Skip to content

octopus.diagnostics

Diagnostics package — interactive study-level diagnostics from saved parquet files.

Provides :class:StudyDiagnostics for exploring predictions, feature importances, and Optuna hyperparameter tuning results across all outer splits and tasks.

No model loading is performed — all data comes from saved parquet artifacts.

Example::

from octopus.diagnostics import StudyDiagnostics

diag = StudyDiagnostics("./studies/my_study/")
diag.plot_feature_importance()
diag.plot_optuna_trials()

StudyDiagnostics

Interactive study-level diagnostics from saved parquet files.

Loads predictions, feature importances, scores, and Optuna results from the study directory structure. No model loading is performed.

Parameters:

Name Type Description Default
study_path str | Path

Path to the study directory.

 required

Raises:

Type Description
FileNotFoundError

If the study directory or study_config.json does not exist.

Example::

from octopus.diagnostics import StudyDiagnostics

diag = StudyDiagnostics("./studies/my_study/")
diag.plot_feature_importance()
diag.plot_optuna_trials()
Source code in octopus/diagnostics/core.py 
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
class StudyDiagnostics:
    """Interactive study-level diagnostics from saved parquet files.

    Loads predictions, feature importances, scores, and Optuna results
    from the study directory structure. No model loading is performed.

    Args:
        study_path: Path to the study directory.

    Raises:
        FileNotFoundError: If the study directory or study_config.json does not exist.

    Example::

        from octopus.diagnostics import StudyDiagnostics

        diag = StudyDiagnostics("./studies/my_study/")
        diag.plot_feature_importance()
        diag.plot_optuna_trials()
    """

    def __init__(self, study_path: str | Path) -> None:
        self._study_path = Path(study_path)
        if not self._study_path.exists():
            raise FileNotFoundError(f"Study path does not exist: {self._study_path}")

        # Load config
        config_path = self._study_path / "study_config.json"
        if config_path.exists():
            with open(config_path) as f:
                self._config: dict[str, Any] = json.load(f)
        else:
            self._config = {}

        # Lazy-loaded DataFrames
        self._predictions: pd.DataFrame | None = None
        self._feature_importances: pd.DataFrame | None = None
        self._optuna: pd.DataFrame | None = None
        self._scores: pd.DataFrame | None = None

    # ── Properties ──────────────────────────────────────────────

    @property
    def study_path(self) -> Path:
        """Path to the study directory."""
        return self._study_path

    @property
    def config(self) -> dict[str, Any]:
        """Study configuration dictionary."""
        return self._config

    @property
    def ml_type(self) -> MLType:
        """Machine learning type (classification, regression, timetoevent)."""
        return MLType(self._config.get("ml_type", ""))

    @property
    def predictions(self) -> pd.DataFrame:
        """All predictions across outersplits and tasks (lazy-loaded)."""
        if self._predictions is None:
            self._predictions = load_predictions(self._study_path)
        return self._predictions

    @property
    def feature_importances(self) -> pd.DataFrame:
        """All feature importances across outersplits and tasks (lazy-loaded)."""
        if self._feature_importances is None:
            self._feature_importances = load_feature_importances(self._study_path)
        return self._feature_importances

    @property
    def optuna_trials(self) -> pd.DataFrame:
        """All Optuna trial results across outersplits and tasks (lazy-loaded)."""
        if self._optuna is None:
            self._optuna = load_optuna(self._study_path)
        return self._optuna

    @property
    def scores(self) -> pd.DataFrame:
        """All scores across outersplits and tasks (lazy-loaded)."""
        if self._scores is None:
            self._scores = load_scores(self._study_path)
        return self._scores

    # ── Filter helpers ──────────────────────────────────────────

    def _get_filter_options(self, df: pd.DataFrame, columns: list[str]) -> dict[str, list[str]]:
        """Extract unique sorted values for each column.

        Args:
            df: DataFrame to extract from.
            columns: Column names.

        Returns:
            Dict mapping column name to sorted unique string values.
        """
        return {col: _unique_sorted(df[col]) for col in columns if col in df.columns}

    # ── Interactive Plots ───────────────────────────────────────

    def plot_feature_importance(
        self,
        outersplit_id: int | None = None,
        task_id: int | None = None,
        training_id: str | None = None,
        fi_method: str | FIResultLabel | None = None,
    ) -> None:
        """Plot feature importance bar chart.

        If ipywidgets is available and parameters are None, shows interactive
        dropdowns. Otherwise uses provided values or defaults.

