# 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), | |
} | |