dlxj commited on
Commit ·
682b227
1
Parent(s): 2517be1
update to support gemma4
Browse filesThis view is limited to 50 files because it contains too many changes. See raw diff
- .devops/cann.Dockerfile +1 -1
- .devops/cpu.Dockerfile +8 -5
- .devops/cuda-new.Dockerfile +0 -95
- .devops/cuda.Dockerfile +8 -5
- .devops/intel.Dockerfile +19 -2
- .devops/llama-cli-cann.Dockerfile +1 -1
- .devops/musa.Dockerfile +1 -1
- .devops/nix/package.nix +7 -2
- .devops/openvino.Dockerfile +1 -1
- .devops/rocm.Dockerfile +6 -6
- .devops/vulkan.Dockerfile +11 -8
- .editorconfig +8 -8
- .gitattributes +17 -7
- .github/ISSUE_TEMPLATE/010-bug-compilation.yml +1 -1
- .github/ISSUE_TEMPLATE/011-bug-results.yml +1 -1
- .github/labeler.yml +13 -0
- .github/pull_request_template.md +3 -1
- .github/workflows/ai-issues.yml +89 -0
- .github/workflows/build-android.yml +23 -62
- .github/workflows/build-cann.yml +1 -1
- .github/workflows/build-cross.yml +1 -0
- .github/workflows/build-msys.yml +1 -1
- .github/workflows/build-riscv.yml +7 -29
- .github/workflows/build-self-hosted.yml +58 -8
- .github/workflows/build-vulkan.yml +3 -2
- .github/workflows/build.yml +130 -73
- .github/workflows/close-issue.yml +1 -1
- .github/workflows/copilot-setup-steps.yml +0 -1
- .github/workflows/docker.yml +298 -83
- .github/workflows/editorconfig.yml +1 -1
- .github/workflows/gguf-publish.yml +3 -3
- .github/workflows/hip-quality-check.yml +82 -0
- .github/workflows/python-lint.yml +1 -1
- .github/workflows/python-type-check.yml +16 -11
- .github/workflows/release.yml +138 -67
- .github/workflows/server-self-hosted.yml +39 -38
- .gitignore +5 -1
- AGENTS.md +3 -81
- CMakeLists.txt +3 -285
- CODEOWNERS +20 -1
- CONTRIBUTING.md +3 -191
- README.md +3 -595
- ci/run.sh +56 -48
- cmake/arm64-linux-clang.cmake +17 -0
- common/CMakeLists.txt +3 -151
- common/arg.cpp +119 -76
- common/build-info.cpp.in +34 -3
- common/build-info.h +11 -0
- common/chat-auto-parser-generator.cpp +149 -125
- common/chat-auto-parser-helpers.cpp +18 -1
.devops/cann.Dockerfile
CHANGED
|
@@ -4,7 +4,7 @@
|
|
| 4 |
|
| 5 |
# Define the CANN base image for easier version updates later
|
| 6 |
ARG CHIP_TYPE=910b
|
| 7 |
-
ARG CANN_BASE_IMAGE=quay.io/ascend/cann:8.
|
| 8 |
|
| 9 |
# ==============================================================================
|
| 10 |
# BUILD STAGE
|
|
|
|
| 4 |
|
| 5 |
# Define the CANN base image for easier version updates later
|
| 6 |
ARG CHIP_TYPE=910b
|
| 7 |
+
ARG CANN_BASE_IMAGE=quay.io/ascend/cann:8.5.0-${CHIP_TYPE}-openeuler24.03-py3.11
|
| 8 |
|
| 9 |
# ==============================================================================
|
| 10 |
# BUILD STAGE
|
.devops/cpu.Dockerfile
CHANGED
|
@@ -1,11 +1,13 @@
|
|
| 1 |
-
ARG UBUNTU_VERSION=
|
| 2 |
|
| 3 |
FROM ubuntu:$UBUNTU_VERSION AS build
|
| 4 |
|
| 5 |
ARG TARGETARCH
|
| 6 |
|
| 7 |
RUN apt-get update && \
|
| 8 |
-
apt-get install -y build-essential git cmake libssl-dev
|
|
|
|
|
|
|
| 9 |
|
| 10 |
WORKDIR /app
|
| 11 |
|
|
@@ -34,7 +36,7 @@ RUN mkdir -p /app/full \
|
|
| 34 |
FROM ubuntu:$UBUNTU_VERSION AS base
|
| 35 |
|
| 36 |
RUN apt-get update \
|
| 37 |
-
&& apt-get install -y libgomp1 curl\
|
| 38 |
&& apt autoremove -y \
|
| 39 |
&& apt clean -y \
|
| 40 |
&& rm -rf /tmp/* /var/tmp/* \
|
|
@@ -55,8 +57,9 @@ RUN apt-get update \
|
|
| 55 |
git \
|
| 56 |
python3 \
|
| 57 |
python3-pip \
|
| 58 |
-
|
| 59 |
-
&& pip install -
|
|
|
|
| 60 |
&& apt autoremove -y \
|
| 61 |
&& apt clean -y \
|
| 62 |
&& rm -rf /tmp/* /var/tmp/* \
|
|
|
|
| 1 |
+
ARG UBUNTU_VERSION=24.04
|
| 2 |
|
| 3 |
FROM ubuntu:$UBUNTU_VERSION AS build
|
| 4 |
|
| 5 |
ARG TARGETARCH
|
| 6 |
|
| 7 |
RUN apt-get update && \
|
| 8 |
+
apt-get install -y gcc-14 g++-14 build-essential git cmake libssl-dev
|
| 9 |
+
|
| 10 |
+
ENV CC=gcc-14 CXX=g++-14
|
| 11 |
|
| 12 |
WORKDIR /app
|
| 13 |
|
|
|
|
| 36 |
FROM ubuntu:$UBUNTU_VERSION AS base
|
| 37 |
|
| 38 |
RUN apt-get update \
|
| 39 |
+
&& apt-get install -y libgomp1 curl \
|
| 40 |
&& apt autoremove -y \
|
| 41 |
&& apt clean -y \
|
| 42 |
&& rm -rf /tmp/* /var/tmp/* \
|
|
|
|
| 57 |
git \
|
| 58 |
python3 \
|
| 59 |
python3-pip \
|
| 60 |
+
python3-wheel \
|
| 61 |
+
&& pip install --break-system-packages --upgrade setuptools \
|
| 62 |
+
&& pip install --break-system-packages -r requirements.txt \
|
| 63 |
&& apt autoremove -y \
|
| 64 |
&& apt clean -y \
|
| 65 |
&& rm -rf /tmp/* /var/tmp/* \
|
.devops/cuda-new.Dockerfile
DELETED
|
@@ -1,95 +0,0 @@
|
|
| 1 |
-
ARG UBUNTU_VERSION=24.04
|
| 2 |
-
# This needs to generally match the container host's environment.
|
| 3 |
-
ARG CUDA_VERSION=13.1.0
|
| 4 |
-
# Target the CUDA build image
|
| 5 |
-
ARG BASE_CUDA_DEV_CONTAINER=nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}
|
| 6 |
-
|
| 7 |
-
ARG BASE_CUDA_RUN_CONTAINER=nvidia/cuda:${CUDA_VERSION}-runtime-ubuntu${UBUNTU_VERSION}
|
| 8 |
-
|
| 9 |
-
FROM ${BASE_CUDA_DEV_CONTAINER} AS build
|
| 10 |
-
|
| 11 |
-
# CUDA architecture to build for (defaults to all supported archs)
|
| 12 |
-
ARG CUDA_DOCKER_ARCH=default
|
| 13 |
-
|
| 14 |
-
RUN apt-get update && \
|
| 15 |
-
apt-get install -y build-essential cmake python3 python3-pip git libssl-dev libgomp1
|
| 16 |
-
|
| 17 |
-
WORKDIR /app
|
| 18 |
-
|
| 19 |
-
COPY . .
|
| 20 |
-
|
| 21 |
-
RUN if [ "${CUDA_DOCKER_ARCH}" != "default" ]; then \
|
| 22 |
-
export CMAKE_ARGS="-DCMAKE_CUDA_ARCHITECTURES=${CUDA_DOCKER_ARCH}"; \
|
| 23 |
-
fi && \
|
| 24 |
-
cmake -B build -DGGML_NATIVE=OFF -DGGML_CUDA=ON -DGGML_BACKEND_DL=ON -DGGML_CPU_ALL_VARIANTS=ON -DLLAMA_BUILD_TESTS=OFF ${CMAKE_ARGS} -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined . && \
|
| 25 |
-
cmake --build build --config Release -j$(nproc)
|
| 26 |
-
|
| 27 |
-
RUN mkdir -p /app/lib && \
|
| 28 |
-
find build -name "*.so*" -exec cp -P {} /app/lib \;
|
| 29 |
-
|
| 30 |
-
RUN mkdir -p /app/full \
|
| 31 |
-
&& cp build/bin/* /app/full \
|
| 32 |
-
&& cp *.py /app/full \
|
| 33 |
-
&& cp -r gguf-py /app/full \
|
| 34 |
-
&& cp -r requirements /app/full \
|
| 35 |
-
&& cp requirements.txt /app/full \
|
| 36 |
-
&& cp .devops/tools.sh /app/full/tools.sh
|
| 37 |
-
|
| 38 |
-
## Base image
|
| 39 |
-
FROM ${BASE_CUDA_RUN_CONTAINER} AS base
|
| 40 |
-
|
| 41 |
-
RUN apt-get update \
|
| 42 |
-
&& apt-get install -y libgomp1 curl\
|
| 43 |
-
&& apt autoremove -y \
|
| 44 |
-
&& apt clean -y \
|
| 45 |
-
&& rm -rf /tmp/* /var/tmp/* \
|
| 46 |
-
&& find /var/cache/apt/archives /var/lib/apt/lists -not -name lock -type f -delete \
|
| 47 |
-
&& find /var/cache -type f -delete
|
| 48 |
-
|
| 49 |
-
COPY --from=build /app/lib/ /app
|
| 50 |
-
|
| 51 |
-
### Full
|
| 52 |
-
FROM base AS full
|
| 53 |
-
|
| 54 |
-
COPY --from=build /app/full /app
|
| 55 |
-
|
| 56 |
-
WORKDIR /app
|
| 57 |
-
|
| 58 |
-
RUN apt-get update \
|
| 59 |
-
&& apt-get install -y \
|
| 60 |
-
git \
|
| 61 |
-
python3 \
|
| 62 |
-
python3-pip \
|
| 63 |
-
python3-wheel \
|
| 64 |
-
&& pip install --break-system-packages --upgrade setuptools \
|
| 65 |
-
&& pip install --break-system-packages -r requirements.txt \
|
| 66 |
-
&& apt autoremove -y \
|
| 67 |
-
&& apt clean -y \
|
| 68 |
-
&& rm -rf /tmp/* /var/tmp/* \
|
| 69 |
-
&& find /var/cache/apt/archives /var/lib/apt/lists -not -name lock -type f -delete \
|
| 70 |
-
&& find /var/cache -type f -delete
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
ENTRYPOINT ["/app/tools.sh"]
|
| 74 |
-
|
| 75 |
-
### Light, CLI only
|
| 76 |
-
FROM base AS light
|
| 77 |
-
|
| 78 |
-
COPY --from=build /app/full/llama-cli /app/full/llama-completion /app
|
| 79 |
-
|
| 80 |
-
WORKDIR /app
|
| 81 |
-
|
| 82 |
-
ENTRYPOINT [ "/app/llama-cli" ]
|
| 83 |
-
|
| 84 |
-
### Server, Server only
|
| 85 |
-
FROM base AS server
|
| 86 |
-
|
| 87 |
-
ENV LLAMA_ARG_HOST=0.0.0.0
|
| 88 |
-
|
| 89 |
-
COPY --from=build /app/full/llama-server /app
|
| 90 |
-
|
| 91 |
-
WORKDIR /app
|
| 92 |
-
|
| 93 |
-
HEALTHCHECK CMD [ "curl", "-f", "http://localhost:8080/health" ]
|
| 94 |
-
|
| 95 |
-
ENTRYPOINT [ "/app/llama-server" ]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
.devops/cuda.Dockerfile
CHANGED
|
@@ -1,6 +1,6 @@
|
|
| 1 |
-
ARG UBUNTU_VERSION=
|
| 2 |
# This needs to generally match the container host's environment.
|
| 3 |
-
ARG CUDA_VERSION=12.
|
| 4 |
# Target the CUDA build image
|
| 5 |
ARG BASE_CUDA_DEV_CONTAINER=nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}
|
| 6 |
|
|
@@ -12,7 +12,9 @@ FROM ${BASE_CUDA_DEV_CONTAINER} AS build
|
|
| 12 |
ARG CUDA_DOCKER_ARCH=default
|
| 13 |
|
| 14 |
RUN apt-get update && \
|
| 15 |
-
apt-get install -y build-essential cmake python3 python3-pip git libssl-dev libgomp1
|
|
|
|
|
|
|
| 16 |
|
| 17 |
WORKDIR /app
|
| 18 |
|
|
@@ -39,7 +41,7 @@ RUN mkdir -p /app/full \
|
|
| 39 |
FROM ${BASE_CUDA_RUN_CONTAINER} AS base
|
| 40 |
|
| 41 |
RUN apt-get update \
|
| 42 |
-
&& apt-get install -y libgomp1 curl\
|
| 43 |
&& apt autoremove -y \
|
| 44 |
&& apt clean -y \
|
| 45 |
&& rm -rf /tmp/* /var/tmp/* \
|
|
@@ -60,7 +62,8 @@ RUN apt-get update \
|
|
| 60 |
git \
|
| 61 |
python3 \
|
| 62 |
python3-pip \
|
| 63 |
-
|
|
|
|
| 64 |
&& pip install --break-system-packages -r requirements.txt \
|
| 65 |
&& apt autoremove -y \
|
| 66 |
&& apt clean -y \
|
|
|
|
| 1 |
+
ARG UBUNTU_VERSION=24.04
|
| 2 |
# This needs to generally match the container host's environment.
|
| 3 |
+
ARG CUDA_VERSION=12.8.1
|
| 4 |
# Target the CUDA build image
|
| 5 |
ARG BASE_CUDA_DEV_CONTAINER=nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}
|
| 6 |
|
|
|
|
| 12 |
ARG CUDA_DOCKER_ARCH=default
|
| 13 |
|
| 14 |
RUN apt-get update && \
|
| 15 |
+
apt-get install -y gcc-14 g++-14 build-essential cmake python3 python3-pip git libssl-dev libgomp1
|
| 16 |
+
|
| 17 |
+
ENV CC=gcc-14 CXX=g++-14 CUDAHOSTCXX=g++-14
|
| 18 |
|
| 19 |
WORKDIR /app
|
| 20 |
|
|
|
|
| 41 |
FROM ${BASE_CUDA_RUN_CONTAINER} AS base
|
| 42 |
|
| 43 |
RUN apt-get update \
|
| 44 |
+
&& apt-get install -y libgomp1 curl \
|
| 45 |
&& apt autoremove -y \
|
| 46 |
&& apt clean -y \
|
| 47 |
&& rm -rf /tmp/* /var/tmp/* \
|
|
|
|
| 62 |
git \
|
| 63 |
python3 \
|
| 64 |
python3-pip \
|
| 65 |
+
python3-wheel \
|
| 66 |
+
&& pip install --break-system-packages --upgrade setuptools \
|
| 67 |
&& pip install --break-system-packages -r requirements.txt \
|
| 68 |
&& apt autoremove -y \
|
| 69 |
&& apt clean -y \
|
.devops/intel.Dockerfile
CHANGED
|
@@ -1,4 +1,4 @@
|
|
| 1 |
-
ARG ONEAPI_VERSION=2025.
|
| 2 |
|
| 3 |
## Build Image
|
| 4 |
|
|
@@ -33,8 +33,25 @@ RUN mkdir -p /app/full \
|
|
| 33 |
|
| 34 |
FROM intel/deep-learning-essentials:$ONEAPI_VERSION AS base
|
| 35 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
RUN apt-get update \
|
| 37 |
-
&& apt-get install -y libgomp1 curl\
|
| 38 |
&& apt autoremove -y \
|
| 39 |
&& apt clean -y \
|
| 40 |
&& rm -rf /tmp/* /var/tmp/* \
|
|
|
|
| 1 |
+
ARG ONEAPI_VERSION=2025.3.2-0-devel-ubuntu24.04
|
| 2 |
|
| 3 |
## Build Image
|
| 4 |
|
|
|
|
| 33 |
|
| 34 |
FROM intel/deep-learning-essentials:$ONEAPI_VERSION AS base
|
| 35 |
|
| 36 |
+
ARG IGC_VERSION=v2.30.1
|
| 37 |
+
ARG IGC_VERSION_FULL=2_2.30.1+20950
|
| 38 |
+
ARG COMPUTE_RUNTIME_VERSION=26.09.37435.1
|
| 39 |
+
ARG COMPUTE_RUNTIME_VERSION_FULL=26.09.37435.1-0
|
| 40 |
+
ARG IGDGMM_VERSION=22.9.0
|
| 41 |
+
RUN mkdir /tmp/neo/ && cd /tmp/neo/ \
|
| 42 |
+
&& wget https://github.com/intel/intel-graphics-compiler/releases/download/$IGC_VERSION/intel-igc-core-${IGC_VERSION_FULL}_amd64.deb \
|
| 43 |
+
&& wget https://github.com/intel/intel-graphics-compiler/releases/download/$IGC_VERSION/intel-igc-opencl-${IGC_VERSION_FULL}_amd64.deb \
|
| 44 |
+
&& wget https://github.com/intel/compute-runtime/releases/download/$COMPUTE_RUNTIME_VERSION/intel-ocloc-dbgsym_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.ddeb \
|
| 45 |
+
&& wget https://github.com/intel/compute-runtime/releases/download/$COMPUTE_RUNTIME_VERSION/intel-ocloc_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.deb \
|
| 46 |
+
&& wget https://github.com/intel/compute-runtime/releases/download/$COMPUTE_RUNTIME_VERSION/intel-opencl-icd-dbgsym_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.ddeb \
|
| 47 |
+
&& wget https://github.com/intel/compute-runtime/releases/download/$COMPUTE_RUNTIME_VERSION/intel-opencl-icd_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.deb \
|
| 48 |
+
&& wget https://github.com/intel/compute-runtime/releases/download/$COMPUTE_RUNTIME_VERSION/libigdgmm12_${IGDGMM_VERSION}_amd64.deb \
|
| 49 |
+
&& wget https://github.com/intel/compute-runtime/releases/download/$COMPUTE_RUNTIME_VERSION/libze-intel-gpu1-dbgsym_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.ddeb \
|
| 50 |
+
&& wget https://github.com/intel/compute-runtime/releases/download/$COMPUTE_RUNTIME_VERSION/libze-intel-gpu1_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.deb \
|
| 51 |
+
&& dpkg --install *.deb
|
| 52 |
+
|
| 53 |
RUN apt-get update \
|
| 54 |
+
&& apt-get install -y libgomp1 curl \
|
| 55 |
&& apt autoremove -y \
|
| 56 |
&& apt clean -y \
|
| 57 |
&& rm -rf /tmp/* /var/tmp/* \
|
.devops/llama-cli-cann.Dockerfile
CHANGED
|
@@ -1,4 +1,4 @@
|
|
| 1 |
-
ARG ASCEND_VERSION=8.
|
| 2 |
|
| 3 |
FROM ascendai/cann:$ASCEND_VERSION AS build
|
| 4 |
|
|
|
|
| 1 |
+
ARG ASCEND_VERSION=8.5.0-910b-openeuler22.03-py3.10
|
| 2 |
|
| 3 |
FROM ascendai/cann:$ASCEND_VERSION AS build
|
| 4 |
|
.devops/musa.Dockerfile
CHANGED
|
@@ -46,7 +46,7 @@ RUN mkdir -p /app/full \
|
|
| 46 |
FROM ${BASE_MUSA_RUN_CONTAINER} AS base
|
| 47 |
|
| 48 |
RUN apt-get update \
|
| 49 |
-
&& apt-get install -y libgomp1 curl\
|
| 50 |
&& apt autoremove -y \
|
| 51 |
&& apt clean -y \
|
| 52 |
&& rm -rf /tmp/* /var/tmp/* \
|
|
|
|
| 46 |
FROM ${BASE_MUSA_RUN_CONTAINER} AS base
|
| 47 |
|
| 48 |
RUN apt-get update \
|
| 49 |
+
&& apt-get install -y libgomp1 curl \
|
| 50 |
&& apt autoremove -y \
|
| 51 |
&& apt clean -y \
|
| 52 |
&& rm -rf /tmp/* /var/tmp/* \
|
.devops/nix/package.nix
CHANGED
|
@@ -16,8 +16,9 @@
|
|
| 16 |
rocmPackages,
|
| 17 |
vulkan-headers,
|
| 18 |
vulkan-loader,
|
| 19 |
-
|
| 20 |
shaderc,
|
|
|
|
| 21 |
useBlas ?
|
| 22 |
builtins.all (x: !x) [
|
| 23 |
useCuda
|
|
@@ -41,6 +42,7 @@
|
|
| 41 |
effectiveStdenv ? if useCuda then cudaPackages.backendStdenv else stdenv,
|
| 42 |
enableStatic ? effectiveStdenv.hostPlatform.isStatic,
|
| 43 |
precompileMetalShaders ? false,
|
|
|
|
| 44 |
}:
|
| 45 |
|
| 46 |
let
|
|
@@ -144,6 +146,7 @@ effectiveStdenv.mkDerivation (finalAttrs: {
|
|
| 144 |
ninja
|
| 145 |
pkg-config
|
| 146 |
git
|
|
|
|
| 147 |
]
|
| 148 |
++ optionals useCuda [
|
| 149 |
cudaPackages.cuda_nvcc
|
|
@@ -159,11 +162,13 @@ effectiveStdenv.mkDerivation (finalAttrs: {
|
|
| 159 |
++ optionals useMpi [ mpi ]
|
| 160 |
++ optionals useRocm rocmBuildInputs
|
| 161 |
++ optionals useBlas [ blas ]
|
| 162 |
-
++ optionals useVulkan vulkanBuildInputs
|
|
|
|
| 163 |
|
| 164 |
cmakeFlags =
|
| 165 |
[
|
| 166 |
(cmakeBool "LLAMA_BUILD_SERVER" true)
|
|
|
|
| 167 |
(cmakeBool "BUILD_SHARED_LIBS" (!enableStatic))
|
| 168 |
(cmakeBool "CMAKE_SKIP_BUILD_RPATH" true)
|
| 169 |
(cmakeBool "GGML_NATIVE" false)
|
|
|
|
| 16 |
rocmPackages,
|
| 17 |
vulkan-headers,
|
| 18 |
vulkan-loader,
|
| 19 |
+
openssl,
|
| 20 |
shaderc,
|
| 21 |
+
spirv-headers,
|
| 22 |
useBlas ?
|
| 23 |
builtins.all (x: !x) [
|
| 24 |
useCuda
|
|
|
|
| 42 |
effectiveStdenv ? if useCuda then cudaPackages.backendStdenv else stdenv,
|
| 43 |
enableStatic ? effectiveStdenv.hostPlatform.isStatic,
|
| 44 |
precompileMetalShaders ? false,
|
| 45 |
+
useWebUi ? true,
|
| 46 |
}:
|
| 47 |
|
| 48 |
let
|
|
|
|
| 146 |
ninja
|
| 147 |
pkg-config
|
| 148 |
git
|
| 149 |
+
spirv-headers
|
| 150 |
]
|
| 151 |
++ optionals useCuda [
|
| 152 |
cudaPackages.cuda_nvcc
|
|
|
|
| 162 |
++ optionals useMpi [ mpi ]
|
| 163 |
++ optionals useRocm rocmBuildInputs
|
| 164 |
++ optionals useBlas [ blas ]
|
| 165 |
+
++ optionals useVulkan vulkanBuildInputs
|
| 166 |
+
++ [ openssl ];
|
| 167 |
|
| 168 |
cmakeFlags =
|
| 169 |
[
|
| 170 |
(cmakeBool "LLAMA_BUILD_SERVER" true)
|
| 171 |
+
(cmakeBool "LLAMA_BUILD_WEBUI" useWebUi)
|
| 172 |
(cmakeBool "BUILD_SHARED_LIBS" (!enableStatic))
|
| 173 |
(cmakeBool "CMAKE_SKIP_BUILD_RPATH" true)
|
| 174 |
(cmakeBool "GGML_NATIVE" false)
|
.devops/openvino.Dockerfile
CHANGED
|
@@ -78,7 +78,7 @@ ARG http_proxy
|
|
| 78 |
ARG https_proxy
|
| 79 |
|
| 80 |
RUN apt-get update \
|
| 81 |
-
&& apt-get install -y libgomp1 libtbb12 curl\
|
| 82 |
&& apt autoremove -y \
|
| 83 |
&& apt clean -y \
|
| 84 |
&& rm -rf /tmp/* /var/tmp/* \
|
|
|
|
| 78 |
ARG https_proxy
|
| 79 |
|
| 80 |
RUN apt-get update \
|
| 81 |
+
&& apt-get install -y libgomp1 libtbb12 curl \
|
| 82 |
&& apt autoremove -y \
|
| 83 |
&& apt clean -y \
|
| 84 |
&& rm -rf /tmp/* /var/tmp/* \
|
.devops/rocm.Dockerfile
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
ARG UBUNTU_VERSION=24.04
|
| 2 |
|
| 3 |
# This needs to generally match the container host's environment.
|
| 4 |
-
ARG ROCM_VERSION=7.2
|
| 5 |
-
ARG AMDGPU_VERSION=7.2
|
| 6 |
|
| 7 |
# Target the ROCm build image
|
| 8 |
ARG BASE_ROCM_DEV_CONTAINER=rocm/dev-ubuntu-${UBUNTU_VERSION}:${ROCM_VERSION}-complete
|
|
@@ -12,11 +12,11 @@ FROM ${BASE_ROCM_DEV_CONTAINER} AS build
|
|
| 12 |
|
| 13 |
# Unless otherwise specified, we make a fat build.
|
| 14 |
# This is mostly tied to rocBLAS supported archs.
|
| 15 |
-
# check https://rocm.docs.amd.com/projects/install-on-linux/en/docs-7.2.
|
| 16 |
# check https://rocm.docs.amd.com/projects/radeon-ryzen/en/latest/docs/compatibility/compatibilityrad/native_linux/native_linux_compatibility.html
|
| 17 |
# check https://rocm.docs.amd.com/projects/radeon-ryzen/en/latest/docs/compatibility/compatibilityryz/native_linux/native_linux_compatibility.html
|
| 18 |
|
| 19 |
-
ARG ROCM_DOCKER_ARCH='gfx908;gfx90a;gfx942;gfx1030;gfx1100;gfx1101;gfx1151;gfx1150;gfx1200;gfx1201'
