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

Training loop #10

Open
arthursw opened this issue Sep 10, 2020 · 3 comments
Open

Training loop #10

arthursw opened this issue Sep 10, 2020 · 3 comments

Comments

@arthursw
Copy link

Hello,

My model has been training for two weeks now (it seems doing well) and it keeps restarting from scratch.
I am not sure to understand how does the training loop work?
When will it stop?

Thanks!

@HuahongZhang
Copy link
Collaborator

Yeah, at default, it will do 5-fold cross-validation and you will get 5 models after successful training.

@arthursw
Copy link
Author

arthursw commented Sep 11, 2020

All right, thank you!

It trained until epoch 250 for the 1st validation, 200 epochs for the 2nd validation, and then 530 for the third validation ; that is a big difference so I thought there might be something I misunderstood.
The training is supposed to stop if it does not improves over 160 epochs, maybe I was just unlucky with the third training.

How important is it to run the five validations? Will it drastically improve the results compared to a one shot training?

@HuahongZhang
Copy link
Collaborator

I would suggest running all five-folds and then use the five models as an ensemble. But it is OK to stop running after some of them, e.g. 3 of 5, are successfully trained and only use those checkpoints.

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

2 participants