The model has been trained on the second version of the SQuAD_es database. It is a question-answering dataset automatically translated from SQUAD to Spanish. This version includes the possibility that the answer does not exist within the context.
The pretrained model used is "dccuchile/bert-base-spanish-wwm-cased", also called as BETO, pretrained on a big spanish corpus.
METRICS
F1: 62.70 EM: 54.60
- Downloads last month
- 42
Dataset used to train inigopm/beto-base-spanish-squades2
Evaluation results
- f1 on squad_es v2.0.0self-reported62.700
- exact match on squad_es v2.0.0self-reported54.600