Release 1.45.0

CHANGELOG.md

@@ -1,6 +1,78 @@
 # Release History
 
-## 1.44.0
+## 1.45.0
+
+### New Features
+
+* Feature Store: `FeatureView` now supports an `initialization_warehouse` that is used for the initial build and any
+  subsequent reinitializations of the backing dynamic table (a full scan of the source data), while `warehouse`
+  continues to drive the lighter incremental refreshes. This mirrors the dynamic table `INITIALIZATION_WAREHOUSE`
+  knob, lets you pair a larger warehouse for initialization with a smaller one for steady-state refresh, and is also
+  used for the one-time backfill of streaming feature views. It can be set at registration, changed via
+  `update_feature_view(initialization_warehouse=...)`, and is surfaced by `list_feature_views(verbose=True)`.
+
+```python
+draft_fv = FeatureView(
+    name="F_TRIP",
+    entities=[entity],
+    feature_df=feature_df,
+    refresh_freq="1d",
+    warehouse="SMALL_WH",            # incremental refreshes
+    initialization_warehouse="LARGE_WH",  # initial build / reinitialization
+)
+fv = fs.register_feature_view(draft_fv, version="1.0")
+```
+
+* Registry: LLM models deployed with the OpenAI chat signatures now support structured outputs
+  through an optional `response_format` param matching the OpenAI Chat Completions API
+  (`{"type": "json_schema", "json_schema": {"name": "...", "schema": {...}}}`), letting callers
+  constrain model output to a JSON Schema.
+
+```python
+from pydantic import BaseModel
+import pandas as pd
+
+class CityCountry(BaseModel):
+    city: str
+    country: str
+
+response_format = {
+    "type": "json_schema",
+    "json_schema": {
+        "name": "city_country",
+        "schema": CityCountry.model_json_schema(),
+    },
+}
+
+x_df = pd.DataFrame.from_records(
+    [
+        {
+            "messages": [
+                {
+                    "role": "user",
+                    "content": [
+                        {"type": "text", "text": "What is the capital of France?"},
+                    ],
+                },
+            ],
+        }
+    ]
+)
+
+mv.run(
+    X=x_df,
+    params={"response_format": response_format},
+    service_name=...,
+)
+```
+
+### Bug Fixes
+
+### Behavior Changes
+
+### Deprecations
+
+## 1.44.0 (2026-06-23)
 
 ### New Features
 

MODULE.bazel

@@ -73,6 +73,7 @@ use_repo(
     "com_github_apache_arrow_go_v18",
     "com_github_buger_jsonparser",
     "com_github_caarlos0_env_v11",
+    "com_github_goccy_go_json",
     "com_github_snowflakedb_gosnowflake",
     "in_gopkg_natefinch_lumberjack_v2",
     "in_gopkg_yaml_v2",

bazel/environments/conda-optional-dependency-llm.yml

@@ -6,7 +6,7 @@ channels:
   - https://repo.anaconda.com/pkgs/snowflake
   - nodefaults
 dependencies:
-  - mlflow>=2.16.0, <4
+  - mlflow>=2.16.0, <3
   - pytorch>=2.0.1,<3
   - sentence-transformers>=2.7.0,<6
   - sentencepiece>=0.1.95,<0.3

bazel/environments/conda-optional-dependency-ml.yml

@@ -9,6 +9,6 @@ dependencies:
   - altair>=5,<7
   - catboost>=1.2.0, <2
   - lightgbm>=4.1.0, <5
-  - mlflow>=2.16.0, <4
+  - mlflow>=2.16.0, <3
   - prophet>=1.1.0, <2
   - streamlit>=1.30.0,<2

bazel/environments/conda-optional-dependency-torch.yml

@@ -6,7 +6,7 @@ channels:
   - https://repo.anaconda.com/pkgs/snowflake
   - nodefaults
 dependencies:
-  - mlflow>=2.16.0, <4
+  - mlflow>=2.16.0, <3
   - pytorch>=2.0.1,<3
   - sentence-transformers>=2.7.0,<6
   - sentencepiece>=0.1.95,<0.3

ci/conda_recipe/meta.yaml

@@ -17,7 +17,7 @@ build:
   noarch: python
 package:
   name: snowflake-ml-python
-  version: 1.44.0
+  version: 1.45.0
 requirements:
   build:
     - python
@@ -58,7 +58,7 @@ requirements:
     - catboost>=1.2.0, <2
     - keras>=2.0.0,<4
     - lightgbm>=4.1.0, <5
-    - mlflow>=2.16.0, <4
+    - mlflow>=2.16.0, <3
     - prophet>=1.1.0, <2
     - pytorch>=2.0.1,<3
     - sentence-transformers>=2.7.0,<6

codegen/sklearn_wrapper_generator.py

@@ -292,6 +292,7 @@ def _is_deterministic(class_object: tuple[str, type]) -> bool:
             WrapperGeneratorFactory._is_class_of_type(class_object[1], "LinearDiscriminantAnalysis")
             or WrapperGeneratorFactory._is_class_of_type(class_object[1], "BernoulliRBM")
             or WrapperGeneratorFactory._is_class_of_type(class_object[1], "TSNE")
+            or WrapperGeneratorFactory._is_class_of_type(class_object[1], "MDS")
         )
 
