Closed nlpofwhat closed 4 years ago
可以对比一下代码 以及论文中的细节 看看有没有不同的地方 我展示的例子就是训练的结果
我只是发现你的代码和原来那个在实现attention的时候少了个偏置其他的没发现什么问题🤣------------------ 原始邮件 ------------------ 发件人: "Tobias Lee"notifications@github.com 发送时间: 2019年11月11日(星期一) 中午11:33 收件人: "TobiasLee/MTA-LSTM-TensorFlow"MTA-LSTM-TensorFlow@noreply.github.com; 抄送: "learnby"952167295@qq.com;"Author"author@noreply.github.com; 主题: Re: [TobiasLee/MTA-LSTM-TensorFlow] composition数据集上测试 (#9)
可以对比一下代码 以及论文中的细节 看看有没有不同的地方 我展示的例子就是训练的结果
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
大佬可以加个好友吗?想请教您一些问题
------------------ 原始邮件 ------------------ 发件人: Tobias Lee <notifications@github.com> 发送时间: 2019年11月20日 08:50 收件人: TobiasLee/MTA-LSTM-TensorFlow <MTA-LSTM-TensorFlow@noreply.github.com> 抄送: learnby <952167295@qq.com>, Author <author@noreply.github.com> 主题: 回复:[TobiasLee/MTA-LSTM-TensorFlow] composition数据集上测试 (#9)
Closed #9.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
为什么我用composition数据集训练效果非常差啊,没有一个句子能看的