maxframe 0.1.0b4__cp39-cp39-win32.whl → 1.0.0rc1__cp39-cp39-win32.whl

maxframe/learn/contrib/xgboost/train.py

@@ -0,0 +1,121 @@
+# Copyright 1999-2024 Alibaba Group Holding Ltd.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import logging
+from collections import OrderedDict
+
+from .... import opcodes as OperandDef
+from ....core import OutputType
+from ....core.operator.base import Operator
+from ....core.operator.core import TileableOperatorMixin
+from ....serialization.serializables import (
+    AnyField,
+    BoolField,
+    DictField,
+    FieldTypes,
+    FunctionField,
+    Int64Field,
+    KeyField,
+    ListField,
+)
+from .dmatrix import ToDMatrix, to_dmatrix
+
+logger = logging.getLogger(__name__)
+
+
+def _on_serialize_evals(evals_val):
+    if evals_val is None:
+        return None
+    return [list(x) for x in evals_val]
+
+
+class XGBTrain(Operator, TileableOperatorMixin):
+    _op_type_ = OperandDef.XGBOOST_TRAIN
+
+    params = DictField("params", key_type=FieldTypes.string, default=None)
+    dtrain = KeyField("dtrain", default=None)
+    evals = ListField("evals", on_serialize=_on_serialize_evals, default=None)
+    obj = FunctionField("obj", default=None)
+    feval = FunctionField("obj", default=None)
+    maximize = BoolField("maximize", default=None)
+    early_stopping_rounds = Int64Field("early_stopping_rounds", default=None)
+    verbose_eval = AnyField("verbose_eval", default=None)
+    xgb_model = AnyField("xgb_model", default=None)
+    callbacks = ListField(
+        "callbacks", field_type=FunctionField.field_type, default=None
+    )
+    custom_metric = FunctionField("custom_metric", default=None)
+    num_boost_round = Int64Field("num_boost_round", default=10)
+    num_class = Int64Field("num_class", default=None)
+
+    # Store evals_result in local to store the remote evals_result
+    evals_result: dict = None
+
+    def __init__(self, gpu=None, **kw):
+        super().__init__(gpu=gpu, **kw)
+        if self.output_types is None:
+            self.output_types = [OutputType.object]
+
+    def _set_inputs(self, inputs):
+        super()._set_inputs(inputs)
+        self.dtrain = self._inputs[0]
+        rest = self._inputs[1:]
+        if self.evals is not None:
+            evals_dict = OrderedDict(self.evals)
+            new_evals_dict = OrderedDict()
+            for new_key, val in zip(rest, evals_dict.values()):
+                new_evals_dict[new_key] = val
+            self.evals = list(new_evals_dict.items())
+
+    def __call__(self):
+        inputs = [self.dtrain]
+        if self.evals is not None:
+            inputs.extend(e[0] for e in self.evals)
+        return self.new_tileable(inputs)
+
+
+def train(params, dtrain, evals=None, evals_result=None, num_class=None, **kwargs):
+    """
+    Train XGBoost model in Mars manner.
+
+    Parameters
+    ----------
+    Parameters are the same as `xgboost.train`.
+
+    Returns
+    -------
+    results: Booster
+    """
+
+    evals_result = evals_result or dict()
+    evals = None or ()
+
+    processed_evals = []
+    if evals:
+        for eval_dmatrix, name in evals:
+            if not isinstance(name, str):
+                raise TypeError("evals must a list of pairs (DMatrix, string)")
+            if hasattr(eval_dmatrix, "op") and isinstance(eval_dmatrix.op, ToDMatrix):
+                processed_evals.append((eval_dmatrix, name))
+            else:
+                processed_evals.append((to_dmatrix(eval_dmatrix), name))
+
+    return XGBTrain(
+        params=params,
+        dtrain=dtrain,
+        evals=processed_evals,
+        evals_result=evals_result,
+        num_class=num_class,
+        **kwargs
+    )()

