📘 ReasoningAnswerPipelineRootFilter

ReasoningAnswerPipelineRootFilter is the root node of the answer processing pipeline. It is responsible for distributing the input data to different processing branches based on whether a ground truth (GT) label is present.

__init__

__init__(self)

init Parameters

Parameter	Type	Default	Description
-	-	-	No parameters.

Prompt Template Descriptions

run

run(self, storage, input_answer_key="output", input_gt_key="golden_answer")

This function executes the main logic of the operator. It reads a DataFrame, splits it into two based on the presence of a ground truth key, and writes them back to storage.

Parameters

Name	Type	Default	Description
storage	DataFlowStorage	Required	The DataFlow storage instance for reading and writing data.
input_answer_key	str	"output"	The column name for the model's generated answer.
input_gt_key	str	"golden_answer"	The column name for the ground truth (GT) label, used to split the data.

🧠 Example Usage

from dataflow.operators.reasoning import ReasoningAnswerPipelineRootFilter
from dataflow.utils.storage import FileStorage
from dataflow.core import LLMServingABC

class ReasoningAnswerPipelineRootFilterTest():
    def __init__(self, llm_serving: LLMServingABC = None):
        
        self.storage = FileStorage(
            first_entry_file_name="example.json",
            cache_path="./cache_local",
            file_name_prefix="dataflow_cache_step",
            cache_type="jsonl",
        )
        
        self.operator = ReasoningAnswerPipelineRootFilter()   
        
    def forward(self):
        self.operator.run(
            storage = self.storage.step(),
            input_answer_key="output",
            input_gt_key="golden_answer"    
        )

if __name__ == "__main__":
    pl = ReasoningAnswerPipelineRootFilterTest()
    pl.forward()

🧾 Default Output Format (Output Format)

This operator does not change the column structure of the original data, but splits the data flow into two outputs based on whether the input_gt_key column has valid values.

Field	Type	Description
...	...	All original fields from the input data.
[input_gt_key]	str	The processed ground truth label column. For data without initial labels, it may be extracted and filled from the input_answer_key column.