|
| 20 |
|
| 21 |
# Set ROCm architectures
|
| 22 |
ENV AMDGPU_TARGETS=${ROCM_DOCKER_ARCH}
|
|
@@ -58,7 +58,7 @@ RUN mkdir -p /app/full \
|
|
| 58 |
FROM ${BASE_ROCM_DEV_CONTAINER} AS base
|
| 59 |
|
| 60 |
RUN apt-get update \
|
| 61 |
-
&& apt-get install -y libgomp1 curl\
|
| 62 |
&& apt autoremove -y \
|
| 63 |
&& apt clean -y \
|
| 64 |
&& rm -rf /tmp/* /var/tmp/* \
|
|
@@ -79,7 +79,7 @@ RUN apt-get update \
|
|
| 79 |
git \
|
| 80 |
python3-pip \
|
| 81 |
python3 \
|
| 82 |
-
python3-wheel\
|
| 83 |
&& pip install --break-system-packages --upgrade setuptools \
|
| 84 |
&& pip install --break-system-packages -r requirements.txt \
|
| 85 |
&& apt autoremove -y \
|
|
|
|
| 1 |
ARG UBUNTU_VERSION=24.04
|
| 2 |
|
| 3 |
# This needs to generally match the container host's environment.
|
| 4 |
+
ARG ROCM_VERSION=7.2.1
|
| 5 |
+
ARG AMDGPU_VERSION=7.2.1
|
| 6 |
|
| 7 |
# Target the ROCm build image
|
| 8 |
ARG BASE_ROCM_DEV_CONTAINER=rocm/dev-ubuntu-${UBUNTU_VERSION}:${ROCM_VERSION}-complete
|
|
|
|
| 12 |
|
| 13 |
# Unless otherwise specified, we make a fat build.
|
| 14 |
# This is mostly tied to rocBLAS supported archs.
|
| 15 |
+
# check https://rocm.docs.amd.com/projects/install-on-linux/en/docs-7.2.1/reference/system-requirements.html
|
| 16 |
# check https://rocm.docs.amd.com/projects/radeon-ryzen/en/latest/docs/compatibility/compatibilityrad/native_linux/native_linux_compatibility.html
|
| 17 |
# check https://rocm.docs.amd.com/projects/radeon-ryzen/en/latest/docs/compatibility/compatibilityryz/native_linux/native_linux_compatibility.html
|
| 18 |
|
| 19 |
+
ARG ROCM_DOCKER_ARCH='gfx908;gfx90a;gfx942;gfx1030;gfx1100;gfx1101;gfx1102;gfx1151;gfx1150;gfx1200;gfx1201'
|
| 20 |
|
| 21 |
# Set ROCm architectures
|
| 22 |
ENV AMDGPU_TARGETS=${ROCM_DOCKER_ARCH}
|
|
|
|
| 58 |
FROM ${BASE_ROCM_DEV_CONTAINER} AS base
|
| 59 |
|
| 60 |
RUN apt-get update \
|
| 61 |
+
&& apt-get install -y libgomp1 curl \
|
| 62 |
&& apt autoremove -y \
|
| 63 |
&& apt clean -y \
|
| 64 |
&& rm -rf /tmp/* /var/tmp/* \
|
|
|
|
| 79 |
git \
|
| 80 |
python3-pip \
|
| 81 |
python3 \
|
| 82 |
+
python3-wheel \
|
| 83 |
&& pip install --break-system-packages --upgrade setuptools \
|
| 84 |
&& pip install --break-system-packages -r requirements.txt \
|
| 85 |
&& apt autoremove -y \
|
.devops/vulkan.Dockerfile
CHANGED
|
@@ -7,7 +7,7 @@ RUN apt update && apt install -y git build-essential cmake wget xz-utils
|
|
| 7 |
|
| 8 |
# Install SSL and Vulkan SDK dependencies
|
| 9 |
RUN apt install -y libssl-dev curl \
|
| 10 |
-
libxcb-xinput0 libxcb-xinerama0 libxcb-cursor-dev libvulkan-dev glslc
|
| 11 |
|
| 12 |
# Build it
|
| 13 |
WORKDIR /app
|
|
@@ -49,17 +49,20 @@ COPY --from=build /app/full /app
|
|
| 49 |
|
| 50 |
WORKDIR /app
|
| 51 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 52 |
RUN apt-get update \
|
| 53 |
&& apt-get install -y \
|
| 54 |
build-essential \
|
|
|
|
| 55 |
git \
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
&&
|
| 61 |
-
&& pip install --break-system-packages --upgrade setuptools \
|
| 62 |
-
&& pip install --break-system-packages -r requirements.txt \
|
| 63 |
&& apt autoremove -y \
|
| 64 |
&& apt clean -y \
|
| 65 |
&& rm -rf /tmp/* /var/tmp/* \
|
|
|
|
| 7 |
|
| 8 |
# Install SSL and Vulkan SDK dependencies
|
| 9 |
RUN apt install -y libssl-dev curl \
|
| 10 |
+
libxcb-xinput0 libxcb-xinerama0 libxcb-cursor-dev libvulkan-dev glslc spirv-headers
|
| 11 |
|
| 12 |
# Build it
|
| 13 |
WORKDIR /app
|
|
|
|
| 49 |
|
| 50 |
WORKDIR /app
|
| 51 |
|
| 52 |
+
ENV PATH="/root/.venv/bin:/root/.local/bin:${PATH}"
|
| 53 |
+
|
| 54 |
+
# Flag for compatibility with pip
|
| 55 |
+
ARG UV_INDEX_STRATEGY="unsafe-best-match"
|
| 56 |
RUN apt-get update \
|
| 57 |
&& apt-get install -y \
|
| 58 |
build-essential \
|
| 59 |
+
curl \
|
| 60 |
git \
|
| 61 |
+
ca-certificates \
|
| 62 |
+
&& curl -LsSf https://astral.sh/uv/install.sh | sh \
|
| 63 |
+
&& uv python install 3.13 \
|
| 64 |
+
&& uv venv --python 3.13 /root/.venv \
|
| 65 |
+
&& uv pip install --python /root/.venv/bin/python -r requirements.txt \
|
|
|
|
|
|
|
| 66 |
&& apt autoremove -y \
|
| 67 |
&& apt clean -y \
|
| 68 |
&& rm -rf /tmp/* /var/tmp/* \
|
.editorconfig
CHANGED
|
@@ -21,14 +21,6 @@ indent_style = tab
|
|
| 21 |
[prompts/*.txt]
|
| 22 |
insert_final_newline = unset
|
| 23 |
|
| 24 |
-
[tools/server/public/*]
|
| 25 |
-
indent_size = 2
|
| 26 |
-
|
| 27 |
-
[tools/server/public/deps_*]
|
| 28 |
-
trim_trailing_whitespace = unset
|
| 29 |
-
indent_style = unset
|
| 30 |
-
indent_size = unset
|
| 31 |
-
|
| 32 |
[tools/server/deps_*]
|
| 33 |
trim_trailing_whitespace = unset
|
| 34 |
indent_style = unset
|
|
@@ -61,6 +53,14 @@ charset = unset
|
|
| 61 |
trim_trailing_whitespace = unset
|
| 62 |
insert_final_newline = unset
|
| 63 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 64 |
[benches/**]
|
| 65 |
indent_style = unset
|
| 66 |
indent_size = unset
|
|
|
|
| 21 |
[prompts/*.txt]
|
| 22 |
insert_final_newline = unset
|
| 23 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
[tools/server/deps_*]
|
| 25 |
trim_trailing_whitespace = unset
|
| 26 |
indent_style = unset
|
|
|
|
| 53 |
trim_trailing_whitespace = unset
|
| 54 |
insert_final_newline = unset
|
| 55 |
|
| 56 |
+
[tools/server/public/**]
|
| 57 |
+
indent_style = unset
|
| 58 |
+
indent_size = unset
|
| 59 |
+
end_of_line = unset
|
| 60 |
+
charset = unset
|
| 61 |
+
trim_trailing_whitespace = unset
|
| 62 |
+
insert_final_newline = unset
|
| 63 |
+
|
| 64 |
[benches/**]
|
| 65 |
indent_style = unset
|
| 66 |
indent_size = unset
|
.gitattributes
CHANGED
|
@@ -1,9 +1,12 @@
|
|
| 1 |
-
|
|
|
|
|
|
|
|
|
|
| 2 |
*.gguf filter=lfs diff=lfs merge=lfs -text
|
| 3 |
*.key filter=lfs diff=lfs merge=lfs -text
|
| 4 |
-
*.
|
|
|
|
| 5 |
*.arrow filter=lfs diff=lfs merge=lfs -text
|
| 6 |
-
*.avro filter=lfs diff=lfs merge=lfs -text
|
| 7 |
*.bin filter=lfs diff=lfs merge=lfs -text
|
| 8 |
*.bz2 filter=lfs diff=lfs merge=lfs -text
|
| 9 |
*.ckpt filter=lfs diff=lfs merge=lfs -text
|
|
@@ -13,7 +16,6 @@
|
|
| 13 |
*.joblib filter=lfs diff=lfs merge=lfs -text
|
| 14 |
*.lfs.* filter=lfs diff=lfs merge=lfs -text
|
| 15 |
*.lz4 filter=lfs diff=lfs merge=lfs -text
|
| 16 |
-
*.mds filter=lfs diff=lfs merge=lfs -text
|
| 17 |
*.mlmodel filter=lfs diff=lfs merge=lfs -text
|
| 18 |
*.model filter=lfs diff=lfs merge=lfs -text
|
| 19 |
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
|
@@ -58,6 +60,14 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 58 |
*.jpg filter=lfs diff=lfs merge=lfs -text
|
| 59 |
*.jpeg filter=lfs diff=lfs merge=lfs -text
|
| 60 |
*.webp filter=lfs diff=lfs merge=lfs -text
|
| 61 |
-
|
| 62 |
-
*.
|
| 63 |
-
*.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Treat the generated single-file WebUI build as binary for diff purposes.
|
| 2 |
+
# Git's pack-file delta compression still works (byte-level), but this prevents
|
| 3 |
+
# git diff from printing the entire minified file on every change.
|
| 4 |
+
tools/server/public/index.html -diff
|
| 5 |
*.gguf filter=lfs diff=lfs merge=lfs -text
|
| 6 |
*.key filter=lfs diff=lfs merge=lfs -text
|
| 7 |
+
*.7z filter=lfs diff=lfs merge=lfs -text
|
| 8 |
+
*.txt filter=lfs diff=lfs merge=lfs -text
|
| 9 |
*.arrow filter=lfs diff=lfs merge=lfs -text
|
|
|
|
| 10 |
*.bin filter=lfs diff=lfs merge=lfs -text
|
| 11 |
*.bz2 filter=lfs diff=lfs merge=lfs -text
|
| 12 |
*.ckpt filter=lfs diff=lfs merge=lfs -text
|
|
|
|
| 16 |
*.joblib filter=lfs diff=lfs merge=lfs -text
|
| 17 |
*.lfs.* filter=lfs diff=lfs merge=lfs -text
|
| 18 |
*.lz4 filter=lfs diff=lfs merge=lfs -text
|
|
|
|
| 19 |
*.mlmodel filter=lfs diff=lfs merge=lfs -text
|
| 20 |
*.model filter=lfs diff=lfs merge=lfs -text
|
| 21 |
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
|
|
|
| 60 |
*.jpg filter=lfs diff=lfs merge=lfs -text
|
| 61 |
*.jpeg filter=lfs diff=lfs merge=lfs -text
|
| 62 |
*.webp filter=lfs diff=lfs merge=lfs -text
|
| 63 |
+
*.pdf filter=lfs diff=lfs merge=lfs -text
|
| 64 |
+
*.epub filter=lfs diff=lfs merge=lfs -text
|
| 65 |
+
*.djvu filter=lfs diff=lfs merge=lfs -text
|
| 66 |
+
*.doc filter=lfs diff=lfs merge=lfs -text
|
| 67 |
+
*.mobi filter=lfs diff=lfs merge=lfs -text
|
| 68 |
+
*.azw3 filter=lfs diff=lfs merge=lfs -text
|
| 69 |
+
*.txt filter=lfs diff=lfs merge=lfs -text
|
| 70 |
+
*.md filter=lfs diff=lfs merge=lfs -text
|
| 71 |
+
*.docx filter=lfs diff=lfs merge=lfs -text
|
| 72 |
+
*.azw3 filter=lfs diff=lfs merge=lfs -text
|
| 73 |
+
*.djvu filter=lfs diff=lfs merge=lfs -text
|
.github/ISSUE_TEMPLATE/010-bug-compilation.yml
CHANGED
|
@@ -41,7 +41,7 @@ body:
|
|
| 41 |
attributes:
|
| 42 |
label: GGML backends
|
| 43 |
description: Which GGML backends do you know to be affected?
|
| 44 |
-
options: [AMX, BLAS, CANN, CPU, CUDA, Hexagon, HIP, Metal, Musa, OpenCL, RPC, SYCL, VirtGPU, Vulkan, WebGPU, zDNN, ZenDNN]
|
| 45 |
multiple: true
|
| 46 |
validations:
|
| 47 |
required: true
|
|
|
|
| 41 |
attributes:
|
| 42 |
label: GGML backends
|
| 43 |
description: Which GGML backends do you know to be affected?
|
| 44 |
+
options: [AMX, BLAS, CANN, CPU, CUDA, Hexagon, HIP, Metal, Musa, OpenCL, OpenVINO, RPC, SYCL, VirtGPU, Vulkan, WebGPU, zDNN, ZenDNN]
|
| 45 |
multiple: true
|
| 46 |
validations:
|
| 47 |
required: true
|
.github/ISSUE_TEMPLATE/011-bug-results.yml
CHANGED
|
@@ -42,7 +42,7 @@ body:
|
|
| 42 |
attributes:
|
| 43 |
label: GGML backends
|
| 44 |
description: Which GGML backends do you know to be affected?
|
| 45 |
-
options: [AMX, BLAS, CANN, CPU, CUDA, Hexagon, HIP, Metal, Musa, OpenCL, RPC, SYCL, VirtGPU, Vulkan, WebGPU, zDNN, ZenDNN]
|
| 46 |
multiple: true
|
| 47 |
validations:
|
| 48 |
required: true
|
|
|
|
| 42 |
attributes:
|
| 43 |
label: GGML backends
|
| 44 |
description: Which GGML backends do you know to be affected?
|
| 45 |
+
options: [AMX, BLAS, CANN, CPU, CUDA, Hexagon, HIP, Metal, Musa, OpenCL, OpenVINO, RPC, SYCL, VirtGPU, Vulkan, WebGPU, zDNN, ZenDNN]
|
| 46 |
multiple: true
|
| 47 |
validations:
|
| 48 |
required: true
|
.github/labeler.yml
CHANGED
|
@@ -27,6 +27,11 @@ IBM zDNN:
|
|
| 27 |
- any-glob-to-any-file:
|
| 28 |
- ggml/include/ggml-zdnn.h
|
| 29 |
- ggml/src/ggml-zdnn/**
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 30 |
documentation:
|
| 31 |
- changed-files:
|
| 32 |
- any-glob-to-any-file:
|
|
@@ -68,10 +73,18 @@ android:
|
|
| 68 |
- changed-files:
|
| 69 |
- any-glob-to-any-file:
|
| 70 |
- examples/llama.android/**
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 71 |
server:
|
| 72 |
- changed-files:
|
| 73 |
- any-glob-to-any-file:
|
| 74 |
- tools/server/**
|
|
|
|
|
|
|
|
|
|
| 75 |
ggml:
|
| 76 |
- changed-files:
|
| 77 |
- any-glob-to-any-file:
|
|
|
|
| 27 |
- any-glob-to-any-file:
|
| 28 |
- ggml/include/ggml-zdnn.h
|
| 29 |
- ggml/src/ggml-zdnn/**
|
| 30 |
+
AMD ZenDNN:
|
| 31 |
+
- changed-files:
|
| 32 |
+
- any-glob-to-any-file:
|
| 33 |
+
- ggml/include/ggml-zendnn.h
|
| 34 |
+
- ggml/src/ggml-zendnn/**
|
| 35 |
documentation:
|
| 36 |
- changed-files:
|
| 37 |
- any-glob-to-any-file:
|
|
|
|
| 73 |
- changed-files:
|
| 74 |
- any-glob-to-any-file:
|
| 75 |
- examples/llama.android/**
|
| 76 |
+
server/webui:
|
| 77 |
+
- changed-files:
|
| 78 |
+
- any-glob-to-any-file:
|
| 79 |
+
- tools/server/webui/**
|
| 80 |
+
- tools/server/public/**
|
| 81 |
server:
|
| 82 |
- changed-files:
|
| 83 |
- any-glob-to-any-file:
|
| 84 |
- tools/server/**
|
| 85 |
+
|
| 86 |
+
|
| 87 |
+
|
| 88 |
ggml:
|
| 89 |
- changed-files:
|
| 90 |
- any-glob-to-any-file:
|
.github/pull_request_template.md
CHANGED
|
@@ -1 +1,3 @@
|
|
| 1 |
-
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cc3613996e5efd54aa6cfa2bf782d034bb40a7cce82e93850b7d7ac0fc3231ff
|
| 3 |
+
size 757
|
.github/workflows/ai-issues.yml
ADDED
|
@@ -0,0 +1,89 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
name: AI review (issues)
|
| 2 |
+
|
| 3 |
+
on:
|
| 4 |
+
issues:
|
| 5 |
+
types: [opened]
|
| 6 |
+
|
| 7 |
+
jobs:
|
| 8 |
+
find-related:
|
| 9 |
+
if: github.event.action == 'opened'
|
| 10 |
+
runs-on: [self-hosted, opencode]
|
| 11 |
+
|
| 12 |
+
permissions:
|
| 13 |
+
contents: read
|
| 14 |
+
issues: write
|
| 15 |
+
|
| 16 |
+
steps:
|
| 17 |
+
- name: Checkout repository
|
| 18 |
+
uses: actions/checkout@v6
|
| 19 |
+
with:
|
| 20 |
+
fetch-depth: 1
|
| 21 |
+
|
| 22 |
+
- name: Find related
|
| 23 |
+
env:
|
| 24 |
+
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
|
| 25 |
+
OPENCODE_PERMISSION: |
|
| 26 |
+
{
|
| 27 |
+
"bash": {
|
| 28 |
+
"*": "deny",
|
| 29 |
+
"gh issue view*": "allow",
|
| 30 |
+
"gh issue list*": "allow",
|
| 31 |
+
"gh issue comment*": "allow",
|
| 32 |
+
"gh search issues*": "allow"
|
| 33 |
+
},
|
| 34 |
+
"webfetch": "deny"
|
| 35 |
+
}
|
| 36 |
+
run: |
|
| 37 |
+
rm AGENTS.md
|
| 38 |
+
rm CLAUDE.md
|
| 39 |
+
|
| 40 |
+
timeout 5m opencode run -m llama.cpp-dgx/ai-review-issues-find-similar --thinking "A new issue has been created:
|
| 41 |
+
|
| 42 |
+
Issue number: ${{ github.event.issue.number }}
|
| 43 |
+
|
| 44 |
+
Lookup the contents of the issue using the following 'gh' command:
|
| 45 |
+
|
| 46 |
+
gh issue view ${{ github.event.issue.number }} --json title,body,url,number
|
| 47 |
+
|
| 48 |
+
Next, perform the following task and then post a SINGLE comment (if needed).
|
| 49 |
+
|
| 50 |
+
---
|
| 51 |
+
|
| 52 |
+
TASK : FIND RELATED ISSUES
|
| 53 |
+
|
| 54 |
+
Using the 'gh' CLI tool, search through existing issues on Github.
|
| 55 |
+
Find related or similar issues to the newly created one and list them.
|
| 56 |
+
Do not list the new issue itself (it is #${{ github.event.issue.number }}).
|
| 57 |
+
|
| 58 |
+
Consider:
|
| 59 |
+
1. Similar titles or descriptions
|
| 60 |
+
2. Same error messages or symptoms
|
| 61 |
+
3. Related functionality or components
|
| 62 |
+
4. Similar feature requests
|
| 63 |
+
|
| 64 |
+
---
|
| 65 |
+
|
| 66 |
+
POSTING YOUR COMMENT:
|
| 67 |
+
|
| 68 |
+
Based on your findings, post a SINGLE comment on issue #${{ github.event.issue.number }}. Build the comment as follows:
|
| 69 |
+
|
| 70 |
+
- If no related issues were found, do NOT comment at all.
|
| 71 |
+
- If related issues were found, include a section listing them with links using the following format:
|
| 72 |
+
|
| 73 |
+
[comment]
|
| 74 |
+
This issue might be similar or related to the following issue(s):
|
| 75 |
+
|
| 76 |
+
- #12942: [brief description of how they are related]
|
| 77 |
+
- #11234: [brief description of how they are related]
|
| 78 |
+
...
|
| 79 |
+
|
| 80 |
+
_This comment was auto-generated locally using **$GA_ENGINE** on **$GA_MACHINE**_
|
| 81 |
+
[/comment]
|
| 82 |
+
|
| 83 |
+
Remember:
|
| 84 |
+
- Do not include the comment tags in your actual comment.
|
| 85 |
+
- Post at most ONE comment combining all findings.
|
| 86 |
+
- If you didn't find issues that are related enough, post nothing.
|
| 87 |
+
- You have access only to the 'gh' CLI tool - don't try to use other tools.
|
| 88 |
+
- If the output from a tool call is too long, try to limit down the search.
|
| 89 |
+
"
|
.github/workflows/build-android.yml
CHANGED
|
@@ -40,13 +40,9 @@ jobs:
|
|
| 40 |
steps:
|
| 41 |
- name: Clone
|
| 42 |
uses: actions/checkout@v6
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
# uses: ggml-org/ccache-action@v1.2.16
|
| 47 |
-
# with:
|
| 48 |
-
# key: android-build
|
| 49 |
-
# evict-old-files: 1d
|
| 50 |
|
| 51 |
- name: Set up JDK
|
| 52 |
uses: actions/setup-java@v5
|
|
@@ -55,7 +51,7 @@ jobs:
|
|
| 55 |
distribution: zulu
|
| 56 |
|
| 57 |
- name: Setup Android SDK
|
| 58 |
-
uses: android-actions/setup-android@
|
| 59 |
with:
|
| 60 |
log-accepted-android-sdk-licenses: false
|
| 61 |
|
|
@@ -66,10 +62,11 @@ jobs:
|
|
| 66 |
|
| 67 |
android-ndk:
|
| 68 |
runs-on: ubuntu-latest
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
|
|
|
| 73 |
strategy:
|
| 74 |
matrix:
|
| 75 |
include:
|
|
@@ -82,59 +79,23 @@ jobs:
|
|
| 82 |
- name: Clone
|
| 83 |
id: checkout
|
| 84 |
uses: actions/checkout@v6
|
|
|
|
|
|
|
|
|
|
| 85 |
|
| 86 |
-
- name:
|
| 87 |
-
id:
|
| 88 |
-
if: ${{ matrix.build == 'arm64-snapdragon' }}
|
| 89 |
-
run: |
|
| 90 |
-
mkdir opencl
|
| 91 |
-
curl -L -o opencl/clhpp.tar.gz https://github.com/KhronosGroup/OpenCL-CLHPP/archive/refs/tags/v${OPENCL_VERSION}.tar.gz
|
| 92 |
-
curl -L -o opencl/headers.tar.gz https://github.com/KhronosGroup/OpenCL-Headers/archive/refs/tags/v${OPENCL_VERSION}.tar.gz
|
| 93 |
-
curl -L -o opencl/icd-loader.tar.gz https://github.com/KhronosGroup/OpenCL-ICD-Loader/archive/refs/tags/v${OPENCL_VERSION}.tar.gz
|
| 94 |
-
tar -xaf opencl/headers.tar.gz -C opencl
|
| 95 |
-
tar -xaf opencl/clhpp.tar.gz -C opencl
|
| 96 |
-
tar -xaf opencl/icd-loader.tar.gz -C opencl
|
| 97 |
-
sudo cp -r opencl/OpenCL-Headers-${OPENCL_VERSION}/CL ${ANDROID_NDK_ROOT}/toolchains/llvm/prebuilt/linux-x86_64/sysroot/usr/include
|
| 98 |
-
sudo cp -r opencl/OpenCL-CLHPP-${OPENCL_VERSION}/include/CL/* ${ANDROID_NDK_ROOT}/toolchains/llvm/prebuilt/linux-x86_64/sysroot/usr/include/CL
|
| 99 |
-
cd opencl/OpenCL-ICD-Loader-${OPENCL_VERSION}
|
| 100 |
-
cmake -B build -G Ninja -DCMAKE_BUILD_TYPE=Release -DCMAKE_TOOLCHAIN_FILE=${ANDROID_NDK_ROOT}/build/cmake/android.toolchain.cmake -DOPENCL_ICD_LOADER_HEADERS_DIR=${ANDROID_NDK_ROOT}/toolchains/llvm/prebuilt/linux-x86_64/sysroot/usr/include -DANDROID_ABI=arm64-v8a -DANDROID_PLATFORM=31 -DANDROID_STL=c++_shared
|
| 101 |
-
cmake --build build
|
| 102 |
-
sudo cp build/libOpenCL.so ${ANDROID_NDK_ROOT}/toolchains/llvm/prebuilt/linux-x86_64/sysroot/usr/lib/aarch64-linux-android
|
| 103 |
-
rm -rf opencl
|
| 104 |
-
|
| 105 |
-
- name: Install Hexagon SDK
|
| 106 |
-
id: install_hexsdk
|
| 107 |
-
if: ${{ matrix.build == 'arm64-snapdragon' }}
|
| 108 |
-
env:
|
| 109 |
-
HEXSDK_VER: 6.4.0.2
|
| 110 |
-
HEXTLS_VER: 19.0.04
|
| 111 |
-
run: |
|
| 112 |
-
curl -L -o hex-sdk.tar.gz https://github.com/snapdragon-toolchain/hexagon-sdk/releases/download/v$HEXSDK_VER/hexagon-sdk-v$HEXSDK_VER-amd64-lnx.tar.xz
|
| 113 |
-
mkdir hex-sdk
|
| 114 |
-
tar -xaf hex-sdk.tar.gz -C hex-sdk
|
| 115 |
-
ls -l hex-sdk
|
| 116 |
-
sudo mv hex-sdk /opt/hexagon
|
| 117 |
-
echo "HEXAGON_SDK_ROOT=/opt/hexagon/$HEXSDK_VER" >> "$GITHUB_ENV"
|
| 118 |
-
echo "HEXAGON_TOOLS_ROOT=/opt/hexagon/$HEXSDK_VER/tools/HEXAGON_Tools/$HEXTLS_VER" >> "$GITHUB_ENV"
|
| 119 |
-
echo "DEFAULT_HLOS_ARCH=64" >> "$GITHUB_ENV"
|
| 120 |
-
echo "DEFAULT_TOOLS_VARIANT=toolv19" >> "$GITHUB_ENV"
|
| 121 |
-
echo "DEFAULT_NO_QURT_INC=0" >> "$GITHUB_ENV"
|
| 122 |
-
echo "DEFAULT_DSP_ARCH=v73" >> "$GITHUB_ENV"
|
| 123 |
-
|
| 124 |
-
- name: Update CMake presets
|
| 125 |
-
id: update_presets
|
| 126 |
-
if: ${{ matrix.build == 'arm64-snapdragon' }}
|
| 127 |
-
run: |
|
| 128 |
-
cp docs/backend/snapdragon/CMakeUserPresets.json .
|
| 129 |
-
|
| 130 |
-
- name: Build
|
| 131 |
-
id: ndk_build
|
| 132 |
run: |
|
|
|
|
|
|
|
|
|
|
| 133 |
cmake ${{ matrix.defines }} -B build
|
| 134 |
cmake --build build
|
| 135 |
cmake --install build --prefix pkg-adb/llama.cpp
|
| 136 |
|
| 137 |
-
- name:
|
| 138 |
-
|
| 139 |
-
|
| 140 |
-
|
|
|
|
|
|
|
|
|
| 40 |
steps:
|
| 41 |
- name: Clone
|
| 42 |
uses: actions/checkout@v6
|
| 43 |
+
with:
|
| 44 |
+
fetch-depth: 0
|
| 45 |
+
lfs: false
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
|
| 47 |
- name: Set up JDK
|
| 48 |
uses: actions/setup-java@v5
|
|
|
|
| 51 |
distribution: zulu
|
| 52 |
|
| 53 |
- name: Setup Android SDK
|
| 54 |
+
uses: android-actions/setup-android@40fd30fb8d7440372e1316f5d1809ec01dcd3699 # v4.0.1
|
| 55 |
with:
|
| 56 |
log-accepted-android-sdk-licenses: false
|
| 57 |
|
|
|
|
| 62 |
|
| 63 |
android-ndk:
|
| 64 |
runs-on: ubuntu-latest
|
| 65 |
+
container:
|
| 66 |
+
image: 'ghcr.io/snapdragon-toolchain/arm64-android:v0.3'
|
| 67 |
+
defaults:
|
| 68 |
+
run:
|
| 69 |
+
shell: bash
|
| 70 |
strategy:
|
| 71 |
matrix:
|
| 72 |
include:
|
|
|
|
| 79 |
- name: Clone
|
| 80 |
id: checkout
|
| 81 |
uses: actions/checkout@v6
|
| 82 |
+
with:
|
| 83 |
+
fetch-depth: 0
|
| 84 |
+
lfs: false
|
| 85 |
|
| 86 |
+
- name: Build Llama.CPP for Hexagon Android
|
| 87 |
+
id: build_llama_cpp_hexagon_android
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 88 |
run: |
|
| 89 |
+
if [[ "${{ matrix.build }}" == "arm64-snapdragon" ]]; then
|
| 90 |
+
cp docs/backend/snapdragon/CMakeUserPresets.json .