        Args:
            outersplit_id: Outer split to filter on.
            task_id: Task to filter on.
            training_id: Training ID to filter on.
            fi_method: FI method to filter on.
        """
        df = self.feature_importances
        if df.empty:
            print("No feature importance data found.")
            return

        if _has_ipywidgets() and outersplit_id is None:
            from ipywidgets import Dropdown, interact  # noqa: PLC0415

            opts = self._get_filter_options(df, ["outersplit_id", "task_id", "training_id", "fi_method"])

            @interact(
                outersplit_id=Dropdown(options=opts.get("outersplit_id", ["0"]), description="Outersplit:"),
                task_id=Dropdown(options=opts.get("task_id", ["0"]), description="Task:"),
                training_id=Dropdown(options=opts.get("training_id", [""]), description="Training:"),
                fi_method=Dropdown(options=opts.get("fi_method", [""]), description="FI Method:"),
            )
            def _plot(outersplit_id: str, task_id: str, training_id: str, fi_method: str) -> None:
                fig = plot_feature_importance_chart(
                    df, outersplit_id=outersplit_id, task_id=task_id, training_id=training_id, fi_method=fi_method
                )
                fig.show()
        else:
            fig = plot_feature_importance_chart(
                df,
                outersplit_id=outersplit_id or 0,
                task_id=task_id or 0,
                training_id=training_id or "",
                fi_method=fi_method or "",
            )
            fig.show()

    def plot_confusion_matrix(
        self,
        outersplit_id: int | None = None,
        task_id: int | None = None,
        training_id: str | None = None,
    ) -> None:
        """Plot confusion matrix heatmap (classification only).

        Args:
            outersplit_id: Outer split to filter on.
            task_id: Task to filter on.
            training_id: Inner split / training ID to filter on.
        """
        df = self.predictions
        if df.empty:
            print("No prediction data found.")
            return

        if _has_ipywidgets() and outersplit_id is None:
            from ipywidgets import Dropdown, interact  # noqa: PLC0415

            opts = self._get_filter_options(df, ["outersplit_id", "task_id", "inner_split_id"])

            @interact(
                outersplit_id=Dropdown(options=opts.get("outersplit_id", ["0"]), description="Outersplit:"),
                task_id=Dropdown(options=opts.get("task_id", ["0"]), description="Task:"),
                training_id=Dropdown(options=opts.get("inner_split_id", [""]), description="Training:"),
            )
            def _plot(outersplit_id: str, task_id: str, training_id: str) -> None:
                fig = plot_confusion_matrix_chart(
                    df, outersplit_id=outersplit_id, task_id=task_id, training_id=training_id
                )
                fig.show()
        else:
            fig = plot_confusion_matrix_chart(
                df,
                outersplit_id=outersplit_id or 0,
                task_id=task_id or 0,
                training_id=training_id or "",
            )
            fig.show()

    def plot_predictions_vs_truth(
        self,
        outersplit_id: int | None = None,
        task_id: int | None = None,
        training_id: str | None = None,
    ) -> None:
        """Plot prediction vs ground truth scatter (regression only).

        Args:
            outersplit_id: Outer split to filter on.
            task_id: Task to filter on.
            training_id: Inner split / training ID to filter on.
        """
        df = self.predictions
        if df.empty:
            print("No prediction data found.")
            return

        if _has_ipywidgets() and outersplit_id is None:
            from ipywidgets import Dropdown, interact  # noqa: PLC0415

            opts = self._get_filter_options(df, ["outersplit_id", "task_id", "inner_split_id"])

            @interact(
                outersplit_id=Dropdown(options=opts.get("outersplit_id", ["0"]), description="Outersplit:"),
                task_id=Dropdown(options=opts.get("task_id", ["0"]), description="Task:"),
                training_id=Dropdown(options=opts.get("inner_split_id", [""]), description="Training:"),
            )
            def _plot(outersplit_id: str, task_id: str, training_id: str) -> None:
                fig = plot_predictions_vs_truth_chart(
                    df, outersplit_id=outersplit_id, task_id=task_id, training_id=training_id
                )
                fig.show()
        else:
            fig = plot_predictions_vs_truth_chart(
                df,
                outersplit_id=outersplit_id or 0,
                task_id=task_id or 0,
                training_id=training_id or "",
            )
            fig.show()

    def plot_optuna_trial_counts(self) -> None:
        """Plot bar chart of unique trial counts per model type."""
        df = self.optuna_trials
        if df.empty:
            print("No Optuna data found.")
            return
        fig = plot_optuna_trial_counts_chart(df)
        fig.show()

    def plot_optuna_trials(
        self,
        outersplit_id: int | None = None,
        task_id: int | None = None,
        direction: MetricDirection = MetricDirection.MINIMIZE,
    ) -> None:
        """Plot Optuna trial scatter + cumulative best line.

