GE2E_loss-attention These two models are improved from "Generalized End-to-End Loss for Speaker Verification". I try to add convolution layer in advaned of original model (GE2E). It give robustness for model, and the use attention layer to approve the precision.