Visualizing Attention in Transformer-Based Language Representation Models

We present an open-source tool for visualizing self-attention in Transformer-based language representation models. The tool extends earlier work by visualizing attention at three levels of granularity: the attention-head level, the model level, and the neuron level. We show how each of these views can help to interpret the model, and we demonstrate the tool on BERT and OpenAI GPT-2. We also present three use cases for analyzing GPT-2: detecting model bias, identifying recurring patterns, and linking neurons to model behavior.