upgini 1.1.244a4__py3-none-any.whl → 1.1.244a6__py3-none-any.whl

upgini/features_enricher.py

@@ -40,6 +40,7 @@ from upgini.metadata import (
     EVAL_SET_INDEX,
     ORIGINAL_INDEX,
     RENAMED_INDEX,
+    SORT_ID,
     SYSTEM_RECORD_ID,
     TARGET,
     CVType,
@@ -1299,7 +1300,11 @@ class FeaturesEnricher(TransformerMixin):
             c
             for c in X_sampled.columns.to_list()
             if c
-            not in (excluding_search_keys + list(self.fit_dropped_features) + [DateTimeSearchKeyConverter.DATETIME_COL])
+            not in (
+                excluding_search_keys
+                + list(self.fit_dropped_features)
+                + [DateTimeSearchKeyConverter.DATETIME_COL, SYSTEM_RECORD_ID]
+            )
         ]
 
         filtered_enriched_features = self.__filtered_enriched_features(
@@ -1307,10 +1312,10 @@ class FeaturesEnricher(TransformerMixin):
             max_features,
         )
 
-        X_sorted, y_sorted = self._sort_by_keys(X_sampled, y_sampled, search_keys, self.cv)
-        enriched_X_sorted, enriched_y_sorted = self._sort_by_keys(enriched_X, y_sampled, search_keys, self.cv)
+        X_sorted, y_sorted = self._sort_by_system_record_id(X_sampled, y_sampled, self.cv)
+        enriched_X_sorted, enriched_y_sorted = self._sort_by_system_record_id(enriched_X, y_sampled, self.cv)
 
-        group_columns = sorted(self._get_group_columns(search_keys))
+        group_columns = sorted(self._get_group_columns(enriched_X_sorted, search_keys))
         groups = (
             None
             if not group_columns or self.cv != CVType.group_k_fold
@@ -1332,9 +1337,9 @@ class FeaturesEnricher(TransformerMixin):
         fitting_eval_set_dict = dict()
         for idx, eval_tuple in eval_set_sampled_dict.items():
             eval_X_sampled, enriched_eval_X, eval_y_sampled = eval_tuple
-            eval_X_sorted, eval_y_sorted = self._sort_by_keys(eval_X_sampled, eval_y_sampled, search_keys, self.cv)
-            enriched_eval_X_sorted, enriched_eval_y_sorted = self._sort_by_keys(
-                enriched_eval_X, eval_y_sampled, search_keys, self.cv
+            eval_X_sorted, eval_y_sorted = self._sort_by_system_record_id(eval_X_sampled, eval_y_sampled, self.cv)
+            enriched_eval_X_sorted, enriched_eval_y_sorted = self._sort_by_system_record_id(
+                enriched_eval_X, eval_y_sampled, self.cv
             )
             fitting_eval_X = eval_X_sorted[client_features].copy()
             fitting_enriched_eval_X = enriched_eval_X_sorted[
@@ -1470,19 +1475,15 @@ class FeaturesEnricher(TransformerMixin):
             self.df_with_original_index,
             fit_features,
             rows_to_drop=rows_to_drop,
+            drop_system_record_id=False,
         )
 
-        original_df_sampled = self.df_with_original_index[
-            self.df_with_original_index[SYSTEM_RECORD_ID].isin(fit_features[SYSTEM_RECORD_ID])
-        ]
-        enriched_X = drop_existing_columns(enriched_Xy, TARGET)
-        if EVAL_SET_INDEX in original_df_sampled.columns:
-            Xy_sampled = original_df_sampled.query(f"{EVAL_SET_INDEX} == 0")
-        else:
-            Xy_sampled = original_df_sampled
-        X_sampled = drop_existing_columns(Xy_sampled, [SYSTEM_RECORD_ID, EVAL_SET_INDEX, TARGET])
+        x_columns = [c for c in self.df_with_original_index.columns if c not in [EVAL_SET_INDEX, TARGET]]
+        X_sampled = enriched_Xy[x_columns].copy()
+        y_sampled = enriched_Xy[TARGET].copy()
+        enriched_X = drop_existing_columns(enriched_Xy, [TARGET, EVAL_SET_INDEX])
+
         search_keys = self.fit_search_keys
-        y_sampled = Xy_sampled[TARGET].copy()
 
         self.logger.info(f"Shape of enriched_X: {enriched_X.shape}")
         self.logger.info(f"Shape of X after sampling: {X_sampled.shape}")
@@ -1495,10 +1496,9 @@ class FeaturesEnricher(TransformerMixin):
                 )
 
