By the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC)
language: el
licence: apache-2.0
dataset: ~5GB of Greek corpora
model: GPT2 (12-layer, 768-hidden, 12-heads, 117M parameters. OpenAI GPT-2 English model, finetuned for the Greek language)
pre-processing: tokenization + BPE segmentation
Model description
A text generation (autoregressive) model, using Huggingface transformers and fastai based on the English GPT-2(small).
Finetuned with gradual layer unfreezing. This is a more efficient and sustainable alternative compared to training from scratch, especially for low-resource languages.
from transformers import pipeline
model = "lighteternal/gpt2-finetuned-greek-small"
generator = pipeline(
'text-generation',
device=0,
model=f'{model}',
tokenizer=f'{model}')
text = "Μια φορά κι έναν καιρό"
print("\\\\
".join([x.get("generated_text") for x in generator(
text,
max_length=len(text.split(" "))+15,
do_sample=True,
top_k=50,
repetition_penalty = 1.2,
add_special_tokens=False,
num_return_sequences=5,
temperature=0.95,
top_p=0.95)]))
Training data
We used a small (~5GB) sample from a consolidated Greek corpus based on CC100, Wikimatrix, Tatoeba, Books, SETIMES and GlobalVoices. A bigger corpus is expected to provide better results (T0D0).
Acknowledgement
The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)
