mrm8488/bert-uncased-finetuned-qnli

Using a Bert model fine tuned on SQUAD v2 and then fine tuned it on QNLI using compression (with a constant replacing rate) The model was trained on a Tesla P100 GPU and 25GB of RAM with the following command:.python /content/BERT-of-Theseus/run_glue.py . The models were trained using a model that was fine tuned using compression and a constant replacement rate of 0.7 . The data was downloaded to a new version of the Bert model and the model was used to train the downstream task (QNLI) using a dataset that was downloaded from the same source as the previous version of this article . The model,