Is it possible to show feature importance using an attention matrix or something similar?"

[Cftq]

In a Transformer model, it is possible to visualize feature importance using self-attention. Is this implemented in this model? Alternatively, is it possible to visualize feature importance with a simple piece of code?"