maxframe 1.0.0rc4__cp310-cp310-win32.whl → 1.1.0__cp310-cp310-win32.whl

maxframe/io/odpsio/tableio.py

@@ -18,10 +18,8 @@ from abc import ABC, abstractmethod
 from contextlib import contextmanager
 from typing import Dict, List, Optional, Union
 
-import numpy as np
 import pyarrow as pa
 from odps import ODPS
-from odps import __version__ as pyodps_version
 from odps.apis.storage_api import (
     StorageApiArrowClient,
     TableBatchScanResponse,
@@ -29,6 +27,7 @@ from odps.apis.storage_api import (
 )
 from odps.tunnel import TableTunnel
 from odps.types import OdpsSchema, PartitionSpec, timestamp_ntz
+from odps.utils import call_with_retry
 
 try:
     import pyarrow.compute as pac
@@ -37,20 +36,18 @@ except ImportError:
 
 from ...config import options
 from ...env import ODPS_STORAGE_API_ENDPOINT
-from ...lib.version import Version
 from ...utils import sync_pyodps_options
 from .schema import odps_schema_to_arrow_schema
 
 PartitionsType = Union[List[str], str, None]
 
 _DEFAULT_ROW_BATCH_SIZE = 4096
-_need_patch_batch = Version(pyodps_version) < Version("0.12.0")
 
 
 class ODPSTableIO(ABC):
     def __new__(cls, odps: ODPS):
         if cls is ODPSTableIO:
-            if options.use_common_table:
+            if options.use_common_table or ODPS_STORAGE_API_ENDPOINT in os.environ:
                 return HaloTableIO(odps)
             else:
                 return TunnelTableIO(odps)
@@ -132,7 +129,12 @@ class TunnelMultiPartitionReader:
         self._cur_partition_id = -1
         self._reader_start_pos = 0
 
-        if partitions is None or isinstance(partitions, str):
+        if partitions is None:
+            if not self._table.table_schema.partitions:
+                self._partitions = [None]
+            else:
+                self._partitions = [str(pt) for pt in self._table.partitions]
+        elif isinstance(partitions, str):
             self._partitions = [partitions]
         else:
             self._partitions = partitions
@@ -160,17 +162,14 @@ class TunnelMultiPartitionReader:
             self._cur_partition_id += 1
 
             part_str = self._partitions[self._cur_partition_id]
-
-            # todo make this more formal when PyODPS 0.12.0 is released
-            req_columns = self._columns
-            if not _need_patch_batch:
-                req_columns = self._schema.names
+            req_columns = self._schema.names
             with sync_pyodps_options():
                 self._cur_reader = self._table.open_reader(
                     part_str,
                     columns=req_columns,
                     arrow=True,
                     download_id=self._partition_to_download_ids.get(part_str),
+                    append_partitions=True,
                 )
             if self._cur_reader.count + self._reader_start_pos > self._start:
                 start = self._start - self._reader_start_pos
@@ -186,35 +185,6 @@ class TunnelMultiPartitionReader:
         else:
             self._cur_reader = None
 
-    def _fill_batch_partition(self, batch: pa.RecordBatch) -> pa.RecordBatch:
-        pt_spec = PartitionSpec(self._partitions[self._cur_partition_id])
-
-        names = list(batch.schema.names)
-        arrays = []
-        for idx in range(batch.num_columns):
-            col = batch.column(idx)
-            if isinstance(col.type, pa.TimestampType):
-                if col.type.tz is not None:
-                    target_type = pa.timestamp(
-                        self._schema.types[idx].unit, col.type.tz
-                    )
-                    arrays.append(col.cast(target_type))
-                else:
-                    target_type = pa.timestamp(
-                        self._schema.types[idx].unit, options.local_timezone
-                    )
-                    pd_col = col.to_pandas().dt.tz_localize(options.local_timezone)
-                    arrays.append(pa.Array.from_pandas(pd_col).cast(target_type))
-            else:
-                arrays.append(batch.column(idx))
-
-        for part_col in self._partition_cols or []:
-            names.append(part_col)
-            col_type = self._schema.field_by_name(part_col).type
-            pt_col = np.repeat([pt_spec[part_col]], batch.num_rows)
-            arrays.append(pa.array(pt_col).cast(col_type))
-        return pa.RecordBatch.from_arrays(arrays, names)
-
     def read(self):
         with sync_pyodps_options():
             if self._cur_reader is None:
@@ -227,10 +197,7 @@ class TunnelMultiPartitionReader:
                     if batch is not None:
                         if self._row_left is not None:
                             self._row_left -= batch.num_rows
-                        if _need_patch_batch:
-                            return self._fill_batch_partition(batch)
-                        else:
-                            return batch
+                        return batch
                 except StopIteration:
                     self._open_next_reader()
             return None
@@ -247,34 +214,6 @@ class TunnelMultiPartitionReader:
         return pa.Table.from_batches(batches)
 
