SitongGong's picture
merge code dirs 1031 (part 3)
4c30587 verified
Raw History Blame Contribute Delete
1.09 kB
"""
Evaluation module for Agentic Memory.
This module provides evaluators for different datasets with a unified interface.
The evaluator handles prompt building, LLM calls, and metric computation.
Usage:
from src.eval import get_evaluator
# Get evaluator for a specific dataset
evaluator = get_evaluator("locomo", args)
# Evaluate on QA list
summary = evaluator.evaluate(qa_list, memory_bank)
# Access results
print(f"Average F1: {summary.avg_f1}")
"""
from .base import (
Evaluator,
EvalResult,
EvalSummary,
get_evaluator,
register_evaluator,
list_evaluators,
)
from .locomo import LoCoMoEvaluator
from .longmemeval import LongMemEvalEvaluator
from .hotpotqa import HotpotQAEvaluator
from .alfworld import ALFWorldEvaluator
__all__ = [
# Base classes
"Evaluator",
"EvalResult",
"EvalSummary",
# Registry functions
"get_evaluator",
"register_evaluator",
"list_evaluators",
# Concrete evaluators
"LoCoMoEvaluator",
"LongMemEvalEvaluator",
"HotpotQAEvaluator",
"ALFWorldEvaluator",
]