Codes for ML mini-paper
My work is associated with Question-Answering tasks, trying to improve the Bert models' performance by applying Text-Classification to it. But it doesn't work. So I try to analyze the attention mechanism to find out why it do not help.