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

Resume sweep and edit batch size #271

Open
NataliaAlves13 opened this issue Sep 3, 2024 · 3 comments
Open

Resume sweep and edit batch size #271

NataliaAlves13 opened this issue Sep 3, 2024 · 3 comments
Assignees

Comments

@NataliaAlves13
Copy link

NataliaAlves13 commented Sep 3, 2024

My job was interrupted while running the sweep command. I would like to resume the sweep without overwriting all files already completed. Is this possible?
Also, how can we increase the batch size manually? I tried changing the plans.pkl file but it just gets overwritten to 4 when training is resumed.

@NataliaAlves13 NataliaAlves13 changed the title Resume sweep Resume sweepv and edit batch size Sep 3, 2024
@NataliaAlves13 NataliaAlves13 changed the title Resume sweepv and edit batch size Resume sweep and edit batch size Sep 3, 2024
@mibaumgartner
Copy link
Collaborator

Dear @NataliaAlves13 ,

there is a nndet_sweep command which will run the sweep - it is also possible to skip the inference step if all cases were already predicted successfully.

You can only overwrite the batch size for new trainings, modifying parameters while training is not supported.

Best,
Michael

@mibaumgartner mibaumgartner self-assigned this Sep 4, 2024
@NataliaAlves13
Copy link
Author

Thank you for the reply Michael. The nndet_sweep command was interrupted after 1 day but not all cases were predicted yet. Is it possible to resume the inference only for cases that haven't been predicted? In any case, how can the inference step be skipped if all cases are predicted successfully?
I do want to overwrite the batch size for a new training, how can I do that?

@mibaumgartner
Copy link
Collaborator

Hi,

  1. It is currently not implemented to only predict cases which are missing in the sweep step.
  2. Sorry, just noticed that we haven't introduced the flag in the github version yet .. You can simply modify this line:
    run_prediction=True, # TODO: add commmand line arg
  3. The most robust way is to change the plan file.

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