BERT Base NLI Mean Tokens

A BERT-based sentence embedding model trained on NLI data, using mean pooling of token embeddings.