Open jnulzl opened 6 years ago
@huangyangyu Hi, your "example/train_val.prototxt" has the "SmoothMarginInnerProduct" layer , but your caffe hasn't it.
I'm sorry to reply so late. It's ok to use MarginInnerProductLayer, the purpose of it is to ignore the label of sequence dataset whose label is negative.
@huangyangyu Hi, your "example/train_val.prototxt" has the "SmoothMarginInnerProduct" layer , but your caffe hasn't it.