mrwu-mac / DIFNet

This repository is for the paper ``DIFNet: Boosting Visual Information Flow for Image Captioning'' (CVPR 2022).
BSD 3-Clause "New" or "Revised" License
20 stars 7 forks source link

关于loss大小问题的咨询 #7

Closed Markkk111 closed 1 year ago

Markkk111 commented 1 year ago

你好,请问这个训练的损失大致要到达多少啊?我的到达0.007了,可以减少epoch吗?我的情况是:epoch到达20之后的每次epoch要长达2~3h。

mrwu-mac commented 1 year ago

你好,感谢你对我们的工作感兴趣。epoch到达20之后为强化学习阶段(用来解决交叉熵训练阶段的exposure bias问题),该阶段损失很小,暂无参考性。建议你通过tensorboard查看log文件里的详细指标变化情况作为参考。

------------------ 原始邮件 ------------------ 发件人: "mrwu-mac/DIFNet" @.>; 发送时间: 2023年3月2日(星期四) 晚上8:26 @.>; @.***>; 主题: [mrwu-mac/DIFNet] 关于loss大小问题的咨询 (Issue #7)

你好,请问这个训练的损失大致要到达多少啊?我的到达0.007了,可以减少epoch吗?我的情况是:epoch到达20之后的每次epoch要长达2~3h。

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: @.***>

Markkk111 commented 1 year ago

谢谢您的回答,有问题再来讨教,,salute!