| datasets: | |
| - oscar | |
| - srwac | |
| - leipzig | |
| - cc100 | |
| - hrwac | |
| language: | |
| - hr | |
| - sr | |
| - multilingual | |
| tags: | |
| - masked-lm | |
| widget: | |
| - text: "Ovo je početak <mask>." | |
| license: apache-2.0 | |
| # Transformer language model for Croatian and Serbian | |
| Trained on 28GB datasets that contain Croatian and Serbian language for one epochs (3 mil. steps). | |
| Leipzig Corpus, OSCAR, srWac, hrWac, cc100-hr and cc100-sr datasets | |
| | Model | #params | Arch. | Training data | | |
| |--------------------------------|--------------------------------|-------|-----------------------------------| | |
| | `Andrija/SRoBERTa-XL` | 80M | Forth | Leipzig Corpus, OSCAR, srWac, hrWac, cc100-hr and cc100-sr (28 GB of text) | |