maxframe/learn/utils/__init__.py

@@ -0,0 +1,15 @@
+# Copyright 1999-2024 Alibaba Group Holding Ltd.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from .core import convert_to_tensor_or_dataframe

maxframe/learn/utils/core.py

@@ -0,0 +1,29 @@
+# Copyright 1999-2024 Alibaba Group Holding Ltd.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import pandas as pd
+
+from ...dataframe import DataFrame, Series
+from ...dataframe.core import DATAFRAME_TYPE, SERIES_TYPE
+from ...tensor import tensor as astensor
+
+
+def convert_to_tensor_or_dataframe(item):
+    if isinstance(item, (DATAFRAME_TYPE, pd.DataFrame)):
+        item = DataFrame(item)
+    elif isinstance(item, (SERIES_TYPE, pd.Series)):
+        item = Series(item)
+    else:
+        item = astensor(item)
+    return item

maxframe/lib/mmh3.pyi

@@ -0,0 +1,43 @@
+# Copyright 1999-2024 Alibaba Group Holding Ltd.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import Tuple
+
+def hash(key, seed=0, signed=True) -> int:
+    """
+    Return a 32 bit integer.
+    """
+
+def hash_from_buffer(key, seed=0, signed=True) -> int:
+    """
+    Return a 32 bit integer. Designed for large memory-views such as numpy arrays.
+    """
+
+def hash64(key, seed=0, x64arch=True, signed=True) -> Tuple[int, int]:
+    """
+    Return a tuple of two 64 bit integers for a string. Optimized for
+    the x64 bit architecture when x64arch=True, otherwise for x86.
+    """
+
+def hash128(key, seed=0, x64arch=True, signed=False) -> int:
+    """
+    Return a 128 bit long integer. Optimized for the x64 bit architecture
+    when x64arch=True, otherwise for x86.
+    """
+
+def hash_bytes(key, seed=0, x64arch=True) -> bytes:
+    """
+    Return a 128 bit hash value as bytes for a string. Optimized for the
+    x64 bit architecture when x64arch=True, otherwise for the x86.
+    """

maxframe/lib/wrapped_pickle.py

@@ -120,7 +120,8 @@ class _UnpickleSwitch:
             @functools.wraps(func)
             async def wrapped(*args, **kwargs):
                 with _UnpickleSwitch(forbidden=self._forbidden):
-                    return await func(*args, **kwargs)
+                    ret = await func(*args, **kwargs)
+                return ret
 
         else:
 

maxframe/odpsio/arrow.py

@@ -17,10 +17,9 @@ from typing import Any, Tuple, Union
 import pandas as pd
 import pyarrow as pa
 
-import maxframe.tensor as mt
-
 from ..core import OutputType
 from ..protocol import DataFrameTableMeta
+from ..tensor.core import TENSOR_TYPE
 from ..typing_ import ArrowTableType, PandasObjectTypes
 from .schema import build_dataframe_table_meta
 
@@ -83,7 +82,7 @@ def pandas_to_arrow(
         df = df.to_frame(name=names[0] if len(names) == 1 else names)
     elif table_meta.type == OutputType.scalar:
         names = ["_idx_0"]
-        if isinstance(df, mt.Tensor):
+        if isinstance(df, TENSOR_TYPE):
             df = pd.DataFrame([], columns=names).astype({names[0]: df.dtype})
         else:
             df = pd.DataFrame([[df]], columns=names)

maxframe/odpsio/tableio.py

@@ -183,6 +183,28 @@ class HaloTableIO(MCTableIO):
             for pt in partitions
         ]
 