             for idx in range(len(eval_set)):
-                enriched_eval_X = drop_existing_columns(enriched_eval_sets[idx + 1], TARGET)
-                eval_Xy_sampled = original_df_sampled.query(f"{EVAL_SET_INDEX} == {idx + 1}")
-                eval_X_sampled = drop_existing_columns(eval_Xy_sampled, [SYSTEM_RECORD_ID, EVAL_SET_INDEX, TARGET])
-                eval_y_sampled = eval_Xy_sampled[TARGET].copy()
+                eval_X_sampled = enriched_eval_sets[idx + 1][x_columns].copy()
+                eval_y_sampled = enriched_eval_sets[idx + 1][TARGET].copy()
+                enriched_eval_X = enriched_eval_sets[idx + 1].drop(columns=[TARGET, EVAL_SET_INDEX])
                 eval_set_sampled_dict[idx] = (eval_X_sampled, enriched_eval_X, eval_y_sampled)
 
         self.__cached_sampled_datasets = (X_sampled, y_sampled, enriched_X, eval_set_sampled_dict, search_keys)
@@ -1542,28 +1542,23 @@ class FeaturesEnricher(TransformerMixin):
                     n=Dataset.FIT_SAMPLE_WITH_EVAL_SET_ROWS, random_state=self.random_state
                 )
 
-            X_sampled = (
-                df_with_eval_set_index.query(f"{EVAL_SET_INDEX} == 0").copy().drop(columns=[EVAL_SET_INDEX, TARGET])
-            )
-            X_sampled, search_keys = self._extend_x(X_sampled, is_demo_dataset)
-            y_sampled = df_with_eval_set_index.query(f"{EVAL_SET_INDEX} == 0").copy()[TARGET]
+            df_extended, search_keys = self._extend_x(df_with_eval_set_index, is_demo_dataset)
+
             eval_set_sampled_dict = dict()
-            for idx in range(len(eval_set)):
-                eval_x_sampled = (
-                    df_with_eval_set_index[df_with_eval_set_index[EVAL_SET_INDEX] == (idx + 1)]
-                    .copy()
-                    .drop(columns=[EVAL_SET_INDEX, TARGET])
-                )
-                eval_x_sampled, _ = self._extend_x(eval_x_sampled, is_demo_dataset)
-                eval_y_sampled = df_with_eval_set_index[df_with_eval_set_index[EVAL_SET_INDEX] == (idx + 1)].copy()[
-                    TARGET
-                ]
-                eval_set_sampled_dict[idx] = (eval_x_sampled, eval_y_sampled)
 
-            df_with_eval_set_index.drop(columns=TARGET, inplace=True)
+            df_extended = self.__add_fit_system_record_id(df_extended, dict(), search_keys)
+
+            x_columns = [
+                c
+                for c in df_extended.columns
+                if c not in [TARGET, EVAL_SET_INDEX, DateTimeSearchKeyConverter.DATETIME_COL]
+            ]
+
+            tmp_target_name = "__target"
+            df_extended.rename(columns={SYSTEM_RECORD_ID: SORT_ID, TARGET: tmp_target_name}, inplace=True)
 
-            enriched = self.transform(
-                df_with_eval_set_index,
+            enriched_df = self.transform(
+                df_extended,
                 exclude_features_sources=exclude_features_sources,
                 silent_mode=True,
                 trace_id=trace_id,
@@ -1571,16 +1566,21 @@ class FeaturesEnricher(TransformerMixin):
                 progress_bar=progress_bar,
                 progress_callback=progress_callback,
             )
-            if enriched is None:
+            if enriched_df is None:
                 return None
 
-            enriched_X = enriched[enriched[EVAL_SET_INDEX] == 0].copy()
-            enriched_X.drop(columns=EVAL_SET_INDEX, inplace=True)
+            enriched_df.rename(columns={SORT_ID: SYSTEM_RECORD_ID, tmp_target_name: TARGET}, inplace=True)
+
+            enriched_Xy = enriched_df.query(f"{EVAL_SET_INDEX} == 0")
+            X_sampled = enriched_Xy[x_columns].copy()
+            y_sampled = enriched_Xy[TARGET].copy()
+            enriched_X = enriched_Xy.drop(columns=[TARGET, EVAL_SET_INDEX])
 
