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

Suggestion: Latest pkl resume #18

Open
chrsmlls333 opened this issue Jan 16, 2021 · 2 comments
Open

Suggestion: Latest pkl resume #18

chrsmlls333 opened this issue Jan 16, 2021 · 2 comments

Comments

@chrsmlls333
Copy link

I've come across @RoyWheels discussing non-square training on reddit and noticed he had reimplemented auto-discovery of the latest pkl in results folder, as well as parsing the kimg number from that filename.

Relevant Commits:

@dvschultz Would you be open to a pull request of this? Is there a reason its not already implemented, as I remember your stylegan2 fork having this.

@dvschultz
Copy link
Owner

You’re welcome to PR this. I personally found people more likely got confused with the auto-discovery so haven’t implemented it myself.

I just pushed an update so that you can do all of this manually if you want this control without the auto-discovery

--nkimg: set start kimg count
--initstrength: set start augmentation strength
--resume: path to pkl

@chrsmlls333
Copy link
Author

Okay no worries, that's why I asked. I won't change the arg default.

I'm a little confused on --initstrength, what is ideal behaviour to make resuming as close to continuous training as possible?

Testing feature branch now, I will shoot the PR soon. This will be my first one so bear with me, thanks.

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