len = len(self._dag.nodes)
--> 351 self._check_default_configuration()

File ~/miniconda3/envs/auto-pytorch/lib/python3.8/site-packages/ConfigSpace/configuration_space.py:915, in ConfigurationSpace._check_default_configuration(self)
912 else:
913 values[hp_name] = hp.default_value
--> 915 return Configuration(self, values=values)

File ~/miniconda3/envs/auto-pytorch/lib/python3.8/site-packages/ConfigSpace/configuration.py:126, in Configuration.init(self, configuration_space, values, vector, allow_inactive_with_values, origin, config_id)
123 self._values[key] = value
124 self._vector[i] = hp.to_vector(value) # type: ignore
--> 126 self.check_valid_configuration()
128 elif vector is not None:
129 if not isinstance(vector, np.ndarray):

File ~/miniconda3/envs/auto-pytorch/lib/python3.8/site-packages/ConfigSpace/configuration.py:160, in Configuration.check_valid_configuration(self)
153 """Check if the object is a valid.
154
155 Raises:
156 ValueError: If configuration is not valid.
157 """
158 from ConfigSpace.util import check_configuration
--> 160 check_configuration(
161 self.config_space,
162 self._vector,
163 allow_inactive_with_values=self.allow_inactive_with_values,
164 )

File ~/miniconda3/envs/auto-pytorch/lib/python3.8/site-packages/ConfigSpace/util.py:593, in check_configuration(space, vector, allow_inactive_with_values)
591 hp_name = space.at[idx]
592 hp = space[hp_name]
--> 593 raise ActiveHyperparameterNotSetError(hp)
595 for hp_idx, hp_node in cnode.unique_children.items():
596 # OPTIM: We bypass the larger safety checking of the hp and access
597 # the underlying transformer directly
598 transformer = hp_node.hp._transformer

ActiveHyperparameterNotSetError: Hyperparameter is active but has no value set.
SelectRatesClassification:mode, Type: Categorical, Choices: {fpr, fdr, fwe, percentile}, Default: fpr