             for idx in range(len(eval_set)):
-                enriched_eval_x = enriched[enriched[EVAL_SET_INDEX] == (idx + 1)].copy()
-                enriched_eval_x.drop(columns=EVAL_SET_INDEX, inplace=True)
-                eval_x_sampled, eval_y_sampled = eval_set_sampled_dict[idx]
+                enriched_eval_xy = enriched_df.query(f"{EVAL_SET_INDEX} == {idx + 1}")
+                eval_x_sampled = enriched_eval_xy[x_columns].copy()
+                eval_y_sampled = enriched_eval_xy[TARGET].copy()
+                enriched_eval_x = enriched_eval_xy.drop(columns=[TARGET, EVAL_SET_INDEX])
                 eval_set_sampled_dict[idx] = (eval_x_sampled, enriched_eval_x, eval_y_sampled)
         else:
             self.logger.info("Transform without eval_set")
@@ -1592,14 +1592,17 @@ class FeaturesEnricher(TransformerMixin):
                 self.logger.info(f"Downsampling from {num_samples} to {Dataset.FIT_SAMPLE_ROWS}")
                 df = df.sample(n=Dataset.FIT_SAMPLE_ROWS, random_state=self.random_state)
 
-            X_sampled = df.copy().drop(columns=TARGET)
-            X_sampled, search_keys = self._extend_x(X_sampled, is_demo_dataset)
-            y_sampled = df.copy()[TARGET]
+            df_extended, search_keys = self._extend_x(df, is_demo_dataset)
+
+            df_extended = self.__add_fit_system_record_id(df_extended, dict(), search_keys)
 
-            df.drop(columns=TARGET, inplace=True)
+            x_columns = [c for c in df_extended if c not in [TARGET, DateTimeSearchKeyConverter.DATETIME_COL]]
 
-            enriched_X = self.transform(
-                df,
+            tmp_target_name = "__target"
+            df_extended.rename(columns={SYSTEM_RECORD_ID: SORT_ID, TARGET: tmp_target_name}, inplace=True)
+
+            enriched_Xy = self.transform(
+                df_extended,
                 exclude_features_sources=exclude_features_sources,
                 silent_mode=True,
                 trace_id=trace_id,
@@ -1607,9 +1610,15 @@ class FeaturesEnricher(TransformerMixin):
                 progress_bar=progress_bar,
                 progress_callback=progress_callback,
             )
-            if enriched_X is None:
+            if enriched_Xy is None:
                 return None
 
+            enriched_Xy.rename(columns={SORT_ID: SYSTEM_RECORD_ID, tmp_target_name: TARGET}, inplace=True)
+
+            X_sampled = enriched_Xy[x_columns].copy()  # check that all columns are present
+            y_sampled = enriched_Xy[TARGET].copy()
+            enriched_X = enriched_Xy.drop(columns=TARGET)
+
         self.__cached_sampled_datasets = (X_sampled, y_sampled, enriched_X, eval_set_sampled_dict, search_keys)
 
         return self.__mk_sampled_data_tuple(X_sampled, y_sampled, enriched_X, eval_set_sampled_dict, search_keys)
@@ -2153,7 +2162,7 @@ class FeaturesEnricher(TransformerMixin):
 
         dataset = Dataset(
             "tds_" + str(uuid.uuid4()),
-            df=df,  # type: ignore
+            df=df,
             model_task_type=model_task_type,
             date_format=self.date_format,
             random_state=self.random_state,
@@ -2323,7 +2332,7 @@ class FeaturesEnricher(TransformerMixin):
         elif (
             self.cv is None
             and model_task_type != ModelTaskType.REGRESSION
-            and self._get_group_columns(self.fit_search_keys)
+            and self._get_group_columns(df, self.fit_search_keys)
         ):
             msg = bundle.get("group_k_fold_in_classification")
             self.__override_cv(CVType.group_k_fold, msg, print_warning=self.cv is not None)
@@ -2508,6 +2517,23 @@ class FeaturesEnricher(TransformerMixin):
         Xy = pd.merge(Xy, enriched_X, left_index=True, right_index=True, how="inner", suffixes=("", "enriched"))
         return Xy[X.columns].copy(), Xy[TARGET].copy()
 
