wavetrainer 0.1.17__tar.gz → 0.1.19__tar.gz

{wavetrainer-0.1.17/wavetrainer.egg-info → wavetrainer-0.1.19}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: wavetrainer
-Version: 0.1.17
+Version: 0.1.19
 Summary: A library for automatically finding the optimal model within feature and hyperparameter space.
 Home-page: https://github.com/8W9aG/wavetrainer
 Author: Will Sackfield

{wavetrainer-0.1.17 → wavetrainer-0.1.19}/setup.py

@@ -23,7 +23,7 @@ def install_requires() -> typing.List[str]:
 
 setup(
     name='wavetrainer',
-    version='0.1.17',
+    version='0.1.19',
     description='A library for automatically finding the optimal model within feature and hyperparameter space.',
     long_description=long_description,
     long_description_content_type='text/markdown',

{wavetrainer-0.1.17 → wavetrainer-0.1.19}/tests/trainer_test.py

@@ -40,7 +40,7 @@ class TestTrainer(unittest.TestCase):
 
     def test_trainer_dt_column(self):
         with tempfile.TemporaryDirectory() as tmpdir:
-            trainer = Trainer(tmpdir, walkforward_timedelta=datetime.timedelta(days=7), trials=5, dt_column="dt_column")
+            trainer = Trainer(tmpdir, walkforward_timedelta=datetime.timedelta(days=7), trials=5, dt_column="dt_column", allowed_models={"catboost"})
             x_data = [i for i in range(100)]
             x_index = [datetime.datetime(2022, 1, 1) + datetime.timedelta(days=i) for i in range(len(x_data))]
             df = pd.DataFrame(
@@ -55,6 +55,7 @@ class TestTrainer(unittest.TestCase):
                 },
                 index=df.index,
             )
+            y["y"] = y["y"].astype(bool)
             trainer.fit(df, y=y)
             df = trainer.transform(df)
             print("df:")

{wavetrainer-0.1.17 → wavetrainer-0.1.19}/wavetrainer/__init__.py

@@ -2,5 +2,5 @@
 
 from .create import create
 
-__VERSION__ = "0.1.17"
+__VERSION__ = "0.1.19"
 __all__ = ("create",)

{wavetrainer-0.1.17 → wavetrainer-0.1.19}/wavetrainer/calibrator/vennabers_calibrator.py

@@ -5,6 +5,7 @@ import os
 from typing import Self
 
 import joblib  # type: ignore
+import numpy as np
 import optuna
 import pandas as pd
 from venn_abers import VennAbers  # type: ignore
@@ -55,18 +56,26 @@ class VennabersCalibrator(Calibrator):
             raise ValueError("vennabers is null")
         if y is None:
             raise ValueError("y is null")
-        prob_columns = [
-            x for x in df.columns.values if x.startswith(PROBABILITY_COLUMN_PREFIX)
-        ]
+        prob_columns = sorted(
+            [x for x in df.columns.values if x.startswith(PROBABILITY_COLUMN_PREFIX)]
+        )
+        probs = df[prob_columns].to_numpy()
         try:
-            vennabers.fit(df[prob_columns].to_numpy(), y.to_numpy())
+            vennabers.fit(probs, y.to_numpy())
         except IndexError:
             logging.error(df)
             raise
         return self
 
     def transform(self, df: pd.DataFrame) -> pd.DataFrame:
-        p_prime, _ = self._vennabers.predict_proba(df.to_numpy())
+        prob_columns = sorted(
+            [x for x in df.columns.values if x.startswith(PROBABILITY_COLUMN_PREFIX)]
+        )
+        probs = df[prob_columns].to_numpy()
+        p_prime, _ = self._vennabers.predict_proba(probs)
+        if np.mean(p_prime[:, 1] > 0.5) > 0.5 and np.mean(probs[:, 0] > 0.5) > 0.5:
+            print("⚠️ Warning: calibration seems inverted — flipping p_prime")
+            p_prime = p_prime[:, ::-1]
         for i in range(p_prime.shape[1]):
             prob = p_prime[:, i]
             df[f"{PROBABILITY_COLUMN_PREFIX}{i}"] = prob

{wavetrainer-0.1.17 → wavetrainer-0.1.19}/wavetrainer/reducer/combined_reducer.py

@@ -41,8 +41,8 @@ class CombinedReducer(Reducer):
             UnseenReducer(),
             NonNumericReducer(),
             PCAReducer(embedding_cols),
-            ConstantReducer(),
-            DuplicateReducer(),
+            # ConstantReducer(),
+            # DuplicateReducer(),
             CorrelationReducer(correlation_chunk_size=correlation_chunk_size),
             SmartCorrelationReducer(),
             # SelectBySingleFeaturePerformanceReducer(),

{wavetrainer-0.1.17 → wavetrainer-0.1.19}/wavetrainer/reducer/correlation_reducer.py

@@ -8,7 +8,6 @@ from typing import Self
 import numpy as np
 import optuna
 import pandas as pd
-from memory_profiler import profile  # type: ignore
 
 from .non_categorical_numeric_columns import \
     find_non_categorical_numeric_columns
@@ -18,7 +17,6 @@ _CORRELATION_REDUCER_FILENAME = "correlation_reducer.json"
 _CORRELATION_REDUCER_THRESHOLD = "correlation_reducer_threshold"
 
 
-@profile
 def _get_correlated_features_to_drop_chunked(
     df: pd.DataFrame,
     threshold: float = 0.85,
@@ -119,7 +117,7 @@ class CorrelationReducer(Reducer):
         eval_y: pd.Series | pd.DataFrame | None = None,
     ) -> Self:
         drop_features = _get_correlated_features_to_drop_chunked(
-            df.copy(),
+            df,
             threshold=self._threshold,
             chunk_size=self._correlation_chunk_size,
         )

{wavetrainer-0.1.17 → wavetrainer-0.1.19/wavetrainer.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: wavetrainer
-Version: 0.1.17
+Version: 0.1.19
 Summary: A library for automatically finding the optimal model within feature and hyperparameter space.
 Home-page: https://github.com/8W9aG/wavetrainer
 Author: Will Sackfield