Trained on 6GB datasets that contain Croatian and Serbian language for two epochs (500k steps). Leipzig, OSCAR and srWac datasets
Andrija/SRoBERTa-L