+    @staticmethod
+    def _sort_by_system_record_id(
+        X: pd.DataFrame, y: pd.Series, cv: Optional[CVType]
+    ) -> Tuple[pd.DataFrame, pd.Series]:
+        if cv not in [CVType.time_series, CVType.blocked_time_series]:
+            Xy = X.copy()
+            Xy[TARGET] = y
+            Xy = Xy.sort_values(by=SYSTEM_RECORD_ID).reset_index(drop=True)
+            X = Xy.drop(columns=TARGET)
+            y = Xy[TARGET].copy()
+
+        if DateTimeSearchKeyConverter.DATETIME_COL in X.columns:
+            X.drop(columns=DateTimeSearchKeyConverter.DATETIME_COL, inplace=True)
+
+        return X, y
+
+    # Deprecated
     @staticmethod
     def _sort_by_keys(
         X: pd.DataFrame, y: pd.Series, search_keys: Dict[str, SearchKey], cv: Optional[CVType]
@@ -2646,8 +2672,12 @@ class FeaturesEnricher(TransformerMixin):
                 return col
 
     @staticmethod
-    def _get_group_columns(search_keys: Dict[str, SearchKey]) -> List[str]:
-        return [col for col, t in search_keys.items() if t not in [SearchKey.DATE, SearchKey.DATETIME]]
+    def _get_group_columns(df: pd.DataFrame, search_keys: Dict[str, SearchKey]) -> List[str]:
+        return [
+            col
+            for col, t in search_keys.items()
+            if t not in [SearchKey.DATE, SearchKey.DATETIME] and df[col].nunique() > 1
+        ]
 
     @staticmethod
     def _get_email_column(search_keys: Dict[str, SearchKey]) -> Optional[str]:
@@ -2684,7 +2714,15 @@ class FeaturesEnricher(TransformerMixin):
                 date_column = self._get_date_column(search_keys)
             sort_columns = [date_column] if date_column is not None else []
 
-            other_search_keys = sorted([sk for sk in search_keys.keys() if sk != date_column and sk in df.columns])
+            other_search_keys = sorted(
+                [
+                    sk
+                    for sk, key_type in search_keys.items()
+                    if key_type not in [SearchKey.DATE, SearchKey.DATETIME]
+                    and sk in df.columns
+                    and df[sk].nunique() > 1  # don't use constant keys for hash
+                ]
+            )
 
             search_keys_hash = "search_keys_hash"
             if len(other_search_keys) > 0:
@@ -2751,6 +2789,7 @@ class FeaturesEnricher(TransformerMixin):
         X: Optional[pd.DataFrame] = None,
         is_transform=False,
         rows_to_drop: Optional[pd.DataFrame] = None,
+        drop_system_record_id=True,
     ) -> Tuple[pd.DataFrame, Dict[int, pd.DataFrame]]:
         if result_features is None:
             self.logger.error(f"result features not found by search_task_id: {self.get_search_id()}")

upgini/metadata.py

@@ -4,13 +4,14 @@ from typing import Dict, List, Optional, Set
 from pydantic import BaseModel
 
 SYSTEM_RECORD_ID = "system_record_id"
+SORT_ID = "sort_id"
 EVAL_SET_INDEX = "eval_set_index"
 TARGET = "target"
 COUNTRY = "country_iso_code"
 RENAMED_INDEX = "index_col"
 DEFAULT_INDEX = "index"
 ORIGINAL_INDEX = "original_index"
-SYSTEM_COLUMNS = {SYSTEM_RECORD_ID, EVAL_SET_INDEX, TARGET, COUNTRY}
+SYSTEM_COLUMNS = {SYSTEM_RECORD_ID, EVAL_SET_INDEX, TARGET, COUNTRY, SORT_ID}
 
 
 class FileColumnMeaningType(Enum):

upgini/metrics.py

@@ -318,6 +318,7 @@ class EstimatorWrapper:
         X: pd.DataFrame,
         scoring: Union[Callable, str, None] = None,
         cat_features: Optional[List[str]] = None,
+        text_features: Optional[List[str]] = None,
         add_params: Optional[Dict[str, Any]] = None,
         groups: Optional[List[str]] = None,
     ) -> "EstimatorWrapper":

{upgini-1.1.244a4.dist-info → upgini-1.1.244a6.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: upgini
-Version: 1.1.244a4
+Version: 1.1.244a6
 Summary: Intelligent data search & enrichment for Machine Learning
 Home-page: https://upgini.com/
 Author: Upgini Developers

