Skip to content
This repository has been archived by the owner on Oct 16, 2023. It is now read-only.

Infinity generation text #3

Open
4EPBEED opened this issue Mar 29, 2023 · 5 comments
Open

Infinity generation text #3

4EPBEED opened this issue Mar 29, 2023 · 5 comments

Comments

@4EPBEED
Copy link

4EPBEED commented Mar 29, 2023

Why pygg infinity generation text? The text is't even related to what I wrote.

@AlpinDale
Copy link
Owner

Why pygg infinity generation text? The text is't even related to what I wrote.

I'm aware of the issue. I've been looking into it.

Problem is, C++ isn't my forte so it'll take a while unless a more skilled person comes along and makes a PR.

@4EPBEED
Copy link
Author

4EPBEED commented Mar 29, 2023

Thanks for the answer. Still, your work is amazing. Hope everything works out for you!

@4EPBEED
Copy link
Author

4EPBEED commented Mar 31, 2023

I think I understand what the problem of infinite generation is. In the original, Pygmalion catches the "/nYou:" token and stops generation and passes the input to the user. It is necessary either to make changes in the dataset, or to enter a stop list of tokens.

@HRNPH
Copy link

HRNPH commented Apr 6, 2023

I Agree with "/nYou:" token catching
It'll be nice to have a one-time inference script too since I think using it through a Python wrapper will be easier just like Tesseract
I'm looking forward to implementing this into my AIWaifu_Project
Good work btw 👍

@4EPBEED
Copy link
Author

4EPBEED commented Apr 6, 2023

In some cases, I still managed to stop the infinite generation, but not often it didn’t work very well. Where |end of text| it stops there. Now I'm thinking about how to intercept the token to stop and pass the input to the user.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants