metadata
language: eu
license: cc-by-sa-4.0
datasets:
- cc100
- oscar
widget:
- text: Zein da zure
- text: Euria egingo
- text: Nola dakizu ?
GPT2 Basque small model Version 2 (Uncased)
Prerequisites
transformers==4.19.2
Model architecture
This model uses approximately half the size of GPT2 base model parameters.
Tokenizer
Using BPE tokenizer with vocabulary size 50,000.
Training Data
Usage
from transformers import pipeline
generator = pipeline('text-generation', model='ClassCat/gpt2-small-basque-v2')
generator("Zein da zure ", max_length=50, num_return_sequences=5)