Files
buddha-gpt/scripts/prepare_train.py
2026-08-17 13:43:32 -07:00

13 lines
505 B
Python

import json, random
from pathlib import Path
rows = [json.loads(l) for l in Path("data/instructions.jsonl").read_text().splitlines()]
random.seed(7); random.shuffle(rows)
n_val = max(200, len(rows) // 20)
Path("data/train").mkdir(parents=True, exist_ok=True)
for name, part in [("valid", rows[:n_val]), ("train", rows[n_val:])]:
with Path(f"data/train/{name}.jsonl").open("w") as f:
for r in part:
f.write(json.dumps(r) + "\n")
print(len(rows) - n_val, "train /", n_val, "valid")