TransferNet

TransferNet copied to clipboard

有部分数据集的链接打不开了

8

Movie QA WebQSP 和Glove的链接打不开了，能分享一下吗？谢谢了

123wangyu244

MovieQA的数据集打不开，能够重新提供下链接吗

1

MoviceQA的数据集打不开，链接外网也打不开，请问有其他的链接吗

justyyau

Neither CWQ nor WebQSP can meet the results reported in your pape

1

Hello, first of all, I would like to express my gratitude for your open source work. After I downloaded your source code and ran it on my own machine, neither...

nenuzhangkuan

Hi, This code can't reproduce your paper results on webqsp. I find some other researchers also can't reproduce. Please give the solutions. Thanks!

JBoRu

你好，想请问一下实验中CWQ数据集跑到48.6的实验设定是怎样的。 我使用代码中的设置直接跑起来，多次尝试，测试效果都只有0.415左右

sitaocheng

感谢分享代码！在您的论文中loss使用的是欧几里得距离，但我看不懂您代码中计算loss的部分，特别是参数entity_range和weight是什么，以及公式 torch.sum(entity_range * weight * torch.pow(last_e - answers, 2)) / torch.sum(entity_range * weight)的具体含义，请问您可以讲讲吗？感谢！

2213500360

hi @shijx12 I see you use `entity_range` in the project to add special weight to entities in the sub_graph of the topic entity. But I have a question why do...

mug2mag

@shijx12 hi Thanks for sharing the great job. I think the model proposed in your project may have a high complexity of computation and large memory storage problems because you...

mug2mag

Only part of the data can be accessed

3

Hello, I found that I can only access part of the linked data, I can't access MetaQA,MovieQA and GloVe 300d pretrained vector, is it convenient to provide other forms of...

CreaterLL

WebquestionSP result lower than 0.714

2

 This is the experimental result on WebQSP(No code changes) My question is that you said the WebQSP had 572 predicates in the paper. And your code...

xianghuisun