DeBERTa v3 (large) fine tuned to Multi NLI (MNLI) This model is for Textual Entailment (aka NLI), i.e., predict whether textA is supported by textB . More specifically, it's a 2 way classification where the relationship between textA and textB can be entail, neutral, contradict . Input: ( textA , textB ) Output: prob(entail), prob(contradict) Note that during training, all 3 labels (entail, neural, contradict) were used. But for this model, the neural output head has been removed. Model Details Base model: deberta v3 large Training data: MNLI Training details: num epochs = 3, batch size = 16, textA=hypothesis , textB=premise Example Citation
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy