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

Add approach to flexibly inject system_messages #1087

Draft
wants to merge 2 commits into
base: develop
Choose a base branch
from

Conversation

davidberenstein1957
Copy link
Member

@davidberenstein1957 davidberenstein1957 commented Dec 24, 2024

I saw the use_system_prompt argument was supposed to be removed in the 1.5 release and it seemed like a good excuse to allow flexible system prompt injection across all tasks.

For the synthetic data generator, I noticed this was a limitation for a more dynamic and diverse generation.

I am unsure if it is relevant for things like URIAL, and some system prompts seem to rely on .format, so we could disable it there too, but let's iterate a bit on this PR.

We can now always do.

task = RandomTask(
    system_prompt="I generally want you to focus on this"
)
task.load()
task.process(inputs=[
    [{"role": "system", "content": "focus more on this"}],
    [{"role": "system", "content": "focus more on that"}]
])

Copy link

Documentation for this PR has been built. You can view it at: https://distilabel.argilla.io/pr-1087/

@davidberenstein1957
Copy link
Member Author

cc @davanstrien I believe you also vouched for something like this a while ago.

@davidberenstein1957 davidberenstein1957 changed the title add approach to flexibly inject system_messages Add approach to flexibly inject system_messages Dec 24, 2024
Copy link

codspeed-hq bot commented Dec 24, 2024

CodSpeed Performance Report

Merging #1087 will not alter performance

Comparing feat/align-usage-system-prompts-across-tasks (f88ded9) with develop (f1f7d77)

Summary

✅ 1 untouched benchmarks

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.

1 participant