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 Tests for extend generation logic #1184

Closed

Conversation

mikeedjones
Copy link
Contributor

@mikeedjones mikeedjones commented Jun 21, 2024

Added tests for extend generation logic to avoid any further breaking changes (#920) which must then be reverted (#1169)

I'm not sure the behaviour currently exhbitied is intended? I have included the generation logs as a comment in each test to illustrate how the lm is being prompted to show the DummyLM responses in context.

Adds tests to illustrate issue #1232 and to test against any solution for that issue.

Cheers

@arnavsinghvi11
Copy link
Collaborator

Hi @mikeedjones , thanks for the PR. can you rebase to main and push to retrieve the latest tests?

@mikeedjones
Copy link
Contributor Author

Hi @arnavsinghvi11 - updated - thanks!

@mikeedjones
Copy link
Contributor Author

Hi @arnavsinghvi11 Is there anything else which needs to be added to this?

@mikeedjones
Copy link
Contributor Author

@arnavsinghvi11 - Anything which needs to be done to merge these tests?

@mikeedjones
Copy link
Contributor Author

mikeedjones commented Oct 3, 2024

Hi, @arnavsinghvi11, it looks like the use case these tests apply to is no longer supported in dspy 2.5? A ValueError is thrown by ChatAdapter if the LM doesn't reproduce every output field. Is there a plan to put support for that back in to a future release? I know that extend-generation capability has been used by a few of the optimizers in the past.

@okhat okhat closed this Oct 10, 2024
@okhat okhat reopened this Oct 10, 2024
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