pytorch文本分类,pytorch多标签分类代码

defbceloss (输出， target ) : positive _ prob=f.log sigmoid (output ) negative _ prob=f.log sigmoid (-output ) loss=-positive _ prob ule ) : def _ init _ (self ) : super (BCE loss，self ) )。 target ) : positive _ prob=f.log sigmoid (pred ) negative _ prob=f.log sigmoid (-1 * pred ) loss=-positive _ prob