+    def get_table_record_count(
+        self, full_table_name: str, partitions: PartitionsType = None
+    ):
+        from odps.apis.storage_api import SplitOptions, TableBatchScanRequest
+
+        table = self._odps.get_table(full_table_name)
+        client = StorageApiArrowClient(
+            self._odps, table, rest_endpoint=self._storage_api_endpoint
+        )
+
+        split_option = SplitOptions.SplitMode.SIZE
+
+        scan_kw = {
+            "required_partitions": self._convert_partitions(partitions),
+            "split_options": SplitOptions.get_default_options(split_option),
+        }
+
+        # todo add more options for partition column handling
+        req = TableBatchScanRequest(**scan_kw)
+        resp = client.create_read_session(req)
+        return resp.record_count
+
     @contextmanager
     def open_reader(
         self,

maxframe/odpsio/tests/test_schema.py

@@ -30,20 +30,23 @@ from ..schema import (
 )
 
 
-def _wrap_maxframe_obj(obj, wrap=True):
-    if not wrap:
+def _wrap_maxframe_obj(obj, wrap="no"):
+    if wrap == "no":
         return obj
     if isinstance(obj, pd.DataFrame):
-        return md.DataFrame(obj)
+        obj = md.DataFrame(obj)
     elif isinstance(obj, pd.Series):
-        return md.Series(obj)
+        obj = md.Series(obj)
     elif isinstance(obj, pd.Index):
-        return md.Index(obj)
+        obj = md.Index(obj)
     else:
-        return mt.scalar(obj)
+        obj = mt.scalar(obj)
+    if wrap == "data":
+        return obj.data
+    return obj
 
 
-@pytest.mark.parametrize("wrap_obj", [False, True])
+@pytest.mark.parametrize("wrap_obj", ["no", "yes", "data"])
 def test_pandas_to_odps_schema_dataframe(wrap_obj):
     data = pd.DataFrame(np.random.rand(100, 5), columns=list("ABCDE"))
 
@@ -94,7 +97,7 @@ def test_pandas_to_odps_schema_dataframe(wrap_obj):
     assert meta.pd_index_level_names == [None, None]
 
 
-@pytest.mark.parametrize("wrap_obj", [False, True])
+@pytest.mark.parametrize("wrap_obj", ["no", "yes", "data"])
 def test_pandas_to_odps_schema_series(wrap_obj):
     data = pd.Series(np.random.rand(100))
 
@@ -135,7 +138,7 @@ def test_pandas_to_odps_schema_series(wrap_obj):
     assert meta.pd_index_level_names == ["c1", "c2"]
 
 
-@pytest.mark.parametrize("wrap_obj", [False, True])
+@pytest.mark.parametrize("wrap_obj", ["no", "yes", "data"])
 def test_pandas_to_odps_schema_index(wrap_obj):
     data = pd.Index(np.random.randint(0, 100, 100))
 
@@ -167,11 +170,13 @@ def test_pandas_to_odps_schema_index(wrap_obj):
     assert meta.pd_index_level_names == ["c1", "c2"]
 
 
-@pytest.mark.parametrize("wrap_obj", [False, True])
+@pytest.mark.parametrize("wrap_obj", ["no", "yes", "data"])
 def test_pandas_to_odps_schema_scalar(wrap_obj):
     data = 1234.56
 
     test_scalar = _wrap_maxframe_obj(data, wrap=wrap_obj)
+    if wrap_obj != "no":
+        test_scalar.op.data = None
     schema, meta = pandas_to_odps_schema(test_scalar, unknown_as_string=True)
     assert schema.columns[0].name == "_idx_0"
     assert schema.columns[0].type.name == "double"
@@ -279,7 +284,7 @@ def test_build_column_name():
     assert build_table_column_name(4, ("A", 1), records) == "a_1"
 
 
-@pytest.mark.parametrize("wrap_obj", [False, True])
+@pytest.mark.parametrize("wrap_obj", ["no", "yes", "data"])
 def test_build_table_meta(wrap_obj):
     data = pd.DataFrame(
         np.random.rand(100, 7),

maxframe/opcodes.py

@@ -386,6 +386,9 @@ DATAFRAME_EVAL = 738
 DUPLICATED = 739
 DELETE = 740
 ALIGN = 741
+CASE_WHEN = 742
+PIVOT = 743
+PIVOT_TABLE = 744
 