 
-class TunnelWrappedWriter:
-    def __init__(self, nested_writer):
-        self._writer = nested_writer
-
-    def write(self, data: Union[pa.RecordBatch, pa.Table]):
-        if not any(isinstance(tp, pa.TimestampType) for tp in data.schema.types):
-            self._writer.write(data)
-            return
-        pa_type = type(data)
-        arrays = []
-        for idx in range(data.num_columns):
-            name = data.schema.names[idx]
-            col = data.column(idx)
-            if not isinstance(col.type, pa.TimestampType):
-                arrays.append(col)
-                continue
-            if self._writer.schema[name].type == timestamp_ntz:
-                col = HaloTableArrowWriter._localize_timezone(col, "UTC")
-            else:
-                col = HaloTableArrowWriter._localize_timezone(col)
-            arrays.append(col)
-        data = pa_type.from_arrays(arrays, names=data.schema.names)
-        self._writer.write(data)
-
-    def __getattr__(self, item):
-        return getattr(self._writer, item)
-
-
 class TunnelTableIO(ODPSTableIO):
     @contextmanager
     def open_reader(
@@ -360,13 +299,7 @@ class TunnelTableIO(ODPSTableIO):
                 create_partition=partition is not None,
                 overwrite=overwrite,
             ) as writer:
-                # fixme should yield writer directly once pyodps fixes
-                #  related arrow timestamp bug when provided schema and
-                #  table schema is identical.
-                if _need_patch_batch:
-                    yield TunnelWrappedWriter(writer)
-                else:
-                    yield writer
+                yield writer
 
 
 class HaloTableArrowReader:
@@ -422,7 +355,7 @@ class HaloTableArrowReader:
             split_index=self._cur_split_id + 1,
             **read_rows_kw,
         )
-        self._cur_reader = self._client.read_rows_arrow(req)
+        self._cur_reader = call_with_retry(self._client.read_rows_arrow, req)
         self._cur_split_id += 1
 
     def _convert_timezone(self, batch: pa.RecordBatch) -> pa.RecordBatch:
@@ -494,8 +427,9 @@ class HaloTableArrowWriter:
     def open(self):
         from odps.apis.storage_api import WriteRowsRequest
 
-        self._writer = self._client.write_rows_arrow(
-            WriteRowsRequest(self._write_info.session_id)
+        self._writer = call_with_retry(
+            self._client.write_rows_arrow,
+            WriteRowsRequest(self._write_info.session_id),
         )
 
     @classmethod
@@ -566,28 +500,6 @@ class HaloTableIO(ODPSTableIO):
             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,
@@ -631,12 +543,12 @@ class HaloTableIO(ODPSTableIO):
 
         # todo add more options for partition column handling
         req = TableBatchScanRequest(**scan_kw)
-        resp = client.create_read_session(req)
+        resp = call_with_retry(client.create_read_session, req)
 
         session_id = resp.session_id
         status = resp.session_status
         while status == SessionStatus.INIT:
-            resp = client.get_read_session(SessionRequest(session_id))
+            resp = call_with_retry(client.get_read_session, SessionRequest(session_id))
             status = resp.session_status
             time.sleep(1.0)
 
@@ -691,7 +603,7 @@ class HaloTableIO(ODPSTableIO):
         part_strs = self._convert_partitions(partition)
         part_str = part_strs[0] if part_strs else None
         req = TableBatchWriteRequest(partition_spec=part_str, overwrite=overwrite)
-        resp = client.create_write_session(req)
+        resp = call_with_retry(client.create_write_session, req)
 
         session_id = resp.session_id
         writer = HaloTableArrowWriter(client, resp, table.table_schema)
@@ -700,9 +612,13 @@ class HaloTableIO(ODPSTableIO):
         yield writer
 
         commit_msg = writer.close()
-        resp = client.commit_write_session(
-            SessionRequest(session_id=session_id), [commit_msg]
+        resp = call_with_retry(
+            client.commit_write_session,
+            SessionRequest(session_id=session_id),
+            [commit_msg],
         )
         while resp.session_status == SessionStatus.COMMITTING:
-            resp = client.get_write_session(SessionRequest(session_id=session_id))
+            resp = call_with_retry(
+                client.get_write_session, SessionRequest(session_id=session_id)
+            )
         assert resp.session_status == SessionStatus.COMMITTED

