fixed mmlu generative response extraction #2503
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
RawthiL
force-pushed
the
mmlu_generative_fix
branch
from
1b7bae7 to
97dccd3
Compare
|
Hi! looks good, to me! Just need you to increment the version. Also you can add some args to |
|
Done, also added the arguments to ignore punctuation and case, which can be useful. |
RawthiL
force-pushed
the
mmlu_generative_fix
branch
from
578dbdb to
eeaf4ce
Compare
|
I rebased to main, but I see some linting errors on files that are not part of my commit, also a failed test but I'm not sure if the problem is on my end |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Related to #2279
The
generativemodality filtering is too strict resulting in zero accuracy on many models. This PR adds a filter to recover cases like:Also, the sub-task metric was changed from
acctoexact_match, to match the aggregated metric.An example of this PR in execution can be seen in this comment #2279 (comment)