UPSTREAM PR #18102: ggml-cuda: Delta-Net linear attention for Qwen3-Next#593
Open
UPSTREAM PR #18102: ggml-cuda: Delta-Net linear attention for Qwen3-Next#593
Conversation
loci-dev
had a problem deploying
to
PROD__AL_DEMO
GitHub Actions
Failure
loci-dev
temporarily deployed
to
PROD__AL_DEMO
GitHub Actions
Inactive
|
Explore the complete analysis inside the Version Insights |
loci-dev
force-pushed
the
main
branch
23 times, most recently
from
ab5b02c to
2f30a3d
Compare
loci-dev
force-pushed
the
main
branch
30 times, most recently
from
086f09a to
15838f1
Compare
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
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Mirrored from ggml-org/llama.cpp#18102
cuda kernel for delta-net linear attention layers in qwen3next.
adds GGML_OP_DELTA_NET + recurrent kernel for decode, blackwell path (sm12.0+) for prefill with 64k shmem. also improved solve_tri for the chunked prefill path.
getting ~45-55 t/s on q4/mxfp4 and ~40 t/s bf16 on 80B-A3B (blackwell). pre-blackwell cards get ~38-40 t/s from solve_tri improvements (baseline was the original ~20 t/s).
Edit: omitted some small bits.