|
| 91 |
+
fi
|
| 92 |
cmake ${{ matrix.defines }} -B build
|
| 93 |
cmake --build build
|
| 94 |
cmake --install build --prefix pkg-adb/llama.cpp
|
| 95 |
|
| 96 |
+
- name: Upload Llama.CPP Hexagon Android Build Artifact
|
| 97 |
+
if: ${{ always() && steps.build_llama_cpp_hexagon_android.outcome == 'success' }}
|
| 98 |
+
uses: actions/upload-artifact@v6
|
| 99 |
+
with:
|
| 100 |
+
name: llama-cpp-android-${{ matrix.build }}
|
| 101 |
+
path: pkg-adb/llama.cpp
|
.github/workflows/build-cann.yml
CHANGED
|
@@ -63,7 +63,7 @@ jobs:
|
|
| 63 |
- name: Set container image
|
| 64 |
id: cann-image
|
| 65 |
run: |
|
| 66 |
-
image="ascendai/cann:${{ matrix.chip_type == '910b' && '8.
|
| 67 |
echo "image=${image}" >> "${GITHUB_OUTPUT}"
|
| 68 |
|
| 69 |
- name: Pull container image
|
|
|
|
| 63 |
- name: Set container image
|
| 64 |
id: cann-image
|
| 65 |
run: |
|
| 66 |
+
image="ascendai/cann:${{ matrix.chip_type == '910b' && '8.5.0-910b-openeuler24.03-py3.11' || '8.5.0-310p-openeuler24.03-py3.11' }}"
|
| 67 |
echo "image=${image}" >> "${GITHUB_OUTPUT}"
|
| 68 |
|
| 69 |
- name: Pull container image
|
.github/workflows/build-cross.yml
CHANGED
|
@@ -246,6 +246,7 @@ jobs:
|
|
| 246 |
apt-get install -y --no-install-recommends \
|
| 247 |
build-essential \
|
| 248 |
glslc \
|
|
|
|
| 249 |
gcc-14-loongarch64-linux-gnu \
|
| 250 |
g++-14-loongarch64-linux-gnu \
|
| 251 |
libvulkan-dev:loong64
|
|
|
|
| 246 |
apt-get install -y --no-install-recommends \
|
| 247 |
build-essential \
|
| 248 |
glslc \
|
| 249 |
+
spirv-headers \
|
| 250 |
gcc-14-loongarch64-linux-gnu \
|
| 251 |
g++-14-loongarch64-linux-gnu \
|
| 252 |
libvulkan-dev:loong64
|
.github/workflows/build-msys.yml
CHANGED
|
@@ -43,7 +43,7 @@ jobs:
|
|
| 43 |
# save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
|
| 44 |
|
| 45 |
- name: Setup ${{ matrix.sys }}
|
| 46 |
-
uses: msys2/setup-msys2@v2
|
| 47 |
with:
|
| 48 |
update: true
|
| 49 |
msystem: ${{matrix.sys}}
|
|
|
|
| 43 |
# save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
|
| 44 |
|
| 45 |
- name: Setup ${{ matrix.sys }}
|
| 46 |
+
uses: msys2/setup-msys2@cafece8e6baf9247cf9b1bf95097b0b983cc558d # v2
|
| 47 |
with:
|
| 48 |
update: true
|
| 49 |
msystem: ${{matrix.sys}}
|
.github/workflows/build-riscv.yml
CHANGED
|
@@ -35,7 +35,7 @@ env:
|
|
| 35 |
|
| 36 |
jobs:
|
| 37 |
ubuntu-riscv64-native-sanitizer:
|
| 38 |
-
runs-on:
|
| 39 |
|
| 40 |
continue-on-error: true
|
| 41 |
|
|
@@ -47,20 +47,9 @@ jobs:
|
|
| 47 |
steps:
|
| 48 |
- name: Install dependencies
|
| 49 |
run: |
|
| 50 |
-
sudo apt-get update
|
| 51 |
-
|
| 52 |
-
# Install necessary packages
|
| 53 |
-
sudo apt-get install -y libatomic1 libtsan2 gcc-14 g++-14 rustup cmake build-essential wget ccache git-lfs
|
| 54 |
-
|
| 55 |
# Set gcc-14 and g++-14 as the default compilers
|
| 56 |
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-14 100
|
| 57 |
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-14 100
|
| 58 |
-
sudo ln -sf /usr/bin/gcc-14 /usr/bin/gcc
|
| 59 |
-
sudo ln -sf /usr/bin/g++-14 /usr/bin/g++
|
| 60 |
-
|
| 61 |
-
# Install Rust stable version
|
| 62 |
-
rustup install stable
|
| 63 |
-
rustup default stable
|
| 64 |
|
| 65 |
git lfs install
|
| 66 |
|
|
@@ -73,23 +62,12 @@ jobs:
|
|
| 73 |
id: checkout
|
| 74 |
uses: actions/checkout@v6
|
| 75 |
|
| 76 |
-
- name:
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
# Configure ccache
|
| 83 |
-
ccache --set-config=max_size=5G
|
| 84 |
-
ccache --set-config=compression=true
|
| 85 |
-
ccache --set-config=compression_level=6
|
| 86 |
-
ccache --set-config=cache_dir="$CCACHE_DIR"
|
| 87 |
-
ccache --set-config=sloppiness=file_macro,time_macros,include_file_mtime,include_file_ctime
|
| 88 |
-
ccache --set-config=hash_dir=false
|
| 89 |
-
|
| 90 |
-
# Export for subsequent steps
|
| 91 |
-
echo "CCACHE_DIR=$CCACHE_DIR" >> $GITHUB_ENV
|
| 92 |
-
echo "PATH=/usr/lib/ccache:$PATH" >> $GITHUB_ENV
|
| 93 |
|
| 94 |
- name: Build
|
| 95 |
id: cmake_build
|
|
|
|
| 35 |
|
| 36 |
jobs:
|
| 37 |
ubuntu-riscv64-native-sanitizer:
|
| 38 |
+
runs-on: ubuntu-24.04-riscv
|
| 39 |
|
| 40 |
continue-on-error: true
|
| 41 |
|
|
|
|
| 47 |
steps:
|
| 48 |
- name: Install dependencies
|
| 49 |
run: |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 50 |
# Set gcc-14 and g++-14 as the default compilers
|
| 51 |
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-14 100
|
| 52 |
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-14 100
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
|
| 54 |
git lfs install
|
| 55 |
|
|
|
|
| 62 |
id: checkout
|
| 63 |
uses: actions/checkout@v6
|
| 64 |
|
| 65 |
+
- name: ccache
|
| 66 |
+
uses: ggml-org/ccache-action@afde29e5b5422e5da23cb1f639e8baecadeadfc3 # https://github.com/ggml-org/ccache-action/pull/1
|
| 67 |
+
with:
|
| 68 |
+
key: ubuntu-riscv64-native-sanitizer-${{ matrix.sanitizer }}-${{ matrix.build_type }}
|
| 69 |
+
evict-old-files: 1d
|
| 70 |
+
save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 71 |
|
| 72 |
- name: Build
|
| 73 |
id: cmake_build
|
.github/workflows/build-self-hosted.yml
CHANGED
|
@@ -97,6 +97,36 @@ jobs:
|
|
| 97 |
vulkaninfo --summary
|
| 98 |
GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
|
| 99 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 100 |
# TODO: provision AMX-compatible machine
|
| 101 |
#ggml-ci-cpu-amx:
|
| 102 |
# runs-on: [self-hosted, Linux, CPU, AMX]
|
|
@@ -165,16 +195,15 @@ jobs:
|
|
| 165 |
- name: Dawn Dependency
|
| 166 |
id: dawn-depends
|
| 167 |
run: |
|
| 168 |
-
DAWN_VERSION="
|
| 169 |
-
DAWN_OWNER="
|
| 170 |
DAWN_REPO="dawn"
|
| 171 |
-
DAWN_ASSET_NAME="Dawn-
|
| 172 |
-
echo "Fetching release asset from https://github.com/
|
| 173 |
-
curl -L -o artifact.
|
| 174 |
-
"https://github.com/
|
| 175 |
mkdir dawn
|
| 176 |
-
|
| 177 |
-
tar -xvf ${DAWN_ASSET_NAME}.tar.gz -C dawn --strip-components=1
|
| 178 |
|
| 179 |
- name: Test
|
| 180 |
id: ggml-ci
|
|
@@ -212,6 +241,27 @@ jobs:
|
|
| 212 |
vulkaninfo --summary
|
| 213 |
GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
|
| 214 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 215 |
ggml-ci-intel-openvino-gpu-low-perf:
|
| 216 |
runs-on: [self-hosted, Linux, Intel, OpenVINO]
|
| 217 |
|
|
|
|
| 97 |
vulkaninfo --summary
|
| 98 |
GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
|
| 99 |
|
| 100 |
+
# TODO: investigate slight precision issues in some operations for test-backend-ops on the WebGPU backend.
|
| 101 |
+
#ggml-ci-nvidia-webgpu:
|
| 102 |
+
# runs-on: [self-hosted, Linux, NVIDIA]
|
| 103 |
+
|
| 104 |
+
# steps:
|
| 105 |
+
# - name: Clone
|
| 106 |
+
# id: checkout
|
| 107 |
+
# uses: actions/checkout@v6
|
| 108 |
+
|
| 109 |
+
# - name: Dawn Dependency
|
| 110 |
+
# id: dawn-depends
|
| 111 |
+
# run: |
|
| 112 |
+
# DAWN_VERSION="v20260317.182325"
|
| 113 |
+
# DAWN_OWNER="google"
|
| 114 |
+
# DAWN_REPO="dawn"
|
| 115 |
+
# DAWN_ASSET_NAME="Dawn-18eb229ef5f707c1464cc581252e7603c73a3ef0-ubuntu-latest-Release"
|
| 116 |
+
# echo "Fetching release asset from https://github.com/google/dawn/releases/download/${DAWN_VERSION}/${DAWN_ASSET_NAME}.tar.gz"
|
| 117 |
+
# curl -L -o artifact.tar.gz \
|
| 118 |
+
# "https://github.com/google/dawn/releases/download/${DAWN_VERSION}/${DAWN_ASSET_NAME}.tar.gz"
|
| 119 |
+
# mkdir dawn
|
| 120 |
+
# tar -xvf artifact.tar.gz -C dawn --strip-components=1
|
| 121 |
+
|
| 122 |
+
# - name: Test
|
| 123 |
+
# id: ggml-ci
|
| 124 |
+
# run: |
|
| 125 |
+
# GG_BUILD_WEBGPU=1 \
|
| 126 |
+
# GG_BUILD_WEBGPU_DAWN_PREFIX="$GITHUB_WORKSPACE/dawn" \
|
| 127 |
+
# GG_BUILD_WEBGPU_DAWN_DIR="$GITHUB_WORKSPACE/dawn/lib64/cmake/Dawn" \
|
| 128 |
+
# bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
|
| 129 |
+
|
| 130 |
# TODO: provision AMX-compatible machine
|
| 131 |
#ggml-ci-cpu-amx:
|
| 132 |
# runs-on: [self-hosted, Linux, CPU, AMX]
|
|
|
|
| 195 |
- name: Dawn Dependency
|
| 196 |
id: dawn-depends
|
| 197 |
run: |
|
| 198 |
+
DAWN_VERSION="v20260317.182325"
|
| 199 |
+
DAWN_OWNER="google"
|
| 200 |
DAWN_REPO="dawn"
|
| 201 |
+
DAWN_ASSET_NAME="Dawn-18eb229ef5f707c1464cc581252e7603c73a3ef0-macos-latest-Release"
|
| 202 |
+
echo "Fetching release asset from https://github.com/google/dawn/releases/download/${DAWN_VERSION}/${DAWN_ASSET_NAME}.tar.gz"
|
| 203 |
+
curl -L -o artifact.tar.gz \
|
| 204 |
+
"https://github.com/google/dawn/releases/download/${DAWN_VERSION}/${DAWN_ASSET_NAME}.tar.gz"
|
| 205 |
mkdir dawn
|
| 206 |
+
tar -xvf artifact.tar.gz -C dawn --strip-components=1
|
|
|
|
| 207 |
|
| 208 |
- name: Test
|
| 209 |
id: ggml-ci
|
|
|
|
| 241 |
vulkaninfo --summary
|
| 242 |
GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
|
| 243 |
|
| 244 |
+
ggml-ci-win-intel-vulkan:
|
| 245 |
+
runs-on: [self-hosted, Windows, X64, Intel]
|
| 246 |
+
|
| 247 |
+
steps:
|
| 248 |
+
- name: Clone
|
| 249 |
+
id: checkout
|
| 250 |
+
uses: actions/checkout@v6
|
| 251 |
+
|
| 252 |
+
- name: Test
|
| 253 |
+
id: ggml-ci
|
| 254 |
+
shell: C:\msys64\usr\bin\bash.exe --noprofile --norc -eo pipefail "{0}"
|
| 255 |
+
env:
|
| 256 |
+
MSYSTEM: UCRT64
|
| 257 |
+
CHERE_INVOKING: 1
|
| 258 |
+
PATH: C:\msys64\ucrt64\bin;C:\msys64\usr\bin;C:\Windows\System32;${{ env.PATH }}
|
| 259 |
+
run: |
|
| 260 |
+
vulkaninfo --summary
|
| 261 |
+
# Skip python related tests with GG_BUILD_LOW_PERF=1 since Windows MSYS2 UCRT64 currently fails to create
|
| 262 |
+
# a valid python environment for testing
|
| 263 |
+
LLAMA_FATAL_WARNINGS=OFF GG_BUILD_NINJA=1 GG_BUILD_VULKAN=1 GG_BUILD_LOW_PERF=1 ./ci/run.sh ./results/llama.cpp ./mnt/llama.cpp
|
| 264 |
+
|
| 265 |
ggml-ci-intel-openvino-gpu-low-perf:
|
| 266 |
runs-on: [self-hosted, Linux, Intel, OpenVINO]
|
| 267 |
|
.github/workflows/build-vulkan.yml
CHANGED
|
@@ -72,7 +72,7 @@ jobs:
|
|
| 72 |
|
| 73 |
- name: Setup Vulkan SDK
|
| 74 |
if: steps.cache-sdk.outputs.cache-hit != 'true'
|
| 75 |
-
uses: ./.github/actions/linux-setup-vulkan
|
| 76 |
with:
|
| 77 |
path: ./vulkan_sdk
|
| 78 |
version: ${{ env.VULKAN_SDK_VERSION }}
|
|
@@ -93,4 +93,5 @@ jobs:
|
|
| 93 |
export GGML_VK_DISABLE_F16=1
|
| 94 |
export GGML_VK_DISABLE_COOPMAT=1
|
| 95 |
# This is using llvmpipe and runs slower than other backends
|
| 96 |
-
|
|
|
|
|
|
| 72 |
|
| 73 |
- name: Setup Vulkan SDK
|
| 74 |
if: steps.cache-sdk.outputs.cache-hit != 'true'
|
| 75 |
+
uses: ./.github/actions/linux-setup-vulkan
|
| 76 |
with:
|
| 77 |
path: ./vulkan_sdk
|
| 78 |
version: ${{ env.VULKAN_SDK_VERSION }}
|
|
|
|
| 93 |
export GGML_VK_DISABLE_F16=1
|
| 94 |
export GGML_VK_DISABLE_COOPMAT=1
|
| 95 |
# This is using llvmpipe and runs slower than other backends
|
| 96 |
+
# test-backend-ops is too slow on llvmpipe, skip it
|
| 97 |
+
ctest -L main -E test-backend-ops --verbose --timeout 900
|
.github/workflows/build.yml
CHANGED
|
@@ -87,7 +87,7 @@ jobs:
|
|
| 87 |
-DGGML_METAL_EMBED_LIBRARY=OFF \
|
| 88 |
-DGGML_METAL_SHADER_DEBUG=ON \
|
| 89 |
-DGGML_RPC=ON
|
| 90 |
-
cmake --build build --config Release -j $(sysctl -n hw.logicalcpu)
|
| 91 |
leaks -atExit -- ./build/bin/test-thread-safety -hf ggml-org/gemma-3-270m-qat-GGUF -ngl 99 -p "$(printf 'hello %.0s' {1..128})" -n 16 -c 512 -ub 32 -np 2 -t 2 -lv 1
|
| 92 |
|
| 93 |
- name: Test
|
|
@@ -124,7 +124,7 @@ jobs:
|
|
| 124 |
-DGGML_METAL=OFF \
|
| 125 |
-DGGML_RPC=ON \
|
| 126 |
-DCMAKE_OSX_DEPLOYMENT_TARGET=13.3
|
| 127 |
-
cmake --build build --config Release -j $(sysctl -n hw.logicalcpu)
|
| 128 |
|
| 129 |
- name: Test
|
| 130 |
id: cmake_test
|
|
@@ -150,23 +150,22 @@ jobs:
|
|
| 150 |
- name: Dawn Dependency
|
| 151 |
id: dawn-depends
|
| 152 |
run: |
|
| 153 |
-
DAWN_VERSION="
|
| 154 |
-
DAWN_OWNER="
|
| 155 |
DAWN_REPO="dawn"
|
| 156 |
-
DAWN_ASSET_NAME="Dawn-
|
| 157 |
-
echo "Fetching release asset from https://github.com/
|
| 158 |
-
curl -L -o artifact.
|
| 159 |
-
"https://github.com/
|
| 160 |
mkdir dawn
|
| 161 |
-
|
| 162 |
-
tar -xvf ${DAWN_ASSET_NAME}.tar.gz -C dawn --strip-components=1
|
| 163 |
|
| 164 |
- name: Build
|
| 165 |
id: cmake_build
|
| 166 |
run: |
|
| 167 |
export CMAKE_PREFIX_PATH=dawn
|
| 168 |
-
cmake -B build -DGGML_WEBGPU=ON -DGGML_METAL=OFF -DGGML_BLAS=OFF
|
| 169 |
-
cmake --build build --config Release -j $(sysctl -n hw.logicalcpu)
|
| 170 |
|
| 171 |
- name: Test
|
| 172 |
id: cmake_test
|
|
@@ -181,7 +180,7 @@ jobs:
|
|
| 181 |
- build: 'x64'
|
| 182 |
os: ubuntu-22.04
|
| 183 |
- build: 'arm64'
|
| 184 |
-
os: ubuntu-
|
| 185 |
- build: 's390x'
|
| 186 |
os: ubuntu-24.04-s390x
|
| 187 |
- build: 'ppc64le'
|
|
@@ -207,14 +206,22 @@ jobs:
|
|
| 207 |
run: |
|
| 208 |
sudo apt-get update
|
| 209 |
sudo apt-get install -y --no-install-recommends \
|
| 210 |
-
python3 python3-pip python3-dev \
|
| 211 |
libjpeg-dev build-essential libssl-dev \
|
| 212 |
git-lfs
|
| 213 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 214 |
- name: Python Dependencies
|
| 215 |
id: python_depends
|
| 216 |
run: |
|
| 217 |
-
|
|
|
|
| 218 |
pip3 install ./gguf-py
|
| 219 |
|
| 220 |
- name: Swap Endianness
|
|
@@ -231,7 +238,7 @@ jobs:
|
|
| 231 |
cmake -B build \
|
| 232 |
-DLLAMA_FATAL_WARNINGS=ON \
|
| 233 |
-DGGML_RPC=ON
|
| 234 |
-
cmake --build build --config Release -j $(nproc)
|
| 235 |
|
| 236 |
- name: Test
|
| 237 |
id: cmake_test
|
|
@@ -260,6 +267,56 @@ jobs:
|
|
| 260 |
wget https://huggingface.co/ggml-org/models/resolve/main/tinyllamas/stories260K-be.gguf
|
| 261 |
./bin/llama-completion -m stories260K-be.gguf -p "One day, Lily met a Shoggoth" -n 500 -c 256
|
| 262 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 263 |
ubuntu-latest-rpc:
|
| 264 |
runs-on: ubuntu-latest
|
| 265 |
|
|
@@ -274,14 +331,16 @@ jobs:
|
|
| 274 |
id: depends
|
| 275 |
run: |
|
| 276 |
sudo apt-get update
|
| 277 |
-
sudo apt-get install build-essential libssl-dev
|
| 278 |
|
| 279 |
- name: Build
|
| 280 |
id: cmake_build
|
| 281 |
run: |
|
| 282 |
cmake -B build \
|
|
|
|
|
|
|
| 283 |
-DGGML_RPC=ON
|
| 284 |
-
cmake --build build --config Release -j $(nproc)
|
| 285 |
|
| 286 |
- name: Test
|
| 287 |
id: cmake_test
|
|
@@ -290,7 +349,15 @@ jobs:
|
|
| 290 |
ctest -L main --verbose
|
| 291 |
|
| 292 |
ubuntu-24-vulkan:
|
| 293 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 294 |
|
| 295 |
steps:
|
| 296 |
- name: Clone
|
|
@@ -300,12 +367,16 @@ jobs:
|
|
| 300 |
- name: Dependencies
|
| 301 |
id: depends
|
| 302 |
run: |
|
| 303 |
-
sudo apt-get
|
|
|
|
|
|
|
|
|
|
| 304 |
|
| 305 |
- name: Configure
|
| 306 |
id: cmake_configure
|
| 307 |
run: |
|
| 308 |
cmake -B build \
|
|
|
|
| 309 |
-DCMAKE_BUILD_TYPE=RelWithDebInfo \
|
| 310 |
-DGGML_BACKEND_DL=ON \
|
| 311 |
-DGGML_CPU_ALL_VARIANTS=ON \
|
|
@@ -314,7 +385,7 @@ jobs:
|
|
| 314 |
- name: Build
|
| 315 |
id: cmake_build
|
| 316 |
run: |
|
| 317 |
-
cmake --build build -j $(nproc)
|
| 318 |
|
| 319 |
ubuntu-24-webgpu:
|
| 320 |
runs-on: ubuntu-24.04
|
|
@@ -336,7 +407,8 @@ jobs:
|
|
| 336 |
run: |
|
| 337 |
sudo add-apt-repository -y ppa:kisak/kisak-mesa
|
| 338 |
sudo apt-get update -y
|
| 339 |
-
sudo apt-get install -y build-essential mesa-vulkan-drivers
|
|
|
|
| 340 |
|
| 341 |
- name: Get latest Vulkan SDK version
|
| 342 |
id: vulkan_sdk_version
|
|
@@ -361,16 +433,15 @@ jobs:
|
|
| 361 |
id: dawn-depends
|
| 362 |
run: |
|
| 363 |
sudo apt-get install -y libxrandr-dev libxinerama-dev libxcursor-dev mesa-common-dev libx11-xcb-dev libxi-dev
|
| 364 |
-
DAWN_VERSION="
|
| 365 |
-
DAWN_OWNER="
|
| 366 |
DAWN_REPO="dawn"
|
| 367 |
-
DAWN_ASSET_NAME="Dawn-
|
| 368 |
-
echo "Fetching release asset from https://github.com/
|
| 369 |
-
curl -L -o artifact.
|
| 370 |
-
"https://github.com/
|
| 371 |
mkdir dawn
|
| 372 |
-
|
| 373 |
-
tar -xvf ${DAWN_ASSET_NAME}.tar.gz -C dawn --strip-components=1
|
| 374 |
|
| 375 |
- name: Build
|
| 376 |
id: cmake_build
|
|
@@ -378,7 +449,7 @@ jobs:
|
|
| 378 |
export Dawn_DIR=dawn/lib64/cmake/Dawn
|
| 379 |
cmake -B build \
|
| 380 |
-DGGML_WEBGPU=ON
|
| 381 |
-
cmake --build build --config Release -j $(nproc)
|
| 382 |
|
| 383 |
- name: Test
|
| 384 |
id: cmake_test
|
|
@@ -404,7 +475,7 @@ jobs:
|
|
| 404 |
|
| 405 |
- name: Fetch emdawnwebgpu
|
| 406 |
run: |
|
| 407 |
-
DAWN_TAG="
|
| 408 |
EMDAWN_PKG="emdawnwebgpu_pkg-${DAWN_TAG}.zip"
|
| 409 |
echo "Downloading ${EMDAWN_PKG}"
|
| 410 |
curl -L -o emdawn.zip \
|
|
@@ -415,11 +486,13 @@ jobs:
|
|
| 415 |
run: |
|
| 416 |
source emsdk/emsdk_env.sh
|
| 417 |
emcmake cmake -B build-wasm \
|
|
|
|
|
|
|
| 418 |
-DGGML_WEBGPU=ON \
|
| 419 |
-DLLAMA_OPENSSL=OFF \
|
| 420 |
-DEMDAWNWEBGPU_DIR=emdawnwebgpu_pkg
|
| 421 |
|
| 422 |
-
cmake --build build-wasm --target test-backend-ops -j $(nproc)
|
| 423 |
|
| 424 |
ubuntu-22-hip:
|
| 425 |
runs-on: ubuntu-22.04
|
|
@@ -449,6 +522,7 @@ jobs:
|
|
| 449 |
cmake -B build -S . \
|
| 450 |
-DCMAKE_HIP_COMPILER="$(hipconfig -l)/clang" \
|
| 451 |
-DGGML_HIP_ROCWMMA_FATTN=ON \
|
|
|
|
| 452 |
-DGGML_HIP=ON
|
| 453 |
cmake --build build --config Release -j $(nproc)
|
| 454 |
|
|
@@ -479,7 +553,7 @@ jobs:
|
|
| 479 |
run: |
|
| 480 |
cmake -B build -S . \
|
| 481 |
-DGGML_MUSA=ON
|
| 482 |
-
cmake --build build --config Release -j $(nproc)
|
| 483 |
|
| 484 |
ubuntu-22-sycl:
|
| 485 |
runs-on: ubuntu-22.04
|
|
@@ -528,7 +602,7 @@ jobs:
|
|
| 528 |
-DGGML_SYCL=ON \
|
| 529 |
-DCMAKE_C_COMPILER=icx \
|
| 530 |
-DCMAKE_CXX_COMPILER=icpx
|
| 531 |
-
cmake --build build --config Release -j $(nproc)
|
| 532 |
|
| 533 |
ubuntu-22-sycl-fp16:
|
| 534 |
runs-on: ubuntu-22.04
|
|
@@ -551,7 +625,7 @@ jobs:
|
|
| 551 |
shell: bash
|
| 552 |
run: |
|
| 553 |
sudo apt update
|
| 554 |
-
sudo apt install intel-oneapi-compiler-dpcpp-cpp libssl-dev
|
| 555 |
|
| 556 |
- name: install oneAPI MKL library
|
| 557 |
shell: bash
|
|
@@ -574,11 +648,13 @@ jobs:
|
|
| 574 |
run: |
|
| 575 |
source /opt/intel/oneapi/setvars.sh
|
| 576 |
cmake -B build \
|
|
|
|
|
|
|
| 577 |
-DGGML_SYCL=ON \
|
| 578 |
-DCMAKE_C_COMPILER=icx \
|
| 579 |
-DCMAKE_CXX_COMPILER=icpx \
|
| 580 |
-DGGML_SYCL_F16=ON
|
| 581 |
-
cmake --build build --config Release -j $(nproc)
|
| 582 |
|
| 583 |
ubuntu-24-openvino:
|
| 584 |
name: ubuntu-24-openvino-${{ matrix.openvino_device }}
|
|
@@ -648,7 +724,7 @@ jobs:
|
|
| 648 |
cmake -B build/ReleaseOV -G Ninja \
|
| 649 |
-DCMAKE_BUILD_TYPE=Release \
|
| 650 |
-DGGML_OPENVINO=ON
|
| 651 |
-
cmake --build build/ReleaseOV --config Release -j $(nproc)
|
| 652 |
|
| 653 |
- name: Test
|
| 654 |
id: cmake_test
|
|
@@ -916,7 +992,7 @@ jobs:
|
|
| 916 |
- name: Grab rocWMMA package
|
| 917 |
id: grab_rocwmma
|
| 918 |
run: |
|
| 919 |
-
curl -o rocwmma.deb "https://repo.radeon.com/rocm/apt/7.2/pool/main/r/rocwmma-dev/rocwmma-dev_2.2.0.
