server: do not remove whitespace at the start of a completion chunk #7524
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
mofosyne
added
Apple Metal
mofosyne
approved these changes
May 25, 2024
mofosyne
added
the
merge ready
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Outdated suggestions cannot be applied.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.
When using Completion mode of the server in recent builds, the first token of the predicted text is typically appended to the prompt without a space character. This can lead to problems when manually editing the text and then predicting again, because the missing space will appear in the new prompt.
Leading spaces are explicitly trimmed in
CompletionControls(), which appears to be unnecessary after #2810 because even the initial prompt will not contain any extra leading space, and all further spaces should be resulting from predictions.So, this PR proposes removing this trimming in Completion mode (but not in Chat mode, where it is still done in
runLlama()and looks necessary to avoid an extra space between the role name and the prompt).