sorry
liyunlongaaa opened this issue · comments
YangGaoBin commented
sorry, I can't understand the code for using the annotated label in evaluate_label_efficient, it seems that the training process is the same as no label process. Can you talk some key about it?
Théo Lepage commented
Hi @liyunlongaaa,
Indeed, the training process is similar. The only difference is that we use SupervisedSampler.py
to make sure that each sample in the training mini-batch is from a unique speaker. This corresponds to a metric learning paradygm (see In defence of metric learning for speaker recognition).