parquet

MPCC.py

@@ -1,72 +0,0 @@
-import os
-import base64
-import io
-from PIL import Image
-import datasets
-
-_DESCRIPTION = """MPCC: A benchmark for Multimodal Planning under Complex Constraints. This dataset includes Calendar Planning, Flight Planning, and Meeting Planning, each with three difficulty levels: easy, medium, hard."""
-
-_TASKS = {
-    "calendar_planning": "Calendar Planning",
-    "flight_planning": "Flight Planning",
-    "meeting_planning": "Meeting Planning"
-}
-
-_LEVELS = ["easy", "medium", "hard"]
-
-class MPCCConfig(datasets.BuilderConfig):
-    def __init__(self, task_key, **kwargs):
-        super().__init__(**kwargs)
-        self.task_key = task_key
-
-class MPCC(datasets.GeneratorBasedBuilder):
-    BUILDER_CONFIGS = [
-        MPCCConfig(
-            name=task_key,
-            task_key=task_key,
-            version=datasets.Version("1.0.0"),
-            description=f"{task_name} subset of MPCC"
-        )
-        for task_key, task_name in _TASKS.items()
-    ]
-
-    def _info(self):
-        return datasets.DatasetInfo(
-            description=_DESCRIPTION,
-            features=datasets.Features({
-                "image": datasets.Image(),
-                "text": datasets.Value("string")
-            }),
-            supervised_keys=None,
-        )
-
-    def _split_generators(self, dl_manager):
-        task_dir = _TASKS[self.config.task_key]
-        base_name = self.config.task_key.replace("_planning", "")
-        file_template = {
-            "easy": f"{base_name}_plan_easy.tsv",
-            "medium": f"{base_name}_plan_medium.tsv",
-            "hard": f"{base_name}_plan_hard.tsv"
-        }
-
-        return [
-            datasets.SplitGenerator(
-                name=level,
-                gen_kwargs={
-                    "filepath": os.path.join(task_dir, fname)
-                }
-            )
-            for level, fname in file_template.items()
-        ]
-
-    def _generate_examples(self, filepath):
-        with open(filepath, encoding="utf-8") as f:
-            header = f.readline().strip().split("\t")
-            for idx, line in enumerate(f):
-                fields = line.strip().split("\t")
-                row = dict(zip(header, fields))
-                image = Image.open(io.BytesIO(base64.b64decode(row["image_base64"]))).convert("RGB")
-                yield idx, {
-                    "image": image,
-                    "text": row["text"]
-                }