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.

Sign up for GitHub

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

Jump to bottom

flickr 8k 数据集bleu-4复现问题？ #2

Open
baiyuting opened this issue Nov 12, 2022 · 6 comments
Open

flickr 8k 数据集bleu-4复现问题？ #2

baiyuting opened this issue Nov 12, 2022 · 6 comments

Comments

@baiyuting
Copy link

image
我跑的时候，仅仅使用 flickr 8k，得到如上结果，是不是过小了？其他设置没变，我看论文中是0.1549
@xu-shitong
Copy link
Owner

是的，这个值应该至少在0.1以上。你运行的是CLIP-DDPM.py吗，summary输出的trial name是什么？

@baiyuting
Copy link
Author

我运行的是 CLIP-DDPM.ipynb, trial_name 是 epoch15_lossseries_sum_sample_mean_lr1E-04-5E-05_schedulerlinspace_round3E-01_dynamic-1_clipconcat_class_weight0E+00_class_prob2E-01_train-embedFalse_samplesize100_x_0_predictTrue_X_INTERVAL100_use_x_tTrue_use_x_1True_use_probTrue

@baiyuting
Copy link
Author

我把 CLIP-DDPM.ipynb 中 flickr30 的地方注释了
image

@xu-shitong
Copy link
Owner

你可以尝试下重新输出trial name看超参数是不是正确的，因为如果你有load模型进行预测那超参数不同会导致预测和训练结果差很多。或者检查summary文件中的train和evaluation loss，metrics evaluation 第一个cell有示范plot函数。

@baiyuting
Copy link
Author

我检查了一下超参数，batch_size 设置为2了，原因是 gpu 不太够了，batch_size 会不会非常影响性能？

@xu-shitong
Copy link
Owner

这我就不太清楚了，实验的时候没有尝试过这么小的batch size

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@baiyuting @xu-shitong