maxframe/io/odpsio/tests/test_schema.py

@@ -21,6 +21,7 @@ from odps import types as odps_types
 from .... import dataframe as md
 from .... import tensor as mt
 from ....core import OutputType
+from ....utils import pd_release_version
 from ..schema import (
     arrow_schema_to_odps_schema,
     build_dataframe_table_meta,
@@ -292,3 +293,42 @@ def test_build_table_meta(wrap_obj):
     table_meta = build_dataframe_table_meta(test_df)
     expected_cols = ["a_2", "a_3", "a_0", "a_1_0", "a_1_1", "b", "c"]
     assert table_meta.table_column_names == expected_cols
+
+
+@pytest.mark.skipif(
+    pd_release_version[0] < 2, reason="only run under pandas 2.0 or greater"
+)
+def test_table_meta_with_datetime():
+    raw_df = pd.DataFrame(
+        [
+            [1, "abc", "2024-10-01 11:23:12"],
+            [3, "uvw", "2024-10-02 22:55:13"],
+        ],
+        columns=["col1", "col2", "col3"],
+    )
+    df = md.DataFrame(raw_df).astype({"col3": "datetime64[ms]"})
+    schema, _ = pandas_to_odps_schema(df, unknown_as_string=True)
+    assert schema.columns[3].type == odps_types.datetime
+
+    raw_series = pd.Series(
+        ["2024-10-01 11:23:12", "2024-10-02 22:55:13"], dtype="datetime64[ms]"
+    )
+    s = md.Series(raw_series)
+    schema, _ = pandas_to_odps_schema(s, unknown_as_string=True)
+    assert schema.columns[1].type == odps_types.datetime
+
+    raw_index = pd.Index(
+        ["2024-10-01 11:23:12", "2024-10-02 22:55:13"], dtype="datetime64[ms]"
+    )
+    idx = md.Index(raw_index)
+    schema, _ = pandas_to_odps_schema(idx, unknown_as_string=True)
+    assert schema.columns[0].type == odps_types.datetime
+
+    src_df = pd.DataFrame(
+        [[1, "2024-10-01 11:23:12"], [3, "2024-10-02 22:55:13"]],
+        columns=["A", "B"],
+    ).astype({"B": "datetime64[ms]"})
+    raw_multiindex = pd.MultiIndex.from_frame(src_df)
+    multiidx = md.Index(raw_multiindex)
+    schema, _ = pandas_to_odps_schema(multiidx, unknown_as_string=True)
+    assert schema.columns[1].type == odps_types.datetime

maxframe/io/odpsio/tests/test_tableio.py

@@ -31,7 +31,7 @@ def switch_table_io(request):
     old_use_common_table = options.use_common_table
     try:
         options.use_common_table = request.param
-        yield
+        yield request.param
     finally:
         options.use_common_table = old_use_common_table
 
@@ -45,7 +45,7 @@ def test_empty_table_io(switch_table_io):
     table_io = ODPSTableIO(o)
 
     # test read from empty table
-    empty_table_name = tn("test_empty_table_halo_read")
+    empty_table_name = tn("test_empty_table_halo_read_" + str(switch_table_io).lower())
     o.delete_table(empty_table_name, if_exists=True)
     tb = o.create_table(empty_table_name, "col1 string", lifecycle=1)
 
@@ -65,7 +65,7 @@ def test_table_io_without_parts(switch_table_io):
     table_io = ODPSTableIO(o)
 
     # test read and write tables without partition
-    no_part_table_name = tn("test_no_part_halo_write")
+    no_part_table_name = tn("test_no_part_halo_write_" + str(switch_table_io).lower())
     o.delete_table(no_part_table_name, if_exists=True)
     col_desc = ",".join(f"{c} double" for c in "abcde") + ", f datetime"
     tb = o.create_table(no_part_table_name, col_desc, lifecycle=1)
@@ -99,7 +99,7 @@ def test_table_io_with_range_reader(switch_table_io):
     table_io = ODPSTableIO(o)
 
     # test read and write tables without partition
-    no_part_table_name = tn("test_no_part_halo_write")
+    no_part_table_name = tn("test_halo_write_range_" + str(switch_table_io).lower())
     o.delete_table(no_part_table_name, if_exists=True)
     tb = o.create_table(
         no_part_table_name, ",".join(f"{c} double" for c in "abcde"), lifecycle=1
@@ -139,7 +139,7 @@ def test_table_io_with_parts(switch_table_io):
     table_io = ODPSTableIO(o)
 
