cmu-arctic-xvectors / cmu-arctic-xvectors.py
Matthijs Hollemans
create dataset
0109099
# coding=utf-8
import os
import numpy as np
import datasets
_DATA_URL = "https://huggingface.co/datasets/Matthijs/cmu-arctic-xvectors/resolve/main/spkrec-xvect.zip"
class ArcticXvectors(datasets.GeneratorBasedBuilder):
BUILDER_CONFIGS = [
datasets.BuilderConfig(
name="default",
version=datasets.Version("0.0.1", ""),
description="",
)
]
def _info(self):
return datasets.DatasetInfo(
features=datasets.Features(
{
"filename": datasets.Value("string"),
"xvector": datasets.Sequence(feature=datasets.Value(dtype="float32"), length=512),
}
),
)
def _split_generators(self, dl_manager):
archive = os.path.join(dl_manager.download_and_extract(_DATA_URL), "spkrec-xvect")
return [
datasets.SplitGenerator(
name=datasets.Split.VALIDATION, gen_kwargs={"files": dl_manager.iter_files(archive)}
),
]
def _generate_examples(self, files):
for i, file in enumerate(sorted(files)):
if os.path.basename(file).endswith(".npy"):
yield str(i), {
"filename": os.path.basename(file)[:-4], # strip off .npy
"xvector": np.load(file),
}