        Args:
            outersplit_id: Outer split to filter on.
            task_id: Task to filter on.
            direction: Optimization direction ('minimize' or 'maximize').
        """
        df = self.optuna_trials
        if df.empty:
            print("No Optuna data found.")
            return

        if _has_ipywidgets() and outersplit_id is None:
            from ipywidgets import Dropdown, interact  # noqa: PLC0415

            opts = self._get_filter_options(df, ["outersplit_id", "task_id"])

            @interact(
                outersplit_id=Dropdown(options=opts.get("outersplit_id", ["0"]), description="Outersplit:"),
                task_id=Dropdown(options=opts.get("task_id", ["0"]), description="Task:"),
            )
            def _plot(outersplit_id: str, task_id: str) -> None:
                fig = plot_optuna_trials_chart(df, outersplit_id=outersplit_id, task_id=task_id, direction=direction)
                fig.show()
        else:
            fig = plot_optuna_trials_chart(
                df,
                outersplit_id=outersplit_id or 0,
                task_id=task_id or 0,
                direction=direction,
            )
            fig.show()

    def plot_optuna_hyperparameters(
        self,
        outersplit_id: int | None = None,
        task_id: int | None = None,
        model_type: str | None = None,
    ) -> None:
        """Plot Optuna hyperparameter scatter plots.

        Args:
            outersplit_id: Outer split to filter on.
            task_id: Task to filter on.
            model_type: Model type to filter on.
        """
        df = self.optuna_trials
        if df.empty:
            print("No Optuna data found.")
            return

        if _has_ipywidgets() and outersplit_id is None:
            from ipywidgets import Dropdown, interact  # noqa: PLC0415

            opts = self._get_filter_options(df, ["outersplit_id", "task_id", "model_type"])

            @interact(
                outersplit_id=Dropdown(options=opts.get("outersplit_id", ["0"]), description="Outersplit:"),
                task_id=Dropdown(options=opts.get("task_id", ["0"]), description="Task:"),
                model_type=Dropdown(options=opts.get("model_type", [""]), description="Model:"),
            )
            def _plot(outersplit_id: str, task_id: str, model_type: str) -> None:
                fig = plot_optuna_hyperparameters_chart(
                    df, outersplit_id=outersplit_id, task_id=task_id, model_type=model_type
                )
                fig.show()
        else:
            fig = plot_optuna_hyperparameters_chart(
                df,
                outersplit_id=outersplit_id or 0,
                task_id=task_id or 0,
                model_type=model_type or "",
            )
            fig.show()

config property

Study configuration dictionary.

feature_importances property

All feature importances across outersplits and tasks (lazy-loaded).

ml_type property

Machine learning type (classification, regression, timetoevent).

optuna_trials property

All Optuna trial results across outersplits and tasks (lazy-loaded).

predictions property

All predictions across outersplits and tasks (lazy-loaded).

scores property

All scores across outersplits and tasks (lazy-loaded).

study_path property

Path to the study directory.

plot_confusion_matrix(outersplit_id=None, task_id=None, training_id=None)

Plot confusion matrix heatmap (classification only).

Parameters:

Name Type Description Default
outersplit_id int | None

Outer split to filter on.

 None
task_id int | None

Task to filter on.

 None
training_id str | None

Inner split / training ID to filter on.

 None
Source code in octopus/diagnostics/core.py 
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
def plot_confusion_matrix(
    self,
    outersplit_id: int | None = None,
    task_id: int | None = None,
    training_id: str | None = None,
) -> None:
    """Plot confusion matrix heatmap (classification only).