{upgini-1.1.244a4.dist-info → upgini-1.1.244a6.dist-info}/RECORD

@@ -2,11 +2,11 @@ upgini/__init__.py,sha256=asENHgEVHQBIkV-e_0IhE_ZWqkCG6398U3ZLrNzAH6k,407
 upgini/ads.py,sha256=mre6xn44wcC_fg63iLT_kTh4mViZqR9AKRJZAtpQz8Y,2592
 upgini/dataset.py,sha256=1AEKFg2ooGnlBzmxX6sw-sDJdDoT8HfHWNYacwGHZGI,50023
 upgini/errors.py,sha256=pdzQl3MKuK52yvncxMWMRWeSIOGhUFzpQoszoRFBOk0,958
-upgini/features_enricher.py,sha256=HZfwDvZvq4IIkguc5MoRpbLeFr3ZKwyYejZfhBteQf4,162494
+upgini/features_enricher.py,sha256=r5k_djjmoVwTJ6Xg9-k9jZfAKTcr1h_wWkI64vkO34E,163572
 upgini/fingerprint.js,sha256=VygVIQlN1v4NGZfjHqtRogOw8zjTnnMNJg_f7M5iGQU,33442
 upgini/http.py,sha256=eSG4gOpmCGlXmB6KIPNzAG8tRZNUjyYpMeUeHw_2li4,42264
-upgini/metadata.py,sha256=FZ5CQluLLWrfrBVThSIes1SW6wcs7n50aNZwzYnHiF0,9584
-upgini/metrics.py,sha256=cb-IY4QyNWq2F8i3vCEDPN8wyblknjVf61JnZKsMY50,23777
+upgini/metadata.py,sha256=55t0uQI910tzTcnwxZCUL1413BhTiSm8oqiwp-94NyA,9613
+upgini/metrics.py,sha256=GFAZNu5V-xrALTiju0vMMdM-ckysnF23ogUtLJRNV08,23828
 upgini/search_task.py,sha256=5n4qGJmtu48s0-FHAtF3L5qVLMd1JVW3FJlM8dFbh-s,17063
 upgini/spinner.py,sha256=Dm1dQ5F_z_Ua2odLxZX7OypcOX9tSx_vE5MGaKtUmfw,1118
 upgini/version_validator.py,sha256=rDIncP6BEko4J2F2hUcMOtKm_vZbI4ICWcNcw8hrwM4,1400
@@ -54,8 +54,8 @@ upgini/utils/sklearn_ext.py,sha256=IMx2La70AXAggApVpT7sMEjWqVWon5AMZt4MARDsIMQ,4
 upgini/utils/target_utils.py,sha256=qyj-bGsIEl9X2Vc5gwXtsuRaocvId8bn46F7mZ9dy9A,1707
 upgini/utils/track_info.py,sha256=EPcJ13Jqa17_T0JjM37Ac9kWDz5Zk0GVsIZKutOb8aU,5207
 upgini/utils/warning_counter.py,sha256=dIWBB4dI5XRRJZudvIlqlIYKEiwLLPcXarsZuYRt338,227
-upgini-1.1.244a4.dist-info/LICENSE,sha256=5RRzgvdJUu3BUDfv4bzVU6FqKgwHlIay63pPCSmSgzw,1514
-upgini-1.1.244a4.dist-info/METADATA,sha256=V7zuNuOujnaTi46Qfpe1FMhpEyyrWEAsy1roqyp0ctc,48264
-upgini-1.1.244a4.dist-info/WHEEL,sha256=pkctZYzUS4AYVn6dJ-7367OJZivF2e8RA9b_ZBjif18,92
-upgini-1.1.244a4.dist-info/top_level.txt,sha256=OFhTGiDIWKl5gFI49qvWq1R9IKflPaE2PekcbDXDtx4,7
-upgini-1.1.244a4.dist-info/RECORD,,
+upgini-1.1.244a6.dist-info/LICENSE,sha256=5RRzgvdJUu3BUDfv4bzVU6FqKgwHlIay63pPCSmSgzw,1514
+upgini-1.1.244a6.dist-info/METADATA,sha256=s3qLjYZ-bw2atIM2spsBKOwRW-Hq6LEGRSp0eR8WOc0,48264
+upgini-1.1.244a6.dist-info/WHEEL,sha256=pkctZYzUS4AYVn6dJ-7367OJZivF2e8RA9b_ZBjif18,92
+upgini-1.1.244a6.dist-info/top_level.txt,sha256=OFhTGiDIWKl5gFI49qvWq1R9IKflPaE2PekcbDXDtx4,7
+upgini-1.1.244a6.dist-info/RECORD,,