Skip to content

Conversation

@kyteinsky
Copy link
Contributor

@kyteinsky kyteinsky commented Nov 6, 2025

resolves #76

@kyteinsky
Copy link
Contributor Author

kyteinsky commented Nov 6, 2025

needs testing and admin docs update

@kyteinsky kyteinsky requested a review from marcelklehr November 6, 2025 15:20
@marcelklehr
Copy link
Member

Shall we do a major version bump for this change to make sure people realize that the default model might not fit on their GPU anymore?

@marcelklehr
Copy link
Member

ping?

@kyteinsky
Copy link
Contributor Author

ah sorry, missed this. The bigger model did not have much improvement in quality as I hoped for, maybe the examples were not good given my own language limitations :)
also, it sometimes goes generating on and on, and exhausts my GPU's 16G VRAM even when the input is not that big. The same thing happened when \n was there in the input, we sanitise the output for that, maybe there are more things offending it, or maybe the input was large but I doubt that, it was just one paragraph from wikipedia.
if the input was large, we might need chunking before we merge this.

@marcelklehr
Copy link
Member

damn, ok.

Copy link
Member

@marcelklehr marcelklehr left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The bigger model did not have much improvement in quality

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

Successfully merging this pull request may close these issues.

Duplicated results.

3 participants