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

Incorrect identify system prompt while generate SDG #406

Open
alfandindarahmawan opened this issue Nov 22, 2024 · 5 comments
Open

Incorrect identify system prompt while generate SDG #406

alfandindarahmawan opened this issue Nov 22, 2024 · 5 comments

Comments

@alfandindarahmawan
Copy link

Hi team,
While i running SDG used mixtral-8x7b as the teacher and the model for student i used granint-3.0-2b-instruct. the system prompt on sdg result like this .

I am a Red Hat® Instruct Model, an AI language model developed by Red Hat and IBM Research based on the granite-3.0-8b-base model. My primary role is to serve as a chat assistant.

Which is incorrect because i used granint-3.0-2b-instruct but the data result form sdg set the system prompt is granite-3.0-8b-base .

@mairin
Copy link
Member

mairin commented Nov 22, 2024

@alfandindarahmawan also found the likely place in the code causing this:

https://github.com/instructlab/instructlab/blob/main/src%2Finstructlab%2Fcommon.py

@cdoern
Copy link
Contributor

cdoern commented Nov 22, 2024

I think, generally, that sys prompt is for any granite 3.0 architecture model.

If there is a specific one per-model in the granite 3 family, this won't scale well in adding support for new models, but @jaideepr97 can comment more here.

@jaideepr97
Copy link
Member

I think, generally, that sys prompt is for any granite 3.0 architecture model.

If there is a specific one per-model in the granite 3 family, this won't scale well in adding support for new models, but @jaideepr97 can comment more here.

what @cdoern said
currently we only have 1 sys prompt for all granite based models

@mairin
Copy link
Member

mairin commented Nov 22, 2024 via email

@Maxusmusti
Copy link
Contributor

Maxusmusti commented Nov 22, 2024

Yes, this is what we wanted to do for earlier versions, but had run into some initial push back. This will definitely be revisited soon.

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

5 participants