snowflake-ml-python 1.4.0__py3-none-any.whl → 1.5.0__py3-none-any.whl

snowflake/ml/modeling/_internal/snowpark_implementations/snowpark_trainer.py

@@ -23,20 +23,26 @@ from snowflake.ml._internal.utils.temp_file_utils import (
     cleanup_temp_files,
     get_temp_file_path,
 )
+from snowflake.ml.modeling._internal.estimator_utils import handle_inference_result
 from snowflake.ml.modeling._internal.model_specifications import (
     ModelSpecifications,
     ModelSpecificationsBuilder,
 )
-from snowflake.snowpark import DataFrame, Session, exceptions as snowpark_exceptions
+from snowflake.snowpark import (
+    DataFrame,
+    Session,
+    exceptions as snowpark_exceptions,
+    functions as F,
+)
 from snowflake.snowpark._internal.utils import (
     TempObjectType,
     random_name_for_temp_object,
 )
-from snowflake.snowpark.functions import sproc
 from snowflake.snowpark.stored_procedure import StoredProcedure
 
 cp.register_pickle_by_value(inspect.getmodule(get_temp_file_path))
 cp.register_pickle_by_value(inspect.getmodule(identifier.get_inferred_name))
+cp.register_pickle_by_value(inspect.getmodule(handle_inference_result))
 
 _PROJECT = "ModelDevelopment"
 
@@ -122,7 +128,7 @@ class SnowparkModelTrainer:
             project=_PROJECT,
             subproject=self._subproject,
             function_name=telemetry.get_statement_params_full_func_name(inspect.currentframe(), self._class_name),
-            api_calls=[sproc],
+            api_calls=[F.sproc],
             custom_tags=dict([("autogen", True)]) if self._autogenerated else None,
         )
         # Put locally serialized transform on stage.
@@ -292,7 +298,7 @@ class SnowparkModelTrainer:
         """
         imports = model_spec.imports  # In order for the sproc to not resolve this reference in snowflake.ml
 
-        def fit_wrapper_function(
+        def fit_predict_wrapper_function(
             session: Session,
             sql_queries: List[str],
             stage_transform_file_name: str,
@@ -329,7 +335,7 @@ class SnowparkModelTrainer:
             with open(local_transform_file_path, mode="r+b") as local_transform_file_obj:
                 estimator = cp.load(local_transform_file_obj)
 
-            fit_predict_result = estimator.fit_predict(df[input_cols])
+            fit_predict_result = estimator.fit_predict(X=df[input_cols])
 
             local_result_file_name = get_temp_file_path()
 
@@ -349,8 +355,16 @@ class SnowparkModelTrainer:
                 fit_predict_result_pd = pd.DataFrame(data=fit_predict_result, columns=expected_output_cols_list)
             else:
                 df = df.copy()
-                fit_predict_result_pd = pd.DataFrame(data=fit_predict_result, columns=expected_output_cols_list)
-                fit_predict_result_pd = pd.concat([df, fit_predict_result_pd], axis=1)
+                # in case the output column name overlap with the input column names,
+                # remove the ones in input column names
+                remove_dataset_col_name_exist_in_output_col = list(set(df.columns) - set(expected_output_cols_list))
+                fit_predict_result_pd = pd.concat(
+                    [
+                        df[remove_dataset_col_name_exist_in_output_col],
+                        pd.DataFrame(data=fit_predict_result, columns=expected_output_cols_list),
+                    ],
+                    axis=1,
+                )
 
             # write into a temp table in sproc and load the table from outside
             session.write_pandas(
@@ -361,17 +375,150 @@ class SnowparkModelTrainer:
             # to pass debug information to the caller.
             return str(os.path.basename(local_result_file_name))
 