     @staticmethod

codegen/transformer_autogen_test_template.py_template

@@ -85,16 +85,6 @@ class {transform.test_class_name}(TestCase):
             fit_with_sproc: bool = True,
             inference_with_udf: bool = True
     ) -> None:
-        # sklearn _parallel_pairwise bug (https://github.com/scikit-learn/scikit-learn/issues/33877):
-        # when effective_n_jobs(None) > 1 (warehouse environment), Y is chunked but Y_norm_squared
-        # is passed through unchanged, causing a dimension mismatch in Birch._predict.
-        # Fixed in https://github.com/scikit-learn/scikit-learn/pull/33876 (sklearn 1.9.0).
-        # TODO(SNOW-952252): Re-enable when test env sklearn is bumped to >= 1.9.0.
-        if "{transform.original_class_name}" == "Birch" and (fit_with_sproc or inference_with_udf):  # type: ignore[comparison-overlap,unused-ignore]
-            import joblib
-            if joblib.effective_n_jobs(None) > 1:
-                self.skipTest("sklearn _parallel_pairwise bug causes dimension mismatch when n_jobs > 1")
-
         input_df_pandas = {transform.test_dataset_func}(as_frame=True).frame
         cols = [inflection.parameterize(c, "_").upper() for c in input_df_pandas.columns if not c.startswith("target")]
         cols_half_1, cols_half_2 = cols[:int(len(cols)/2)], cols[int(len(cols)/2)+1:]

requirements.yml

@@ -150,7 +150,7 @@
   version_requirements: '>=0.46.0,<1'
 - name: mlflow
   dev_version: 2.16.2
-  version_requirements: '>=2.16.0, <4'
+  version_requirements: '>=2.16.0, <3'
   requirements_extra_tags:
     - mlflow
     - llm

snowflake/ml/_internal/file_utils.py

@@ -313,6 +313,38 @@ def _resolve_stage_dir_path(
         return str((stage_path / relative_path).parent)
 
 
+def upload_file_to_stage(
+    session: snowpark.Session,
+    local_file_path: Union[str, pathlib.Path],
+    stage_dir_path: str,
+    *,
+    statement_params: Optional[dict[str, Any]] = None,
+) -> None:
+    """Upload a single local file to a stage directory.
+
+    Args:
+        session: Snowpark Session.
+        local_file_path: Local path of the file to upload.
+        stage_dir_path: Destination directory path in the stage.
+        statement_params: Statement Params.
+    """
+    import retrying
+
+    file_operation = snowpark.FileOperation(session=session)
+    retrying.retry(
+        retry_on_exception=_retry_on_sql_error,
+        stop_max_attempt_number=5,
+        wait_exponential_multiplier=100,
+        wait_exponential_max=10000,
+    )(file_operation.put)(
+        str(local_file_path),
+        str(stage_dir_path),
+        auto_compress=False,
+        overwrite=False,
+        statement_params=statement_params,
+    )
+
+
 def upload_directory_to_stage(
     session: snowpark.Session,
     local_path: pathlib.Path,
@@ -328,27 +360,16 @@ def upload_directory_to_stage(
         stage_path: Base path in the stage.
         statement_params: Statement Params.
     """
-    import retrying
-
-    file_operation = snowpark.FileOperation(session=session)
-
     for root, _, filenames in os.walk(local_path):
         root_path = pathlib.Path(root)
         for filename in filenames:
             local_file_path = root_path / filename
             relative_path = pathlib.PurePosixPath(local_file_path.relative_to(local_path).as_posix())
             stage_dir_path = _resolve_stage_dir_path(stage_path, relative_path)
-
-            retrying.retry(
-                retry_on_exception=_retry_on_sql_error,
-                stop_max_attempt_number=5,
-                wait_exponential_multiplier=100,
-                wait_exponential_max=10000,
-            )(file_operation.put)(
-                str(local_file_path),
-                str(stage_dir_path),
-                auto_compress=False,
-                overwrite=False,
+            upload_file_to_stage(
+                session,
+                local_file_path,
+                stage_dir_path,
                 statement_params=statement_params,
             )
 

snowflake/ml/experiment/_client/experiment_tracking_sql_client.py

@@ -118,11 +118,20 @@ def drop_experiment(
         ).validate()
 
