Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

evaluate.py #2

Open
lugeluge opened this issue May 22, 2019 · 5 comments
Open

evaluate.py #2

lugeluge opened this issue May 22, 2019 · 5 comments

Comments

@lugeluge
Copy link

你好,能提供预训练模型和evaluate.py文件吗

@HongChow
Copy link

你好,能提供预训练模型和evaluate.py文件吗

我正在训练, 你这边训练好了么??

@hengyeliu
Copy link

你好,能提供预训练模型和evaluate.py文件吗

我正在训练, 你这边训练好了么??

你好 你这边训练的模型能达到论文的精度吗?

@luopeixiang
Copy link
Owner

代码更新了。效果在README中有写。比起论文的精度要差一点。没有对超参数进行微调,并且解码的结果中还存在exposure bias的问题,我尝试加入schedule sampling解决,后续有进展会继续更新。

@hengyeliu
Copy link

hengyeliu commented Sep 26, 2019 via email

@luopeixiang
Copy link
Owner

font{ line-height: 1.6; } ul,ol{ padding-left: 20px; list-style-position: inside; } 您好:   非常感谢您更新代码,然后我这边现在遇到一个问题就是我在用哈佛的脚本(https://github.com/harvardnlp/im2markup)处理标签的时候遇到很多问题,例如:原本的标签是f,\tilde{f}\to%E5%A4%84%E7%90%86%E6%A0%87%E7%AD%BE%E7%9A%84%E6%97%B6%E5%80%99%E9%81%87%E5%88%B0%E5%BE%88%E5%A4%9A%E9%97%AE%E9%A2%98%EF%BC%8C%E4%BE%8B%E5%A6%82%EF%BC%9A%E5%8E%9F%E6%9C%AC%E7%9A%84%E6%A0%87%E7%AD%BE%E6%98%AFf,%5Ctilde%7Bf%7D%5Cto) 1,\hspace{3mm} a,\tilde{a}\to 1\hspace{3mm} (r\to \infty). 对应的公式是 然后用哈佛的脚本正则化之后的标签是 f , \tilde { f } \to 1 , \hspace { 3 m m } a , \tilde { a } \to 1 \hspace { 3 m m } ( r \to \infty ) . 对应的公式却是也就是说哈佛的脚本处理标签还是会有很多问题,所以我想请问一下您这边是怎么处理标签以及图片数据的。谢谢。祝好 刘恒业 On 9/26/2019 19:15,luopeixiangnotifications@github.com wrote: 代码更新了。效果在README中有写。比起论文的精度要差一点。没有对超参数进行微调,并且解码的结果中还存在exposure bias的问题,我尝试加入schedule sampling解决,后续有进展会继续更新。 —You are receiving this because you commented.Reply to this email directly, view it on GitHub, or mute the thread.

http://lstm.seas.harvard.edu/latex/data/
你可以去这个网址下处理好的数据。
我的模型就是根据这些数据训练的。

Repository owner deleted a comment from liuzhebaba Feb 23, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

4 participants