|
| 920 |
7z x rocwmma.deb
|
| 921 |
7z x data.tar
|
| 922 |
|
|
@@ -959,42 +1035,32 @@ jobs:
|
|
| 959 |
cmake -G "Unix Makefiles" -B build -S . `
|
| 960 |
-DCMAKE_C_COMPILER="${env:HIP_PATH}\bin\clang.exe" `
|
| 961 |
-DCMAKE_CXX_COMPILER="${env:HIP_PATH}\bin\clang++.exe" `
|
| 962 |
-
-DCMAKE_CXX_FLAGS="-I$($PWD.Path.Replace('\', '/'))/opt/rocm-7.2.
|
| 963 |
-DCMAKE_BUILD_TYPE=Release `
|
| 964 |
-DLLAMA_BUILD_BORINGSSL=ON `
|
| 965 |
-DROCM_DIR="${env:HIP_PATH}" `
|
| 966 |
-DGGML_HIP=ON `
|
| 967 |
-DGGML_HIP_ROCWMMA_FATTN=ON `
|
|
|
|
| 968 |
-DGGML_RPC=ON
|
| 969 |
cmake --build build -j ${env:NUMBER_OF_PROCESSORS}
|
| 970 |
|
| 971 |
ubuntu-cpu-riscv64-native:
|
| 972 |
-
runs-on:
|
| 973 |
|
| 974 |
steps:
|
| 975 |
- name: Install dependencies
|
| 976 |
run: |
|
| 977 |
-
sudo apt-get update
|
| 978 |
-
|
| 979 |
# Install necessary packages
|
| 980 |
-
sudo apt-get
|
|
|
|
| 981 |
|
| 982 |
# Set gcc-14 and g++-14 as the default compilers
|
| 983 |
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-14 100
|
| 984 |
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-14 100
|
| 985 |
-
sudo ln -sf /usr/bin/gcc-14 /usr/bin/gcc
|
| 986 |
-
sudo ln -sf /usr/bin/g++-14 /usr/bin/g++
|
| 987 |
-
|
| 988 |
-
# Install Rust stable version
|
| 989 |
-
rustup install stable
|
| 990 |
-
rustup default stable
|
| 991 |
|
| 992 |
git lfs install
|
| 993 |
|
| 994 |
-
- name: Clone
|
| 995 |
-
id: checkout
|
| 996 |
-
uses: actions/checkout@v6
|
| 997 |
-
|
| 998 |
- name: Check environment
|
| 999 |
run: |
|
| 1000 |
uname -a
|
|
@@ -1004,25 +1070,16 @@ jobs:
|
|
| 1004 |
cmake --version
|
| 1005 |
rustc --version
|
| 1006 |
|
| 1007 |
-
- name:
|
| 1008 |
-
|
| 1009 |
-
|
| 1010 |
-
export CCACHE_DIR="$HOME/.ccache/cpu-cmake-rv64-native"
|
| 1011 |
-
mkdir -p "$CCACHE_DIR"
|
| 1012 |
-
|
| 1013 |
-
# Configure ccache for optimal performance
|
| 1014 |
-
ccache --set-config=max_size=5G
|
| 1015 |
-
ccache --set-config=compression=true
|
| 1016 |
-
ccache --set-config=compression_level=6
|
| 1017 |
-
ccache --set-config=cache_dir="$CCACHE_DIR"
|
| 1018 |
-
|
| 1019 |
-
# Enable more aggressive caching
|
| 1020 |
-
ccache --set-config=sloppiness=file_macro,time_macros,include_file_mtime,include_file_ctime
|
| 1021 |
-
ccache --set-config=hash_dir=false
|
| 1022 |
|
| 1023 |
-
|
| 1024 |
-
|
| 1025 |
-
|
|
|
|
|
|
|
|
|
|
| 1026 |
|
| 1027 |
- name: Build
|
| 1028 |
id: cmake_build
|
|
@@ -1039,7 +1096,7 @@ jobs:
|
|
| 1039 |
-DCMAKE_C_COMPILER=riscv64-linux-gnu-gcc-14 \
|
| 1040 |
-DCMAKE_CXX_COMPILER=riscv64-linux-gnu-g++-14
|
| 1041 |
|
| 1042 |
-
cmake --build build --config Release -j $(nproc)
|
| 1043 |
|
| 1044 |
- name: Test
|
| 1045 |
id: cmake_test
|
|
|
|
| 87 |
-DGGML_METAL_EMBED_LIBRARY=OFF \
|
| 88 |
-DGGML_METAL_SHADER_DEBUG=ON \
|
| 89 |
-DGGML_RPC=ON
|
| 90 |
+
time cmake --build build --config Release -j $(sysctl -n hw.logicalcpu)
|
| 91 |
leaks -atExit -- ./build/bin/test-thread-safety -hf ggml-org/gemma-3-270m-qat-GGUF -ngl 99 -p "$(printf 'hello %.0s' {1..128})" -n 16 -c 512 -ub 32 -np 2 -t 2 -lv 1
|
| 92 |
|
| 93 |
- name: Test
|
|
|
|
| 124 |
-DGGML_METAL=OFF \
|
| 125 |
-DGGML_RPC=ON \
|
| 126 |
-DCMAKE_OSX_DEPLOYMENT_TARGET=13.3
|
| 127 |
+
time cmake --build build --config Release -j $(sysctl -n hw.logicalcpu)
|
| 128 |
|
| 129 |
- name: Test
|
| 130 |
id: cmake_test
|
|
|
|
| 150 |
- name: Dawn Dependency
|
| 151 |
id: dawn-depends
|
| 152 |
run: |
|
| 153 |
+
DAWN_VERSION="v20260317.182325"
|
| 154 |
+
DAWN_OWNER="google"
|
| 155 |
DAWN_REPO="dawn"
|
| 156 |
+
DAWN_ASSET_NAME="Dawn-18eb229ef5f707c1464cc581252e7603c73a3ef0-macos-latest-Release"
|
| 157 |
+
echo "Fetching release asset from https://github.com/google/dawn/releases/download/${DAWN_VERSION}/${DAWN_ASSET_NAME}.tar.gz"
|
| 158 |
+
curl -L -o artifact.tar.gz \
|
| 159 |
+
"https://github.com/google/dawn/releases/download/${DAWN_VERSION}/${DAWN_ASSET_NAME}.tar.gz"
|
| 160 |
mkdir dawn
|
| 161 |
+
tar -xvf artifact.tar.gz -C dawn --strip-components=1
|
|
|
|
| 162 |
|
| 163 |
- name: Build
|
| 164 |
id: cmake_build
|
| 165 |
run: |
|
| 166 |
export CMAKE_PREFIX_PATH=dawn
|
| 167 |
+
cmake -B build -G "Ninja" -DCMAKE_BUILD_TYPE=Release -DGGML_WEBGPU=ON -DGGML_METAL=OFF -DGGML_BLAS=OFF
|
| 168 |
+
time cmake --build build --config Release -j $(sysctl -n hw.logicalcpu)
|
| 169 |
|
| 170 |
- name: Test
|
| 171 |
id: cmake_test
|
|
|
|
| 180 |
- build: 'x64'
|
| 181 |
os: ubuntu-22.04
|
| 182 |
- build: 'arm64'
|
| 183 |
+
os: ubuntu-24.04-arm
|
| 184 |
- build: 's390x'
|
| 185 |
os: ubuntu-24.04-s390x
|
| 186 |
- build: 'ppc64le'
|
|
|
|
| 206 |
run: |
|
| 207 |
sudo apt-get update
|
| 208 |
sudo apt-get install -y --no-install-recommends \
|
| 209 |
+
python3 python3-pip python3-dev python3-wheel \
|
| 210 |
libjpeg-dev build-essential libssl-dev \
|
| 211 |
git-lfs
|
| 212 |
|
| 213 |
+
- name: Toolchain workaround (GCC 14)
|
| 214 |
+
if: ${{ contains(matrix.os, 'ubuntu-24.04') }}
|
| 215 |
+
run: |
|
| 216 |
+
sudo apt-get install -y gcc-14 g++-14
|
| 217 |
+
echo "CC=gcc-14" >> "$GITHUB_ENV"
|
| 218 |
+
echo "CXX=g++-14" >> "$GITHUB_ENV"
|
| 219 |
+
|
| 220 |
- name: Python Dependencies
|
| 221 |
id: python_depends
|
| 222 |
run: |
|
| 223 |
+
export PIP_BREAK_SYSTEM_PACKAGES="1"
|
| 224 |
+
python3 -m pip install --upgrade pip setuptools
|
| 225 |
pip3 install ./gguf-py
|
| 226 |
|
| 227 |
- name: Swap Endianness
|
|
|
|
| 238 |
cmake -B build \
|
| 239 |
-DLLAMA_FATAL_WARNINGS=ON \
|
| 240 |
-DGGML_RPC=ON
|
| 241 |
+
time cmake --build build --config Release -j $(nproc)
|
| 242 |
|
| 243 |
- name: Test
|
| 244 |
id: cmake_test
|
|
|
|
| 267 |
wget https://huggingface.co/ggml-org/models/resolve/main/tinyllamas/stories260K-be.gguf
|
| 268 |
./bin/llama-completion -m stories260K-be.gguf -p "One day, Lily met a Shoggoth" -n 500 -c 256
|
| 269 |
|
| 270 |
+
android-arm64:
|
| 271 |
+
runs-on: ubuntu-latest
|
| 272 |
+
|
| 273 |
+
env:
|
| 274 |
+
NDK_VERSION: "29.0.14206865"
|
| 275 |
+
|
| 276 |
+
steps:
|
| 277 |
+
- name: Clone
|
| 278 |
+
id: checkout
|
| 279 |
+
uses: actions/checkout@v6
|
| 280 |
+
|
| 281 |
+
- name: ccache
|
| 282 |
+
uses: ggml-org/ccache-action@v1.2.21
|
| 283 |
+
with:
|
| 284 |
+
key: android-arm64
|
| 285 |
+
evict-old-files: 1d
|
| 286 |
+
save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
|
| 287 |
+
|
| 288 |
+
- name: Set up JDK
|
| 289 |
+
uses: actions/setup-java@v5
|
| 290 |
+
with:
|
| 291 |
+
java-version: 17
|
| 292 |
+
distribution: temurin
|
| 293 |
+
|
| 294 |
+
- name: Setup Android SDK
|
| 295 |
+
uses: android-actions/setup-android@40fd30fb8d7440372e1316f5d1809ec01dcd3699 # v4.0.1
|
| 296 |
+
with:
|
| 297 |
+
log-accepted-android-sdk-licenses: false
|
| 298 |
+
|
| 299 |
+
- name: Install NDK
|
| 300 |
+
run: |
|
| 301 |
+
sdkmanager "ndk;${{ env.NDK_VERSION }}"
|
| 302 |
+
echo "ANDROID_NDK=${ANDROID_SDK_ROOT}/ndk/${{ env.NDK_VERSION }}" >> $GITHUB_ENV
|
| 303 |
+
|
| 304 |
+
- name: Build
|
| 305 |
+
id: cmake_build
|
| 306 |
+
run: |
|
| 307 |
+
cmake -B build \
|
| 308 |
+
-DCMAKE_TOOLCHAIN_FILE=${ANDROID_NDK}/build/cmake/android.toolchain.cmake \
|
| 309 |
+
-DANDROID_ABI=arm64-v8a \
|
| 310 |
+
-DANDROID_PLATFORM=android-28 \
|
| 311 |
+
-DLLAMA_FATAL_WARNINGS=ON \
|
| 312 |
+
-DGGML_BACKEND_DL=ON \
|
| 313 |
+
-DGGML_NATIVE=OFF \
|
| 314 |
+
-DGGML_CPU_ALL_VARIANTS=ON \
|
| 315 |
+
-DGGML_OPENMP=OFF \
|
| 316 |
+
-DLLAMA_BUILD_BORINGSSL=ON \
|
| 317 |
+
-DGGML_RPC=ON
|
| 318 |
+
time cmake --build build --config Release -j $(nproc)
|
| 319 |
+
|
| 320 |
ubuntu-latest-rpc:
|
| 321 |
runs-on: ubuntu-latest
|
| 322 |
|
|
|
|
| 331 |
id: depends
|
| 332 |
run: |
|
| 333 |
sudo apt-get update
|
| 334 |
+
sudo apt-get install build-essential libssl-dev ninja-build
|
| 335 |
|
| 336 |
- name: Build
|
| 337 |
id: cmake_build
|
| 338 |
run: |
|
| 339 |
cmake -B build \
|
| 340 |
+
-G "Ninja" \
|
| 341 |
+
-DCMAKE_BUILD_TYPE=Release \
|
| 342 |
-DGGML_RPC=ON
|
| 343 |
+
time cmake --build build --config Release -j $(nproc)
|
| 344 |
|
| 345 |
- name: Test
|
| 346 |
id: cmake_test
|
|
|
|
| 349 |
ctest -L main --verbose
|
| 350 |
|
| 351 |
ubuntu-24-vulkan:
|
| 352 |
+
strategy:
|
| 353 |
+
matrix:
|
| 354 |
+
include:
|
| 355 |
+
- build: 'x64'
|
| 356 |
+
os: ubuntu-24.04
|
| 357 |
+
- build: 'arm64'
|
| 358 |
+
os: ubuntu-24.04-arm
|
| 359 |
+
|
| 360 |
+
runs-on: ${{ matrix.os }}
|
| 361 |
|
| 362 |
steps:
|
| 363 |
- name: Clone
|
|
|
|
| 367 |
- name: Dependencies
|
| 368 |
id: depends
|
| 369 |
run: |
|
| 370 |
+
sudo apt-get update
|
| 371 |
+
sudo apt-get install -y gcc-14 g++-14 build-essential glslc libvulkan-dev spirv-headers libssl-dev ninja-build
|
| 372 |
+
echo "CC=gcc-14" >> "$GITHUB_ENV"
|
| 373 |
+
echo "CXX=g++-14" >> "$GITHUB_ENV"
|
| 374 |
|
| 375 |
- name: Configure
|
| 376 |
id: cmake_configure
|
| 377 |
run: |
|
| 378 |
cmake -B build \
|
| 379 |
+
-G "Ninja" \
|
| 380 |
-DCMAKE_BUILD_TYPE=RelWithDebInfo \
|
| 381 |
-DGGML_BACKEND_DL=ON \
|
| 382 |
-DGGML_CPU_ALL_VARIANTS=ON \
|
|
|
|
| 385 |
- name: Build
|
| 386 |
id: cmake_build
|
| 387 |
run: |
|
| 388 |
+
time cmake --build build -j $(nproc)
|
| 389 |
|
| 390 |
ubuntu-24-webgpu:
|
| 391 |
runs-on: ubuntu-24.04
|
|
|
|
| 407 |
run: |
|
| 408 |
sudo add-apt-repository -y ppa:kisak/kisak-mesa
|
| 409 |
sudo apt-get update -y
|
| 410 |
+
sudo apt-get install -y build-essential mesa-vulkan-drivers \
|
| 411 |
+
libxcb-xinput0 libxcb-xinerama0 libxcb-cursor-dev libssl-dev
|
| 412 |
|
| 413 |
- name: Get latest Vulkan SDK version
|
| 414 |
id: vulkan_sdk_version
|
|
|
|
| 433 |
id: dawn-depends
|
| 434 |
run: |
|
| 435 |
sudo apt-get install -y libxrandr-dev libxinerama-dev libxcursor-dev mesa-common-dev libx11-xcb-dev libxi-dev
|
| 436 |
+
DAWN_VERSION="v20260317.182325"
|
| 437 |
+
DAWN_OWNER="google"
|
| 438 |
DAWN_REPO="dawn"
|
| 439 |
+
DAWN_ASSET_NAME="Dawn-18eb229ef5f707c1464cc581252e7603c73a3ef0-ubuntu-latest-Release"
|
| 440 |
+
echo "Fetching release asset from https://github.com/google/dawn/releases/download/${DAWN_VERSION}/${DAWN_ASSET_NAME}.tar.gz"
|
| 441 |
+
curl -L -o artifact.tar.gz \
|
| 442 |
+
"https://github.com/google/dawn/releases/download/${DAWN_VERSION}/${DAWN_ASSET_NAME}.tar.gz"
|
| 443 |
mkdir dawn
|
| 444 |
+
tar -xvf artifact.tar.gz -C dawn --strip-components=1
|
|
|
|
| 445 |
|
| 446 |
- name: Build
|
| 447 |
id: cmake_build
|
|
|
|
| 449 |
export Dawn_DIR=dawn/lib64/cmake/Dawn
|
| 450 |
cmake -B build \
|
| 451 |
-DGGML_WEBGPU=ON
|
| 452 |
+
time cmake --build build --config Release -j $(nproc)
|
| 453 |
|
| 454 |
- name: Test
|
| 455 |
id: cmake_test
|
|
|
|
| 475 |
|
| 476 |
- name: Fetch emdawnwebgpu
|
| 477 |
run: |
|
| 478 |
+
DAWN_TAG="v20260317.182325"
|
| 479 |
EMDAWN_PKG="emdawnwebgpu_pkg-${DAWN_TAG}.zip"
|
| 480 |
echo "Downloading ${EMDAWN_PKG}"
|
| 481 |
curl -L -o emdawn.zip \
|
|
|
|
| 486 |
run: |
|
| 487 |
source emsdk/emsdk_env.sh
|
| 488 |
emcmake cmake -B build-wasm \
|
| 489 |
+
-G "Ninja" \
|
| 490 |
+
-DCMAKE_BUILD_TYPE=Release \
|
| 491 |
-DGGML_WEBGPU=ON \
|
| 492 |
-DLLAMA_OPENSSL=OFF \
|
| 493 |
-DEMDAWNWEBGPU_DIR=emdawnwebgpu_pkg
|
| 494 |
|
| 495 |
+
time cmake --build build-wasm --config Release --target test-backend-ops -j $(nproc)
|
| 496 |
|
| 497 |
ubuntu-22-hip:
|
| 498 |
runs-on: ubuntu-22.04
|
|
|
|
| 522 |
cmake -B build -S . \
|
| 523 |
-DCMAKE_HIP_COMPILER="$(hipconfig -l)/clang" \
|
| 524 |
-DGGML_HIP_ROCWMMA_FATTN=ON \
|
| 525 |
+
-DGPU_TARGETS="gfx1030" \
|
| 526 |
-DGGML_HIP=ON
|
| 527 |
cmake --build build --config Release -j $(nproc)
|
| 528 |
|
|
|
|
| 553 |
run: |
|
| 554 |
cmake -B build -S . \
|
| 555 |
-DGGML_MUSA=ON
|
| 556 |
+
time cmake --build build --config Release -j $(nproc)
|
| 557 |
|
| 558 |
ubuntu-22-sycl:
|
| 559 |
runs-on: ubuntu-22.04
|
|
|
|
| 602 |
-DGGML_SYCL=ON \
|
| 603 |
-DCMAKE_C_COMPILER=icx \
|
| 604 |
-DCMAKE_CXX_COMPILER=icpx
|
| 605 |
+
time cmake --build build --config Release -j $(nproc)
|
| 606 |
|
| 607 |
ubuntu-22-sycl-fp16:
|
| 608 |
runs-on: ubuntu-22.04
|
|
|
|
| 625 |
shell: bash
|
| 626 |
run: |
|
| 627 |
sudo apt update
|
| 628 |
+
sudo apt install intel-oneapi-compiler-dpcpp-cpp libssl-dev ninja-build
|
| 629 |
|
| 630 |
- name: install oneAPI MKL library
|
| 631 |
shell: bash
|
|
|
|
| 648 |
run: |
|
| 649 |
source /opt/intel/oneapi/setvars.sh
|
| 650 |
cmake -B build \
|
| 651 |
+
-G "Ninja" \
|
| 652 |
+
-DCMAKE_BUILD_TYPE=Release \
|
| 653 |
-DGGML_SYCL=ON \
|
| 654 |
-DCMAKE_C_COMPILER=icx \
|
| 655 |
-DCMAKE_CXX_COMPILER=icpx \
|
| 656 |
-DGGML_SYCL_F16=ON
|
| 657 |
+
time cmake --build build --config Release -j $(nproc)
|
| 658 |
|
| 659 |
ubuntu-24-openvino:
|
| 660 |
name: ubuntu-24-openvino-${{ matrix.openvino_device }}
|
|
|
|
| 724 |
cmake -B build/ReleaseOV -G Ninja \
|
| 725 |
-DCMAKE_BUILD_TYPE=Release \
|
| 726 |
-DGGML_OPENVINO=ON
|
| 727 |
+
time cmake --build build/ReleaseOV --config Release -j $(nproc)
|
| 728 |
|
| 729 |
- name: Test
|
| 730 |
id: cmake_test
|
|
|
|
| 992 |
- name: Grab rocWMMA package
|
| 993 |
id: grab_rocwmma
|
| 994 |
run: |
|
| 995 |
+
curl -o rocwmma.deb "https://repo.radeon.com/rocm/apt/7.2.1/pool/main/r/rocwmma-dev/rocwmma-dev_2.2.0.70201-81~24.04_amd64.deb"
|
| 996 |
7z x rocwmma.deb
|
| 997 |
7z x data.tar
|
| 998 |
|
|
|
|
| 1035 |
cmake -G "Unix Makefiles" -B build -S . `
|
| 1036 |
-DCMAKE_C_COMPILER="${env:HIP_PATH}\bin\clang.exe" `
|
| 1037 |
-DCMAKE_CXX_COMPILER="${env:HIP_PATH}\bin\clang++.exe" `
|
| 1038 |
+
-DCMAKE_CXX_FLAGS="-I$($PWD.Path.Replace('\', '/'))/opt/rocm-7.2.1/include/" `
|
| 1039 |
-DCMAKE_BUILD_TYPE=Release `
|
| 1040 |
-DLLAMA_BUILD_BORINGSSL=ON `
|
| 1041 |
-DROCM_DIR="${env:HIP_PATH}" `
|
| 1042 |
-DGGML_HIP=ON `
|
| 1043 |
-DGGML_HIP_ROCWMMA_FATTN=ON `
|
| 1044 |
+
-DGPU_TARGETS="gfx1100" `
|
| 1045 |
-DGGML_RPC=ON
|
| 1046 |
cmake --build build -j ${env:NUMBER_OF_PROCESSORS}
|
| 1047 |
|
| 1048 |
ubuntu-cpu-riscv64-native:
|
| 1049 |
+
runs-on: ubuntu-24.04-riscv
|
| 1050 |
|
| 1051 |
steps:
|
| 1052 |
- name: Install dependencies
|
| 1053 |
run: |
|
|
|
|
|
|
|
| 1054 |
# Install necessary packages
|
| 1055 |
+
sudo apt-get update
|
| 1056 |
+
sudo apt-get install -y libssl-dev
|
| 1057 |
|
| 1058 |
# Set gcc-14 and g++-14 as the default compilers
|
| 1059 |
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-14 100
|
| 1060 |
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-14 100
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1061 |
|
| 1062 |
git lfs install
|
| 1063 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1064 |
- name: Check environment
|
| 1065 |
run: |
|
| 1066 |
uname -a
|
|
|
|
| 1070 |
cmake --version
|
| 1071 |
rustc --version
|
| 1072 |
|
| 1073 |
+
- name: Clone
|
| 1074 |
+
id: checkout
|
| 1075 |
+
uses: actions/checkout@v6
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1076 |
|
| 1077 |
+
- name: ccache
|
| 1078 |
+
uses: ggml-org/ccache-action@afde29e5b5422e5da23cb1f639e8baecadeadfc3 # https://github.com/ggml-org/ccache-action/pull/1
|
| 1079 |
+
with:
|
| 1080 |
+
key: ubuntu-cpu-riscv64-native
|
| 1081 |
+
evict-old-files: 1d
|
| 1082 |
+
save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
|
| 1083 |
|
| 1084 |
- name: Build
|
| 1085 |
id: cmake_build
|
|
|
|
| 1096 |
-DCMAKE_C_COMPILER=riscv64-linux-gnu-gcc-14 \
|
| 1097 |
-DCMAKE_CXX_COMPILER=riscv64-linux-gnu-g++-14
|
| 1098 |
|
| 1099 |
+
time cmake --build build --config Release -j $(nproc)
|
| 1100 |
|
| 1101 |
- name: Test
|
| 1102 |
id: cmake_test
|
.github/workflows/close-issue.yml
CHANGED
|
@@ -17,7 +17,7 @@ jobs:
|
|
| 17 |
steps:
|
| 18 |
- uses: actions/stale@v10
|
| 19 |
with:
|
| 20 |
-
exempt-issue-labels: "refactoring,help wanted,good first issue,research 🔬,bug,roadmap"
|
| 21 |
days-before-issue-stale: 30
|
| 22 |
days-before-issue-close: 14
|
| 23 |
stale-issue-label: "stale"
|
|
|
|
| 17 |
steps:
|
| 18 |
- uses: actions/stale@v10
|
| 19 |
with:
|
| 20 |
+
exempt-issue-labels: "refactoring,help wanted,good first issue,research 🔬,bug,roadmap,security"
|
| 21 |
days-before-issue-stale: 30
|
| 22 |
days-before-issue-close: 14
|
| 23 |
stale-issue-label: "stale"
|
.github/workflows/copilot-setup-steps.yml
CHANGED
|
@@ -54,4 +54,3 @@ jobs:
|
|
| 54 |
python3 -m venv .venv
|
| 55 |
source .venv/bin/activate
|
| 56 |
pip install -r requirements/requirements-all.txt -r tools/server/tests/requirements.txt
|
| 57 |
-
pip install flake8 pyright pre-commit
|
|
|
|
| 54 |
python3 -m venv .venv
|
| 55 |
source .venv/bin/activate
|
| 56 |
pip install -r requirements/requirements-all.txt -r tools/server/tests/requirements.txt
|
|
|
.github/workflows/docker.yml
CHANGED
|
@@ -25,64 +25,146 @@ permissions:
|
|
| 25 |
packages: write
|
| 26 |
|
| 27 |
jobs:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
push_to_registry:
|
| 29 |
-
name: Push Docker image to Docker
|
|
|
|
| 30 |
|
| 31 |
runs-on: ${{ matrix.config.runs_on }}
|
| 32 |
-
env:
|
| 33 |
-
COMMIT_SHA: ${{ github.sha }}
|
| 34 |
strategy:
|
| 35 |
fail-fast: false
|
| 36 |
matrix:
|
| 37 |
-
config:
|
| 38 |
-
# Multi-stage build
|
| 39 |
-
# Note: the arm64 images are failing, which prevents the amd64 images from being built
|
| 40 |
-
# https://github.com/ggml-org/llama.cpp/issues/11888
|
| 41 |
-
#- { tag: "cpu", dockerfile: ".devops/cpu.Dockerfile", platforms: "linux/amd64,linux/arm64", full: true, light: true, server: true, free_disk_space: false }
|
| 42 |
-
- { tag: "cpu", dockerfile: ".devops/cpu.Dockerfile", platforms: "linux/amd64", full: true, light: true, server: true, free_disk_space: false, runs_on: "ubuntu-22.04" }
|
| 43 |
-
- { tag: "cuda cuda12", dockerfile: ".devops/cuda.Dockerfile", platforms: "linux/amd64", full: true, light: true, server: true, free_disk_space: true, runs_on: "ubuntu-22.04", cuda_version: "12.4.0", ubuntu_version: "22.04" }
|
| 44 |
-
- { tag: "cuda13", dockerfile: ".devops/cuda-new.Dockerfile", platforms: "linux/amd64", full: true, light: true, server: true, free_disk_space: true, runs_on: "ubuntu-22.04", cuda_version: "13.1.0", ubuntu_version: "24.04" }
|
| 45 |
-
- { tag: "musa", dockerfile: ".devops/musa.Dockerfile", platforms: "linux/amd64", full: true, light: true, server: true, free_disk_space: true, runs_on: "ubuntu-22.04" }
|
| 46 |
-
- { tag: "intel", dockerfile: ".devops/intel.Dockerfile", platforms: "linux/amd64", full: true, light: true, server: true, free_disk_space: true, runs_on: "ubuntu-22.04" }
|
| 47 |
-
- { tag: "vulkan", dockerfile: ".devops/vulkan.Dockerfile", platforms: "linux/amd64", full: true, light: true, server: true, free_disk_space: false, runs_on: "ubuntu-22.04" }
|
| 48 |
-
- { tag: "s390x", dockerfile: ".devops/s390x.Dockerfile", platforms: "linux/s390x", full: true, light: true, server: true, free_disk_space: false, runs_on: "ubuntu-22.04-s390x" }
|
| 49 |
-
- { tag: "rocm", dockerfile: ".devops/rocm.Dockerfile", platforms: "linux/amd64", full: true, light: true, server: true, free_disk_space: true, runs_on: "ubuntu-22.04" }
|
| 50 |
-
- { tag: "openvino", dockerfile: ".devops/openvino.Dockerfile", platforms: "linux/amd64", full: true, light: true, server: true, free_disk_space: false, runs_on: "ubuntu-22.04" }
|
| 51 |
steps:
|
| 52 |
- name: Check out the repo
|
| 53 |
uses: actions/checkout@v6
|
| 54 |
with:
|
| 55 |
-
fetch-depth: 0
|
|
|
|
| 56 |
|
| 57 |
- name: Set up QEMU
|
| 58 |
-
if: ${{ matrix.config.
