Add a two-class variant of the MNLI task#976
Merged
sleepinyourhat merged 4 commits intomasterfrom Dec 20, 2019
Merged
Conversation
sleepinyourhat
added
low-priority
sleepinyourhat
requested review from
W4ngatang,
iftenney and
pruksmhc
as code owners
W4ngatang
approved these changes
Dec 20, 2019
Collaborator
W4ngatang
left a comment
W4ngatang
left a comment
There was a problem hiding this comment.
I think a more robust way to implement this would be to have a variable (and global, so we can share between NLI tasks) label indexing instead of a specific scorer, but seems like it'll work as is.
phu-pmh
pushed a commit
that referenced
this pull request
Apr 17, 2020
* Part I * Add metric * Clarifications * Update tasks.py
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
This adds the option to train a model on the three-class MNLI dataset and evaluate it on comparable two-class datasets like RTE or HANS. It does this by overriding the usual accuracy metric with an NLI-specific accuracy metric that collapses 'neutral' and 'contradiction'.