This model is based on the norkart/mt5-large-no checkpoint and then trained for another 2000 steps on the squad_v2 dataset, then 1000 steps on the norwegian split of alexandrainst/scandi-qa.
Given a question and a context, the model can find the answer in the context. The answer does not need to be stated verbatim in the context.
Format: "question: 'your question' context: 'context to the question'"
- Downloads last month
- 1