SpidR VP-20 is a SpidR model pretrained pretrained on a subset of 6k hours and 20 languages of VoxPopuli
(all EU languages except English, French, and German)
for the
DiscoPhon benchmark.
It was pretrained using the
spidr library.
1from spidr.models import SpidR
2from torch.hub import load_state_dict_from_url
3
4state_dict = load_state_dict_from_url("https://huggingface.co/coml/spidr-vp20/resolve/main/final.pt")
5model = SpidR().eval()
6model.load_state_dict(state_dict)
1@misc{poli2026discophon,
2 title={{DiscoPhon}: Benchmarking the Unsupervised Discovery of Phoneme Inventories With Discrete Speech Units},
3 author={Maxime Poli and Manel Khentout and Angelo Ortiz Tandazo and Ewan Dunbar and Emmanuel Chemla and Emmanuel Dupoux},
4 year={2026},
5 eprint={2603.18612},
6 archivePrefix={arXiv},
7 primaryClass={cs.CL},
8 url={https://arxiv.org/abs/2603.18612},
9}
