Skip to content

Latest commit

 

History

History
32 lines (20 loc) · 693 Bytes

File metadata and controls

32 lines (20 loc) · 693 Bytes

A PyTorch implementation of mutil_label text classification

其中损失函数采用了sigmoid+BCE
苏神提出的——multilabel_crossentropy

将“softmax+交叉熵”推广到多标签分类问题

用法

python train_bert_mutillabel_classification.py

结果

BCE     train_acc:1.0000 val_acc:0.9720------best_acc:0.9720
MLCE    train_acc:1.0000 val_acc:0.9790------best_acc:0.9790