    Args:
        outersplit_id: Outer split to filter on.
        task_id: Task to filter on.
        training_id: Inner split / training ID to filter on.
    """
    df = self.predictions
    if df.empty:
        print("No prediction data found.")
        return

    if _has_ipywidgets() and outersplit_id is None:
        from ipywidgets import Dropdown, interact  # noqa: PLC0415

        opts = self._get_filter_options(df, ["outersplit_id", "task_id", "inner_split_id"])

        @interact(
            outersplit_id=Dropdown(options=opts.get("outersplit_id", ["0"]), description="Outersplit:"),
            task_id=Dropdown(options=opts.get("task_id", ["0"]), description="Task:"),
            training_id=Dropdown(options=opts.get("inner_split_id", [""]), description="Training:"),
        )
        def _plot(outersplit_id: str, task_id: str, training_id: str) -> None:
            fig = plot_confusion_matrix_chart(
                df, outersplit_id=outersplit_id, task_id=task_id, training_id=training_id
            )
            fig.show()
    else:
        fig = plot_confusion_matrix_chart(
            df,
            outersplit_id=outersplit_id or 0,
            task_id=task_id or 0,
            training_id=training_id or "",
        )
        fig.show()

plot_feature_importance(outersplit_id=None, task_id=None, training_id=None, fi_method=None)

Plot feature importance bar chart.

If ipywidgets is available and parameters are None, shows interactive dropdowns. Otherwise uses provided values or defaults.

Parameters:

Name Type Description Default
outersplit_id int | None

Outer split to filter on.

 None
task_id int | None

Task to filter on.

 None
training_id str | None

Training ID to filter on.

 None
fi_method str | FIResultLabel | None

FI method to filter on.

 None
Source code in octopus/diagnostics/core.py 
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
def plot_feature_importance(
    self,
    outersplit_id: int | None = None,
    task_id: int | None = None,
    training_id: str | None = None,
    fi_method: str | FIResultLabel | None = None,
) -> None:
    """Plot feature importance bar chart.

    If ipywidgets is available and parameters are None, shows interactive
    dropdowns. Otherwise uses provided values or defaults.

    Args:
        outersplit_id: Outer split to filter on.
        task_id: Task to filter on.
        training_id: Training ID to filter on.
        fi_method: FI method to filter on.
    """
    df = self.feature_importances
    if df.empty:
        print("No feature importance data found.")
        return

    if _has_ipywidgets() and outersplit_id is None:
        from ipywidgets import Dropdown, interact  # noqa: PLC0415

        opts = self._get_filter_options(df, ["outersplit_id", "task_id", "training_id", "fi_method"])

        @interact(
            outersplit_id=Dropdown(options=opts.get("outersplit_id", ["0"]), description="Outersplit:"),
            task_id=Dropdown(options=opts.get("task_id", ["0"]), description="Task:"),
            training_id=Dropdown(options=opts.get("training_id", [""]), description="Training:"),
            fi_method=Dropdown(options=opts.get("fi_method", [""]), description="FI Method:"),
        )
        def _plot(outersplit_id: str, task_id: str, training_id: str, fi_method: str) -> None:
            fig = plot_feature_importance_chart(
                df, outersplit_id=outersplit_id, task_id=task_id, training_id=training_id, fi_method=fi_method
            )
            fig.show()
    else:
        fig = plot_feature_importance_chart(
            df,
            outersplit_id=outersplit_id or 0,
            task_id=task_id or 0,
            training_id=training_id or "",
            fi_method=fi_method or "",
        )
        fig.show()

plot_optuna_hyperparameters(outersplit_id=None, task_id=None, model_type=None)

Plot Optuna hyperparameter scatter plots.

Parameters:

Name Type Description Default
outersplit_id int | None

Outer split to filter on.

 None
task_id int | None

Task to filter on.

 None
model_type str | None

Model type to filter on.

 None
Source code in octopus/diagnostics/core.py 
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
def plot_optuna_hyperparameters(
    self,
    outersplit_id: int | None = None,
    task_id: int | None = None,
    model_type: str | None = None,
) -> None:
    """Plot Optuna hyperparameter scatter plots.

    Args:
        outersplit_id: Outer split to filter on.
        task_id: Task to filter on.
        model_type: Model type to filter on.
    """
    df = self.optuna_trials
    if df.empty:
        print("No Optuna data found.")
        return

    if _has_ipywidgets() and outersplit_id is None:
        from ipywidgets import Dropdown, interact  # noqa: PLC0415

        opts = self._get_filter_options(df, ["outersplit_id", "task_id", "model_type"])

        @interact(
            outersplit_id=Dropdown(options=opts.get("outersplit_id", ["0"]), description="Outersplit:"),
            task_id=Dropdown(options=opts.get("task_id", ["0"]), description="Task:"),
            model_type=Dropdown(options=opts.get("model_type", [""]), description="Model:"),
        )
        def _plot(outersplit_id: str, task_id: str, model_type: str) -> None:
            fig = plot_optuna_hyperparameters_chart(
                df, outersplit_id=outersplit_id, task_id=task_id, model_type=model_type
            )
            fig.show()
    else:
        fig = plot_optuna_hyperparameters_chart(
            df,
            outersplit_id=outersplit_id or 0,
            task_id=task_id or 0,
            model_type=model_type or "",
        )
        fig.show()

plot_optuna_trial_counts()

Plot bar chart of unique trial counts per model type.

