Repository navigation
common: fix draft-mtp with embeddings (#26352, #27299) - #27400
Merged
Merged
Conversation
Contributor
Author
|
Hey @CISC , would you have a look at the PR? |
CISC
approved these changes
Aug 21, 2026
pwilkin
approved these changes
Aug 21, 2026
Contributor
Author
therealkenc
pushed a commit
to therealkenc/llama.cpp
that referenced
this pull request
Aug 24, 2026
… (ggml-org#27400) * common: fix draft-mtp with embeddings (ggml-org#26352) * --whitespace --------- Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
thecodacus
pushed a commit
to thecodacus/llama.cpp
that referenced
this pull request
Sep 7, 2026
… (ggml-org#27400) * common: fix draft-mtp with embeddings (ggml-org#26352) * --whitespace --------- Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
zbrad
pushed a commit
to zbrad/llama.cpp
that referenced
this pull request
Sep 10, 2026
… (ggml-org#27400) * common: fix draft-mtp with embeddings (ggml-org#26352) * --whitespace --------- Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@huggingface.co>
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Overview
Fixes an issue where embedding and pooling_type settings from the main context were incorrectly inherited by the MTP context. The patch disables embeddings and pooling for the speculative context.
Additional information
The MTP context is used as a decoder/token-prediction context and should not produce embeddings, so embedding should be set to false. This prevents build_pooling() from being called with a null tensor and explicitly decouples the draft context's embedding/pooling settings from those requested for the main model.
Requirements