File size: 4,364 Bytes
ab3806f 58e063b ab3806f |
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 |
import json
import os
import logging
import argparse
from PIL import Image
from datasets import Dataset
import io
# Configure logging for detailed output
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
def load_questions_from_meta_qa(meta_qa_file):
with open(meta_qa_file, "r") as f:
questions = [line.strip() for line in f if line.strip()]
return questions
def process_parquet_files(data_dir, output_jsonl, meta_qa_file=None, output_imgs=None, process_qa=False):
"""
Process Parquet files to generate a JSONL file with optional image export and QA list creation.
Args:
data_dir (str): Directory containing Parquet files.
output_jsonl (str): Output JSONL file path.
meta_qa_file (str, optional): Path to the meta_qa_en.txt file for QA list creation.
output_imgs (str, optional): Directory path to save images. If None, images are not saved.
process_qa (bool): Whether to process and include QA pairs in the output.
Returns:
None
"""
if output_imgs and not os.path.exists(output_imgs):
os.makedirs(output_imgs)
# Load questions only if QA processing is enabled
questions = None
if process_qa and meta_qa_file:
questions = load_questions_from_meta_qa(meta_qa_file)
jsonl_data = []
parquet_files = [os.path.join(data_dir, f) for f in os.listdir(data_dir) if f.endswith(".parquet")]
for parquet_file in parquet_files:
dataset = Dataset.from_parquet(parquet_file)
for row in dataset:
json_item = {
"internal_id": row["internal_id"],
"url": row["url"],
"annotation": row["annotation"],
"meta_result": row["meta_result"],
"meta_mask": row["meta_mask"],
}
# Optionally save images
if output_imgs:
img_data = row["image"]
img_path = os.path.join(output_imgs, f"{row['internal_id']}.jpg")
try:
with open(img_path, "wb") as img_file:
img_file.write(img_data)
json_item["image_path"] = img_path
except Exception as e:
logger.error(f"Error saving image for internal_id {row['internal_id']}: {e}")
# Optionally process QA pairs
if process_qa and questions:
qa_list = []
meta_result = row["meta_result"]
meta_mask = row["meta_mask"]
for idx, mask in enumerate(meta_mask):
if mask == 1: # Process questions only if the mask is 1
question = questions[idx]
answer = 'yes' if meta_result[idx] == 1 else 'no'
qa_list.append({"question": question, "answer": answer})
json_item["qa_list"] = qa_list
jsonl_data.append(json_item)
with open(output_jsonl, "w") as outfile:
for json_item in jsonl_data:
outfile.write(json.dumps(json_item) + "\n")
logger.info(f"Finished writing JSONL file with {len(jsonl_data)} items.")
if __name__ == "__main__":
parser = argparse.ArgumentParser(description="Convert VisionReward Parquet dataset files to JSONL format with optional image extraction and QA list generation.")
parser.add_argument("--data_dir", type=str, default='data', help="Directory containing Parquet files.")
parser.add_argument("--output_jsonl", type=str, default='annotation.jsonl', help="Path to the output JSONL file.")
parser.add_argument("--meta_qa_file", type=str, default="meta_qa_en.txt", help="Optional: Path to the meta_qa_en.txt file for QA list generation.")
parser.add_argument("--save_imgs", action="store_true", help="Optional: Whether to save images.")
parser.add_argument("--process_qa", action="store_true", help="Optional: Process and include QA pairs in the output.")
args = parser.parse_args()
output_imgs = 'imgs' if args.save_imgs else None
process_parquet_files(
data_dir=args.data_dir,
output_jsonl=args.output_jsonl,
meta_qa_file=args.meta_qa_file,
output_imgs=output_imgs,
process_qa=args.process_qa
) |