Skip to content

fix: Use LARGE models for responses#853

Merged
jkbrooks merged 1 commit into
mainfrom
shaw/upgrade-models
Dec 5, 2024
Merged

fix: Use LARGE models for responses#853
jkbrooks merged 1 commit into
mainfrom
shaw/upgrade-models

Conversation

@lalalune

@lalalune lalalune commented Dec 5, 2024

Copy link
Copy Markdown
Member

This PR makes the small models big. Apparently we've been using small models for interactions and message responses, but this is wrong.

@lalalune lalalune changed the title Make models big Use LARGE models for responses Dec 5, 2024
@shakkernerd shakkernerd changed the title Use LARGE models for responses fix: Use LARGE models for responses Dec 5, 2024

@mcp97 mcp97 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

LGTM! Thanks so much for doing this :) Was always confused by this lol - appreciate the fix!

@jkbrooks jkbrooks merged commit 05f279a into main Dec 5, 2024
@jkbrooks jkbrooks deleted the shaw/upgrade-models branch December 5, 2024 19:28
dsldsl pushed a commit to dsldsl/eliza that referenced this pull request Dec 7, 2024
Kevin-Mok pushed a commit to Kevin-Mok/eliza that referenced this pull request Dec 11, 2024
@seanbhart

Copy link
Copy Markdown

Why hardcode any values? Hardcoding makes the model settings very limited and forces developers to fork the library just to use a different model. Allowing them to be set in the character file is much more useful. This is what the character file model settings used to look like:

{
    "settings": {
        "model": "medium",
        "temperature": 0.45
    }
}

This PR allows this much more flexible approach:
#845

lalalune pushed a commit that referenced this pull request May 3, 2026
fix: Use LARGE models for responses
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.

4 participants