 FUSE = 801
 

maxframe/protocol.py

@@ -32,6 +32,7 @@ from .serialization.serializables import (
     EnumField,
     FieldTypes,
     Float64Field,
+    Int32Field,
     ListField,
     ReferenceField,
     Serializable,
@@ -71,6 +72,9 @@ class DagStatus(enum.Enum):
     CANCELLING = 4
     CANCELLED = 5
 
+    def is_terminated(self):
+        return self in (DagStatus.CANCELLED, DagStatus.SUCCEEDED, DagStatus.FAILED)
+
 
 class DimensionIndex(Serializable):
     is_slice: bool = BoolField("is_slice", default=None)
@@ -190,9 +194,9 @@ class ErrorInfo(JsonSerializable):
         "error_tracebacks", FieldTypes.list
     )
     raw_error_source: ErrorSource = EnumField(
-        "raw_error_source", ErrorSource, FieldTypes.int8
+        "raw_error_source", ErrorSource, FieldTypes.int8, default=None
     )
-    raw_error_data: Optional[Exception] = AnyField("raw_error_data")
+    raw_error_data: Optional[Exception] = AnyField("raw_error_data", default=None)
 
     @classmethod
     def from_exception(cls, exc: Exception):
@@ -201,20 +205,29 @@ class ErrorInfo(JsonSerializable):
         return cls(messages, tracebacks, ErrorSource.PYTHON, exc)
 
     def reraise(self):
-        if self.raw_error_source == ErrorSource.PYTHON:
+        if (
+            self.raw_error_source == ErrorSource.PYTHON
+            and self.raw_error_data is not None
+        ):
             raise self.raw_error_data
         raise RemoteException(self.error_messages, self.error_tracebacks, [])
 
     @classmethod
     def from_json(cls, serialized: dict) -> "ErrorInfo":
         kw = serialized.copy()
-        kw["raw_error_source"] = ErrorSource(serialized["raw_error_source"])
+        if kw.get("raw_error_source") is not None:
+            kw["raw_error_source"] = ErrorSource(serialized["raw_error_source"])
+        else:
+            kw["raw_error_source"] = None
+
         if kw.get("raw_error_data"):
             bufs = [base64.b64decode(s) for s in kw["raw_error_data"]]
             try:
                 kw["raw_error_data"] = pickle.loads(bufs[0], buffers=bufs[1:])
             except:
-                kw["raw_error_data"] = None
+                # both error source and data shall be None to make sure
+                # RemoteException is raised.
+                kw["raw_error_source"] = kw["raw_error_data"] = None
         return cls(**kw)
 
     def to_json(self) -> dict:
@@ -227,7 +240,12 @@ class ErrorInfo(JsonSerializable):
         if isinstance(self.raw_error_data, (PickleContainer, RemoteException)):
             err_data_bufs = self.raw_error_data.get_buffers()
         elif isinstance(self.raw_error_data, BaseException):
-            err_data_bufs = pickle_buffers(self.raw_error_data)
+            try:
+                err_data_bufs = pickle_buffers(self.raw_error_data)
+            except:
+                err_data_bufs = None
+                ret["raw_error_source"] = None
+
         if err_data_bufs:
             ret["raw_error_data"] = [
                 base64.b64encode(s).decode() for s in err_data_bufs
@@ -249,9 +267,17 @@ class DagInfo(JsonSerializable):
     error_info: Optional[ErrorInfo] = ReferenceField("error_info", default=None)
     start_timestamp: Optional[float] = Float64Field("start_timestamp", default=None)
     end_timestamp: Optional[float] = Float64Field("end_timestamp", default=None)
+    subdag_infos: Dict[str, "SubDagInfo"] = DictField(
+        "subdag_infos",
+        key_type=FieldTypes.string,
+        value_type=FieldTypes.reference,
+        default_factory=dict,
+    )
 
