Yougen/

said_testset

Speaker Recognition (SRE) speech dataset, packed as WebDataset tar shards.

data/ train/ metadata.csv audio/ train-000.tar train-001.tar ... validation/ metadata.csv audio/ validation-000.tar ... test/ metadata.csv audio/ test-000.tar ...

Shard counts:

test: 20 tar shard(s)

Inside each tar, every sample is a pair sharing a unique key: .wav # raw audio bytes… See the full description on the dataset page: https://huggingface.co/datasets/Yougen/said_testset.