How to find the best model

[HFrancisla]

Thank you for your work, I have a question.

Good metrics on the validation set during training don't necessarily guarantee good metrics during testing. When I save so many models (e.g., saving every 4000 iterations), how can I quickly identify which one will perform better during testing?

Previously, my approach was to select several models that performed relatively well on the validation set during training, test all of them, compare the results, and keep the best one. However, this is too tedious. Do you have any better methods?

[Mytttttttt]

这是来自QQ邮箱的假期自动回复邮件。你好，我最近正在休假中，无法亲自回复你的邮件。我将在假期结束后，尽快给你回复。