Skip to content

Conversation

@klopsahlong
Copy link
Collaborator

In this PR, we make a few small updates to the SignatureOptimizer teleprompter:

  1. We fix a bug in the previous implementation, where the prompt_model defaulted to a string, which threw an error. We now default prompt_model to None, and use the default lm as our prompt model when this is the case.
  2. We fix a bug where duplicate prompt candidates were being included in the prompt history.
  3. We add in the ability to track statistics of the quality of prompts found over each iteration of the SignatureOptimizer algorithm with the field track_stats

@okhat okhat merged commit 6b9ec94 into stanfordnlp:main Dec 24, 2023
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.

3 participants