     @classmethod
-    def from_json(cls, serialized: dict) -> "DagInfo":
+    def from_json(cls, serialized: dict) -> Optional["DagInfo"]:
+        if serialized is None:
+            return None
         kw = serialized.copy()
         kw["status"] = DagStatus(kw["status"])
         if kw.get("tileable_to_result_infos"):
@@ -261,6 +287,10 @@ class DagInfo(JsonSerializable):
             }
         if kw.get("error_info"):
             kw["error_info"] = ErrorInfo.from_json(kw["error_info"])
+        if kw.get("subdag_infos"):
+            kw["subdag_infos"] = {
+                k: SubDagInfo.from_json(v) for k, v in kw["subdag_infos"].items()
+            }
         return DagInfo(**kw)
 
     def to_json(self) -> dict:
@@ -279,6 +309,8 @@ class DagInfo(JsonSerializable):
             }
         if self.error_info:
             ret["error_info"] = self.error_info.to_json()
+        if self.subdag_infos:
+            ret["subdag_infos"] = {k: v.to_json() for k, v in self.subdag_infos.items()}
         return ret
 
 
@@ -302,7 +334,9 @@ class SessionInfo(JsonSerializable):
     error_info: Optional[ErrorInfo] = ReferenceField("error_info", default=None)
 
     @classmethod
-    def from_json(cls, serialized: dict) -> "SessionInfo":
+    def from_json(cls, serialized: dict) -> Optional["SessionInfo"]:
+        if serialized is None:
+            return None
         kw = serialized.copy()
         if kw.get("dag_infos"):
             kw["dag_infos"] = {
@@ -320,7 +354,10 @@ class SessionInfo(JsonSerializable):
             "idle_timestamp": self.idle_timestamp,
         }
         if self.dag_infos:
-            ret["dag_infos"] = {k: v.to_json() for k, v in self.dag_infos.items()}
+            ret["dag_infos"] = {
+                k: v.to_json() if v is not None else None
+                for k, v in self.dag_infos.items()
+            }
         if self.error_info:
             ret["error_info"] = self.error_info.to_json()
         return ret
@@ -342,7 +379,25 @@ class ExecuteDagRequest(Serializable):
     )
 
 
-class SubDagInfo(Serializable):
+class SubDagSubmitInstanceInfo(JsonSerializable):
+    submit_reason: str = StringField("submit_reason")
+    instance_id: str = StringField("instance_id")
+    subquery_id: Optional[int] = Int32Field("subquery_id", default=None)
+
+    @classmethod
+    def from_json(cls, serialized: dict) -> "SubDagSubmitInstanceInfo":
+        return SubDagSubmitInstanceInfo(**serialized)
+
+    def to_json(self) -> dict:
+        ret = {
+            "submit_reason": self.submit_reason,
+            "instance_id": self.instance_id,
+            "subquery_id": self.subquery_id,
+        }
+        return ret
+
+
+class SubDagInfo(JsonSerializable):
     subdag_id: str = StringField("subdag_id")
     status: DagStatus = EnumField("status", DagStatus, FieldTypes.int8, default=None)
     progress: float = Float64Field("progress", default=None)
@@ -355,9 +410,52 @@ class SubDagInfo(Serializable):
         FieldTypes.reference,
         default_factory=dict,
     )
+    start_timestamp: Optional[float] = Float64Field("start_timestamp", default=None)
+    end_timestamp: Optional[float] = Float64Field("end_timestamp", default=None)
+    submit_instances: List[SubDagSubmitInstanceInfo] = ListField(
+        "submit_instances",
+        FieldTypes.reference,
+        default_factory=list,
+    )
+
+    @classmethod
+    def from_json(cls, serialized: dict) -> "SubDagInfo":
+        kw = serialized.copy()
+        kw["status"] = DagStatus(kw["status"])
+        if kw.get("tileable_to_result_infos"):
+            kw["tileable_to_result_infos"] = {
+                k: ResultInfo.from_json(s)
+                for k, s in kw["tileable_to_result_infos"].items()
+            }
+        if kw.get("error_info"):
+            kw["error_info"] = ErrorInfo.from_json(kw["error_info"])
+        if kw.get("submit_instances"):
+            kw["submit_instances"] = [
+                SubDagSubmitInstanceInfo.from_json(s) for s in kw["submit_instances"]
+            ]
+        return SubDagInfo(**kw)
+
+    def to_json(self) -> dict:
+        ret = {
+            "subdag_id": self.subdag_id,
+            "status": self.status.value,
+            "progress": self.progress,
+            "start_timestamp": self.start_timestamp,
+            "end_timestamp": self.end_timestamp,
+        }
+        if self.error_info:
+            ret["error_info"] = self.error_info.to_json()
+        if self.tileable_to_result_infos:
+            ret["tileable_to_result_infos"] = {
+                k: v.to_json() for k, v in self.tileable_to_result_infos.items()
+            }
+        if self.submit_instances:
+            ret["submit_instances"] = [i.to_json() for i in self.submit_instances]
+        return ret
 
