choihy/mtg-l3ac

MTG-Jamendo for L3AC

This dataset is a preprocessed subset of the MTG-Jamendo dataset, prepared for neural audio codec experiments (L3AC).

Structure

train/ eval/ test/

Audio

• Sample rate: 16 kHz
• Clip length: 30 seconds

How to Download and Use

using HF library

from datasets import load_dataset

dataset = load_dataset("choihy/mtg-l3ac")

train_ds = dataset["train"]
eval_ds = dataset["eval"]
test_ds = dataset["test"]

Download

pip install huggingface_hub
huggingface-cli download \
  choihy/mtg-l3ac \
  --repo-type dataset \
  --local-dir ./mtg-l3ac \
  --local-dir-use-symlinks False

License

CC BY-NC-SA 4.0
Non-commercial use only.

Citation

@inproceedings{bogdanov2019mtgjamendo,
  title={The MTG-Jamendo Dataset for Music Information Retrieval},
  author={Bogdanov et al.},
  year={2019}
}