hf_generator_vqav2.py

import glob
import json
import os
from tqdm import tqdm
import datasets
import yaml
from PIL import Image as PilImage
from transformers import AutoTokenizer
import os
import shutil
PilImage.MAX_IMAGE_PIXELS = 11178956970
import tarfile
with open("./config.yaml", "r") as file:
    config = yaml.safe_load(file)
    
    
def generate_new_path(old_path, prefix="COCO_train2014_"):
    directory, old_filename = os.path.split(old_path)
    new_filename = f"{prefix}{old_filename}"
    return os.path.join(directory, new_filename)

def read_image(image, image_dir):
    # image = image.replace("vqav2/", "")
    image_path = os.path.join(image_dir, image)
    
    return PilImage.open(generate_new_path(image_path))

_CITATION = """\
@InProceedings{huggingface:dataset,
title = {A great new dataset},
author={huggingface, Inc.},
year={2020}
}
"""
VERSION = "1.0.0"
_HOMEPAGE = ""
_LICENSE = ""
_URLS = {
    "image": config["image_path"],
    "json_files": "./metadata/speech_vqav2",
}

class PDFText(datasets.GeneratorBasedBuilder):
    """PDF text Dataset."""
    
    BUILDER_CONFIGS = [
        datasets.BuilderConfig(
            name="speech_vqav2",
            version=VERSION,
            description="speech_vqav2",
        ),
    ]

    def _info(self):
        description = "This new dataset is designed to solve education problems."
        features = datasets.Features(
            {
                "id": datasets.Value("int64"),
                "question": datasets.Value("string"),
                "answer": datasets.Value("string"),
                "dataset": datasets.Value("string"),
                "task": datasets.Value("string"),
                "image": [datasets.Image()],
            }
        )
        return datasets.DatasetInfo(
            description=description,
            features=features,
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        
        image_data_dir = _URLS["image"]
        json_files_data_dir = _URLS["json_files"]
        
        split_files = [
            os.path.join(json_files_data_dir, f) 
            for f in os.listdir(json_files_data_dir) 
            if f.endswith('.json')
        ]
        split_files.sort()
        split_index = int(len(split_files) * 0.98)
        train_data = split_files[:split_index]
        val_data = split_files[split_index:]
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "filepath": train_data,
                    "image_dir": image_data_dir
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                gen_kwargs={
                    "filepath": val_data,
                    "image_dir": image_data_dir
                },
            ),
        ]

    def _generate_examples(self, filepath, image_dir):
        """Yields examples from all JSON files with globally unique IDs."""
        global_id = 0  # Keep track of global ID across all files
        
        for json_file in filepath:
            # try:
            file_name = os.path.basename(json_file)
            print(file_name)
            with open(json_file, 'r') as file:
                data = json.load(file)
            # data = data[:10]  # Limiting to 10 examples as in original
            
            for item in tqdm(data, desc=f"Processing {os.path.basename(json_file)}"):
                # try:
                    # Try to read the image
                    image = [read_image(item["image"], image_dir)]
                    # if image 
                    # Only yield if image loading succeeds
                    yield global_id, {
                        "id": global_id,
                        "image": image,
                        "task": item["task"],
                        "dataset": "vqav2",
                        "question": item["question"],
                        "answer": item["answer"],
                    }
                    global_id += 1  # Increment only on successful yield