prajjwal1/bert-tiny-mnli

These BERT variants were introduced in the paper Well-Read Students Learn Better: On the Importance of Pre-training Compact Models . These models are trained on MNLI, MNLI and MNLI . MNLI-mm: 61.61%. MNLI: 60%. Sniffle: 61% Sniffles: 61%. MNLI is a pre-trained model from converting Tensorflow to a TensorFlow checkpoint. MNL-mm was trained for 4 epochs.MNLI is the most successful model in the world of pre-training BERT models. The model was found in the official Google BERT repository. It has been trained for four epochs .,