fix1

MMB_dataset.py

@@ -1,173 +0,0 @@
-"""
-Hugging Face datasets loading script for the MMB Counterfactual Dataset.
-
-This dataset contains counterfactual visual question answering examples with:
-- Original images and counterfactual variants
-- Questions for each image variant
-- Answer matrices showing how each image answers each question
-"""
-
-from __future__ import annotations
-
-import csv
-from pathlib import Path
-from typing import Iterator
-
-import datasets
-from datasets import Image, Value
-
-
-_DESCRIPTION = """
-MMB-style counterfactual visual question answering dataset.
-
-This dataset contains scenes with:
-- Original images and counterfactual variants
-- Questions for each image variant
-- Answer matrices showing how each image answers each question
-"""
-
-_HOMEPAGE = "https://huggingface.co/datasets/scholo/MMB_dataset"
-
-_LICENSE = "mit"
-
-_CITATION = """
-@misc{mmb_counterfactual_dataset,
-  title={MMB Counterfactual Dataset},
-  author={MMB-Dataset authors},
-  year={2026}
-}
-"""
-
-
-class MMB_dataset(datasets.GeneratorBasedBuilder):
-    """MMB counterfactual dataset."""
-
-    VERSION = datasets.Version("1.0.0")
-
-    def _info(self) -> datasets.DatasetInfo:
-        return datasets.DatasetInfo(
-            description=_DESCRIPTION,
-            features=datasets.Features(
-                {
-                    # Simple scene identifier derived from the original image filename
-                    "scene_id": Value("string"),
-                    # Image triplet (paths resolved relative to dataset_1k_720p_2/images/)
-                    "original_image": Image(decode=True),
-                    "counterfactual1_image": Image(decode=True),
-                    "counterfactual2_image": Image(decode=True),
-                    # Questions
-                    "original_question": Value("string"),
-                    "counterfactual1_question": Value("string"),
-                    "counterfactual2_question": Value("string"),
-                    # Difficulties
-                    "original_question_difficulty": Value("string"),
-                    "counterfactual1_question_difficulty": Value("string"),
-                    "counterfactual2_question_difficulty": Value("string"),
-                    # Answer matrix (9 entries, image × question)
-                    "original_image_answer_to_original_question": Value("string"),
-                    "original_image_answer_to_cf1_question": Value("string"),
-                    "original_image_answer_to_cf2_question": Value("string"),
-                    "cf1_image_answer_to_original_question": Value("string"),
-                    "cf1_image_answer_to_cf1_question": Value("string"),
-                    "cf1_image_answer_to_cf2_question": Value("string"),
-                    "cf2_image_answer_to_original_question": Value("string"),
-                    "cf2_image_answer_to_cf1_question": Value("string"),
-                    "cf2_image_answer_to_cf2_question": Value("string"),
-                }
-            ),
-            homepage=_HOMEPAGE,
-            license=_LICENSE,
-            citation=_CITATION,
-        )
-
-    def _split_generators(
-        self, dl_manager: datasets.DownloadManager
-    ) -> list[datasets.SplitGenerator]:
-        """We expose a single train split backed by dataset_1k_720p_2/image_mapping_with_questions.csv."""
-        data_dir = Path(__file__).parent.resolve()
-        csv_path = data_dir / "dataset_1k_720p_2" / "image_mapping_with_questions.csv"
-        images_dir = data_dir / "dataset_1k_720p_2" / "images"
-
-        return [
-            datasets.SplitGenerator(
-                name=datasets.Split.TRAIN,
-                gen_kwargs={
-                    "csv_path": str(csv_path),
-                    "images_dir": str(images_dir),
-                },
-            )
-        ]
-
-    def _generate_examples(
-        self, csv_path: str, images_dir: str
-    ) -> Iterator[tuple[int, dict]]:
-        """Yield (key, example) tuples from the wide-format CSV."""
-        csv_path = Path(csv_path)
-        images_root = Path(images_dir)
-
-        if not csv_path.exists():
-            raise FileNotFoundError(f"CSV file not found: {csv_path}")
-
-        with csv_path.open("r", encoding="utf-8", newline="") as f:
-            reader = csv.DictReader(f)
-            for idx, row in enumerate(reader):
-                # Resolve scene_id from original_image filename, e.g. scene_0000_original.png -> scene_0000
-                original_image_name = row.get("original_image", "") or ""
-                scene_id = original_image_name.split("_original", 1)[0]
-
-                def _image_field(col_name: str) -> dict | None:
-                    filename = (row.get(col_name) or "").strip()
-                    if not filename:
-                        return None
-                    path = images_root / filename
-                    if not path.exists():
-                        return None
-                    return {"path": str(path)}
-
-                example = {
-                    "scene_id": scene_id,
-                    "original_image": _image_field("original_image"),
-                    "counterfactual1_image": _image_field("counterfactual1_image"),
-                    "counterfactual2_image": _image_field("counterfactual2_image"),
-                    "original_question": row.get("original_question", ""),
-                    "counterfactual1_question": row.get("counterfactual1_question", ""),
-                    "counterfactual2_question": row.get("counterfactual2_question", ""),
-                    "original_question_difficulty": row.get(
-                        "original_question_difficulty", ""
-                    ),
-                    "counterfactual1_question_difficulty": row.get(
-                        "counterfactual1_question_difficulty", ""
-                    ),
-                    "counterfactual2_question_difficulty": row.get(
-                        "counterfactual2_question_difficulty", ""
-                    ),
-                    "original_image_answer_to_original_question": row.get(
-                        "original_image_answer_to_original_question", ""
-                    ),
-                    "original_image_answer_to_cf1_question": row.get(
-                        "original_image_answer_to_cf1_question", ""
-                    ),
-                    "original_image_answer_to_cf2_question": row.get(
-                        "original_image_answer_to_cf2_question", ""
-                    ),
-                    "cf1_image_answer_to_original_question": row.get(
-                        "cf1_image_answer_to_original_question", ""
-                    ),
-                    "cf1_image_answer_to_cf1_question": row.get(
-                        "cf1_image_answer_to_cf1_question", ""
-                    ),
-                    "cf1_image_answer_to_cf2_question": row.get(
-                        "cf1_image_answer_to_cf2_question", ""
-                    ),
-                    "cf2_image_answer_to_original_question": row.get(
-                        "cf2_image_answer_to_original_question", ""
-                    ),
-                    "cf2_image_answer_to_cf1_question": row.get(
-                        "cf2_image_answer_to_cf1_question", ""
-                    ),
-                    "cf2_image_answer_to_cf2_question": row.get(
-                        "cf2_image_answer_to_cf2_question", ""
-                    ),
-                }
-
-                yield idx, example