|
| 59 |
-
uses: docker/setup-qemu-action@
|
| 60 |
with:
|
| 61 |
-
image: tonistiigi/binfmt:qemu-
|
| 62 |
|
| 63 |
- name: Set up Docker Buildx
|
| 64 |
-
uses: docker/setup-buildx-action@
|
| 65 |
|
| 66 |
-
- name: Log in to Docker
|
| 67 |
-
uses: docker/login-action@
|
| 68 |
with:
|
| 69 |
registry: ghcr.io
|
| 70 |
username: ${{ github.repository_owner }}
|
| 71 |
password: ${{ secrets.GITHUB_TOKEN }}
|
| 72 |
|
| 73 |
-
- name: Determine
|
| 74 |
-
id:
|
| 75 |
-
uses: ./.github/actions/get-tag-name
|
| 76 |
-
env:
|
| 77 |
-
BRANCH_NAME: ${{ github.head_ref || github.ref_name }}
|
| 78 |
-
|
| 79 |
-
- name: Determine image tag name
|
| 80 |
-
id: tag
|
| 81 |
shell: bash
|
| 82 |
run: |
|
|
|
|
|
|
|
| 83 |
REPO_OWNER="${GITHUB_REPOSITORY_OWNER@L}" # to lower case
|
| 84 |
REPO_NAME="${{ github.event.repository.name }}"
|
| 85 |
-
|
|
|
|
|
|
|
|
|
|
| 86 |
|
| 87 |
# list all tags possible
|
| 88 |
tags="${{ matrix.config.tag }}"
|
|
@@ -92,19 +174,16 @@ jobs:
|
|
| 92 |
else
|
| 93 |
TYPE="-$tag"
|
| 94 |
fi
|
| 95 |
-
|
| 96 |
-
FULLTAGS="${FULLTAGS:+$FULLTAGS,}${PREFIX}full${TYPE},${PREFIX}full${TYPE}-${{ steps.srctag.outputs.name }}"
|
| 97 |
-
LIGHTTAGS="${LIGHTTAGS:+$LIGHTTAGS,}${PREFIX}light${TYPE},${PREFIX}light${TYPE}-${{ steps.srctag.outputs.name }}"
|
| 98 |
-
SERVERTAGS="${SERVERTAGS:+$SERVERTAGS,}${PREFIX}server${TYPE},${PREFIX}server${TYPE}-${{ steps.srctag.outputs.name }}"
|
| 99 |
done
|
| 100 |
-
|
| 101 |
-
echo "
|
| 102 |
-
|
| 103 |
-
echo "
|
| 104 |
-
echo "
|
| 105 |
-
echo "
|
| 106 |
-
echo "
|
| 107 |
-
echo "
|
| 108 |
env:
|
| 109 |
GITHUB_REPOSITORY_OWNER: '${{ github.repository_owner }}'
|
| 110 |
|
|
@@ -125,15 +204,14 @@ jobs:
|
|
| 125 |
docker-images: true
|
| 126 |
swap-storage: true
|
| 127 |
|
| 128 |
-
- name: Build and push Full Docker image
|
|
|
|
| 129 |
if: ${{ (github.event_name == 'push' || github.event_name == 'schedule' || github.event_name == 'workflow_dispatch') && matrix.config.full == true }}
|
| 130 |
-
uses: docker/build-push-action@
|
| 131 |
with:
|
| 132 |
context: .
|
| 133 |
-
push: true
|
| 134 |
platforms: ${{ matrix.config.platforms }}
|
| 135 |
-
|
| 136 |
-
tags: ${{ steps.tag.outputs.full_output_tags }}
|
| 137 |
file: ${{ matrix.config.dockerfile }}
|
| 138 |
target: full
|
| 139 |
provenance: false
|
|
@@ -147,18 +225,17 @@ jobs:
|
|
| 147 |
#cache-to: type=local,dest=/tmp/.buildx-cache
|
| 148 |
#cache-from: type=local,src=/tmp/.buildx-cache
|
| 149 |
# using registry cache (no storage limit)
|
| 150 |
-
cache-from: type=registry,ref=${{ steps.
|
| 151 |
-
cache-to: type=registry,ref=${{ steps.
|
| 152 |
|
| 153 |
-
- name: Build and push Light Docker image
|
|
|
|
| 154 |
if: ${{ (github.event_name == 'push' || github.event_name == 'schedule' || github.event_name == 'workflow_dispatch') && matrix.config.light == true }}
|
| 155 |
-
uses: docker/build-push-action@
|
| 156 |
with:
|
| 157 |
context: .
|
| 158 |
-
push: true
|
| 159 |
platforms: ${{ matrix.config.platforms }}
|
| 160 |
-
|
| 161 |
-
tags: ${{ steps.tag.outputs.light_output_tags }}
|
| 162 |
file: ${{ matrix.config.dockerfile }}
|
| 163 |
target: light
|
| 164 |
provenance: false
|
|
@@ -172,18 +249,17 @@ jobs:
|
|
| 172 |
#cache-to: type=local,dest=/tmp/.buildx-cache
|
| 173 |
#cache-from: type=local,src=/tmp/.buildx-cache
|
| 174 |
# using registry cache (no storage limit)
|
| 175 |
-
cache-from: type=registry,ref=${{ steps.
|
| 176 |
-
cache-to: type=registry,ref=${{ steps.
|
| 177 |
|
| 178 |
-
- name: Build and push Server Docker image
|
|
|
|
| 179 |
if: ${{ (github.event_name == 'push' || github.event_name == 'schedule' || github.event_name == 'workflow_dispatch') && matrix.config.server == true }}
|
| 180 |
-
uses: docker/build-push-action@
|
| 181 |
with:
|
| 182 |
context: .
|
| 183 |
-
push: true
|
| 184 |
platforms: ${{ matrix.config.platforms }}
|
| 185 |
-
|
| 186 |
-
tags: ${{ steps.tag.outputs.server_output_tags }}
|
| 187 |
file: ${{ matrix.config.dockerfile }}
|
| 188 |
target: server
|
| 189 |
provenance: false
|
|
@@ -197,31 +273,170 @@ jobs:
|
|
| 197 |
#cache-to: type=local,dest=/tmp/.buildx-cache
|
| 198 |
#cache-from: type=local,src=/tmp/.buildx-cache
|
| 199 |
# using registry cache (no storage limit)
|
| 200 |
-
cache-from: type=registry,ref=${{ steps.
|
| 201 |
-
cache-to: type=registry,ref=${{ steps.
|
| 202 |
|
| 203 |
-
|
| 204 |
-
|
| 205 |
-
|
| 206 |
-
|
| 207 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 208 |
|
| 209 |
steps:
|
| 210 |
-
- name:
|
| 211 |
-
id: checkout
|
| 212 |
uses: actions/checkout@v6
|
| 213 |
with:
|
| 214 |
fetch-depth: 0
|
| 215 |
|
| 216 |
-
- name:
|
| 217 |
-
|
| 218 |
-
|
| 219 |
-
|
| 220 |
-
|
|
|
|
| 221 |
|
| 222 |
-
- name:
|
| 223 |
-
|
| 224 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 225 |
run: |
|
| 226 |
-
|
| 227 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 25 |
packages: write
|
| 26 |
|
| 27 |
jobs:
|
| 28 |
+
create_tag:
|
| 29 |
+
name: Create and push git tag
|
| 30 |
+
runs-on: ubuntu-slim
|
| 31 |
+
permissions:
|
| 32 |
+
contents: write
|
| 33 |
+
outputs:
|
| 34 |
+
source_tag: ${{ steps.srctag.outputs.name }}
|
| 35 |
+
|
| 36 |
+
steps:
|
| 37 |
+
- name: Clone
|
| 38 |
+
id: checkout
|
| 39 |
+
uses: actions/checkout@v6
|
| 40 |
+
with:
|
| 41 |
+
fetch-depth: 0
|
| 42 |
+
|
| 43 |
+
- name: Determine source tag name
|
| 44 |
+
id: srctag
|
| 45 |
+
uses: ./.github/actions/get-tag-name
|
| 46 |
+
env:
|
| 47 |
+
BRANCH_NAME: ${{ github.head_ref || github.ref_name }}
|
| 48 |
+
|
| 49 |
+
- name: Create and push git tag
|
| 50 |
+
env:
|
| 51 |
+
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
|
| 52 |
+
run: |
|
| 53 |
+
git tag ${{ steps.srctag.outputs.name }} || exit 0
|
| 54 |
+
git push origin ${{ steps.srctag.outputs.name }} || exit 0
|
| 55 |
+
|
| 56 |
+
prepare_matrices:
|
| 57 |
+
name: Prepare Docker matrices
|
| 58 |
+
runs-on: ubuntu-24.04
|
| 59 |
+
outputs:
|
| 60 |
+
build_matrix: ${{ steps.matrices.outputs.build_matrix }}
|
| 61 |
+
merge_matrix: ${{ steps.matrices.outputs.merge_matrix }}
|
| 62 |
+
|
| 63 |
+
steps:
|
| 64 |
+
- name: Generate build and merge matrices
|
| 65 |
+
id: matrices
|
| 66 |
+
shell: bash
|
| 67 |
+
run: |
|
| 68 |
+
set -euo pipefail
|
| 69 |
+
|
| 70 |
+
# Keep all build targets in one place and derive merge targets from it.
|
| 71 |
+
cat > build-matrix.json <<'JSON'
|
| 72 |
+
[
|
| 73 |
+
{ "tag": "cpu", "dockerfile": ".devops/cpu.Dockerfile", "platforms": "linux/amd64", "full": true, "light": true, "server": true, "free_disk_space": false, "runs_on": "ubuntu-24.04" },
|
| 74 |
+
{ "tag": "cpu", "dockerfile": ".devops/cpu.Dockerfile", "platforms": "linux/arm64", "full": true, "light": true, "server": true, "free_disk_space": false, "runs_on": "ubuntu-24.04-arm" },
|
| 75 |
+
{ "tag": "cpu", "dockerfile": ".devops/s390x.Dockerfile", "platforms": "linux/s390x", "full": true, "light": true, "server": true, "free_disk_space": false, "runs_on": "ubuntu-24.04-s390x" },
|
| 76 |
+
{ "tag": "cuda cuda12", "dockerfile": ".devops/cuda.Dockerfile", "cuda_version": "12.8.1", "platforms": "linux/amd64", "full": true, "light": true, "server": true, "free_disk_space": true, "runs_on": "ubuntu-24.04" },
|
| 77 |
+
{ "tag": "cuda cuda12", "dockerfile": ".devops/cuda.Dockerfile", "cuda_version": "12.8.1", "platforms": "linux/arm64", "full": true, "light": true, "server": true, "free_disk_space": true, "runs_on": "ubuntu-24.04-arm" },
|
| 78 |
+
{ "tag": "cuda13", "dockerfile": ".devops/cuda.Dockerfile", "cuda_version": "13.1.1", "platforms": "linux/amd64", "full": true, "light": true, "server": true, "free_disk_space": true, "runs_on": "ubuntu-24.04" },
|
| 79 |
+
{ "tag": "cuda13", "dockerfile": ".devops/cuda.Dockerfile", "cuda_version": "13.1.1", "platforms": "linux/arm64", "full": true, "light": true, "server": true, "free_disk_space": true, "runs_on": "ubuntu-24.04-arm" },
|
| 80 |
+
{ "tag": "musa", "dockerfile": ".devops/musa.Dockerfile", "platforms": "linux/amd64", "full": true, "light": true, "server": true, "free_disk_space": true, "runs_on": "ubuntu-24.04" },
|
| 81 |
+
{ "tag": "intel", "dockerfile": ".devops/intel.Dockerfile", "platforms": "linux/amd64", "full": true, "light": true, "server": true, "free_disk_space": true, "runs_on": "ubuntu-24.04" },
|
| 82 |
+
{ "tag": "vulkan", "dockerfile": ".devops/vulkan.Dockerfile", "platforms": "linux/amd64", "full": true, "light": true, "server": true, "free_disk_space": false, "runs_on": "ubuntu-24.04" },
|
| 83 |
+
{ "tag": "vulkan", "dockerfile": ".devops/vulkan.Dockerfile", "platforms": "linux/arm64", "full": true, "light": true, "server": true, "free_disk_space": false, "runs_on": "ubuntu-24.04-arm" },
|
| 84 |
+
{ "tag": "rocm", "dockerfile": ".devops/rocm.Dockerfile", "platforms": "linux/amd64", "full": true, "light": true, "server": true, "free_disk_space": true, "runs_on": "ubuntu-24.04" },
|
| 85 |
+
{ "tag": "openvino", "dockerfile": ".devops/openvino.Dockerfile", "platforms": "linux/amd64", "full": true, "light": true, "server": true, "free_disk_space": false, "runs_on": "ubuntu-24.04" }
|
| 86 |
+
]
|
| 87 |
+
JSON
|
| 88 |
+
|
| 89 |
+
BUILD_MATRIX="$(jq -c . build-matrix.json)"
|
| 90 |
+
MERGE_MATRIX="$(jq -c '
|
| 91 |
+
reduce .[] as $entry ({}; .[$entry.tag] |= (
|
| 92 |
+
. // {
|
| 93 |
+
tag: $entry.tag,
|
| 94 |
+
arches: [],
|
| 95 |
+
full: false,
|
| 96 |
+
light: false,
|
| 97 |
+
server: false
|
| 98 |
+
}
|
| 99 |
+
| .full = (.full or ($entry.full // false))
|
| 100 |
+
| .light = (.light or ($entry.light // false))
|
| 101 |
+
| .server = (.server or ($entry.server // false))
|
| 102 |
+
| .arches += [($entry.platforms | sub("^linux/"; ""))]
|
| 103 |
+
))
|
| 104 |
+
# Backward compatibility: s390x tags are aliases of cpu for the linux/s390x platform.
|
| 105 |
+
| if (has("cpu") and (((.cpu.arches // []) | index("s390x")) != null)) then
|
| 106 |
+
. + {
|
| 107 |
+
s390x: {
|
| 108 |
+
tag: "s390x",
|
| 109 |
+
arches: ["s390x"],
|
| 110 |
+
full: .cpu.full,
|
| 111 |
+
light: .cpu.light,
|
| 112 |
+
server: .cpu.server
|
| 113 |
+
}
|
| 114 |
+
}
|
| 115 |
+
else
|
| 116 |
+
.
|
| 117 |
+
end
|
| 118 |
+
| [.[] | .arches = (.arches | unique | sort | join(" "))]
|
| 119 |
+
' build-matrix.json)"
|
| 120 |
+
|
| 121 |
+
echo "build_matrix=$BUILD_MATRIX" >> "$GITHUB_OUTPUT"
|
| 122 |
+
echo "merge_matrix=$MERGE_MATRIX" >> "$GITHUB_OUTPUT"
|
| 123 |
+
|
| 124 |
push_to_registry:
|
| 125 |
+
name: Push Docker image to Docker Registry
|
| 126 |
+
needs: [prepare_matrices, create_tag]
|
| 127 |
|
| 128 |
runs-on: ${{ matrix.config.runs_on }}
|
|
|
|
|
|
|
| 129 |
strategy:
|
| 130 |
fail-fast: false
|
| 131 |
matrix:
|
| 132 |
+
config: ${{ fromJSON(needs.prepare_matrices.outputs.build_matrix) }}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 133 |
steps:
|
| 134 |
- name: Check out the repo
|
| 135 |
uses: actions/checkout@v6
|
| 136 |
with:
|
| 137 |
+
fetch-depth: 0
|
| 138 |
+
ref: ${{ needs.create_tag.outputs.source_tag }}
|
| 139 |
|
| 140 |
- name: Set up QEMU
|
| 141 |
+
if: ${{ contains(matrix.config.platforms, 'linux/amd64') }}
|
| 142 |
+
uses: docker/setup-qemu-action@ce360397dd3f832beb865e1373c09c0e9f86d70a # v4
|
| 143 |
with:
|
| 144 |
+
image: tonistiigi/binfmt:qemu-v10.2.1
|
| 145 |
|
| 146 |
- name: Set up Docker Buildx
|
| 147 |
+
uses: docker/setup-buildx-action@4d04d5d9486b7bd6fa91e7baf45bbb4f8b9deedd # v4
|
| 148 |
|
| 149 |
+
- name: Log in to Docker Registry
|
| 150 |
+
uses: docker/login-action@b45d80f862d83dbcd57f89517bcf500b2ab88fb2 # v4
|
| 151 |
with:
|
| 152 |
registry: ghcr.io
|
| 153 |
username: ${{ github.repository_owner }}
|
| 154 |
password: ${{ secrets.GITHUB_TOKEN }}
|
| 155 |
|
| 156 |
+
- name: Determine image metadata
|
| 157 |
+
id: meta
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 158 |
shell: bash
|
| 159 |
run: |
|
| 160 |
+
set -euo pipefail
|
| 161 |
+
|
| 162 |
REPO_OWNER="${GITHUB_REPOSITORY_OWNER@L}" # to lower case
|
| 163 |
REPO_NAME="${{ github.event.repository.name }}"
|
| 164 |
+
IMAGE_REPO="ghcr.io/${REPO_OWNER}/${REPO_NAME}"
|
| 165 |
+
PREFIX="${IMAGE_REPO}:"
|
| 166 |
+
PLATFORM="${{ matrix.config.platforms }}"
|
| 167 |
+
ARCH_SUFFIX="${PLATFORM#linux/}"
|
| 168 |
|
| 169 |
# list all tags possible
|
| 170 |
tags="${{ matrix.config.tag }}"
|
|
|
|
| 174 |
else
|
| 175 |
TYPE="-$tag"
|
| 176 |
fi
|
| 177 |
+
CACHETAG="${PREFIX}buildcache${TYPE}-${ARCH_SUFFIX}"
|
|
|
|
|
|
|
|
|
|
| 178 |
done
|
| 179 |
+
|
| 180 |
+
SAFE_TAGS="$(echo "$tags" | tr ' ' '_')"
|
| 181 |
+
|
| 182 |
+
echo "image_repo=$IMAGE_REPO" >> $GITHUB_OUTPUT
|
| 183 |
+
echo "arch_suffix=$ARCH_SUFFIX" >> $GITHUB_OUTPUT
|
| 184 |
+
echo "cache_output_tag=$CACHETAG" >> $GITHUB_OUTPUT
|
| 185 |
+
echo "digest_artifact_suffix=${SAFE_TAGS}-${ARCH_SUFFIX}" >> $GITHUB_OUTPUT
|
| 186 |
+
echo "cache_output_tag=$CACHETAG" # print out for debugging
|
| 187 |
env:
|
| 188 |
GITHUB_REPOSITORY_OWNER: '${{ github.repository_owner }}'
|
| 189 |
|
|
|
|
| 204 |
docker-images: true
|
| 205 |
swap-storage: true
|
| 206 |
|
| 207 |
+
- name: Build and push Full Docker image by digest
|
| 208 |
+
id: build_full
|
| 209 |
if: ${{ (github.event_name == 'push' || github.event_name == 'schedule' || github.event_name == 'workflow_dispatch') && matrix.config.full == true }}
|
| 210 |
+
uses: docker/build-push-action@d08e5c354a6adb9ed34480a06d141179aa583294 # v7
|
| 211 |
with:
|
| 212 |
context: .
|
|
|
|
| 213 |
platforms: ${{ matrix.config.platforms }}
|
| 214 |
+
outputs: type=image,name=${{ steps.meta.outputs.image_repo }},push-by-digest=true,name-canonical=true,push=true
|
|
|
|
| 215 |
file: ${{ matrix.config.dockerfile }}
|
| 216 |
target: full
|
| 217 |
provenance: false
|
|
|
|
| 225 |
#cache-to: type=local,dest=/tmp/.buildx-cache
|
| 226 |
#cache-from: type=local,src=/tmp/.buildx-cache
|
| 227 |
# using registry cache (no storage limit)
|
| 228 |
+
cache-from: type=registry,ref=${{ steps.meta.outputs.cache_output_tag }}
|
| 229 |
+
cache-to: type=registry,ref=${{ steps.meta.outputs.cache_output_tag }},mode=max
|
| 230 |
|
| 231 |
+
- name: Build and push Light Docker image by digest
|
| 232 |
+
id: build_light
|
| 233 |
if: ${{ (github.event_name == 'push' || github.event_name == 'schedule' || github.event_name == 'workflow_dispatch') && matrix.config.light == true }}
|
| 234 |
+
uses: docker/build-push-action@d08e5c354a6adb9ed34480a06d141179aa583294 # v7
|
| 235 |
with:
|
| 236 |
context: .
|
|
|
|
| 237 |
platforms: ${{ matrix.config.platforms }}
|
| 238 |
+
outputs: type=image,name=${{ steps.meta.outputs.image_repo }},push-by-digest=true,name-canonical=true,push=true
|
|
|
|
| 239 |
file: ${{ matrix.config.dockerfile }}
|
| 240 |
target: light
|
| 241 |
provenance: false
|
|
|
|
| 249 |
#cache-to: type=local,dest=/tmp/.buildx-cache
|
| 250 |
#cache-from: type=local,src=/tmp/.buildx-cache
|
| 251 |
# using registry cache (no storage limit)
|
| 252 |
+
cache-from: type=registry,ref=${{ steps.meta.outputs.cache_output_tag }}
|
| 253 |
+
cache-to: type=registry,ref=${{ steps.meta.outputs.cache_output_tag }},mode=max
|
| 254 |
|
| 255 |
+
- name: Build and push Server Docker image by digest
|
| 256 |
+
id: build_server
|
| 257 |
if: ${{ (github.event_name == 'push' || github.event_name == 'schedule' || github.event_name == 'workflow_dispatch') && matrix.config.server == true }}
|
| 258 |
+
uses: docker/build-push-action@d08e5c354a6adb9ed34480a06d141179aa583294 # v7
|
| 259 |
with:
|
| 260 |
context: .