-        return fit_wrapper_function
+        return fit_predict_wrapper_function
+
+    def _build_fit_transform_wrapper_sproc(
+        self,
+        model_spec: ModelSpecifications,
+    ) -> Callable[
+        [
+            Session,
+            List[str],
+            str,
+            str,
+            List[str],
+            Optional[List[str]],
+            Optional[str],
+            Dict[str, str],
+            bool,
+            List[str],
+            str,
+        ],
+        str,
+    ]:
+        """
+        Constructs and returns a python stored procedure function to be used for training model.
+
+        Args:
+            model_spec: ModelSpecifications object that contains model specific information
+                like required imports, package dependencies, etc.
+
+        Returns:
+            A callable that can be registered as a stored procedure.
+        """
+        imports = model_spec.imports  # In order for the sproc to not resolve this reference in snowflake.ml
+
+        def fit_transform_wrapper_function(
+            session: Session,
+            sql_queries: List[str],
+            stage_transform_file_name: str,
+            stage_result_file_name: str,
+            input_cols: List[str],
+            label_cols: Optional[List[str]],
+            sample_weight_col: Optional[str],
+            statement_params: Dict[str, str],
+            drop_input_cols: bool,
+            expected_output_cols_list: List[str],
+            fit_transform_result_name: str,
+        ) -> str:
+            import os
+
+            import cloudpickle as cp
+            import pandas as pd
+
+            for import_name in imports:
+                importlib.import_module(import_name)
+
+            # Execute snowpark queries and obtain the results as pandas dataframe
+            # NB: this implies that the result data must fit into memory.
+            for query in sql_queries[:-1]:
+                _ = session.sql(query).collect(statement_params=statement_params)
+            sp_df = session.sql(sql_queries[-1])
+            df: pd.DataFrame = sp_df.to_pandas(statement_params=statement_params)
+            df.columns = sp_df.columns
+
+            local_transform_file_name = get_temp_file_path()
+
+            session.file.get(stage_transform_file_name, local_transform_file_name, statement_params=statement_params)
+
+            local_transform_file_path = os.path.join(
+                local_transform_file_name, os.listdir(local_transform_file_name)[0]
+            )
+            with open(local_transform_file_path, mode="r+b") as local_transform_file_obj:
+                estimator = cp.load(local_transform_file_obj)
+
+            argspec = inspect.getfullargspec(estimator.fit)
+            args = {"X": df[input_cols]}
+            if label_cols:
+                label_arg_name = "Y" if "Y" in argspec.args else "y"
+                args[label_arg_name] = df[label_cols].squeeze()
+
+            if sample_weight_col is not None and "sample_weight" in argspec.args:
+                args["sample_weight"] = df[sample_weight_col].squeeze()
+
+            fit_transform_result = estimator.fit_transform(**args)
+
+            local_result_file_name = get_temp_file_path()
+
+            with open(local_result_file_name, mode="w+b") as local_result_file_obj:
+                cp.dump(estimator, local_result_file_obj)
+
+            session.file.put(
+                local_result_file_name,
+                stage_result_file_name,
+                auto_compress=False,
+                overwrite=True,
+                statement_params=statement_params,
+            )
+
+            transformed_numpy_array, output_cols = handle_inference_result(
+                inference_res=fit_transform_result,
+                output_cols=expected_output_cols_list,
+                inference_method="fit_transform",
+                within_udf=True,
+            )
+
+            if len(transformed_numpy_array.shape) > 1:
+                if transformed_numpy_array.shape[1] != len(output_cols):
+                    series = pd.Series(transformed_numpy_array.tolist())
+                    transformed_pandas_df = pd.DataFrame(series, columns=output_cols)
+                else:
+                    transformed_pandas_df = pd.DataFrame(transformed_numpy_array.tolist(), columns=output_cols)
+            else:
+                transformed_pandas_df = pd.DataFrame(transformed_numpy_array, columns=output_cols)
+
+            # store the transform output
+            if not drop_input_cols:
+                df = df.copy()
+                # in case the output column name overlap with the input column names,
+                # remove the ones in input column names
+                remove_dataset_col_name_exist_in_output_col = list(set(df.columns) - set(output_cols))
+                transformed_pandas_df = pd.concat(
+                    [df[remove_dataset_col_name_exist_in_output_col], transformed_pandas_df], axis=1
+                )
+
+            # write into a temp table in sproc and load the table from outside
+            session.write_pandas(
+                transformed_pandas_df,
+                fit_transform_result_name,
+                auto_create_table=True,
+                table_type="temp",
+                quote_identifiers=False,
+            )
+
+            return str(os.path.basename(local_result_file_name))
+
+        return fit_transform_wrapper_function
 