     # test read and write tables with partition
-    parted_table_name = tn("test_parted_halo_write")
+    parted_table_name = tn("test_parted_halo_write_" + str(switch_table_io).lower())
     o.delete_table(parted_table_name, if_exists=True)
     tb = o.create_table(
         parted_table_name,

maxframe/io/odpsio/tests/test_volumeio.py

@@ -42,15 +42,33 @@ def create_volume(request, oss_config):
             oss_bucket_name,
             oss_endpoint,
         ) = oss_config.oss_config
-        test_location = "oss://%s:%s@%s/%s/%s" % (
-            oss_access_id,
-            oss_secret_access_key,
-            oss_endpoint,
-            oss_bucket_name,
-            oss_test_dir_name,
-        )
+
+        if "test" in oss_endpoint:
+            # offline config
+            test_location = "oss://%s:%s@%s/%s/%s" % (
+                oss_access_id,
+                oss_secret_access_key,
+                oss_endpoint,
+                oss_bucket_name,
+                oss_test_dir_name,
+            )
+            rolearn = None
+        else:
+            # online config
+            endpoint_parts = oss_endpoint.split(".", 1)
+            if "-internal" not in endpoint_parts[0]:
+                endpoint_parts[0] += "-internal"
+            test_location = "oss://%s/%s/%s" % (
+                ".".join(endpoint_parts),
+                oss_bucket_name,
+                oss_test_dir_name,
+            )
+            rolearn = oss_config.oss_rolearn
+
         oss_config.oss_bucket.put_object(oss_test_dir_name + "/", b"")
-        odps_entry.create_external_volume(test_vol_name, location=test_location)
+        odps_entry.create_external_volume(
+            test_vol_name, location=test_location, rolearn=rolearn
+        )
     try:
         yield test_vol_name
     finally:
@@ -75,13 +93,19 @@ def test_read_write_volume(create_volume):
 
     odps_entry = ODPS.from_environments()
 
-    writer = ODPSVolumeWriter(odps_entry, create_volume, test_vol_dir)
+    writer = ODPSVolumeWriter(
+        odps_entry, create_volume, test_vol_dir, replace_internal_host=True
+    )
 
-    writer = ODPSVolumeWriter(odps_entry, create_volume, test_vol_dir)
+    writer = ODPSVolumeWriter(
+        odps_entry, create_volume, test_vol_dir, replace_internal_host=True
+    )
     writer.write_file("file1", b"content1")
     writer.write_file("file2", b"content2")
 
-    reader = ODPSVolumeReader(odps_entry, create_volume, test_vol_dir)
+    reader = ODPSVolumeReader(
+        odps_entry, create_volume, test_vol_dir, replace_internal_host=True
+    )
     assert reader.read_file("file1") == b"content1"
     assert reader.read_file("file2") == b"content2"
 

maxframe/io/odpsio/volumeio.py

@@ -16,13 +16,25 @@ import inspect
 from typing import Iterator, List, Optional, Union
 
 from odps import ODPS
+from odps import __version__ as pyodps_version
+
+from ...lib.version import Version
+
+_has_replace_internal_host = Version(pyodps_version) >= Version("0.12.0")
 
 
 class ODPSVolumeReader:
-    def __init__(self, odps_entry: ODPS, volume_name: str, volume_dir: str):
+    def __init__(
+        self,
+        odps_entry: ODPS,
+        volume_name: str,
+        volume_dir: str,
+        replace_internal_host: bool = False,
+    ):
         self._odps_entry = odps_entry
         self._volume = odps_entry.get_volume(volume_name)
         self._volume_dir = volume_dir
+        self._replace_internal_host = replace_internal_host
 
     def list_files(self) -> List[str]:
         def _get_file_name(vol_file):
@@ -38,7 +50,12 @@ class ODPSVolumeReader:
         ]
 
     def read_file(self, file_name: str) -> bytes:
-        with self._volume.open_reader(self._volume_dir + "/" + file_name) as reader:
+        kw = {}
+        if _has_replace_internal_host and self._replace_internal_host:
+            kw = {"replace_internal_host": self._replace_internal_host}
+        with self._volume.open_reader(
+            self._volume_dir + "/" + file_name, **kw
+        ) as reader:
             return reader.read()
 
