clay_vector_embeddings_loader.py

"""
Custom HuggingFace dataset loading script for GeoParquet files.

References:
- https://huggingface.co/docs/datasets/v2.15.0/en/dataset_script
- https://github.com/huggingface/datasets/blob/2.15.0/templates/new_dataset_script.py
- https://huggingface.co/docs/datasets/v2.15.0/en/package_reference/builder_classes
- https://huggingface.co/docs/datasets/v2.15.0/en/about_dataset_load
- https://discuss.huggingface.co/t/how-to-tweak-a-dataset-without-a-loading-script/43533/5
"""
import datasets
import pyarrow as pa
import pyarrow.parquet as pq


_URLS = {"32VLM": " 32VLM_v01.gpq"}
_MGRS_TILES = ["32VLM"]


class ClayVectorEmbeddings(datasets.ArrowBasedBuilder):
    """Clay Vector Embeddings in GeoParquet format."""

    # You will be able to load one or the other configurations in the following list with
    # data = datasets.load_dataset('my_dataset', 'MGRS_TILE')
    BUILDER_CONFIGS = [
        datasets.BuilderConfig(
            name=name,
            version=datasets.Version(version="0.0.1"),
            description=f"Clay vector embeddings for MGRS tile {name}",
        )
        for name in _MGRS_TILES
    ]

    # DEFAULT_CONFIG_NAME = "32VLM"

    def _info(self):
        return datasets.DatasetInfo(
            # This is the description that will appear on the datasets page.
            description="Clay Vector Embeddings in GeoParquet format.",
            # This defines the different columns of the dataset and their types
            features=datasets.Features(
                {
                    "source_url": datasets.Value(dtype="string"),
                    "date": datasets.Value(dtype="date32"),
                    "embeddings": datasets.Value("string"),
                    "geometry": datasets.Value("binary"),
                    # These are the features of your dataset like images, labels ...
                }
            ),
        )

    def _split_generators(self, dl_manager: datasets.download.DownloadManager):
        files = _URLS[self.config.name]
        downloaded_files = dl_manager.download(files)
        return [
            datasets.SplitGenerator(
                name=datasets.Split.ALL,
                # These kwargs will be passed to _generate_tables
                gen_kwargs={"filepaths": downloaded_files},
            )
        ]

    # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
    def _generate_tables(self, filepaths: list[str] = ["32VLM_v01.gpq"]):
        for file_idx, filepath in enumerate(filepaths):
            with open(filepath, mode="rb") as f:
                parquet_file = pq.ParquetFile(source=filepath)
                for batch_idx, record_batch in enumerate(parquet_file.iter_batches()):
                    pa_table = pa.Table.from_batches([record_batch])
                    yield f"{file_idx_batch_idx}", pa_table