This repository has been archived by the owner on Nov 3, 2023. It is now read-only.
Friends Dataset Teacher Add speakers
field and flag to exclude speaker labels in text
#4693
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Outdated suggestions cannot be applied.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.
Patch Description
This patch adds a
speakers
field in the message generated from the teacher for the Friends dataset.It also adds an option to exclude speaker ids from the
text
. This option is named--include-speaker-in-context
. The original--include-speaker-in-context
flag, which determines whether speaker labels are added at the end oftext
, is renamed to--add-speaker-to-context-end
.Together, these changes make it convenient to feed the data into downstream models such as BlenderBot2 that require cleaned
text
input without speaker labels, but need to restore the speaker labels later.Sample Output
parlai dd -t friends -n 2 --verbose --include-speaker-in-context True --add-speaker-to-context-end True
This is the default behavior.
parlai dd -t friends -n 2 --verbose --include-speaker-in-context True --add-speaker-to-context-end False
parlai dd -t friends -n 2 --verbose --include-speaker-in-context False --add-speaker-to-context-end True
Notice the empty line which marks an empty sentence, and notice how the current speaker label is also added to the
speakers
field.parlai dd -t friends -n 2 --verbose --include-speaker-in-context False --add-speaker-to-context-end False