File size: 4,044 Bytes
778d47d | 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 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 | """
Build SFT data for a binary correctness classifier (selection agent).
Input: rollout JSONL from run_pipeline_rollouts.py
Output: HF DatasetDict where each row is one (question, schema, candidate_sql, exec_result) trajectory
with a YES/NO label based on whether the final SQL is correct.
The selector at eval time scores each of N candidates with this classifier and picks the highest
yes-probability candidate.
Usage:
python scripts/build_selector_sft_data.py \\
--rollouts data/rollouts/scaleup_bird_train_2stage_K4.jsonl \\
--output_dir data/sft_selector_classifier
"""
import argparse
import json
import os
import random
import re
import sys
ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
os.chdir(ROOT)
PROMPT_TEMPLATE = (
"You are a SQL correctness judge.\n"
"Schema:\n{schema}\n\n"
"Question: {question}\n"
"External knowledge: {evidence}\n\n"
"Candidate SQL:\n{sql}\n\n"
"Execution result:\n{exec_result}\n\n"
"Is this SQL correct for the question? Answer YES or NO."
)
def safe_truncate(s, n=400):
if s is None:
return "(empty)"
s = str(s)
return s if len(s) <= n else s[:n] + "..."
def main():
parser = argparse.ArgumentParser()
parser.add_argument("--rollouts", required=True)
parser.add_argument("--output_dir", required=True)
parser.add_argument("--train_frac", type=float, default=0.95)
args = parser.parse_args()
print(f"Loading {args.rollouts}...")
samples = []
with open(args.rollouts) as f:
for line in f:
line = line.strip()
if not line:
continue
samples.append(json.loads(line))
print(f" {len(samples)} samples")
rows = []
n_correct = n_wrong = 0
for s in samples:
schema = s.get("schema", "")
question = s.get("question", "")
evidence = s.get("evidence", "") or "None"
for t in s.get("trajectories", []):
fixed_sql = t.get("fixed_sql") or t.get("planner_sql")
if not fixed_sql or not fixed_sql.strip():
continue
# Use planner's execution preview as the "execution result" if available
# otherwise fall back to a generic note
exec_response = ""
if t.get("planner_exec_ok"):
exec_response = "OK"
else:
exec_response = "Error / no rows"
label = "YES" if t.get("is_fixed_correct") else "NO"
if label == "YES":
n_correct += 1
else:
n_wrong += 1
prompt = PROMPT_TEMPLATE.format(
schema=safe_truncate(schema, 3000),
question=question,
evidence=evidence,
sql=safe_truncate(fixed_sql, 800),
exec_result=safe_truncate(exec_response, 300),
)
rows.append({
"prompt": prompt,
"completion": label,
"messages": {"prompt": prompt, "completion": label},
"question": question,
"db_id": s.get("db_id", ""),
"label_int": 1 if label == "YES" else 0,
})
print(f"Built {len(rows)} (correct={n_correct} wrong={n_wrong})")
random.seed(42)
indices = list(range(len(rows)))
random.shuffle(indices)
n_train = int(len(rows) * args.train_frac)
train_rows = [rows[i] for i in indices[:n_train]]
test_rows = [rows[i] for i in indices[n_train:]] or [rows[-1]]
from datasets import Dataset, DatasetDict
import shutil
if os.path.exists(args.output_dir):
shutil.rmtree(args.output_dir)
os.makedirs(args.output_dir, exist_ok=True)
ds = DatasetDict({
"train": Dataset.from_list(train_rows),
"test": Dataset.from_list(test_rows),
})
ds.save_to_disk(args.output_dir)
print(f"Saved DatasetDict (train={len(train_rows)}, test={len(test_rows)}) → {args.output_dir}")
if __name__ == "__main__":
main()
|