Add new add_raw_input argument to _Task so we can automatically include the formatted input

[plaguss]

Description

Adds a new attribute to the Tasks to automatically include the formatted input in the distilabel_metadata column.

Example from the tests:

class DummyTask(Task):
    @property
    def inputs(self) -> List[str]:
        return ["instruction", "additional_info"]

    def format_input(self, input: Dict[str, Any]) -> "ChatType":
        return [
            {"role": "system", "content": ""},
            {"role": "user", "content": input["instruction"]},
        ]

    @property
    def outputs(self) -> List[str]:
        return ["output", "info_from_input"]

    def format_output(
        self, output: Union[str, None], input: Union[Dict[str, Any], None] = None
    ) -> Dict[str, Any]:
        return {"output": output, "info_from_input": input["additional_info"]}  # type: ignore


task = DummyTask(
    llm=DummyLLM(),
    add_raw_output=True,
    add_raw_input=True,
)
task.load()
next(task.process([
    {"instruction": "test_0", "additional_info": "additional_info_0"},
]))
# {'additional_info': 'additional_info_0',
#  'distilabel_metadata': {'raw_input_dummy_task_0': [{'content': '',
#                                                      'role': 'system'},
#                                                     {'content': 'test_0',
#                                                      'role': 'user'}],
#                          'raw_output_dummy_task_0': 'output'},
#  'info_from_input': 'additional_info_0',
#  'instruction': 'test_0',
#  'model_name': 'test',
#  'output': 'output'}]

Example in the docs:

Closes #698

[github-actions]

Documentation for this PR has been built. You can view it at: https://distilabel.argilla.io/pr-903/

[codspeed-hq]

CodSpeed Performance Report

Merging #903 will not alter performance

_{Comparing add-raw-input (50d7b4c) with develop (c8df5a9)}

Summary

✅ 1 untouched benchmarks