Hello!
I have a question regarding the visualization of attention maps presented in Figure 4 of your paper.
When visualizing the attention map for each word as shown in the figure, could you please clarify how the timesteps and layers were handled?
Specifically, did you average the attention maps across all layers and timesteps, or were certain layers or timesteps selected for visualization?
Thank you very much for your time and for sharing your excellent work.
Hello!
I have a question regarding the visualization of attention maps presented in Figure 4 of your paper.
When visualizing the attention map for each word as shown in the figure, could you please clarify how the timesteps and layers were handled?
Specifically, did you average the attention maps across all layers and timesteps, or were certain layers or timesteps selected for visualization?
Thank you very much for your time and for sharing your excellent work.