aodiniz/bert_uncased_L-10_H-512_A-8_cord19-200616_squad2

BERT L-10 H-512 CORD-19 (2020/06/16) fine-tuned on SQuAD v2.0.0 with 10 Transformer layers and hidden embedding of size 512 . The model was referenced in Well-Read Students Learn Better: On the Importance of Pre-training Compact Models, on the importance of pre-training compact models . It is fine-tuneed for MLM on Cord-19 dataset (as released on 2020/06 /16) and fine,tuned, for QA on SQAD . The training code is now being used to train the model and predict predictability of the predicted model .,