## Training Data
`agirdi` was fine-tuned on the **LIEPA-3 (Didysis lietuvių kalbos garsynas)** Lithuanian speech corpus.
LIEPA-3 contains approximately 10,000 hours of Lithuanian speech, including read, spontaneous, dialectal, and phonetically annotated speech recorded in a variety of conditions. The corpus was created by Vytautas Magnus University, Vilnius University, and the Institute of the Lithuanian Language.
Dataset: `meldynamics/liepa-3`
LIEPA-3 is released under the **Creative Commons Attribution 4.0 International (CC BY 4.0)** license.
### LIEPA-3 Citation
```bibtex
@misc{liepa3,
title = {Didysis lietuvių kalbos garsynas (LIEPA-3)},
author = {Vytautas Magnus University and Vilnius University and
Institute of the Lithuanian Language},
year = {2026},
note = {Large Lithuanian speech corpus, ~10,000 hours},
howpublished = {CC BY 4.0}
}
```