|
|
|
|
| 261 |
platforms: ${{ matrix.config.platforms }}
|
| 262 |
+
outputs: type=image,name=${{ steps.meta.outputs.image_repo }},push-by-digest=true,name-canonical=true,push=true
|
|
|
|
| 263 |
file: ${{ matrix.config.dockerfile }}
|
| 264 |
target: server
|
| 265 |
provenance: false
|
|
|
|
| 273 |
#cache-to: type=local,dest=/tmp/.buildx-cache
|
| 274 |
#cache-from: type=local,src=/tmp/.buildx-cache
|
| 275 |
# using registry cache (no storage limit)
|
| 276 |
+
cache-from: type=registry,ref=${{ steps.meta.outputs.cache_output_tag }}
|
| 277 |
+
cache-to: type=registry,ref=${{ steps.meta.outputs.cache_output_tag }},mode=max
|
| 278 |
|
| 279 |
+
- name: Export digest metadata
|
| 280 |
+
shell: bash
|
| 281 |
+
run: |
|
| 282 |
+
set -euo pipefail
|
| 283 |
+
|
| 284 |
+
TAGS="${{ matrix.config.tag }}"
|
| 285 |
+
ARCH_SUFFIX="${{ steps.meta.outputs.arch_suffix }}"
|
| 286 |
+
DIGEST_FILE="/tmp/digests/${{ steps.meta.outputs.digest_artifact_suffix }}.tsv"
|
| 287 |
+
mkdir -p /tmp/digests
|
| 288 |
+
|
| 289 |
+
add_digest_rows() {
|
| 290 |
+
local image_type="$1"
|
| 291 |
+
local digest="$2"
|
| 292 |
+
|
| 293 |
+
if [[ -z "$digest" ]]; then
|
| 294 |
+
echo "Missing digest for image_type=${image_type}" >&2
|
| 295 |
+
exit 1
|
| 296 |
+
fi
|
| 297 |
+
|
| 298 |
+
for tag in $TAGS; do
|
| 299 |
+
printf '%s\t%s\t%s\t%s\n' "$tag" "$ARCH_SUFFIX" "$image_type" "$digest" >> "$DIGEST_FILE"
|
| 300 |
+
done
|
| 301 |
+
}
|
| 302 |
+
|
| 303 |
+
if [[ "${{ matrix.config.full }}" == "true" ]]; then
|
| 304 |
+
add_digest_rows "full" "${{ steps.build_full.outputs.digest }}"
|
| 305 |
+
fi
|
| 306 |
+
|
| 307 |
+
if [[ "${{ matrix.config.light }}" == "true" ]]; then
|
| 308 |
+
add_digest_rows "light" "${{ steps.build_light.outputs.digest }}"
|
| 309 |
+
fi
|
| 310 |
+
|
| 311 |
+
if [[ "${{ matrix.config.server }}" == "true" ]]; then
|
| 312 |
+
add_digest_rows "server" "${{ steps.build_server.outputs.digest }}"
|
| 313 |
+
fi
|
| 314 |
+
|
| 315 |
+
- name: Upload digest metadata
|
| 316 |
+
uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f # v7
|
| 317 |
+
with:
|
| 318 |
+
name: digests-${{ steps.meta.outputs.digest_artifact_suffix }}
|
| 319 |
+
path: /tmp/digests/${{ steps.meta.outputs.digest_artifact_suffix }}.tsv
|
| 320 |
+
if-no-files-found: error
|
| 321 |
+
|
| 322 |
+
merge_arch_tags:
|
| 323 |
+
name: Create shared tags from digests
|
| 324 |
+
needs: [prepare_matrices, push_to_registry, create_tag]
|
| 325 |
+
runs-on: ubuntu-24.04
|
| 326 |
+
strategy:
|
| 327 |
+
fail-fast: false
|
| 328 |
+
matrix:
|
| 329 |
+
config: ${{ fromJSON(needs.prepare_matrices.outputs.merge_matrix) }}
|
| 330 |
|
| 331 |
steps:
|
| 332 |
+
- name: Check out the repo
|
|
|
|
| 333 |
uses: actions/checkout@v6
|
| 334 |
with:
|
| 335 |
fetch-depth: 0
|
| 336 |
|
| 337 |
+
- name: Download digest metadata
|
| 338 |
+
uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8
|
| 339 |
+
with:
|
| 340 |
+
pattern: digests-*
|
| 341 |
+
path: /tmp/digests
|
| 342 |
+
merge-multiple: true
|
| 343 |
|
| 344 |
+
- name: Set up Docker Buildx
|
| 345 |
+
uses: docker/setup-buildx-action@4d04d5d9486b7bd6fa91e7baf45bbb4f8b9deedd # v4
|
| 346 |
+
|
| 347 |
+
- name: Log in to Docker Registry
|
| 348 |
+
uses: docker/login-action@b45d80f862d83dbcd57f89517bcf500b2ab88fb2 # v4
|
| 349 |
+
with:
|
| 350 |
+
registry: ghcr.io
|
| 351 |
+
username: ${{ github.repository_owner }}
|
| 352 |
+
password: ${{ secrets.GITHUB_TOKEN }}
|
| 353 |
+
|
| 354 |
+
- name: Create tags from digests
|
| 355 |
+
shell: bash
|
| 356 |
run: |
|
| 357 |
+
set -euo pipefail
|
| 358 |
+
|
| 359 |
+
REPO_OWNER="${GITHUB_REPOSITORY_OWNER@L}" # to lower case
|
| 360 |
+
REPO_NAME="${{ github.event.repository.name }}"
|
| 361 |
+
IMAGE_REPO="ghcr.io/${REPO_OWNER}/${REPO_NAME}"
|
| 362 |
+
PREFIX="${IMAGE_REPO}:"
|
| 363 |
+
SRC_TAG="${{ needs.create_tag.outputs.source_tag }}"
|
| 364 |
+
TAGS="${{ matrix.config.tag }}"
|
| 365 |
+
ARCHES="${{ matrix.config.arches }}"
|
| 366 |
+
DIGEST_GLOB="/tmp/digests/*.tsv"
|
| 367 |
+
|
| 368 |
+
if ! ls ${DIGEST_GLOB} >/dev/null 2>&1; then
|
| 369 |
+
echo "No digest metadata found in /tmp/digests" >&2
|
| 370 |
+
exit 1
|
| 371 |
+
fi
|
| 372 |
+
|
| 373 |
+
if [[ -z "$SRC_TAG" ]]; then
|
| 374 |
+
echo "Missing source tag from create_tag" >&2
|
| 375 |
+
exit 1
|
| 376 |
+
fi
|
| 377 |
+
|
| 378 |
+
find_digest() {
|
| 379 |
+
local tag_name="$1"
|
| 380 |
+
local arch="$2"
|
| 381 |
+
local image_type="$3"
|
| 382 |
+
local digest
|
| 383 |
+
|
| 384 |
+
digest="$(awk -F '\t' -v t="$tag_name" -v a="$arch" -v i="$image_type" '$1 == t && $2 == a && $3 == i { print $4; exit }' ${DIGEST_GLOB})"
|
| 385 |
+
|
| 386 |
+
# Backward compatibility: s390x tags are aliases of cpu for the linux/s390x platform.
|
| 387 |
+
if [[ -z "$digest" && "$tag_name" == "s390x" && "$arch" == "s390x" ]]; then
|
| 388 |
+
digest="$(awk -F '\t' -v t="cpu" -v a="$arch" -v i="$image_type" '$1 == t && $2 == a && $3 == i { print $4; exit }' ${DIGEST_GLOB})"
|
| 389 |
+
fi
|
| 390 |
+
|
| 391 |
+
if [[ -z "$digest" ]]; then
|
| 392 |
+
echo "Missing digest for tag=${tag_name} arch=${arch} image_type=${image_type}" >&2
|
| 393 |
+
exit 1
|
| 394 |
+
fi
|
| 395 |
+
|
| 396 |
+
echo "$digest"
|
| 397 |
+
}
|
| 398 |
+
|
| 399 |
+
create_manifest_tags() {
|
| 400 |
+
local image_type="$1"
|
| 401 |
+
local tag_name="$2"
|
| 402 |
+
local suffix="$3"
|
| 403 |
+
|
| 404 |
+
local merged_tag="${PREFIX}${image_type}${suffix}"
|
| 405 |
+
local merged_versioned_tag="${merged_tag}-${SRC_TAG}"
|
| 406 |
+
|
| 407 |
+
local refs=()
|
| 408 |
+
|
| 409 |
+
for arch in $ARCHES; do
|
| 410 |
+
local digest
|
| 411 |
+
digest="$(find_digest "$tag_name" "$arch" "$image_type")"
|
| 412 |
+
refs+=("${IMAGE_REPO}@${digest}")
|
| 413 |
+
done
|
| 414 |
+
|
| 415 |
+
echo "Creating ${merged_tag} from ${refs[*]}"
|
| 416 |
+
docker buildx imagetools create --tag "${merged_tag}" "${refs[@]}"
|
| 417 |
+
|
| 418 |
+
echo "Creating ${merged_versioned_tag} from ${refs[*]}"
|
| 419 |
+
docker buildx imagetools create --tag "${merged_versioned_tag}" "${refs[@]}"
|
| 420 |
+
}
|
| 421 |
+
|
| 422 |
+
for tag in $TAGS; do
|
| 423 |
+
if [[ "$tag" == "cpu" ]]; then
|
| 424 |
+
TYPE=""
|
| 425 |
+
else
|
| 426 |
+
TYPE="-$tag"
|
| 427 |
+
fi
|
| 428 |
+
|
| 429 |
+
if [[ "${{ matrix.config.full }}" == "true" ]]; then
|
| 430 |
+
create_manifest_tags "full" "$tag" "$TYPE"
|
| 431 |
+
fi
|
| 432 |
+
|
| 433 |
+
if [[ "${{ matrix.config.light }}" == "true" ]]; then
|
| 434 |
+
create_manifest_tags "light" "$tag" "$TYPE"
|
| 435 |
+
fi
|
| 436 |
+
|
| 437 |
+
if [[ "${{ matrix.config.server }}" == "true" ]]; then
|
| 438 |
+
create_manifest_tags "server" "$tag" "$TYPE"
|
| 439 |
+
fi
|
| 440 |
+
done
|
| 441 |
+
env:
|
| 442 |
+
GITHUB_REPOSITORY_OWNER: '${{ github.repository_owner }}'
|
.github/workflows/editorconfig.yml
CHANGED
|
@@ -23,7 +23,7 @@ jobs:
|
|
| 23 |
runs-on: ubuntu-slim
|
| 24 |
steps:
|
| 25 |
- uses: actions/checkout@v6
|
| 26 |
-
- uses: editorconfig-checker/action-editorconfig-checker@v2
|
| 27 |
with:
|
| 28 |
version: v3.0.3
|
| 29 |
- run: editorconfig-checker
|
|
|
|
| 23 |
runs-on: ubuntu-slim
|
| 24 |
steps:
|
| 25 |
- uses: actions/checkout@v6
|
| 26 |
+
- uses: editorconfig-checker/action-editorconfig-checker@840e866d93b8e032123c23bac69dece044d4d84c # v2.2.0
|
| 27 |
with:
|
| 28 |
version: v3.0.3
|
| 29 |
- run: editorconfig-checker
|
.github/workflows/gguf-publish.yml
CHANGED
|
@@ -28,17 +28,17 @@ jobs:
|
|
| 28 |
- name: Set up Python
|
| 29 |
uses: actions/setup-python@v6
|
| 30 |
with:
|
| 31 |
-
python-version: '3.
|
| 32 |
- name: Install dependencies
|
| 33 |
run: |
|
| 34 |
cd gguf-py
|
| 35 |
-
python -m pip install poetry
|
| 36 |
poetry install
|
| 37 |
|
| 38 |
- name: Build package
|
| 39 |
run: cd gguf-py && poetry build
|
| 40 |
- name: Publish package
|
| 41 |
-
uses: pypa/gh-action-pypi-publish@release/v1
|
| 42 |
with:
|
| 43 |
password: ${{ secrets.PYPI_API_TOKEN }}
|
| 44 |
packages-dir: gguf-py/dist
|
|
|
|
| 28 |
- name: Set up Python
|
| 29 |
uses: actions/setup-python@v6
|
| 30 |
with:
|
| 31 |
+
python-version: '3.11'
|
| 32 |
- name: Install dependencies
|
| 33 |
run: |
|
| 34 |
cd gguf-py
|
| 35 |
+
python -m pip install poetry==2.3.2
|
| 36 |
poetry install
|
| 37 |
|
| 38 |
- name: Build package
|
| 39 |
run: cd gguf-py && poetry build
|
| 40 |
- name: Publish package
|
| 41 |
+
uses: pypa/gh-action-pypi-publish@ed0c53931b1dc9bd32cbe73a98c7f6766f8a527e # release/v1
|
| 42 |
with:
|
| 43 |
password: ${{ secrets.PYPI_API_TOKEN }}
|
| 44 |
packages-dir: gguf-py/dist
|
.github/workflows/hip-quality-check.yml
ADDED
|
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
name: HIP quality check
|
| 2 |
+
|
| 3 |
+
on:
|
| 4 |
+
workflow_dispatch: # allows manual triggering
|
| 5 |
+
push:
|
| 6 |
+
branches:
|
| 7 |
+
- master
|
| 8 |
+
paths: [
|
| 9 |
+
'.github/workflows/hip-quality-check.yml',
|
| 10 |
+
'**/*.cu',
|
| 11 |
+
'**/*.cuh',
|
| 12 |
+
'scripts/hip/gcn-cdna-vgpr-check.py'
|
| 13 |
+
]
|
| 14 |
+
|
| 15 |
+
pull_request:
|
| 16 |
+
types: [opened, synchronize, reopened]
|
| 17 |
+
paths: [
|
| 18 |
+
'.github/workflows/hip-quality-check.yml',
|
| 19 |
+
'**/*.cu',
|
| 20 |
+
'**/*.cuh',
|
| 21 |
+
'scripts/hip/gcn-cdna-vgpr-check.py'
|
| 22 |
+
]
|
| 23 |
+
|
| 24 |
+
concurrency:
|
| 25 |
+
group: ${{ github.workflow }}-${{ github.head_ref && github.ref || github.run_id }}
|
| 26 |
+
cancel-in-progress: true
|
| 27 |
+
|
| 28 |
+
env:
|
| 29 |
+
GGML_NLOOP: 3
|
| 30 |
+
GGML_N_THREADS: 1
|
| 31 |
+
LLAMA_LOG_COLORS: 1
|
| 32 |
+
LLAMA_LOG_PREFIX: 1
|
| 33 |
+
LLAMA_LOG_TIMESTAMPS: 1
|
| 34 |
+
|
| 35 |
+
jobs:
|
| 36 |
+
ubuntu-22-hip-quality-check:
|
| 37 |
+
runs-on: ubuntu-22.04
|
| 38 |
+
container: rocm/dev-ubuntu-22.04:7.2.1
|
| 39 |
+
steps:
|
| 40 |
+
- name: Clone
|
| 41 |
+
id: checkout
|
| 42 |
+
uses: actions/checkout@v6
|
| 43 |
+
|
| 44 |
+
- name: Dependencies
|
| 45 |
+
id: depends
|
| 46 |
+
run: |
|
| 47 |
+
sudo apt-get update
|
| 48 |
+
sudo apt-get install -y build-essential git cmake rocblas-dev hipblas-dev libssl-dev python3
|
| 49 |
+
|
| 50 |
+
- name: ccache
|
| 51 |
+
uses: ggml-org/ccache-action@v1.2.21
|
| 52 |
+
with:
|
| 53 |
+
key: ubuntu-22-hip-quality-check
|
| 54 |
+
evict-old-files: 1d
|
| 55 |
+
save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
|
| 56 |
+
|
| 57 |
+
- name: Build with Werror
|
| 58 |
+
id: cmake_build
|
| 59 |
+
run: |
|
| 60 |
+
cmake -B build -S . \
|
| 61 |
+
-DCMAKE_HIP_COMPILER="$(hipconfig -l)/clang" \
|
| 62 |
+
-DGPU_TARGETS=gfx942 \
|
| 63 |
+
-DGGML_HIP=ON \
|
| 64 |
+
-DGGML_HIP_EXPORT_METRICS=Off \
|
| 65 |
+
-DCMAKE_HIP_FLAGS="-Werror -Wno-tautological-compare" \
|
| 66 |
+
-DCMAKE_BUILD_TYPE=Release
|
| 67 |
+
cd build
|
| 68 |
+
make -j $(nproc)
|
| 69 |
+
|
| 70 |
+
- name: Check for major VGPR spills
|
| 71 |
+
id: vgpr_check
|
| 72 |
+
run: |
|
| 73 |
+
cmake -B build -S . \
|
| 74 |
+
-DCMAKE_HIP_COMPILER="$(hipconfig -l)/clang" \
|
| 75 |
+
-DGPU_TARGETS=gfx908 \
|
| 76 |
+
-DGGML_HIP=ON \
|
| 77 |
+
-DGGML_HIP_EXPORT_METRICS=On \
|
| 78 |
+
-DCMAKE_HIP_FLAGS="" \
|
| 79 |
+
-DCMAKE_BUILD_TYPE=Release
|
| 80 |
+
cd build
|
| 81 |
+
make -j $(nproc) 2>&1 | tee metrics.log | grep -v 'Rpass-analysis=kernel-resource-usage\|remark:\|^$'
|
| 82 |
+
python3 ../scripts/hip/gcn-cdna-vgpr-check.py metrics.log
|
.github/workflows/python-lint.yml
CHANGED
|
@@ -31,6 +31,6 @@ jobs:
|
|
| 31 |
with:
|
| 32 |
python-version: "3.11"
|
| 33 |
- name: flake8 Lint
|
| 34 |
-
uses: py-actions/flake8@v2
|
| 35 |
with:
|
| 36 |
plugins: "flake8-no-print"
|
|
|
|
| 31 |
with:
|
| 32 |
python-version: "3.11"
|
| 33 |
- name: flake8 Lint
|
| 34 |
+
uses: py-actions/flake8@84ec6726560b6d5bd68f2a5bed83d62b52bb50ba # v2
|
| 35 |
with:
|
| 36 |
plugins: "flake8-no-print"
|
.github/workflows/python-type-check.yml
CHANGED
|
@@ -4,15 +4,17 @@ on:
|
|
| 4 |
push:
|
| 5 |
paths:
|
| 6 |
- '.github/workflows/python-type-check.yml'
|
| 7 |
-
- '
|
| 8 |
- '**.py'
|
| 9 |
- '**/requirements*.txt'
|
|
|
|
| 10 |
pull_request:
|
| 11 |
paths:
|
| 12 |
- '.github/workflows/python-type-check.yml'
|
| 13 |
-
- '
|
| 14 |
- '**.py'
|
| 15 |
- '**/requirements*.txt'
|
|
|
|
| 16 |
|
| 17 |
concurrency:
|
| 18 |
group: ${{ github.workflow }}-${{ github.head_ref && github.ref || github.run_id }}
|
|
@@ -20,8 +22,8 @@ concurrency:
|
|
| 20 |
|
| 21 |
jobs:
|
| 22 |
python-type-check:
|
| 23 |
-
runs-on: ubuntu-
|
| 24 |
-
name:
|
| 25 |
steps:
|
| 26 |
- name: Check out source repository
|
| 27 |
uses: actions/checkout@v6
|
|
@@ -29,10 +31,13 @@ jobs:
|
|
| 29 |
uses: actions/setup-python@v6
|
| 30 |
with:
|
| 31 |
python-version: "3.11"
|
| 32 |
-
pip-install: -r requirements/requirements-all.txt
|
| 33 |
-
- name: Type-check with Pyright
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
push:
|
| 5 |
paths:
|
| 6 |
- '.github/workflows/python-type-check.yml'
|
| 7 |
+
- 'ty.toml'
|
| 8 |
- '**.py'
|
| 9 |
- '**/requirements*.txt'
|
| 10 |
+
# - 'pyrightconfig.json'
|
| 11 |
pull_request:
|
| 12 |
paths:
|
| 13 |
- '.github/workflows/python-type-check.yml'
|
| 14 |
+
- 'ty.toml'
|
| 15 |
- '**.py'
|
| 16 |
- '**/requirements*.txt'
|
| 17 |
+
# - 'pyrightconfig.json'
|
| 18 |
|
| 19 |
concurrency:
|
| 20 |
group: ${{ github.workflow }}-${{ github.head_ref && github.ref || github.run_id }}
|
|
|
|
| 22 |
|
| 23 |
jobs:
|
| 24 |
python-type-check:
|
| 25 |
+
runs-on: ubuntu-slim
|
| 26 |
+
name: python type-check
|
| 27 |
steps:
|
| 28 |
- name: Check out source repository
|
| 29 |
uses: actions/checkout@v6
|
|
|
|
| 31 |
uses: actions/setup-python@v6
|
| 32 |
with:
|
| 33 |
python-version: "3.11"
|
| 34 |
+
pip-install: -r requirements/requirements-all.txt ty==0.0.26
|
| 35 |
+
# - name: Type-check with Pyright
|
| 36 |
+
# uses: jakebailey/pyright-action@v2
|
| 37 |
+
# with:
|
| 38 |
+
# version: 1.1.382
|
| 39 |
+
# level: warning
|
| 40 |
+
# warnings: true
|
| 41 |
+
- name: Type-check with ty
|
| 42 |
+
run: |
|
| 43 |
+
ty check --output-format=github
|
.github/workflows/release.yml
CHANGED
|
@@ -36,8 +36,26 @@ env:
|
|
| 36 |
CMAKE_ARGS: "-DLLAMA_BUILD_EXAMPLES=OFF -DLLAMA_BUILD_TESTS=OFF -DLLAMA_BUILD_TOOLS=ON -DLLAMA_BUILD_SERVER=ON -DGGML_RPC=ON"
|
| 37 |
|
| 38 |
jobs:
|
| 39 |
-
macOS-
|
| 40 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
|
| 42 |
steps:
|
| 43 |
- name: Clone
|
|
@@ -49,7 +67,7 @@ jobs:
|
|
| 49 |
- name: ccache
|
| 50 |
uses: ggml-org/ccache-action@v1.2.21
|
| 51 |
with:
|
| 52 |
-
key: macOS-latest-
|
| 53 |
evict-old-files: 1d
|
| 54 |
|
| 55 |
- name: Build
|
|
@@ -57,13 +75,11 @@ jobs:
|
|
| 57 |
run: |
|
| 58 |
sysctl -a
|
| 59 |
cmake -B build \
|
|
|
|
| 60 |
-DCMAKE_INSTALL_RPATH='@loader_path' \
|
| 61 |
-DCMAKE_BUILD_WITH_INSTALL_RPATH=ON \
|
| 62 |
-DLLAMA_FATAL_WARNINGS=ON \
|
| 63 |
-DLLAMA_BUILD_BORINGSSL=ON \
|
| 64 |
-
-DGGML_METAL_USE_BF16=ON \
|
| 65 |
-
-DGGML_METAL_EMBED_LIBRARY=ON \
|
| 66 |
-
-DGGML_RPC=ON \
|
| 67 |
${{ env.CMAKE_ARGS }}
|
| 68 |
cmake --build build --config Release -j $(sysctl -n hw.logicalcpu)
|
| 69 |
|
|
@@ -75,16 +91,26 @@ jobs:
|
|
| 75 |
id: pack_artifacts
|
| 76 |
run: |
|
| 77 |
cp LICENSE ./build/bin/
|
| 78 |
-
tar -czvf llama-${{ steps.tag.outputs.name }}-bin-macos-
|
| 79 |
|
| 80 |
- name: Upload artifacts
|
| 81 |
uses: actions/upload-artifact@v6
|
| 82 |
with:
|
| 83 |
-
path: llama-${{ steps.tag.outputs.name }}-bin-macos-
|
| 84 |
-
name: llama-bin-macos-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 85 |
|
| 86 |
-
|
| 87 |
-
runs-on: macos-15-intel
|
| 88 |
|
| 89 |
steps:
|
| 90 |
- name: Clone
|
|
@@ -94,26 +120,37 @@ jobs:
|
|
| 94 |
fetch-depth: 0
|
| 95 |
|
| 96 |
- name: ccache
|
|
|
|
| 97 |
uses: ggml-org/ccache-action@v1.2.21
|
| 98 |
with:
|
| 99 |
-
key:
|
| 100 |
evict-old-files: 1d
|
| 101 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 102 |
- name: Build
|
| 103 |
id: cmake_build
|
| 104 |
run: |
|
| 105 |
-
sysctl -a
|
| 106 |
-
# Metal is disabled due to intermittent failures with Github runners not having a GPU:
|
| 107 |
-
# https://github.com/ggml-org/llama.cpp/actions/runs/8635935781/job/23674807267#step:5:2313
|
| 108 |
cmake -B build \
|
| 109 |
-
-DCMAKE_INSTALL_RPATH='
|
| 110 |
-DCMAKE_BUILD_WITH_INSTALL_RPATH=ON \
|
|
|
|
|
|
|
|
|
|
| 111 |
-DLLAMA_FATAL_WARNINGS=ON \
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
-DGGML_RPC=ON \
|
| 115 |
-
-DCMAKE_OSX_DEPLOYMENT_TARGET=13.3
|
| 116 |
-
cmake --build build --config Release -j $(sysctl -n hw.logicalcpu)
|
| 117 |
|
| 118 |
- name: Determine tag name
|
| 119 |
id: tag
|
|
@@ -123,25 +160,22 @@ jobs:
|
|
| 123 |
id: pack_artifacts
|
| 124 |
run: |
|
| 125 |
cp LICENSE ./build/bin/
|
| 126 |
-
tar -czvf llama-${{ steps.tag.outputs.name }}-bin-
|
| 127 |
|
| 128 |
- name: Upload artifacts
|
| 129 |
uses: actions/upload-artifact@v6
|
| 130 |
with:
|
| 131 |
-
path: llama-${{ steps.tag.outputs.name }}-bin-
|
| 132 |
-
name: llama-bin-
|
| 133 |
|
| 134 |
-
ubuntu-
|
| 135 |
strategy:
|
| 136 |
matrix:
|
| 137 |
include:
|
| 138 |
- build: 'x64'
|
| 139 |
os: ubuntu-22.04
|
| 140 |
-
- build: '
|
| 141 |
-
os: ubuntu-24.04-
|
| 142 |
-
# GGML_BACKEND_DL and GGML_CPU_ALL_VARIANTS are not currently supported on arm
|
| 143 |
-
# - build: 'arm64'
|
| 144 |
-
# os: ubuntu-22.04-arm
|
| 145 |
|
| 146 |
runs-on: ${{ matrix.os }}
|
| 147 |
|
|
@@ -153,17 +187,25 @@ jobs:
|
|
| 153 |
fetch-depth: 0
|
| 154 |
|
| 155 |
- name: ccache
|
| 156 |
-
if: ${{ matrix.build != 's390x' }}
|
| 157 |
uses: ggml-org/ccache-action@v1.2.21
|
| 158 |
with:
|
| 159 |
-
key: ubuntu-
|
| 160 |
evict-old-files: 1d
|
| 161 |
|
| 162 |
- name: Dependencies
|
| 163 |
id: depends
|
| 164 |
run: |
|
| 165 |
-
|
| 166 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 167 |
|
| 168 |
- name: Build
|
| 169 |
id: cmake_build
|
|
@@ -174,7 +216,7 @@ jobs:
|
|
| 174 |
-DGGML_BACKEND_DL=ON \
|
| 175 |
-DGGML_NATIVE=OFF \
|
| 176 |
-DGGML_CPU_ALL_VARIANTS=ON \
|
| 177 |
-
-
|
| 178 |
${{ env.CMAKE_ARGS }}
|
| 179 |
cmake --build build --config Release -j $(nproc)
|
| 180 |
|
|
@@ -186,16 +228,19 @@ jobs:
|
|
| 186 |
id: pack_artifacts
|
| 187 |
run: |
|
| 188 |
cp LICENSE ./build/bin/
|
| 189 |
-
tar -czvf llama-${{ steps.tag.outputs.name }}-bin-ubuntu-${{ matrix.build }}.tar.gz --transform "s,./,llama-${{ steps.tag.outputs.name }}/," -C ./build/bin .
|
| 190 |
|
| 191 |
- name: Upload artifacts
|
| 192 |
uses: actions/upload-artifact@v6
|
| 193 |
with:
|
| 194 |
-
path: llama-${{ steps.tag.outputs.name }}-bin-ubuntu-${{ matrix.build }}.tar.gz
|
| 195 |
-
name: llama-bin-ubuntu-${{ matrix.build }}.tar.gz
|
| 196 |
|
| 197 |
-
|
| 198 |
-
runs-on: ubuntu-
|
|
|
|
|
|
|
|
|
|
| 199 |
|
| 200 |
steps:
|
| 201 |
- name: Clone
|
|
@@ -207,27 +252,40 @@ jobs:
|
|
| 207 |
- name: ccache
|
| 208 |
uses: ggml-org/ccache-action@v1.2.21
|
| 209 |
with:
|
| 210 |
-
key:
|
| 211 |
evict-old-files: 1d
|
| 212 |
|
| 213 |
-
- name:
|
| 214 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 215 |
run: |
|
| 216 |
-
|
| 217 |
-
|
| 218 |
-
sudo apt-get update -y
|
| 219 |
-
sudo apt-get install -y build-essential mesa-vulkan-drivers vulkan-sdk libssl-dev
|
| 220 |
|
| 221 |
- name: Build
|
| 222 |
id: cmake_build
|
| 223 |
run: |
|
| 224 |
cmake -B build \
|
|
|
|
|
|
|
|
|
|
| 225 |
-DCMAKE_INSTALL_RPATH='$ORIGIN' \
|
| 226 |
-DCMAKE_BUILD_WITH_INSTALL_RPATH=ON \
|
| 227 |
-DGGML_BACKEND_DL=ON \
|
| 228 |
-DGGML_NATIVE=OFF \
|
| 229 |
-DGGML_CPU_ALL_VARIANTS=ON \
|
| 230 |
-
-
|
|
|
|
|
|
|
| 231 |
${{ env.CMAKE_ARGS }}
|
| 232 |
cmake --build build --config Release -j $(nproc)
|
| 233 |
|
|
@@ -239,13 +297,13 @@ jobs:
|
|
| 239 |
id: pack_artifacts
|
| 240 |
run: |
|
| 241 |
cp LICENSE ./build/bin/
|
| 242 |
-
tar -czvf llama-${{ steps.tag.outputs.name }}-bin-
|
| 243 |
|
| 244 |
- name: Upload artifacts
|
| 245 |
uses: actions/upload-artifact@v6
|
| 246 |
with:
|
| 247 |
-
path: llama-${{ steps.tag.outputs.name }}-bin-
|
| 248 |
-
name: llama-bin-
|
| 249 |
|
| 250 |
ubuntu-24-openvino:
|
| 251 |
runs-on: ubuntu-24.04
|
|
@@ -618,8 +676,8 @@ jobs:
|
|
| 618 |
strategy:
|
| 619 |
matrix:
|
| 620 |
include:
|
| 621 |
-
- ROCM_VERSION: "7.2"
|
| 622 |
-
gpu_targets: "gfx908;gfx90a;gfx942;gfx1030;gfx1100;gfx1101;gfx1151;gfx1150;gfx1200;gfx1201"
|
| 623 |
build: 'x64'
|
| 624 |
|
| 625 |
steps:
|
|
@@ -629,6 +687,11 @@ jobs:
|
|
| 629 |
with:
|
| 630 |
fetch-depth: 0
|
| 631 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 632 |
- name: ccache
|
| 633 |
uses: ggml-org/ccache-action@v1.2.21
|
| 634 |
with:
|
|
@@ -641,7 +704,7 @@ jobs:
|
|
| 641 |
sudo apt install -y build-essential git cmake wget
|
| 642 |
|
| 643 |
- name: Setup Legacy ROCm
|
| 644 |
-
if: matrix.ROCM_VERSION == '7.2'
|
| 645 |
id: legacy_env
|
| 646 |
run: |
|
| 647 |
sudo mkdir --parents --mode=0755 /etc/apt/keyrings
|
|
@@ -662,7 +725,7 @@ jobs:
|
|
| 662 |
sudo apt-get install -y libssl-dev rocm-hip-sdk
|
| 663 |
|
| 664 |
- name: Setup TheRock
|
| 665 |
-
if: matrix.ROCM_VERSION != '7.2'
|
| 666 |
id: therock_env
|
| 667 |
run: |
|
| 668 |
wget https://repo.amd.com/rocm/tarball/therock-dist-linux-gfx1151-${{ matrix.ROCM_VERSION }}.tar.gz
|
|
@@ -678,7 +741,6 @@ jobs:
|
|
| 678 |
run: |
|
| 679 |
cmake -B build -S . \
|
| 680 |
-DCMAKE_HIP_COMPILER="$(hipconfig -l)/clang" \
|
| 681 |
-
-DCMAKE_HIP_FLAGS="-mllvm --amdgpu-unroll-threshold-local=600" \
|
| 682 |
-DCMAKE_BUILD_TYPE=Release \
|
| 683 |
-DGGML_BACKEND_DL=ON \
|
| 684 |
-DGGML_NATIVE=OFF \
|
|
@@ -696,17 +758,20 @@ jobs:
|
|
| 696 |
id: tag
|
| 697 |
uses: ./.github/actions/get-tag-name
|
| 698 |
|
|
|
|
|
|
|
|
|
|
| 699 |
- name: Pack artifacts
|
| 700 |
id: pack_artifacts
|
| 701 |
run: |
|
| 702 |
cp LICENSE ./build/bin/
|
| 703 |
-
tar -czvf llama-${{ steps.tag.outputs.name }}-bin-ubuntu-rocm-${{
|
| 704 |
|
| 705 |
- name: Upload artifacts
|
| 706 |
uses: actions/upload-artifact@v6
|
| 707 |
with:
|
| 708 |
-
path: llama-${{ steps.tag.outputs.name }}-bin-ubuntu-rocm-${{
|
| 709 |
-
name: llama-bin-ubuntu-rocm-${{
|
| 710 |
|
| 711 |
windows-hip:
|
| 712 |
runs-on: windows-2022
|
|
@@ -728,7 +793,7 @@ jobs:
|
|
| 728 |
- name: Grab rocWMMA package
|
| 729 |
id: grab_rocwmma
|
| 730 |
run: |
|
| 731 |
-
curl -o rocwmma.deb "https://repo.radeon.com/rocm/apt/7.2/pool/main/r/rocwmma-dev/rocwmma-dev_2.2.0.
