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

对于有label的测试集准确率低的问题 #142

Open
uncle-tou opened this issue Apr 20, 2023 · 4 comments
Open

对于有label的测试集准确率低的问题 #142

uncle-tou opened this issue Apr 20, 2023 · 4 comments

Comments

@uncle-tou
Copy link

您好,我将自己的数据集按照格式整理好放入程序中运行,所有数据集都包含了label,测试集格式与训练集相同。运行后训练集准确率可达99.8%,但测试集及验证集准确率均只有0.01甚至更低。请问这是正常的吗?若不正常,我应该如何修改代码呢?

@yao8839836
Copy link
Owner

yao8839836 commented Apr 20, 2023

@uncle-tou 您好,不正常,估计要检查一下处理完的ty变量跟原始测试集label能否对应上

@uncle-tou
Copy link
Author

您好,我检查了我的数据集(为THUCNews数据集),处理后的格式与例程自带的原始数据集一致。我又分别用R8和oshume数据集试了一次,都是会有同样的结果:
Epoch: 0001 train_loss= 2.30258 train_acc= 0.09851 val_loss= 2.30259 val_acc= 0.09850 time= 12.74146
Epoch: 0002 train_loss= 2.29527 train_acc= 0.87257 val_loss= 2.30259 val_acc= 0.09850 time= 10.11644
Epoch: 0003 train_loss= 2.27775 train_acc= 0.86065 val_loss= 2.30259 val_acc= 0.09850 time= 9.97561
Epoch: 0004 train_loss= 2.25075 train_acc= 0.85864 val_loss= 2.30259 val_acc= 0.09850 time= 9.80227
Epoch: 0005 train_loss= 2.21337 train_acc= 0.85752 val_loss= 2.30259 val_acc= 0.09850 time= 10.08559
Epoch: 0006 train_loss= 2.16524 train_acc= 0.86353 val_loss= 2.30259 val_acc= 0.09850 time= 9.75813
Epoch: 0007 train_loss= 2.10707 train_acc= 0.86670 val_loss= 2.30259 val_acc= 0.09850 time= 9.95471
...
即在所有epoch中验证集的损失和准确率均一直保持不变。最后的结果中,测试集的第一种label召回率为1,其余为0。我在build-graph时检查过ty和y,都是没有问题的。
验证集是从训练集中分离出来的(不知道我有没有理解错),那么在训练集没问题的情况下,验证集不应该出现问题呀。

(PS.由于我无法安装TensorFlow2.0以下的版本,为了程序能够运行,我进行了一些必要的语句修改(皆针对TensorFlow报错而修改),不知道这么做是否与这种异常情况相关。

(PPS.另外,我发现本例程的#130问题与我有相似的疑问:https://github.com/yao8839836/text_gcn/issues/130。有可能错误原因跟我一样

@uncle-tou
Copy link
Author

问题解决了,问题就出在TensorFlow的代码修改上,我重新修改了部分语句,现已可成功运行,验证集测试集皆无误。

@tianranlan
Copy link

问题解决了,问题就出在TensorFlow的代码修改上,我重新修改了部分语句,现已可成功运行,验证集测试集皆无误。

你好,可以分享一下经验吗

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

3 participants