lighteternal/gpt2-finetuned-greek-small

A new version trained on 5x more data is available at: https://huggingface.co/lighteternal/gpt2-finetuned-greek (recommended) A bigger corpus of Greek corpora is expected to provide better results (T0D0) We used a small (~5GB sample from a consolidated Greek corpus based on CC100, Wikimatrix, Tatoeba, Books, SETIMES and GlobalVoices . We used Huggingface transformers and fastai based on the English GPT-2(small) model . This is a more efficient and sustainable alternative compared to training from scratch, especially for low-resource,