Closed vanewu closed 5 years ago
It might be more reasonable to add a mask for various attention.
eg. Multihead Attention https://github.com/Tencent/NeuralNLP-NeuralClassifier/blob/master/model/layers.py#L132
Thanks for your good advice, we will consider adding it later.
It might be more reasonable to add a mask for various attention.
eg. Multihead Attention https://github.com/Tencent/NeuralNLP-NeuralClassifier/blob/master/model/layers.py#L132