Spaces:
Sleeping
Sleeping
File size: 1,825 Bytes
f884e6e |
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 |
"""
RAG Evaluation Framework
Comprehensive evaluation system for RAG applications including:
- Retrieval quality metrics (Precision@K, Recall@K, MRR, NDCG)
- Generation quality metrics (BLEU, ROUGE, BERTScore, Faithfulness)
- System performance metrics (Latency, Throughput, Error rates)
- User experience metrics (Satisfaction, Task completion)
"""
"""Lazy evaluation exports to avoid heavy imports at test collection time.
Some evaluation modules (BLEU/ROUGE/BERTScore) import large packages like
`torch` which slow down pytest collection. Expose lightweight accessors that
perform imports only when evaluation functions are actually used.
"""
from .core import BenchmarkResults, EvaluationMetrics, EvaluationResult # noqa: E402
def _lazy_import(name: str):
module = __import__(f"src.evaluation.metrics.{name}", fromlist=["*"])
return module
def get_generation_metrics():
m = _lazy_import("generation_metrics")
return (
m.calculate_bert_score,
m.calculate_bleu_score,
m.calculate_faithfulness_score,
m.calculate_rouge_scores,
)
def get_retrieval_metrics():
m = _lazy_import("retrieval_metrics")
return m.mean_reciprocal_rank, m.ndcg_at_k, m.precision_at_k, m.recall_at_k
def get_system_metrics():
m = _lazy_import("system_metrics")
return m.ErrorTracker, m.LatencyTracker, m.ThroughputTracker
def get_user_metrics():
m = _lazy_import("user_metrics")
return m.CitationAccuracyTracker, m.TaskCompletionTracker, m.UserSatisfactionTracker
def get_runner():
m = _lazy_import("runner")
return m.EvaluationRunner
__all__ = [
"EvaluationMetrics",
"EvaluationResult",
"BenchmarkResults",
"get_generation_metrics",
"get_retrieval_metrics",
"get_system_metrics",
"get_user_metrics",
"get_runner",
]
|