add shuffle func (#451)

* add sample function

* `eval_only` supports summarization

* add shuffle

python/qianfan/dataset/dataset.py

@@ -1053,11 +1053,36 @@ def sample(
                 this value to True
             **kwargs (Any):
                 other arguments
+
+        Returns:
+            Dataset: a sliced dataset
         """
         return super().sample(
             sample_number, start, end, should_create_new_obj, **kwargs
         )
 
+    @_online_except_decorator
+    def shuffle(
+        self,
+        should_create_new_obj: bool = False,
+        **kwargs: Any,
+    ) -> Self:
+        """
+        make a shuffled Dataset
+
+        Args:
+            should_create_new_obj (bool):
+                should a new object be created when mapping terminates.
+                Default to False. In some cases, you may want to set
+                this value to True
+            **kwargs (Any):
+                other arguments
+
+        Returns:
+            Dataset: a sliced dataset
+        """
+        return super().shuffle(should_create_new_obj, **kwargs)
+
     def __getitem__(self, key: Any) -> Any:
         if (
             isinstance(key, int)

python/qianfan/dataset/table.py

@@ -552,6 +552,12 @@ def sample(
         numbers = random.sample(range(start, end + 1), sample_number)
         return self.table.take(numbers)
 
+    def shuffle(self) -> PyarrowTable:
+        indices = list(range(0, self.table.num_rows))
+        random.shuffle(indices)
+
+        return self.table.take(indices)
+
 
 class _PyarrowColumnManipulator(BaseModel, Addable, Listable, Processable):
     """handler for processing of pyarrow table column"""
@@ -1238,6 +1244,15 @@ def sample(
         result_ds = manipulator.sample(sample_number, start, end)
         return self._create_new_obj(result_ds, should_create_new_obj)
 
+    def shuffle(
+        self,
+        should_create_new_obj: bool = False,
+        **kwargs: Any,
+    ) -> Self:
+        manipulator = self._row_op()
+        result_ds = manipulator.shuffle()
+        return self._create_new_obj(result_ds, should_create_new_obj)
+
     def col_map(
         self,
         op: Callable[[Any], Any],

python/qianfan/evaluation/evaluation_manager.py

@@ -42,7 +42,6 @@
 from qianfan.dataset.data_source.utils import (
     _download_file_from_url_streamly,
 )
-from qianfan.dataset.schema import EvaluationSchema
 from qianfan.errors import QianfanError
 from qianfan.evaluation.consts import QianfanRefereeEvaluatorPromptTemplate
 from qianfan.evaluation.evaluation_result import EvaluationResult
@@ -415,9 +414,6 @@ def eval_only(
                 {OldReferenceColumnName: [None for _ in range(len(dataset))]}
             )
 
-        if not EvaluationSchema().validate(dataset):
-            raise ValueError("validate failed before evaluation")
-
         tmp_ds = Dataset.create_from_pyobj(
             self._run_evaluator_locally(dataset, **kwargs)
         )