ggml webgpu: add support for soft_max, optimize rms_norm #16357
Merged
Conversation
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
github-actions
bot
added
testing
ggerganov
approved these changes
Oct 2, 2025
Co-authored-by: Georgi Gerganov <[email protected]>
gabe-l-hart
added a commit
to gabe-l-hart/llama.cpp
that referenced
this pull request
Oct 3, 2025
* origin/master: (124 commits) metal : fix loop bound in ggml_mem_ranges (ggml-org#16412) llama : fix shapes for bert/mpt q/k norm (ggml-org#16409) ggml : fix graph reallocation with multiple chunks (ggml-org#16396) Fix missing messages on sibling navigation (ggml-org#16408) vulkan: Replace uses of maxMemoryAllocationSize and VK_WHOLE_SIZE (ggml-org#16354) vulkan: Fix FA coopmat1 invalid array indexing (ggml-org#16365) ci : change macos-13 to macos-15-intel (ggml-org#16401) Capture model name only after first token (streaming) or completed request (ggml-org#16405) vulkan: in flash attention, bounds check against nem1 (don't rely on GGML_KQ_MASK_PAD) (ggml-org#16316) webui : Fix messages payload sent to chat completions (ggml-org#16402) fix: track viewportHeight via window.innerHeight to avoid unwanted scrolling (ggml-org#16356) test-barrier : do not use more threads than physically available (ggml-org#16389) ggml webgpu: add support for soft_max, optimize rms_norm (ggml-org#16357) model : Apertus model implementation (ggml-org#15852) musa: update compile flags (ggml-org#16265) ci : fix ubuntu-latest-cmake-rpc (disable ccache) (ggml-org#16388) ci: update vulkan ci (ggml-org#16294) ci : fix clean-up of old logs (ggml-org#16381) SYCL: Update to oneAPI 2025.2 (ggml-org#16371) HIP: add IMbackK to codeowner (ggml-org#16375) ...
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.
This PR adds support for SOFT_MAX, and moves RMS_NORM to use a similar split-row approach for efficiency.
ggml_soft_max_ext_inplacefunction so thattest-backend-opscan run inplace tests.