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

Can't properly edit the seed for training #262

Open
Gylthas opened this issue Dec 2, 2024 · 4 comments
Open

Can't properly edit the seed for training #262

Gylthas opened this issue Dec 2, 2024 · 4 comments

Comments

@Gylthas
Copy link

Gylthas commented Dec 2, 2024

I was trying to train a lora again after doing some slight adjustments on the database and I wanted to reuse the same seed that the previous training used.

When I attempted editing the seed on the trainer UI it was impossible to set the seed to the previous one, as it had too many digits and it wasn't accepting because it was bigger than the maximum seed value. I tried editing the .py files to see if that was enough, but whenever I made an edit that was higher than certain values the UI stopped working.

Currently we're stuck not being able to replicate the previous seed and don't know if there's any other way to be able to apply it. Is there a way to actually rise the maximum seed value accepted by the UI? The previous one was randomly generated while using your scripts, so I don't know why it can produce a value that won't be able to be replicated later on.

@derrian-distro
Copy link
Owner

the seed is never randomly generated, it's always default set to "23" if you didn't touch it, that should be the seed. you must have done something really crazy to get a randomly generated seed. that being said, yes I can update that field to integer limit numbers. However it won't be instant, due to lack of time to work on this project

@Jelosus2
Copy link
Collaborator

Jelosus2 commented Dec 2, 2024

Should I make a PR about it? Shouldn't be hard to implement

@derrian-distro
Copy link
Owner

sure, it shouldn't be hard to implement, so there's no harm if you want to pr it

@Gylthas
Copy link
Author

Gylthas commented Dec 8, 2024

Sorry for not answering back earlier. We got the seed randomly generated by setting it to "0".

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

3 participants