{
  "name": "OCR_Finetuning_Dataset",
  "version": "1.0",
  "description": "Dataset for fine tuning prithivMLmods/Qwen2-VL-OCR-2B-Instruct on OCR tasks. Each record contains an image path, a prompt, and a structured response with document-specific fields.",
  "dataset_file": "dataset.json",
  "image_folder": "imgdeer",
  "data_format": "json",
  "fields": {
    "image": "Relative path to the image file (e.g., 'imgdeer/pan.jpg')",
    "prompt": "Instruction text for OCR extraction",
    "response": {
      "predictedclass": "Document type prediction",
      "extract": "Dictionary of extracted fields"
    }
  }
}