Skip to content

vulkan: add noncontiguous GLU support#21081

Merged
0cc4m merged 2 commits into
masterfrom
0cc4m/vulkan-glu-noncontig
Mar 28, 2026
Merged

vulkan: add noncontiguous GLU support#21081
0cc4m merged 2 commits into
masterfrom
0cc4m/vulkan-glu-noncontig

Conversation

@0cc4m
Copy link
Copy Markdown
Contributor

@0cc4m 0cc4m commented Mar 27, 2026

This was missing for Mistral-Small-4 to work fully on Vulkan.

Qwen3.5 122B Q4_0 wrote the code, fully locally. I edited some minor style problems and fixed a compile issue that I ran into. Quite impressive for a local model.

@0cc4m 0cc4m requested a review from a team as a code owner March 27, 2026 18:51
@github-actions github-actions Bot added Vulkan Issues specific to the Vulkan backend ggml changes relating to the ggml tensor library for machine learning labels Mar 27, 2026
@CISC
Copy link
Copy Markdown
Member

CISC commented Mar 27, 2026

Cool, I thought Vulkan already supported this (or at least contiguous_1). :)

I think Qwen3.5 has been extensively trained on llama.cpp, it seems to know a lot about it.

@ggerganov
Copy link
Copy Markdown
Member

Qwen3.5 122B Q4_0 wrote the code, fully locally.

Which client do you use?

@0cc4m
Copy link
Copy Markdown
Contributor Author

0cc4m commented Mar 27, 2026

This was OpenCode, but frankly the default config of that project is really sketchy. Every edit or command is allowed, and it autoformats entire files when they get edited. I don't know what they are thinking. I'll look around for something saner, otherwise I'll have to spend some time checking and configuring everything.

Copy link
Copy Markdown
Member

@ggerganov ggerganov left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah. With native tools in llama-server added soon, we'll just be able to use the built-in WebUI for such tasks.

@0cc4m 0cc4m merged commit 0eb4764 into master Mar 28, 2026
46 of 51 checks passed
@0cc4m 0cc4m deleted the 0cc4m/vulkan-glu-noncontig branch March 28, 2026 07:45
slartibardfast pushed a commit to slartibardfast/llama.cpp that referenced this pull request Apr 12, 2026
* vulkan: add noncontiguous GLU support

* fix compile issue
Seunghhon pushed a commit to Seunghhon/llama.cpp that referenced this pull request Apr 26, 2026
* vulkan: add noncontiguous GLU support

* fix compile issue
rsenthilkumar6 pushed a commit to rsenthilkumar6/llama.cpp that referenced this pull request May 1, 2026
* vulkan: add noncontiguous GLU support

* fix compile issue
ljubomirj pushed a commit to ljubomirj/llama.cpp that referenced this pull request May 6, 2026
* vulkan: add noncontiguous GLU support

* fix compile issue
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ggml changes relating to the ggml tensor library for machine learning Vulkan Issues specific to the Vulkan backend

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants