桐原因

Results 63 comments of 桐原因

请问这个问题,你解决了嘛?

就是将意图识别和槽填充改为自己的模型

I try open this link ``` https://raw.githubusercontent.com/lintool/AnseriniMaven/master/mvn-repo/ ``` ![image](https://user-images.githubusercontent.com/43271630/171412193-5156231c-d1d1-4377-888b-a4a783042463.png) please help!

> Sometimes this is a transient error... can you try again? Thanks for your reply, but it is still a mistake to try many times......

给测试集伪标签吗?然后建图进行预测

就比如:总的标签数量为 [1,2,3,4] 4类,test的数据集的伪标签都选择1,然后进行建图进行训练和预测。

好的,我试试,感谢回复。

你好,我尝试在别的数据集上进行训练,但是发现train的acc和val的acc很快就达到100%, 对照了build_graph.py的文件发现并没有什么不一样,请问可能是哪里出了问题? ![image](https://user-images.githubusercontent.com/43271630/132953084-fc611682-c287-484a-aa5d-3d8665b26d48.png)

model.py ``` #!/usr/bin/env python import torch import torch.nn as nn class GraphConvolution(nn.Module): def __init__(self, input_dim, output_dim,support, \ act_func=None,featureless=False,dropout_rate=0., \ bias=False): super(GraphConvolution, self).__init__() self.support = support self.featureless = featureless for i...