     def _get_fit_predict_wrapper_sproc(self, statement_params: Dict[str, str]) -> StoredProcedure:
         # If the sproc already exists, don't register.
-        if not hasattr(self.session, "_FIT_PRE_WRAPPER_SPROCS"):
-            self.session._FIT_PRE_WRAPPER_SPROCS: Dict[str, StoredProcedure] = {}  # type: ignore[attr-defined, misc]
+        if not hasattr(self.session, "_FIT_WRAPPER_SPROCS"):
+            self.session._FIT_WRAPPER_SPROCS: Dict[str, StoredProcedure] = {}  # type: ignore[attr-defined, misc]
 
         model_spec = ModelSpecificationsBuilder.build(model=self.estimator)
-        fit_predict_sproc_key = model_spec.__class__.__name__
-        if fit_predict_sproc_key in self.session._FIT_PRE_WRAPPER_SPROCS:  # type: ignore[attr-defined]
-            fit_sproc: StoredProcedure = self.session._FIT_PRE_WRAPPER_SPROCS[  # type: ignore[attr-defined]
+        fit_predict_sproc_key = model_spec.__class__.__name__ + "_fit_predict"
+        if fit_predict_sproc_key in self.session._FIT_WRAPPER_SPROCS:  # type: ignore[attr-defined]
+            fit_sproc: StoredProcedure = self.session._FIT_WRAPPER_SPROCS[  # type: ignore[attr-defined]
                 fit_predict_sproc_key
             ]
             return fit_sproc
@@ -392,12 +539,47 @@ class SnowparkModelTrainer:
             statement_params=statement_params,
         )
 
-        self.session._FIT_PRE_WRAPPER_SPROCS[  # type: ignore[attr-defined]
+        self.session._FIT_WRAPPER_SPROCS[  # type: ignore[attr-defined]
             fit_predict_sproc_key
         ] = fit_predict_wrapper_sproc
 
         return fit_predict_wrapper_sproc
 
+    def _get_fit_transform_wrapper_sproc(self, statement_params: Dict[str, str]) -> StoredProcedure:
+        # If the sproc already exists, don't register.
+        if not hasattr(self.session, "_FIT_WRAPPER_SPROCS"):
+            self.session._FIT_WRAPPER_SPROCS: Dict[str, StoredProcedure] = {}  # type: ignore[attr-defined, misc]
+
+        model_spec = ModelSpecificationsBuilder.build(model=self.estimator)
+        fit_transform_sproc_key = model_spec.__class__.__name__ + "_fit_transform"
+        if fit_transform_sproc_key in self.session._FIT_WRAPPER_SPROCS:  # type: ignore[attr-defined]
+            fit_sproc: StoredProcedure = self.session._FIT_WRAPPER_SPROCS[  # type: ignore[attr-defined]
+                fit_transform_sproc_key
+            ]
+            return fit_sproc
+
+        fit_transform_sproc_name = random_name_for_temp_object(TempObjectType.PROCEDURE)
+
+        relaxed_dependencies = pkg_version_utils.get_valid_pkg_versions_supported_in_snowflake_conda_channel(
+            pkg_versions=model_spec.pkgDependencies, session=self.session
+        )
+
+        fit_transform_wrapper_sproc = self.session.sproc.register(
+            func=self._build_fit_transform_wrapper_sproc(model_spec=model_spec),
+            is_permanent=False,
+            name=fit_transform_sproc_name,
+            packages=["snowflake-snowpark-python"] + relaxed_dependencies,  # type: ignore[arg-type]
+            replace=True,
+            session=self.session,
+            statement_params=statement_params,
+        )
+
+        self.session._FIT_WRAPPER_SPROCS[  # type: ignore[attr-defined]
+            fit_transform_sproc_key
+        ] = fit_transform_wrapper_sproc
+
+        return fit_transform_wrapper_sproc
+
     def train(self) -> object:
         """
         Trains the model by pushing down the compute into Snowflake using stored procedures.
@@ -498,10 +680,10 @@ class SnowparkModelTrainer:
             custom_tags=dict([("autogen", True)]) if self._autogenerated else None,
         )
 
