Add new model trained on the spanish subset of suitable images of the 20% of the WIT dataset, with a validation loss of 2.4001
049fefd
edugp
commited on