genshin_piano.py

import os
import random
import datasets
from midi2abc import midi2abc
from data import rm_duplicates_in_folder


_HOMEPAGE = f"https://huggingface.co/datasets/MuGeminorum/{os.path.basename(__file__).split('.')[0]}"

_CITATION = """\
@dataset{genshin_piano,
  author       = {MuGeminorum Studio},
  title        = {Genshin Piano Songs},
  month        = {nov},
  year         = {2023},
  publisher    = {HF},
  version      = {1.1},
  url          = {https://huggingface.co/datasets/MuGeminorum/genshin_piano}
}
"""

_DESCRIPTION = """\
This database contains genshin piano songs downloaded from musescore
"""

_URL = f"{_HOMEPAGE}/resolve/main/data/dataset.zip"


class genshin_piano(datasets.GeneratorBasedBuilder):
    def _info(self):
        return datasets.DatasetInfo(
            features=datasets.Features(
                {
                    "midi": datasets.Value("string"),
                    "abc": datasets.Value("string"),
                    "tag": datasets.Value("string")
                }
            ),
            supervised_keys=("abc", "tags"),
            homepage=_HOMEPAGE,
            license="mit",
            citation=_CITATION,
            description=_DESCRIPTION
        )

    def _split_generators(self, dl_manager):
        data_files = dl_manager.download_and_extract(_URL)
        files = dl_manager.iter_files([data_files])
        dataset = []

        extract_dir = os.path.dirname(data_files[0])
        duplist = rm_duplicates_in_folder(extract_dir)

        for path in files:
            if (not path in duplist) and (os.path.basename(path).endswith(".mid")):
                dataset.append(path)

        random.shuffle(dataset)
        # data_count = len(dataset)
        # p80 = int(data_count * 0.8)
        # p90 = int(data_count * 0.9)

        # return [
        #     datasets.SplitGenerator(
        #         name=datasets.Split.TRAIN,
        #         gen_kwargs={
        #             "files": dataset[:p80]
        #         }
        #     ),
        #     datasets.SplitGenerator(
        #         name=datasets.Split.VALIDATION,
        #         gen_kwargs={
        #             "files": dataset[p80:p90]
        #         }
        #     ),
        #     datasets.SplitGenerator(
        #         name=datasets.Split.TEST,
        #         gen_kwargs={
        #             "files": dataset[p90:]
        #         }
        #     )
        # ]

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "files": dataset
                }
            )
        ]

    def _generate_examples(self, files):
        for i, path in enumerate(files):
            yield i, {
                "midi": path,
                "abc": midi2abc(path),
                "tag": os.path.basename(path)[:-4].encode('cp437').decode('gbk')
            }