 
 class ExecuteSubDagRequest(Serializable):
+    subdag_id: str = StringField("subdag_id")
     dag: TileableGraph = ReferenceField(
         "dag",
         on_serialize=SerializableGraph.from_graph,

maxframe/serialization/core.pxd

@@ -18,6 +18,9 @@ from libc.stdint cimport int32_t, uint64_t
 cdef class Serializer:
     cdef int _serializer_id
 
+    cpdef bint is_public_data_exist(self, dict context, object key)
+    cpdef put_public_data(self, dict context, object key, object value)
+    cpdef get_public_data(self, dict context, object key)
     cpdef serial(self, object obj, dict context)
     cpdef deserial(self, list serialized, dict context, list subs)
     cpdef on_deserial_error(

maxframe/serialization/core.pyi

@@ -0,0 +1,64 @@
+# Copyright 1999-2024 Alibaba Group Holding Ltd.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from concurrent.futures import Executor
+from typing import Any, Callable, Dict, List, TypeVar
+
+def buffered(func: Callable) -> Callable: ...
+def fast_id(obj: Any) -> int: ...
+
+LoadType = TypeVar("LoadType")
+
+def load_type(class_name: str, parent_class: LoadType) -> LoadType: ...
+
+class PickleContainer:
+    def __init__(self, buffers: List[bytes]): ...
+    def get(self) -> Any: ...
+    def get_buffers(self) -> List[bytes]: ...
+
+class Serializer:
+    serializer_id: int
+    def is_public_data_exist(self, context: Dict, key: Any) -> bool: ...
+    def put_public_data(self, context: Dict, key: Any, value: Any) -> None: ...
+    def get_public_data(self, context: Dict, key: Any) -> Any: ...
+    def serial(self, obj: Any, context: Dict): ...
+    def deserial(self, serialized: List, context: Dict, subs: List[Any]): ...
+    def on_deserial_error(
+        self,
+        serialized: List,
+        context: Dict,
+        subs_serialized: List,
+        error_index: int,
+        exc: BaseException,
+    ): ...
+    @classmethod
+    def register(cls, obj_type): ...
+    @classmethod
+    def unregister(cls, obj_type): ...
+
+class Placeholder:
+    id: int
+    callbacks: List[Callable]
+    def __init__(self, id_: int): ...
+    def __hash__(self): ...
+    def __eq__(self, other): ...
+
+def serialize(obj: Any, context: Dict = None): ...
+async def serialize_with_spawn(
+    obj: Any,
+    context: Dict = None,
+    spawn_threshold: int = 100,
+    executor: Executor = None,
+): ...
+def deserialize(headers: List, buffers: List, context: Dict = None): ...