speediedan/deep_classiflie

Explore/analyze contextual embedding interpretations

speediedan opened this issue · 0 comments

e.g. which statements maximize difference in l2 norm of their output embedding after respective ctxt embedding applied