File size: 2,981 Bytes
289c905 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 |
from dataclasses import dataclass
from typing import Any, ClassVar, List
from llmdataparser.base_parser import HuggingFaceDatasetParser, HuggingFaceParseEntry
from llmdataparser.prompts import IFEVAL_SYSTEM_PROMPT # You'll need to create this
@dataclass(frozen=True, kw_only=True, slots=True)
class IFEvalParseEntry(HuggingFaceParseEntry):
"""Custom entry class for IFEval, with fields specific to this dataset parser."""
key: int
instruction_id_list: List[str]
kwargs: dict[str, Any]
@classmethod
def create(
cls,
prompt: str,
answer: str,
raw_question: str,
raw_answer: str,
key: int,
instruction_id_list: List[str],
kwargs: dict[str, Any],
task_name: str,
) -> "IFEvalParseEntry":
return cls(
prompt=prompt,
answer=answer,
raw_question=raw_question,
raw_answer=raw_answer,
key=key,
instruction_id_list=instruction_id_list,
kwargs=kwargs,
task_name=task_name,
)
class IFEvalDatasetParser(HuggingFaceDatasetParser[IFEvalParseEntry]):
"""Parser for the IFEval dataset."""
_data_source: ClassVar[str] = "google/IFEval"
_default_task: ClassVar[str] = "default"
_task_names: ClassVar[list[str]] = ["default"]
_default_system_prompt: ClassVar[str] = IFEVAL_SYSTEM_PROMPT
def process_entry(
self, row: dict[str, Any], task_name: str | None = None, **kwargs: Any
) -> IFEvalParseEntry:
"""Process a single IFEval entry."""
# Extract fields from the row
key = row["key"]
raw_question = row["prompt"] # The prompt is the raw question in this case
instruction_id_list = row["instruction_id_list"]
kwargs_data = row["kwargs"]
# For IFEval, we don't have explicit answers in the dataset
# We'll use empty strings as placeholders
answer = ""
raw_answer = ""
# Combine system prompt with the instruction prompt
prompt = f"{self._system_prompt}\n\n{raw_question}"
# Use task_name if provided, otherwise use default
task = task_name or self._get_current_task(row)
return IFEvalParseEntry.create(
prompt=prompt,
answer=answer,
raw_question=raw_question,
raw_answer=raw_answer,
key=key,
instruction_id_list=instruction_id_list,
kwargs=kwargs_data,
task_name=task,
)
if __name__ == "__main__":
# Example usage
parser = IFEvalDatasetParser()
parser.load()
parser.parse()
parsed_data = parser.get_parsed_data
if parsed_data:
example = parsed_data[0]
print("\nExample parsed entry:")
print(f"Key: {example.key}")
print(f"Prompt: {example.prompt}")
print(f"Instruction IDs: {example.instruction_id_list}")
print(f"kwargs: {example.kwargs}")
|