-        fit_wrapper_sproc = self._get_fit_predict_wrapper_sproc(statement_params=statement_params)
+        fit_predict_wrapper_sproc = self._get_fit_predict_wrapper_sproc(statement_params=statement_params)
         fit_predict_result_name = random_name_for_temp_object(TempObjectType.TABLE)
 
-        sproc_export_file_name: str = fit_wrapper_sproc(
+        sproc_export_file_name: str = fit_predict_wrapper_sproc(
             self.session,
             queries,
             stage_transform_file_name,
@@ -521,3 +703,66 @@ class SnowparkModelTrainer:
         )
 
         return output_result_sp, fitted_estimator
+
+    def train_fit_transform(
+        self,
+        expected_output_cols_list: List[str],
+        drop_input_cols: Optional[bool] = False,
+    ) -> Tuple[Union[DataFrame, pd.DataFrame], object]:
+        """Trains the model by pushing down the compute into Snowflake using stored procedures.
+        This API is different from fit itself because it would also provide the transform
+        output.
+
+        Args:
+            expected_output_cols_list (List[str]): The output columns
+                name as a list. Defaults to None.
+            drop_input_cols (Optional[bool]): Boolean to determine whether to
+                drop the input columns from the output dataset.
+
+        Returns:
+            Tuple[Union[DataFrame, pd.DataFrame], object]: [transformed dataset, estimator]
+        """
+        dataset = snowpark_dataframe_utils.cast_snowpark_dataframe_column_types(self.dataset)
+
+        # Extract query that generated the dataframe. We will need to pass it to the fit procedure.
+        queries = dataset.queries["queries"]
+
+        transform_stage_name = self._create_temp_stage()
+        (stage_transform_file_name, stage_result_file_name) = self._upload_model_to_stage(
+            stage_name=transform_stage_name
+        )
+
+        # Call fit sproc
+        statement_params = telemetry.get_function_usage_statement_params(
+            project=_PROJECT,
+            subproject=self._subproject,
+            function_name=telemetry.get_statement_params_full_func_name(inspect.currentframe(), self._class_name),
+            api_calls=[Session.call],
+            custom_tags=dict([("autogen", True)]) if self._autogenerated else None,
+        )
+
+        fit_transform_wrapper_sproc = self._get_fit_transform_wrapper_sproc(statement_params=statement_params)
+        fit_transform_result_name = random_name_for_temp_object(TempObjectType.TABLE)
+
+        sproc_export_file_name: str = fit_transform_wrapper_sproc(
+            self.session,
+            queries,
+            stage_transform_file_name,
+            stage_result_file_name,
+            self.input_cols,
+            self.label_cols,
+            self.sample_weight_col,
+            statement_params,
+            drop_input_cols,
+            expected_output_cols_list,
+            fit_transform_result_name,
+        )
+
+        output_result_sp = self.session.table(fit_transform_result_name)
+        fitted_estimator = self._fetch_model_from_stage(
+            dir_path=stage_result_file_name,
+            file_name=sproc_export_file_name,
+            statement_params=statement_params,
+        )
+
+        return output_result_sp, fitted_estimator