|
| 732 |
7z x rocwmma.deb
|
| 733 |
7z x data.tar
|
| 734 |
|
|
@@ -785,7 +850,7 @@ jobs:
|
|
| 785 |
cmake -G "Unix Makefiles" -B build -S . `
|
| 786 |
-DCMAKE_C_COMPILER="${env:HIP_PATH}\bin\clang.exe" `
|
| 787 |
-DCMAKE_CXX_COMPILER="${env:HIP_PATH}\bin\clang++.exe" `
|
| 788 |
-
-DCMAKE_CXX_FLAGS="-I$($PWD.Path.Replace('\', '/'))/opt/rocm-7.2.
|
| 789 |
-DCMAKE_BUILD_TYPE=Release `
|
| 790 |
-DGGML_BACKEND_DL=ON `
|
| 791 |
-DGGML_NATIVE=OFF `
|
|
@@ -907,7 +972,7 @@ jobs:
|
|
| 907 |
- name: Set container image
|
| 908 |
id: cann-image
|
| 909 |
run: |
|
| 910 |
-
image="ascendai/cann:${{ matrix.chip_type == '910b' && '8.
|
| 911 |
echo "image=${image}" >> "${GITHUB_OUTPUT}"
|
| 912 |
|
| 913 |
- name: Pull container image
|
|
@@ -977,11 +1042,11 @@ jobs:
|
|
| 977 |
- windows-sycl
|
| 978 |
- windows-hip
|
| 979 |
- ubuntu-22-rocm
|
| 980 |
-
- ubuntu-
|
| 981 |
-
- ubuntu-
|
| 982 |
- ubuntu-24-openvino
|
| 983 |
-
-
|
| 984 |
-
- macOS-
|
| 985 |
- ios-xcode-build
|
| 986 |
- openEuler-cann
|
| 987 |
|
|
@@ -1056,16 +1121,22 @@ jobs:
|
|
| 1056 |
|
| 1057 |
**macOS/iOS:**
|
| 1058 |
- [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-macos-arm64.tar.gz)
|
|
|
|
| 1059 |
- [macOS Intel (x64)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-macos-x64.tar.gz)
|
| 1060 |
- [iOS XCFramework](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-xcframework.zip)
|
| 1061 |
|
| 1062 |
**Linux:**
|
| 1063 |
- [Ubuntu x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-x64.tar.gz)
|
|
|
|
|
|
|
| 1064 |
- [Ubuntu x64 (Vulkan)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-vulkan-x64.tar.gz)
|
|
|
|
| 1065 |
- [Ubuntu x64 (ROCm 7.2)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-rocm-7.2-x64.tar.gz)
|
| 1066 |
-
- [Ubuntu s390x (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-s390x.tar.gz)
|
| 1067 |
- [Ubuntu x64 (OpenVINO)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-openvino-${{ needs.ubuntu-24-openvino.outputs.openvino_version }}-x64.tar.gz)
|
| 1068 |
|
|
|
|
|
|
|
|
|
|
| 1069 |
**Windows:**
|
| 1070 |
- [Windows x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-win-cpu-x64.zip)
|
| 1071 |
- [Windows arm64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-win-cpu-arm64.zip)
|
|
|
|
| 36 |
CMAKE_ARGS: "-DLLAMA_BUILD_EXAMPLES=OFF -DLLAMA_BUILD_TESTS=OFF -DLLAMA_BUILD_TOOLS=ON -DLLAMA_BUILD_SERVER=ON -DGGML_RPC=ON"
|
| 37 |
|
| 38 |
jobs:
|
| 39 |
+
macOS-cpu:
|
| 40 |
+
strategy:
|
| 41 |
+
matrix:
|
| 42 |
+
include:
|
| 43 |
+
- build: 'arm64'
|
| 44 |
+
arch: 'arm64'
|
| 45 |
+
os: macos-14
|
| 46 |
+
defines: "-DGGML_METAL_USE_BF16=ON -DGGML_METAL_EMBED_LIBRARY=ON"
|
| 47 |
+
- build: 'arm64-kleidiai'
|
| 48 |
+
arch: 'arm64'
|
| 49 |
+
os: macos-14
|
| 50 |
+
defines: "-DGGML_METAL_USE_BF16=ON -DGGML_METAL_EMBED_LIBRARY=ON -DGGML_CPU_KLEIDIAI=ON"
|
| 51 |
+
- build: 'x64'
|
| 52 |
+
arch: 'x64'
|
| 53 |
+
os: macos-15-intel
|
| 54 |
+
# Metal is disabled on x64 due to intermittent failures with Github runners not having a GPU:
|
| 55 |
+
# https://github.com/ggml-org/llama.cpp/actions/runs/8635935781/job/23674807267#step:5:2313
|
| 56 |
+
defines: "-DGGML_METAL=OFF -DCMAKE_OSX_DEPLOYMENT_TARGET=13.3"
|
| 57 |
+
|
| 58 |
+
runs-on: ${{ matrix.os }}
|
| 59 |
|
| 60 |
steps:
|
| 61 |
- name: Clone
|
|
|
|
| 67 |
- name: ccache
|
| 68 |
uses: ggml-org/ccache-action@v1.2.21
|
| 69 |
with:
|
| 70 |
+
key: macOS-latest-${{ matrix.arch }}
|
| 71 |
evict-old-files: 1d
|
| 72 |
|
| 73 |
- name: Build
|
|
|
|
| 75 |
run: |
|
| 76 |
sysctl -a
|
| 77 |
cmake -B build \
|
| 78 |
+
${{ matrix.defines }} \
|
| 79 |
-DCMAKE_INSTALL_RPATH='@loader_path' \
|
| 80 |
-DCMAKE_BUILD_WITH_INSTALL_RPATH=ON \
|
| 81 |
-DLLAMA_FATAL_WARNINGS=ON \
|
| 82 |
-DLLAMA_BUILD_BORINGSSL=ON \
|
|
|
|
|
|
|
|
|
|
| 83 |
${{ env.CMAKE_ARGS }}
|
| 84 |
cmake --build build --config Release -j $(sysctl -n hw.logicalcpu)
|
| 85 |
|
|
|
|
| 91 |
id: pack_artifacts
|
| 92 |
run: |
|
| 93 |
cp LICENSE ./build/bin/
|
| 94 |
+
tar -czvf llama-${{ steps.tag.outputs.name }}-bin-macos-${{ matrix.build }}.tar.gz -s ",./,llama-${{ steps.tag.outputs.name }}/," -C ./build/bin .
|
| 95 |
|
| 96 |
- name: Upload artifacts
|
| 97 |
uses: actions/upload-artifact@v6
|
| 98 |
with:
|
| 99 |
+
path: llama-${{ steps.tag.outputs.name }}-bin-macos-${{ matrix.build }}.tar.gz
|
| 100 |
+
name: llama-bin-macos-${{ matrix.build }}.tar.gz
|
| 101 |
+
|
| 102 |
+
ubuntu-cpu:
|
| 103 |
+
strategy:
|
| 104 |
+
matrix:
|
| 105 |
+
include:
|
| 106 |
+
- build: 'x64'
|
| 107 |
+
os: ubuntu-22.04
|
| 108 |
+
- build: 'arm64'
|
| 109 |
+
os: ubuntu-24.04-arm
|
| 110 |
+
- build: 's390x'
|
| 111 |
+
os: ubuntu-24.04-s390x
|
| 112 |
|
| 113 |
+
runs-on: ${{ matrix.os }}
|
|
|
|
| 114 |
|
| 115 |
steps:
|
| 116 |
- name: Clone
|
|
|
|
| 120 |
fetch-depth: 0
|
| 121 |
|
| 122 |
- name: ccache
|
| 123 |
+
if: ${{ matrix.build != 's390x' }}
|
| 124 |
uses: ggml-org/ccache-action@v1.2.21
|
| 125 |
with:
|
| 126 |
+
key: ubuntu-cpu-${{ matrix.build }}
|
| 127 |
evict-old-files: 1d
|
| 128 |
|
| 129 |
+
- name: Dependencies
|
| 130 |
+
id: depends
|
| 131 |
+
run: |
|
| 132 |
+
sudo apt-get update
|
| 133 |
+
sudo apt-get install build-essential libssl-dev
|
| 134 |
+
|
| 135 |
+
- name: Toolchain workaround (GCC 14)
|
| 136 |
+
if: ${{ contains(matrix.os, 'ubuntu-24.04') }}
|
| 137 |
+
run: |
|
| 138 |
+
sudo apt-get install -y gcc-14 g++-14
|
| 139 |
+
echo "CC=gcc-14" >> "$GITHUB_ENV"
|
| 140 |
+
echo "CXX=g++-14" >> "$GITHUB_ENV"
|
| 141 |
+
|
| 142 |
- name: Build
|
| 143 |
id: cmake_build
|
| 144 |
run: |
|
|
|
|
|
|
|
|
|
|
| 145 |
cmake -B build \
|
| 146 |
+
-DCMAKE_INSTALL_RPATH='$ORIGIN' \
|
| 147 |
-DCMAKE_BUILD_WITH_INSTALL_RPATH=ON \
|
| 148 |
+
-DGGML_BACKEND_DL=ON \
|
| 149 |
+
-DGGML_NATIVE=OFF \
|
| 150 |
+
-DGGML_CPU_ALL_VARIANTS=ON \
|
| 151 |
-DLLAMA_FATAL_WARNINGS=ON \
|
| 152 |
+
${{ env.CMAKE_ARGS }}
|
| 153 |
+
cmake --build build --config Release -j $(nproc)
|
|
|
|
|
|
|
|
|
|
| 154 |
|
| 155 |
- name: Determine tag name
|
| 156 |
id: tag
|
|
|
|
| 160 |
id: pack_artifacts
|
| 161 |
run: |
|
| 162 |
cp LICENSE ./build/bin/
|
| 163 |
+
tar -czvf llama-${{ steps.tag.outputs.name }}-bin-ubuntu-${{ matrix.build }}.tar.gz --transform "s,./,llama-${{ steps.tag.outputs.name }}/," -C ./build/bin .
|
| 164 |
|
| 165 |
- name: Upload artifacts
|
| 166 |
uses: actions/upload-artifact@v6
|
| 167 |
with:
|
| 168 |
+
path: llama-${{ steps.tag.outputs.name }}-bin-ubuntu-${{ matrix.build }}.tar.gz
|
| 169 |
+
name: llama-bin-ubuntu-${{ matrix.build }}.tar.gz
|
| 170 |
|
| 171 |
+
ubuntu-vulkan:
|
| 172 |
strategy:
|
| 173 |
matrix:
|
| 174 |
include:
|
| 175 |
- build: 'x64'
|
| 176 |
os: ubuntu-22.04
|
| 177 |
+
- build: 'arm64'
|
| 178 |
+
os: ubuntu-24.04-arm
|
|
|
|
|
|
|
|
|
|
| 179 |
|
| 180 |
runs-on: ${{ matrix.os }}
|
| 181 |
|
|
|
|
| 187 |
fetch-depth: 0
|
| 188 |
|
| 189 |
- name: ccache
|
|
|
|
| 190 |
uses: ggml-org/ccache-action@v1.2.21
|
| 191 |
with:
|
| 192 |
+
key: ubuntu-vulkan-${{ matrix.build }}
|
| 193 |
evict-old-files: 1d
|
| 194 |
|
| 195 |
- name: Dependencies
|
| 196 |
id: depends
|
| 197 |
run: |
|
| 198 |
+
if [[ "${{ matrix.os }}" =~ "ubuntu-22.04" ]]; then
|
| 199 |
+
wget -qO - https://packages.lunarg.com/lunarg-signing-key-pub.asc | sudo apt-key add -
|
| 200 |
+
sudo wget -qO /etc/apt/sources.list.d/lunarg-vulkan-jammy.list https://packages.lunarg.com/vulkan/lunarg-vulkan-jammy.list
|
| 201 |
+
sudo apt-get update -y
|
| 202 |
+
sudo apt-get install -y build-essential mesa-vulkan-drivers vulkan-sdk libssl-dev
|
| 203 |
+
else
|
| 204 |
+
sudo apt-get update -y
|
| 205 |
+
sudo apt-get install -y gcc-14 g++-14 build-essential glslc libvulkan-dev spirv-headers libssl-dev ninja-build
|
| 206 |
+
echo "CC=gcc-14" >> "$GITHUB_ENV"
|
| 207 |
+
echo "CXX=g++-14" >> "$GITHUB_ENV"
|
| 208 |
+
fi
|
| 209 |
|
| 210 |
- name: Build
|
| 211 |
id: cmake_build
|
|
|
|
| 216 |
-DGGML_BACKEND_DL=ON \
|
| 217 |
-DGGML_NATIVE=OFF \
|
| 218 |
-DGGML_CPU_ALL_VARIANTS=ON \
|
| 219 |
+
-DGGML_VULKAN=ON \
|
| 220 |
${{ env.CMAKE_ARGS }}
|
| 221 |
cmake --build build --config Release -j $(nproc)
|
| 222 |
|
|
|
|
| 228 |
id: pack_artifacts
|
| 229 |
run: |
|
| 230 |
cp LICENSE ./build/bin/
|
| 231 |
+
tar -czvf llama-${{ steps.tag.outputs.name }}-bin-ubuntu-vulkan-${{ matrix.build }}.tar.gz --transform "s,./,llama-${{ steps.tag.outputs.name }}/," -C ./build/bin .
|
| 232 |
|
| 233 |
- name: Upload artifacts
|
| 234 |
uses: actions/upload-artifact@v6
|
| 235 |
with:
|
| 236 |
+
path: llama-${{ steps.tag.outputs.name }}-bin-ubuntu-vulkan-${{ matrix.build }}.tar.gz
|
| 237 |
+
name: llama-bin-ubuntu-vulkan-${{ matrix.build }}.tar.gz
|
| 238 |
|
| 239 |
+
android-arm64:
|
| 240 |
+
runs-on: ubuntu-latest
|
| 241 |
+
|
| 242 |
+
env:
|
| 243 |
+
NDK_VERSION: "29.0.14206865"
|
| 244 |
|
| 245 |
steps:
|
| 246 |
- name: Clone
|
|
|
|
| 252 |
- name: ccache
|
| 253 |
uses: ggml-org/ccache-action@v1.2.21
|
| 254 |
with:
|
| 255 |
+
key: android-arm64
|
| 256 |
evict-old-files: 1d
|
| 257 |
|
| 258 |
+
- name: Set up JDK
|
| 259 |
+
uses: actions/setup-java@v5
|
| 260 |
+
with:
|
| 261 |
+
java-version: 17
|
| 262 |
+
distribution: temurin
|
| 263 |
+
|
| 264 |
+
- name: Setup Android SDK
|
| 265 |
+
uses: android-actions/setup-android@40fd30fb8d7440372e1316f5d1809ec01dcd3699 # v4.0.1
|
| 266 |
+
with:
|
| 267 |
+
log-accepted-android-sdk-licenses: false
|
| 268 |
+
|
| 269 |
+
- name: Install NDK
|
| 270 |
run: |
|
| 271 |
+
sdkmanager "ndk;${{ env.NDK_VERSION }}"
|
| 272 |
+
echo "ANDROID_NDK=${ANDROID_SDK_ROOT}/ndk/${{ env.NDK_VERSION }}" >> $GITHUB_ENV
|
|
|
|
|
|
|
| 273 |
|
| 274 |
- name: Build
|
| 275 |
id: cmake_build
|
| 276 |
run: |
|
| 277 |
cmake -B build \
|
| 278 |
+
-DCMAKE_TOOLCHAIN_FILE=${ANDROID_NDK}/build/cmake/android.toolchain.cmake \
|
| 279 |
+
-DANDROID_ABI=arm64-v8a \
|
| 280 |
+
-DANDROID_PLATFORM=android-28 \
|
| 281 |
-DCMAKE_INSTALL_RPATH='$ORIGIN' \
|
| 282 |
-DCMAKE_BUILD_WITH_INSTALL_RPATH=ON \
|
| 283 |
-DGGML_BACKEND_DL=ON \
|
| 284 |
-DGGML_NATIVE=OFF \
|
| 285 |
-DGGML_CPU_ALL_VARIANTS=ON \
|
| 286 |
+
-DLLAMA_FATAL_WARNINGS=ON \
|
| 287 |
+
-DGGML_OPENMP=OFF \
|
| 288 |
+
-DLLAMA_BUILD_BORINGSSL=ON \
|
| 289 |
${{ env.CMAKE_ARGS }}
|
| 290 |
cmake --build build --config Release -j $(nproc)
|
| 291 |
|
|
|
|
| 297 |
id: pack_artifacts
|
| 298 |
run: |
|
| 299 |
cp LICENSE ./build/bin/
|
| 300 |
+
tar -czvf llama-${{ steps.tag.outputs.name }}-bin-android-arm64.tar.gz --transform "s,./,llama-${{ steps.tag.outputs.name }}/," -C ./build/bin .
|
| 301 |
|
| 302 |
- name: Upload artifacts
|
| 303 |
uses: actions/upload-artifact@v6
|
| 304 |
with:
|
| 305 |
+
path: llama-${{ steps.tag.outputs.name }}-bin-android-arm64.tar.gz
|
| 306 |
+
name: llama-bin-android-arm64.tar.gz
|
| 307 |
|
| 308 |
ubuntu-24-openvino:
|
| 309 |
runs-on: ubuntu-24.04
|
|
|
|
| 676 |
strategy:
|
| 677 |
matrix:
|
| 678 |
include:
|
| 679 |
+
- ROCM_VERSION: "7.2.1"
|
| 680 |
+
gpu_targets: "gfx908;gfx90a;gfx942;gfx1030;gfx1100;gfx1101;gfx1102;gfx1151;gfx1150;gfx1200;gfx1201"
|
| 681 |
build: 'x64'
|
| 682 |
|
| 683 |
steps:
|
|
|
|
| 687 |
with:
|
| 688 |
fetch-depth: 0
|
| 689 |
|
| 690 |
+
- name: Free up disk space
|
| 691 |
+
uses: ggml-org/free-disk-space@v1.3.1
|
| 692 |
+
with:
|
| 693 |
+
tool-cache: true
|
| 694 |
+
|
| 695 |
- name: ccache
|
| 696 |
uses: ggml-org/ccache-action@v1.2.21
|
| 697 |
with:
|
|
|
|
| 704 |
sudo apt install -y build-essential git cmake wget
|
| 705 |
|
| 706 |
- name: Setup Legacy ROCm
|
| 707 |
+
if: matrix.ROCM_VERSION == '7.2.1'
|
| 708 |
id: legacy_env
|
| 709 |
run: |
|
| 710 |
sudo mkdir --parents --mode=0755 /etc/apt/keyrings
|
|
|
|
| 725 |
sudo apt-get install -y libssl-dev rocm-hip-sdk
|
| 726 |
|
| 727 |
- name: Setup TheRock
|
| 728 |
+
if: matrix.ROCM_VERSION != '7.2.1'
|
| 729 |
id: therock_env
|
| 730 |
run: |
|
| 731 |
wget https://repo.amd.com/rocm/tarball/therock-dist-linux-gfx1151-${{ matrix.ROCM_VERSION }}.tar.gz
|
|
|
|
| 741 |
run: |
|
| 742 |
cmake -B build -S . \
|
| 743 |
-DCMAKE_HIP_COMPILER="$(hipconfig -l)/clang" \
|
|
|
|
| 744 |
-DCMAKE_BUILD_TYPE=Release \
|
| 745 |
-DGGML_BACKEND_DL=ON \
|
| 746 |
-DGGML_NATIVE=OFF \
|
|
|
|
| 758 |
id: tag
|
| 759 |
uses: ./.github/actions/get-tag-name
|
| 760 |
|
| 761 |
+
- name: Get ROCm short version
|
| 762 |
+
run: echo "ROCM_VERSION_SHORT=$(echo '${{ matrix.ROCM_VERSION }}' | cut -d '.' -f 1,2)" >> $GITHUB_ENV
|
| 763 |
+
|
| 764 |
- name: Pack artifacts
|
| 765 |
id: pack_artifacts
|
| 766 |
run: |
|
| 767 |
cp LICENSE ./build/bin/
|
| 768 |
+
tar -czvf llama-${{ steps.tag.outputs.name }}-bin-ubuntu-rocm-${{ env.ROCM_VERSION_SHORT }}-${{ matrix.build }}.tar.gz --transform "s,./,llama-${{ steps.tag.outputs.name }}/," -C ./build/bin .
|
| 769 |
|
| 770 |
- name: Upload artifacts
|
| 771 |
uses: actions/upload-artifact@v6
|
| 772 |
with:
|
| 773 |
+
path: llama-${{ steps.tag.outputs.name }}-bin-ubuntu-rocm-${{ env.ROCM_VERSION_SHORT }}-${{ matrix.build }}.tar.gz
|
| 774 |
+
name: llama-bin-ubuntu-rocm-${{ env.ROCM_VERSION_SHORT }}-${{ matrix.build }}.tar.gz
|
| 775 |
|
| 776 |
windows-hip:
|
| 777 |
runs-on: windows-2022
|
|
|
|
| 793 |
- name: Grab rocWMMA package
|
| 794 |
id: grab_rocwmma
|
| 795 |
run: |
|
| 796 |
+
curl -o rocwmma.deb "https://repo.radeon.com/rocm/apt/7.2.1/pool/main/r/rocwmma-dev/rocwmma-dev_2.2.0.70201-81~24.04_amd64.deb"
|
| 797 |
7z x rocwmma.deb
|
| 798 |
7z x data.tar
|
| 799 |
|
|
|
|
| 850 |
cmake -G "Unix Makefiles" -B build -S . `
|
| 851 |
-DCMAKE_C_COMPILER="${env:HIP_PATH}\bin\clang.exe" `
|
| 852 |
-DCMAKE_CXX_COMPILER="${env:HIP_PATH}\bin\clang++.exe" `
|
| 853 |
+
-DCMAKE_CXX_FLAGS="-I$($PWD.Path.Replace('\', '/'))/opt/rocm-7.2.1/include/ -Wno-ignored-attributes -Wno-nested-anon-types" `
|
| 854 |
-DCMAKE_BUILD_TYPE=Release `
|
| 855 |
-DGGML_BACKEND_DL=ON `
|
| 856 |
-DGGML_NATIVE=OFF `
|
|
|
|
| 972 |
- name: Set container image
|
| 973 |
id: cann-image
|
| 974 |
run: |
|
| 975 |
+
image="ascendai/cann:${{ matrix.chip_type == '910b' && '8.5.0-910b-openeuler24.03-py3.11' || '8.5.0-310p-openeuler24.03-py3.11' }}"
|
| 976 |
echo "image=${image}" >> "${GITHUB_OUTPUT}"
|
| 977 |
|
| 978 |
- name: Pull container image
|
|
|
|
| 1042 |
- windows-sycl
|
| 1043 |
- windows-hip
|
| 1044 |
- ubuntu-22-rocm
|
| 1045 |
+
- ubuntu-cpu
|
| 1046 |
+
- ubuntu-vulkan
|
| 1047 |
- ubuntu-24-openvino
|
| 1048 |
+
- android-arm64
|
| 1049 |
+
- macOS-cpu
|
| 1050 |
- ios-xcode-build
|
| 1051 |
- openEuler-cann
|
| 1052 |
|
|
|
|
| 1121 |
|
| 1122 |
**macOS/iOS:**
|
| 1123 |
- [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-macos-arm64.tar.gz)
|
| 1124 |
+
- [macOS Apple Silicon (arm64, KleidiAI enabled)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-macos-arm64-kleidiai.tar.gz)
|
| 1125 |
- [macOS Intel (x64)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-macos-x64.tar.gz)
|
| 1126 |
- [iOS XCFramework](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-xcframework.zip)
|
| 1127 |
|
| 1128 |
**Linux:**
|
| 1129 |
- [Ubuntu x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-x64.tar.gz)
|
| 1130 |
+
- [Ubuntu arm64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-arm64.tar.gz)
|
| 1131 |
+
- [Ubuntu s390x (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-s390x.tar.gz)
|
| 1132 |
- [Ubuntu x64 (Vulkan)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-vulkan-x64.tar.gz)
|
| 1133 |
+
- [Ubuntu arm64 (Vulkan)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-vulkan-arm64.tar.gz)
|
| 1134 |
- [Ubuntu x64 (ROCm 7.2)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-rocm-7.2-x64.tar.gz)
|
|
|
|
| 1135 |
- [Ubuntu x64 (OpenVINO)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-openvino-${{ needs.ubuntu-24-openvino.outputs.openvino_version }}-x64.tar.gz)
|
| 1136 |
|
| 1137 |
+
**Android:**
|
| 1138 |
+
- [Android arm64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-android-arm64.tar.gz)
|
| 1139 |
+
|
| 1140 |
**Windows:**
|
| 1141 |
- [Windows x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-win-cpu-x64.zip)
|
| 1142 |
- [Windows arm64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-win-cpu-arm64.zip)
|
.github/workflows/server-self-hosted.yml
CHANGED
|
@@ -84,41 +84,42 @@ jobs:
|
|
| 84 |
export ${{ matrix.extra_args }}
|
| 85 |
pytest -v -x -m "not slow"
|
| 86 |
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
cmake -
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
|
|
|
|
|
| 84 |
export ${{ matrix.extra_args }}
|
| 85 |
pytest -v -x -m "not slow"
|
| 86 |
|
| 87 |
+
# TODO: provision CUDA runner
|
| 88 |
+
# server-cuda:
|
| 89 |
+
# runs-on: [self-hosted, llama-server, Linux, NVIDIA]
|
| 90 |
+
#
|
| 91 |
+
# name: server-cuda (${{ matrix.wf_name }})
|
| 92 |
+
# strategy:
|
| 93 |
+
# matrix:
|
| 94 |
+
# build_type: [Release]
|
| 95 |
+
# wf_name: ["GPUx1"]
|
| 96 |
+
# include:
|
| 97 |
+
# - build_type: Release
|
| 98 |
+
# extra_args: "LLAMA_ARG_BACKEND_SAMPLING=1"
|
| 99 |
+
# wf_name: "GPUx1, backend-sampling"
|
| 100 |
+
# fail-fast: false
|
| 101 |
+
#
|
| 102 |
+
# steps:
|
| 103 |
+
# - name: Clone
|
| 104 |
+
# id: checkout
|
| 105 |
+
# uses: actions/checkout@v6
|
| 106 |
+
# with:
|
| 107 |
+
# fetch-depth: 0
|
| 108 |
+
# ref: ${{ github.event.inputs.sha || github.event.pull_request.head.sha || github.sha || github.head_ref || github.ref_name }}
|
| 109 |
+
#
|
| 110 |
+
# - name: Build
|
| 111 |
+
# id: cmake_build
|
| 112 |
+
# run: |
|
| 113 |
+
# cmake -B build -DGGML_SCHED_NO_REALLOC=ON
|
| 114 |
+
# cmake --build build --config ${{ matrix.build_type }} -j $(sysctl -n hw.logicalcpu) --target llama-server
|
| 115 |
+
#
|
| 116 |
+
# - name: Tests
|
| 117 |
+
# id: server_integration_tests
|
| 118 |
+
# if: ${{ (!matrix.disabled_on_pr || !github.event.pull_request) }}
|
| 119 |
+
# run: |
|
| 120 |
+
# cd tools/server/tests
|
| 121 |
+
# python3 -m venv venv
|
| 122 |
+
# source venv/bin/activate
|
| 123 |
+
# pip install -r requirements.txt
|
| 124 |
+
# export ${{ matrix.extra_args }}
|
| 125 |
+
# pytest -v -x -m "not slow"
|
.gitignore
CHANGED
|
@@ -1,5 +1,7 @@
|
|
| 1 |
# Extensions
|
| 2 |
-
|
|
|
|
|
|
|
| 3 |
*.a
|
| 4 |
*.bat
|
| 5 |
*.bin
|
|
@@ -95,6 +97,8 @@
|
|
| 95 |
# Server Web UI temporary files
|
| 96 |
/tools/server/webui/node_modules
|
| 97 |
/tools/server/webui/dist
|
|
|
|
|
|
|
| 98 |
|
| 99 |
# Python
|
| 100 |
|
|
|
|
| 1 |
# Extensions
|
| 2 |
+
build/
|
| 3 |
+
build_64/
|
| 4 |
+
models/
|
| 5 |
*.a
|
| 6 |
*.bat
|
| 7 |
*.bin
|
|
|
|
| 97 |
# Server Web UI temporary files
|
| 98 |
/tools/server/webui/node_modules
|
| 99 |
/tools/server/webui/dist
|
| 100 |
+
# we no longer use gz for index.html
|
| 101 |
+
/tools/server/public/index.html.gz
|
| 102 |
|
| 103 |
# Python
|
| 104 |
|
AGENTS.md
CHANGED
|
@@ -1,81 +1,3 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
> This project does **not** accept pull requests that are fully or predominantly AI-generated. AI tools may be utilized solely in an assistive capacity.
