File size: 2,088 Bytes
e183645 | 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 | """Minimal usage example for the Hub snapshot.
python examples/quickstart.py # local snapshot (repo root)
python examples/quickstart.py <repo-id> # straight from the Hub
"""
from __future__ import annotations
import sys
from pathlib import Path
ROOT = Path(__file__).resolve().parents[1]
sys.path.insert(0, str(ROOT))
from modeling_megaram import load # noqa: E402
DOCS = [
{
"ext_id": "bm25",
"title": "BM25",
"text": (
"BM25 is a bag-of-words retrieval function that ranks documents by query "
"term frequency with saturation k1 and document length normalisation b. "
"Typical values are k1 equal to 1.2 and b equal to 0.75."
),
},
{
"ext_id": "beir",
"title": "BEIR",
"text": (
"BEIR is a heterogeneous benchmark of 18 information retrieval datasets, "
"introduced by Thakur, Reimers and Gurevych in 2021. It measures zero-shot "
"retrieval with nDCG at 10."
),
},
{
"ext_id": "rgb",
"title": "RGB",
"text": (
"RGB tests noise robustness, negative rejection, information integration "
"and counterfactual robustness in retrieval augmented generation."
),
},
]
def main() -> int:
target = sys.argv[1] if len(sys.argv) > 1 else str(ROOT)
rag = load(target)
print("loaded:", rag.info())
rag.index_documents(DOCS)
for q in [
"What are the typical values of k1 and b in BM25?",
"How many datasets does BEIR contain?",
"Which benchmark tests negative rejection?",
"What is the capital of France?", # answerable from no document
]:
ans = rag.answer(q)
print(f"\nQ: {q}")
print(f" intent={ans.intent} conf={ans.confidence:.3f} abstained={ans.abstained}")
print(f" A: {ans.text}")
for c in ans.citations:
print(f" [{c.doc_id}] {c.quote}")
return 0
if __name__ == "__main__":
raise SystemExit(main())
|