 
@@ -49,13 +66,20 @@ class ODPSVolumeWriter:
         volume_name: str,
         volume_dir: str,
         schema_name: Optional[str] = None,
+        replace_internal_host: bool = False,
     ):
         self._odps_entry = odps_entry
         self._volume = odps_entry.get_volume(volume_name, schema=schema_name)
         self._volume_dir = volume_dir
+        self._replace_internal_host = replace_internal_host
 
     def write_file(self, file_name: str, data: Union[bytes, Iterator[bytes]]):
-        with self._volume.open_writer(self._volume_dir + "/" + file_name) as writer:
+        kw = {}
+        if _has_replace_internal_host and self._replace_internal_host:
+            kw = {"replace_internal_host": self._replace_internal_host}
+        with self._volume.open_writer(
+            self._volume_dir + "/" + file_name, **kw
+        ) as writer:
             if not inspect.isgenerator(data):
                 writer.write(data)
             else:

maxframe/learn/contrib/__init__.py

@@ -12,7 +12,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from . import graph, pytorch
+from . import graph, llm, pytorch
 
-del pytorch
 del graph
+del llm
+del pytorch

maxframe/learn/contrib/llm/__init__.py

@@ -0,0 +1,16 @@
+# 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 . import models, multi_modal, text
+
+del models

maxframe/learn/contrib/llm/core.py

@@ -0,0 +1,54 @@
+# 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 Any, Dict
+
+import numpy as np
+import pandas as pd
+
+from ....core.entity.output_types import OutputType
+from ....core.operator.base import Operator
+from ....core.operator.core import TileableOperatorMixin
+from ....dataframe.utils import parse_index
+from ....serialization.serializables.core import Serializable
+from ....serialization.serializables.field import AnyField, DictField, StringField
+
+
+class LLM(Serializable):
+    name = StringField("name", default=None)
+
+    def validate_params(self, params: Dict[str, Any]):
+        pass
+
+
+class LLMOperator(Operator, TileableOperatorMixin):
+    model = AnyField("model", default=None)
+    prompt_template = AnyField("prompt_template", default=None)
+    params = DictField("params", default=None)
+
+    def __init__(self, output_types=None, **kw):
+        if output_types is None:
+            output_types = [OutputType.dataframe]
+        super().__init__(_output_types=output_types, **kw)
+
+    def __call__(self, data):
+        col_names = ["response", "success"]
+        columns = parse_index(pd.Index(col_names), store_data=True)
+        out_dtypes = pd.Series([np.dtype("O"), np.dtype("bool")], index=col_names)
+        return self.new_tileable(
+            inputs=[data],
+            dtypes=out_dtypes,
+            shape=(data.shape[0], len(col_names)),
+            index_value=data.index_value,
+            columns_value=columns,
+        )

maxframe/learn/contrib/llm/models/__init__.py

@@ -0,0 +1,14 @@
+# 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 .dashscope import DashScopeMultiModalLLM, DashScopeTextLLM

maxframe/learn/contrib/llm/models/dashscope.py

@@ -0,0 +1,73 @@
+# 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 Any, Dict
+
+from ..... import opcodes
+from .....serialization.serializables.core import Serializable
+from .....serialization.serializables.field import StringField
+from ..core import LLMOperator
+from ..multi_modal import MultiModalLLM
+from ..text import TextLLM
+
+
+class DashScopeLLMMixin(Serializable):
+    __slots__ = ()
+
+    _not_supported_params = {"stream", "incremental_output"}
+
+    def validate_params(self, params: Dict[str, Any]):
+        for k in params.keys():
+            if k in self._not_supported_params:
+                raise ValueError(f"{k} is not supported")
+
+
+class DashScopeTextLLM(TextLLM, DashScopeLLMMixin):
+    api_key_resource = StringField("api_key_resource", default=None)
+
+    def generate(
+        self,
+        data,
+        prompt_template: Dict[str, Any],
+        params: Dict[str, Any] = None,
+    ):
+        return DashScopeTextGenerationOperator(
+            model=self,
+            prompt_template=prompt_template,
+            params=params,
+        )(data)
+
+
+class DashScopeMultiModalLLM(MultiModalLLM, DashScopeLLMMixin):
+    api_key_resource = StringField("api_key_resource", default=None)
+
+    def generate(
+        self,
+        data,
+        prompt_template: Dict[str, Any],
+        params: Dict[str, Any] = None,
+    ):
+        # TODO add precheck here
+        return DashScopeMultiModalGenerationOperator(
+            model=self,
+            prompt_template=prompt_template,
+            params=params,
+        )(data)
+
+
+class DashScopeTextGenerationOperator(LLMOperator):
+    _op_type_ = opcodes.DASHSCOPE_TEXT_GENERATION
+
+
+class DashScopeMultiModalGenerationOperator(LLMOperator):
+    _op_type_ = opcodes.DASHSCOPE_MULTI_MODAL_GENERATION