Dynamic Context Pruning for Efficient and Interpretable Autoregressive Transformers

Open in new window