esT5s: A Spanish Model for Text Summarization

Deep Learning models based on the Transformer architecture have revolutionized state of art NLP tasks. As English is language in which most significant advances are made, languages like Spanish require specific training, but this training has a computational cost so high that only big corporations with servers and GPUs capable generating them. This work explored how to create model for from multilingual model. Specifically, aimed at creating text summarization, very common task NLP. The results, concerning quality summarization (ROUGE score), point out these small models, language, achieve similar results than much bigger reasonable terms time required power, significantly faster inference.