Repository navigation
chore(sdk): bump llama.cpp submodule to 94256114c - #1483
Merged
Merged
Conversation
Bumps third-party/llama.cpp from c550d2f60 to upstream master tip 94256114c (49 commits): Hexagon buffer/DMA improvements, server fixes, and general ggml/CUDA/Metal/OpenCL changes. All vendored Hexagon and OpenCL patches apply cleanly against the new tip; no refresh needed. Signed-off-by: Mengsheng Wu <mengshengwu@outlook.com>
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Bumps
third-party/llama.cppfromc550d2f60(merged via #1475) to upstreammastertip94256114c(49 commits): Hexagon buffer/DMA improvements, server fixes, and general ggml/CUDA/Metal/OpenCL changes. All 6 vendored Hexagon/OpenCL patches apply cleanly against the new tip — no refresh needed.Test plan
cmake --preset arm64-linux-snapdragon-release -B build-linux . && cmake --build build-linux -jsucceeds on Linux ARM64 (Docker toolchain container) with all 6 vendored llama.cpp patches applying cleanly against the bumped tipgeniex-benchandllama-server(built with-DLLAMA_BUILD_SERVER=ON) both produce valid aarch64 binaries and run on an IQ-9075 EVK (QCS9075) deviceunsloth/gemma-4-E2B-it-GGUF(Q4_K_M) andunsloth/Qwen3.5-0.8B-GGUF(Q4_K_M) both generate correct image descriptions throughllama-server's/v1/chat/completionswith an embedded image, confirming VLM inference still works after the bump