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

How to find the best parameter #2026

Closed
wenwenshenqihailuo opened this issue Jan 9, 2025 · 3 comments
Closed

How to find the best parameter #2026

wenwenshenqihailuo opened this issue Jan 9, 2025 · 3 comments

Comments

@wenwenshenqihailuo
Copy link

Optimizer have so much parameter inside such as max_bootstrapped_demos, max_labeled_demos so any method or way I can find out the best parameter number is fit for me?

@okhat
Copy link
Collaborator

okhat commented Jan 9, 2025

We will be updating the default parameters in 2.6 to make them "pretty good" out of the box.

Unfortunately the old defaults are not ideal. In general, you can configure them by thinking about what they mean:

  • max_bootstrapped_demos: how many examples do you want to be "bootstrapped" and then their trajectories used as few-shot examples for each modules?
  • max_labeled_demos: how many examples from your training set do you want to be used as-is as few shot examples?

etc.

@wenwenshenqihailuo
Copy link
Author

but case by case some parameter may not fit my program

@wenwenshenqihailuo
Copy link
Author

for me I want do some text generation job

@okhat okhat closed this as completed Jan 13, 2025
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

2 participants