Repository navigation
Pin llama.cpp 706ad1a: ZAYA with several sequences per ubatch - #130
Merged
Merged
Conversation
…ONSTER/llama.cpp #15) ZAYA's CCA conv input was transposed across every sequence of a ubatch, so with more than one sequence (llama-perplexity's default batching, `1bit serve --parallel`) channels and sequences mixed: ZAYA1-8B F16 perplexity went 27.89 -> 58,258 with four sequences per batch. #15 splits the sequences before transposing; one sequence is unchanged. registry/ architectures.json regenerated for the pin. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Docs7 for 1bit-monster/engine
Commit |
|
Preparing review... |
This was referenced Oct 24, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Moves
third_party/llama.cppfrom567d2dc(#129) to706ad1a, the merge of 1bit-MONSTER/llama.cpp#15 and the head of1bit/hrx-vulkan-patched.ZAYA's CCA conv input was transposed across every sequence in a ubatch. With more than one sequence per batch, channels and sequences got mixed: that covers
1bit serve --paralleland llama-perplexity's default batching. #15 splits the sequences before transposing.Checked on Strix Halo, ZAYA1-8B Q4_K_M on Vulkan, 10 × 512-token chunks of the engine docs:
The PR author measured four sequences at 58,258 before the fix.
registry/architectures.jsonis regenerated for the pin.🤖 Generated with Claude Code