     @telemetry.send_api_usage_telemetry(project=telemetry.TelemetryProject.EXPERIMENT_TRACKING.value)
-    def add_run(self, *, experiment_name: sql_identifier.SqlIdentifier, run_name: sql_identifier.SqlIdentifier) -> None:
+    def add_run(
+        self,
+        *,
+        experiment_name: sql_identifier.SqlIdentifier,
+        run_name: sql_identifier.SqlIdentifier,
+        source_info: Optional[str] = None,
+    ) -> None:
         experiment_fqn = self.fully_qualified_object_name(self._database_name, self._schema_name, experiment_name)
-        query_result_checker.SqlResultValidator(
-            self._session, f"ALTER EXPERIMENT {experiment_fqn} ADD RUN {run_name}"
-        ).has_dimensions(expected_rows=1, expected_cols=1).validate()
+        query = f"ALTER EXPERIMENT {experiment_fqn} ADD RUN {run_name}"
+        if source_info:
+            query += f" WITH (SOURCE_INFO = $${source_info}$$)"
+        query_result_checker.SqlResultValidator(self._session, query).has_dimensions(
+            expected_rows=1, expected_cols=1
+        ).validate()
 
     @telemetry.send_api_usage_telemetry(project=telemetry.TelemetryProject.EXPERIMENT_TRACKING.value)
     def commit_run(

snowflake/ml/experiment/_client/experiment_tracking_sql_client_test.py

@@ -169,6 +169,38 @@ def test_add_run(self) -> None:
         )
         self.client.add_run(experiment_name=experiment_name, run_name=run_name)
 
+    def test_add_run_with_source_info(self) -> None:
+        # The caller passes an already-serialized JSON string; the client embeds it verbatim.
+        experiment_name = sql_identifier.SqlIdentifier("TEST_EXPERIMENT")
+        run_name = sql_identifier.SqlIdentifier("TEST_RUN")
+
+        payload = '{"entry_point": "train/main.py", "git": {"commit_hash": "abc"}}'
+        self.m_session.add_mock_sql(
+            "ALTER EXPERIMENT TEST_DB.TEST_SCHEMA.TEST_EXPERIMENT ADD RUN TEST_RUN"
+            f" WITH (SOURCE_INFO = $${payload}$$)",
+            self._create_mock_df(),
+        )
+        self.client.add_run(experiment_name=experiment_name, run_name=run_name, source_info=payload)
+
+    def test_add_run_with_none_source_info_omits_clause(self) -> None:
+        experiment_name = sql_identifier.SqlIdentifier("TEST_EXPERIMENT")
+        run_name = sql_identifier.SqlIdentifier("TEST_RUN")
+
+        self.m_session.add_mock_sql(
+            "ALTER EXPERIMENT TEST_DB.TEST_SCHEMA.TEST_EXPERIMENT ADD RUN TEST_RUN", self._create_mock_df()
+        )
+        self.client.add_run(experiment_name=experiment_name, run_name=run_name, source_info=None)
+
+    def test_add_run_with_empty_source_info_omits_clause(self) -> None:
+        # An empty string is falsy and must be treated like "nothing to send".
+        experiment_name = sql_identifier.SqlIdentifier("TEST_EXPERIMENT")
+        run_name = sql_identifier.SqlIdentifier("TEST_RUN")
+
+        self.m_session.add_mock_sql(
+            "ALTER EXPERIMENT TEST_DB.TEST_SCHEMA.TEST_EXPERIMENT ADD RUN TEST_RUN", self._create_mock_df()
+        )
+        self.client.add_run(experiment_name=experiment_name, run_name=run_name, source_info="")
+
     def test_commit_run(self) -> None:
         # Test committing a run
         experiment_name = sql_identifier.SqlIdentifier("TEST_EXPERIMENT")

snowflake/ml/feature_store/BUILD.bazel

@@ -274,6 +274,18 @@ py_test(
     ],
 )
 
+py_test(
+    name = "feature_store_initialization_warehouse_test",
+    srcs = ["feature_store_initialization_warehouse_test.py"],
+    tags = [
+        "feature:feature_store",
+        "short_regress",
+    ],
+    deps = [
+        ":feature_store_lib",
+    ],
+)
+
 py_test(
     name = "realtime_config_test",
     srcs = ["realtime_config_test.py"],