Source code in octopus/diagnostics/core.py 
285
286
287
288
289
290
291
292
def plot_optuna_trial_counts(self) -> None:
    """Plot bar chart of unique trial counts per model type."""
    df = self.optuna_trials
    if df.empty:
        print("No Optuna data found.")
        return
    fig = plot_optuna_trial_counts_chart(df)
    fig.show()

plot_optuna_trials(outersplit_id=None, task_id=None, direction=MetricDirection.MINIMIZE)

Plot Optuna trial scatter + cumulative best line.

Parameters:

Name Type Description Default
outersplit_id int | None

Outer split to filter on.

 None
task_id int | None

Task to filter on.

 None
direction MetricDirection

Optimization direction ('minimize' or 'maximize').

 MINIMIZE
Source code in octopus/diagnostics/core.py 
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
def plot_optuna_trials(
    self,
    outersplit_id: int | None = None,
    task_id: int | None = None,
    direction: MetricDirection = MetricDirection.MINIMIZE,
) -> None:
    """Plot Optuna trial scatter + cumulative best line.

    Args:
        outersplit_id: Outer split to filter on.
        task_id: Task to filter on.
        direction: Optimization direction ('minimize' or 'maximize').
    """
    df = self.optuna_trials
    if df.empty:
        print("No Optuna data found.")
        return

    if _has_ipywidgets() and outersplit_id is None:
        from ipywidgets import Dropdown, interact  # noqa: PLC0415

        opts = self._get_filter_options(df, ["outersplit_id", "task_id"])

        @interact(
            outersplit_id=Dropdown(options=opts.get("outersplit_id", ["0"]), description="Outersplit:"),
            task_id=Dropdown(options=opts.get("task_id", ["0"]), description="Task:"),
        )
        def _plot(outersplit_id: str, task_id: str) -> None:
            fig = plot_optuna_trials_chart(df, outersplit_id=outersplit_id, task_id=task_id, direction=direction)
            fig.show()
    else:
        fig = plot_optuna_trials_chart(
            df,
            outersplit_id=outersplit_id or 0,
            task_id=task_id or 0,
            direction=direction,
        )
        fig.show()

plot_predictions_vs_truth(outersplit_id=None, task_id=None, training_id=None)

Plot prediction vs ground truth scatter (regression only).

Parameters:

Name Type Description Default
outersplit_id int | None

Outer split to filter on.

 None
task_id int | None

Task to filter on.

 None
training_id str | None

Inner split / training ID to filter on.

 None
Source code in octopus/diagnostics/core.py 
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
def plot_predictions_vs_truth(
    self,
    outersplit_id: int | None = None,
    task_id: int | None = None,
    training_id: str | None = None,
) -> None:
    """Plot prediction vs ground truth scatter (regression only).

    Args:
        outersplit_id: Outer split to filter on.
        task_id: Task to filter on.
        training_id: Inner split / training ID to filter on.
    """
    df = self.predictions
    if df.empty:
        print("No prediction data found.")
        return

    if _has_ipywidgets() and outersplit_id is None:
        from ipywidgets import Dropdown, interact  # noqa: PLC0415

        opts = self._get_filter_options(df, ["outersplit_id", "task_id", "inner_split_id"])

        @interact(
            outersplit_id=Dropdown(options=opts.get("outersplit_id", ["0"]), description="Outersplit:"),
            task_id=Dropdown(options=opts.get("task_id", ["0"]), description="Task:"),
            training_id=Dropdown(options=opts.get("inner_split_id", [""]), description="Training:"),
        )
        def _plot(outersplit_id: str, task_id: str, training_id: str) -> None:
            fig = plot_predictions_vs_truth_chart(
                df, outersplit_id=outersplit_id, task_id=task_id, training_id=training_id
            )
            fig.show()
    else:
        fig = plot_predictions_vs_truth_chart(
            df,
            outersplit_id=outersplit_id or 0,
            task_id=task_id or 0,
            training_id=training_id or "",
        )
        fig.show()