|
| 5 |
-
>
|
| 6 |
-
> Read more: [CONTRIBUTING.md](CONTRIBUTING.md)
|
| 7 |
-
|
| 8 |
-
AI assistance is permissible only when the majority of the code is authored by a human contributor, with AI employed exclusively for corrections or to expand on verbose modifications that the contributor has already conceptualized (see examples below)
|
| 9 |
-
|
| 10 |
-
---
|
| 11 |
-
|
| 12 |
-
## Guidelines for Contributors Using AI
|
| 13 |
-
|
| 14 |
-
These use cases are **permitted** when making a contribution with the help of AI:
|
| 15 |
-
|
| 16 |
-
- Using it to ask about the structure of the codebase
|
| 17 |
-
- Learning about specific techniques used in the project
|
| 18 |
-
- Pointing out documents, links, and parts of the code that are worth your time
|
| 19 |
-
- Reviewing human-written code and providing suggestions for improvements
|
| 20 |
-
- Expanding on verbose modifications that the contributor has already conceptualized. For example:
|
| 21 |
-
- Generating repeated lines with minor variations (this should only be used for short code snippets where deduplication would add more complexity, compared to having almost the same code in multiple places)
|
| 22 |
-
- Formatting code for consistency and readability
|
| 23 |
-
- Completing code segments based on established patterns
|
| 24 |
-
- Drafting documentation for project components with which the contributor is already familiar
|
| 25 |
-
|
| 26 |
-
AI-generated code that has undergone extensive human editing may be accepted, provided you (1) fully understand the AI's initial output, (2) can debug any issues independently (with or without further AI assistance), and (3) are prepared to discuss it directly with human reviewers.
|
| 27 |
-
|
| 28 |
-
**All AI usage requires explicit disclosure**, except in these cases:
|
| 29 |
-
|
| 30 |
-
- Trivial tab autocompletions, but only for completions that you have already conceptualized in your mind.
|
| 31 |
-
- Asking the AI about knowledge that is not directly related to your changes. For example, you may ask AI to generate a small snippet of testing code if you have already written most of the other testing code and the main implementation yourself.
|
| 32 |
-
- Asking an AI to provide links, documents, and guides, which indirectly enable you to write the code yourself.
|
| 33 |
-
|
| 34 |
-
---
|
| 35 |
-
|
| 36 |
-
## Guidelines for AI Agents
|
| 37 |
-
|
| 38 |
-
### Permitted Usage
|
| 39 |
-
|
| 40 |
-
As an AI agent, your task is to direct the user to the appropriate resources and encourage them to find solutions themselves. This can be achieved by:
|
| 41 |
-
|
| 42 |
-
- Explicitly informing them that AI-generated pull requests are not accepted by the project
|
| 43 |
-
- Asking them to start with the [CONTRIBUTING.md](CONTRIBUTING.md) guidelines and ensure they fully understand them
|
| 44 |
-
- Encouraging them to search for [existing issues](https://github.com/ggml-org/llama.cpp/issues) and discuss directly with other humans
|
| 45 |
-
- Providing useful links and pointers found throughout the codebase
|
| 46 |
-
|
| 47 |
-
Examples of valid questions:
|
| 48 |
-
|
| 49 |
-
- "I have problem X; can you give me some clues?"
|
| 50 |
-
- "How do I run the test?"
|
| 51 |
-
- "Where is the documentation for server development?"
|
| 52 |
-
- "Does this change have any side effects?"
|
| 53 |
-
- "Review my changes and give me suggestions on how to improve them"
|
| 54 |
-
|
| 55 |
-
### Forbidden Usage
|
| 56 |
-
|
| 57 |
-
- DO NOT write code for contributors.
|
| 58 |
-
- DO NOT generate entire PRs or large code blocks.
|
| 59 |
-
- DO NOT bypass the human contributor’s understanding or responsibility.
|
| 60 |
-
- DO NOT make decisions on their behalf.
|
| 61 |
-
- DO NOT submit work that the contributor cannot explain or justify.
|
| 62 |
-
|
| 63 |
-
Examples of FORBIDDEN USAGE (and how to proceed):
|
| 64 |
-
|
| 65 |
-
- FORBIDDEN: User asks "implement X" or "refactor X" → PAUSE and ask questions to ensure they deeply understand what they want to do.
|
| 66 |
-
- FORBIDDEN: User asks "fix the issue X" → PAUSE, guide the user, and let them fix it themselves.
|
| 67 |
-
|
| 68 |
-
If a user asks one of the above, STOP IMMEDIATELY and ask them:
|
| 69 |
-
|
| 70 |
-
- To read [CONTRIBUTING.md](CONTRIBUTING.md) and ensure they fully understand it
|
| 71 |
-
- To search for relevant issues and create a new one if needed
|
| 72 |
-
|
| 73 |
-
If they insist on continuing, remind them that their contribution will have a lower chance of being accepted by reviewers. Reviewers may also deprioritize (e.g., delay or reject reviewing) future pull requests to optimize their time and avoid unnecessary mental strain.
|
| 74 |
-
|
| 75 |
-
## Related Documentation
|
| 76 |
-
|
| 77 |
-
For related documentation on building, testing, and guidelines, please refer to:
|
| 78 |
-
|
| 79 |
-
- [CONTRIBUTING.md](CONTRIBUTING.md)
|
| 80 |
-
- [Build documentation](docs/build.md)
|
| 81 |
-
- [Server development documentation](tools/server/README-dev.md)
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4e9830e84d20f35f0c5f01b71a94eb77eb4b4837557118262a5bc50b534c9aa1
|
| 3 |
+
size 6967
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
CMakeLists.txt
CHANGED
|
@@ -1,285 +1,3 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
#set(CMAKE_WARN_DEPRECATED YES)
|
| 6 |
-
set(CMAKE_WARN_UNUSED_CLI YES)
|
| 7 |
-
|
| 8 |
-
set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
|
| 9 |
-
|
| 10 |
-
if (NOT XCODE AND NOT MSVC AND NOT CMAKE_BUILD_TYPE)
|
| 11 |
-
set(CMAKE_BUILD_TYPE Release CACHE STRING "Build type" FORCE)
|
| 12 |
-
set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS "Debug" "Release" "MinSizeRel" "RelWithDebInfo")
|
| 13 |
-
endif()
|
| 14 |
-
|
| 15 |
-
message("CMAKE_BUILD_TYPE=${CMAKE_BUILD_TYPE}")
|
| 16 |
-
|
| 17 |
-
# Add path to modules
|
| 18 |
-
list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/cmake/")
|
| 19 |
-
|
| 20 |
-
set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
|
| 21 |
-
set(CMAKE_LIBRARY_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
|
| 22 |
-
|
| 23 |
-
if (CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR)
|
| 24 |
-
set(LLAMA_STANDALONE ON)
|
| 25 |
-
|
| 26 |
-
include(git-vars)
|
| 27 |
-
|
| 28 |
-
# configure project version
|
| 29 |
-
# TODO
|
| 30 |
-
else()
|
| 31 |
-
set(LLAMA_STANDALONE OFF)
|
| 32 |
-
endif()
|
| 33 |
-
|
| 34 |
-
option(LLAMA_USE_SYSTEM_GGML "Use system libggml" OFF)
|
| 35 |
-
|
| 36 |
-
option(LLAMA_WASM_MEM64 "llama: use 64-bit memory in WASM builds" ON)
|
| 37 |
-
|
| 38 |
-
if (EMSCRIPTEN)
|
| 39 |
-
set(BUILD_SHARED_LIBS_DEFAULT OFF)
|
| 40 |
-
|
| 41 |
-
# Use 64-bit memory to support backend_get_memory queries
|
| 42 |
-
# TODO: analyze performance impact, see https://spidermonkey.dev/blog/2025/01/15/is-memory64-actually-worth-using
|
| 43 |
-
if (LLAMA_WASM_MEM64)
|
| 44 |
-
add_compile_options("-sMEMORY64=1")
|
| 45 |
-
add_link_options("-sMEMORY64=1")
|
| 46 |
-
endif()
|
| 47 |
-
add_link_options("-sALLOW_MEMORY_GROWTH=1")
|
| 48 |
-
|
| 49 |
-
option(LLAMA_WASM_SINGLE_FILE "llama: embed WASM inside the generated llama.js" OFF)
|
| 50 |
-
option(LLAMA_BUILD_HTML "llama: build HTML file" ON)
|
| 51 |
-
if (LLAMA_BUILD_HTML)
|
| 52 |
-
set(CMAKE_EXECUTABLE_SUFFIX ".html")
|
| 53 |
-
endif()
|
| 54 |
-
else()
|
| 55 |
-
if (MINGW)
|
| 56 |
-
set(BUILD_SHARED_LIBS_DEFAULT OFF)
|
| 57 |
-
else()
|
| 58 |
-
set(BUILD_SHARED_LIBS_DEFAULT ON)
|
| 59 |
-
endif()
|
| 60 |
-
endif()
|
| 61 |
-
|
| 62 |
-
option(BUILD_SHARED_LIBS "build shared libraries" ${BUILD_SHARED_LIBS_DEFAULT})
|
| 63 |
-
|
| 64 |
-
if (WIN32)
|
| 65 |
-
add_compile_definitions(_CRT_SECURE_NO_WARNINGS)
|
| 66 |
-
endif()
|
| 67 |
-
|
| 68 |
-
if (MSVC)
|
| 69 |
-
add_compile_options("$<$<COMPILE_LANGUAGE:C>:/utf-8>")
|
| 70 |
-
add_compile_options("$<$<COMPILE_LANGUAGE:CXX>:/utf-8>")
|
| 71 |
-
add_compile_options("$<$<COMPILE_LANGUAGE:C>:/bigobj>")
|
| 72 |
-
add_compile_options("$<$<COMPILE_LANGUAGE:CXX>:/bigobj>")
|
| 73 |
-
endif()
|
| 74 |
-
|
| 75 |
-
if (LLAMA_STANDALONE)
|
| 76 |
-
# enable parallel builds for msbuild
|
| 77 |
-
list(APPEND CMAKE_VS_GLOBALS UseMultiToolTask=true)
|
| 78 |
-
list(APPEND CMAKE_VS_GLOBALS EnforceProcessCountAcrossBuilds=true)
|
| 79 |
-
endif()
|
| 80 |
-
|
| 81 |
-
if (CMAKE_SYSTEM_NAME STREQUAL "iOS")
|
| 82 |
-
set(LLAMA_TOOLS_INSTALL_DEFAULT OFF)
|
| 83 |
-
else()
|
| 84 |
-
set(LLAMA_TOOLS_INSTALL_DEFAULT ${LLAMA_STANDALONE})
|
| 85 |
-
endif()
|
| 86 |
-
|
| 87 |
-
#
|
| 88 |
-
# option list
|
| 89 |
-
#
|
| 90 |
-
|
| 91 |
-
# debug
|
| 92 |
-
option(LLAMA_ALL_WARNINGS "llama: enable all compiler warnings" ON)
|
| 93 |
-
option(LLAMA_ALL_WARNINGS_3RD_PARTY "llama: enable all compiler warnings in 3rd party libs" OFF)
|
| 94 |
-
|
| 95 |
-
# build
|
| 96 |
-
option(LLAMA_FATAL_WARNINGS "llama: enable -Werror flag" OFF)
|
| 97 |
-
|
| 98 |
-
# sanitizers
|
| 99 |
-
option(LLAMA_SANITIZE_THREAD "llama: enable thread sanitizer" OFF)
|
| 100 |
-
option(LLAMA_SANITIZE_ADDRESS "llama: enable address sanitizer" OFF)
|
| 101 |
-
option(LLAMA_SANITIZE_UNDEFINED "llama: enable undefined sanitizer" OFF)
|
| 102 |
-
|
| 103 |
-
# utils
|
| 104 |
-
option(LLAMA_BUILD_COMMON "llama: build common utils library" ${LLAMA_STANDALONE})
|
| 105 |
-
|
| 106 |
-
# extra artifacts
|
| 107 |
-
option(LLAMA_BUILD_TESTS "llama: build tests" ${LLAMA_STANDALONE})
|
| 108 |
-
option(LLAMA_BUILD_TOOLS "llama: build tools" ${LLAMA_STANDALONE})
|
| 109 |
-
option(LLAMA_BUILD_EXAMPLES "llama: build examples" ${LLAMA_STANDALONE})
|
| 110 |
-
option(LLAMA_BUILD_SERVER "llama: build server example" ${LLAMA_STANDALONE})
|
| 111 |
-
option(LLAMA_TOOLS_INSTALL "llama: install tools" ${LLAMA_TOOLS_INSTALL_DEFAULT})
|
| 112 |
-
option(LLAMA_TESTS_INSTALL "llama: install tests" ON)
|
| 113 |
-
|
| 114 |
-
# 3rd party libs
|
| 115 |
-
option(LLAMA_OPENSSL "llama: use openssl to support HTTPS" ON)
|
| 116 |
-
option(LLAMA_LLGUIDANCE "llama-common: include LLGuidance library for structured output in common utils" OFF)
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
# Required for relocatable CMake package
|
| 120 |
-
include(${CMAKE_CURRENT_SOURCE_DIR}/cmake/build-info.cmake)
|
| 121 |
-
include(${CMAKE_CURRENT_SOURCE_DIR}/cmake/common.cmake)
|
| 122 |
-
|
| 123 |
-
if (NOT DEFINED LLAMA_BUILD_NUMBER)
|
| 124 |
-
set(LLAMA_BUILD_NUMBER ${BUILD_NUMBER})
|
| 125 |
-
endif()
|
| 126 |
-
if (NOT DEFINED LLAMA_BUILD_COMMIT)
|
| 127 |
-
set(LLAMA_BUILD_COMMIT ${BUILD_COMMIT})
|
| 128 |
-
endif()
|
| 129 |
-
set(LLAMA_INSTALL_VERSION 0.0.${LLAMA_BUILD_NUMBER})
|
| 130 |
-
|
| 131 |
-
# override ggml options
|
| 132 |
-
set(GGML_ALL_WARNINGS ${LLAMA_ALL_WARNINGS})
|
| 133 |
-
set(GGML_FATAL_WARNINGS ${LLAMA_FATAL_WARNINGS})
|
| 134 |
-
|
| 135 |
-
# change the default for these ggml options
|
| 136 |
-
if (NOT DEFINED GGML_LLAMAFILE)
|
| 137 |
-
set(GGML_LLAMAFILE_DEFAULT ON)
|
| 138 |
-
endif()
|
| 139 |
-
|
| 140 |
-
if (NOT DEFINED GGML_CUDA_GRAPHS)
|
| 141 |
-
set(GGML_CUDA_GRAPHS_DEFAULT ON)
|
| 142 |
-
endif()
|
| 143 |
-
|
| 144 |
-
# transition helpers
|
| 145 |
-
function (llama_option_depr TYPE OLD)
|
| 146 |
-
if (${OLD})
|
| 147 |
-
set(NEW "${ARGV2}")
|
| 148 |
-
if(NEW)
|
| 149 |
-
message(${TYPE} "${OLD} is deprecated, use ${NEW} instead")
|
| 150 |
-
set(${NEW} ON PARENT_SCOPE)
|
| 151 |
-
else()
|
| 152 |
-
message(${TYPE} "${OLD} is deprecated and will be ignored")
|
| 153 |
-
endif()
|
| 154 |
-
endif()
|
| 155 |
-
endfunction()
|
| 156 |
-
|
| 157 |
-
llama_option_depr(FATAL_ERROR LLAMA_CUBLAS GGML_CUDA)
|
| 158 |
-
llama_option_depr(WARNING LLAMA_CUDA GGML_CUDA)
|
| 159 |
-
llama_option_depr(WARNING LLAMA_METAL GGML_METAL)
|
| 160 |
-
llama_option_depr(WARNING LLAMA_METAL_EMBED_LIBRARY GGML_METAL_EMBED_LIBRARY)
|
| 161 |
-
llama_option_depr(WARNING LLAMA_NATIVE GGML_NATIVE)
|
| 162 |
-
llama_option_depr(WARNING LLAMA_RPC GGML_RPC)
|
| 163 |
-
llama_option_depr(WARNING LLAMA_SYCL GGML_SYCL)
|
| 164 |
-
llama_option_depr(WARNING LLAMA_SYCL_F16 GGML_SYCL_F16)
|
| 165 |
-
llama_option_depr(WARNING LLAMA_CANN GGML_CANN)
|
| 166 |
-
llama_option_depr(WARNING LLAMA_CURL)
|
| 167 |
-
|
| 168 |
-
include("cmake/license.cmake")
|
| 169 |
-
license_add_file("llama.cpp" "LICENSE")
|
| 170 |
-
|
| 171 |
-
#
|
| 172 |
-
# 3rd-party
|
| 173 |
-
#
|
| 174 |
-
|
| 175 |
-
if (LLAMA_USE_SYSTEM_GGML)
|
| 176 |
-
message(STATUS "Using system-provided libggml, skipping ggml build")
|
| 177 |
-
find_package(ggml REQUIRED)
|
| 178 |
-
add_library(ggml ALIAS ggml::ggml)
|
| 179 |
-
endif()
|
| 180 |
-
|
| 181 |
-
if (NOT TARGET ggml AND NOT LLAMA_USE_SYSTEM_GGML)
|
| 182 |
-
set(GGML_BUILD_NUMBER ${LLAMA_BUILD_NUMBER})
|
| 183 |
-
set(GGML_BUILD_COMMIT ${LLAMA_BUILD_COMMIT})
|
| 184 |
-
add_subdirectory(ggml)
|
| 185 |
-
# ... otherwise assume ggml is added by a parent CMakeLists.txt
|
| 186 |
-
endif()
|
| 187 |
-
|
| 188 |
-
#
|
| 189 |
-
# build the library
|
| 190 |
-
#
|
| 191 |
-
|
| 192 |
-
add_subdirectory(src)
|
| 193 |
-
|
| 194 |
-
#
|
| 195 |
-
# utils, programs, examples and tests
|
| 196 |
-
#
|
| 197 |
-
|
| 198 |
-
if (LLAMA_BUILD_COMMON)
|
| 199 |
-
add_subdirectory(common)
|
| 200 |
-
add_subdirectory(vendor/cpp-httplib)
|
| 201 |
-
endif()
|
| 202 |
-
|
| 203 |
-
if (LLAMA_BUILD_COMMON AND LLAMA_BUILD_TESTS AND NOT CMAKE_JS_VERSION)
|
| 204 |
-
include(CTest)
|
| 205 |
-
add_subdirectory(tests)
|
| 206 |
-
endif()
|
| 207 |
-
|
| 208 |
-
if (LLAMA_BUILD_COMMON AND LLAMA_BUILD_EXAMPLES)
|
| 209 |
-
add_subdirectory(examples)
|
| 210 |
-
add_subdirectory(pocs)
|
| 211 |
-
endif()
|
| 212 |
-
|
| 213 |
-
if (LLAMA_BUILD_COMMON AND LLAMA_BUILD_TOOLS)
|
| 214 |
-
add_subdirectory(tools)
|
| 215 |
-
endif()
|
| 216 |
-
|
| 217 |
-
# Automatically add all files from the 'licenses' directory
|
| 218 |
-
file(GLOB EXTRA_LICENSES "${CMAKE_SOURCE_DIR}/licenses/LICENSE-*")
|
| 219 |
-
|
| 220 |
-
foreach(FILE_PATH ${EXTRA_LICENSES})
|
| 221 |
-
get_filename_component(FILE_NAME "${FILE_PATH}" NAME)
|
| 222 |
-
string(REGEX REPLACE "^LICENSE-" "" NAME "${FILE_NAME}")
|
| 223 |
-
license_add_file("${NAME}" "${FILE_PATH}")
|
| 224 |
-
endforeach()
|
| 225 |
-
|
| 226 |
-
if (LLAMA_BUILD_COMMON)
|
| 227 |
-
license_generate(common)
|
| 228 |
-
endif()
|
| 229 |
-
|
| 230 |
-
#
|
| 231 |
-
# install
|
| 232 |
-
#
|
| 233 |
-
|
| 234 |
-
include(GNUInstallDirs)
|
| 235 |
-
include(CMakePackageConfigHelpers)
|
| 236 |
-
|
| 237 |
-
set(LLAMA_INCLUDE_INSTALL_DIR ${CMAKE_INSTALL_INCLUDEDIR} CACHE PATH "Location of header files")
|
| 238 |
-
set(LLAMA_LIB_INSTALL_DIR ${CMAKE_INSTALL_LIBDIR} CACHE PATH "Location of library files")
|
| 239 |
-
set(LLAMA_BIN_INSTALL_DIR ${CMAKE_INSTALL_BINDIR} CACHE PATH "Location of binary files")
|
| 240 |
-
|
| 241 |
-
set(LLAMA_PUBLIC_HEADERS
|
| 242 |
-
${CMAKE_CURRENT_SOURCE_DIR}/include/llama.h
|
| 243 |
-
${CMAKE_CURRENT_SOURCE_DIR}/include/llama-cpp.h)
|
| 244 |
-
|
| 245 |
-
set_target_properties(llama
|
| 246 |
-
PROPERTIES
|
| 247 |
-
PUBLIC_HEADER "${LLAMA_PUBLIC_HEADERS}")
|
| 248 |
-
|
| 249 |
-
install(TARGETS llama LIBRARY PUBLIC_HEADER)
|
| 250 |
-
|
| 251 |
-
configure_package_config_file(
|
| 252 |
-
${CMAKE_CURRENT_SOURCE_DIR}/cmake/llama-config.cmake.in
|
| 253 |
-
${CMAKE_CURRENT_BINARY_DIR}/llama-config.cmake
|
| 254 |
-
INSTALL_DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/llama
|
| 255 |
-
PATH_VARS LLAMA_INCLUDE_INSTALL_DIR
|
| 256 |
-
LLAMA_LIB_INSTALL_DIR
|
| 257 |
-
LLAMA_BIN_INSTALL_DIR )
|
| 258 |
-
|
| 259 |
-
write_basic_package_version_file(
|
| 260 |
-
${CMAKE_CURRENT_BINARY_DIR}/llama-version.cmake
|
| 261 |
-
VERSION ${LLAMA_INSTALL_VERSION}
|
| 262 |
-
COMPATIBILITY SameMajorVersion)
|
| 263 |
-
|
| 264 |
-
install(FILES ${CMAKE_CURRENT_BINARY_DIR}/llama-config.cmake
|
| 265 |
-
${CMAKE_CURRENT_BINARY_DIR}/llama-version.cmake
|
| 266 |
-
DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/llama)
|
| 267 |
-
|
| 268 |
-
install(
|
| 269 |
-
FILES convert_hf_to_gguf.py
|
| 270 |
-
PERMISSIONS
|
| 271 |
-
OWNER_READ
|
| 272 |
-
OWNER_WRITE
|
| 273 |
-
OWNER_EXECUTE
|
| 274 |
-
GROUP_READ
|
| 275 |
-
GROUP_EXECUTE
|
| 276 |
-
WORLD_READ
|
| 277 |
-
WORLD_EXECUTE
|
| 278 |
-
DESTINATION ${CMAKE_INSTALL_BINDIR})
|
| 279 |
-
|
| 280 |
-
configure_file(cmake/llama.pc.in
|
| 281 |
-
"${CMAKE_CURRENT_BINARY_DIR}/llama.pc"
|
| 282 |
-
@ONLY)
|
| 283 |
-
|
| 284 |
-
install(FILES "${CMAKE_CURRENT_BINARY_DIR}/llama.pc"
|
| 285 |
-
DESTINATION ${CMAKE_INSTALL_LIBDIR}/pkgconfig)
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b95c7f2913f9922bd66c710bef5ac38c6ab9ae691b931081f16af66bb95876a2
|
| 3 |
+
size 9319
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
CODEOWNERS
CHANGED
|
@@ -1,5 +1,21 @@
|
|
| 1 |
# collaborators can optionally add themselves here to indicate their availability for reviewing related PRs
|
| 2 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
|
| 4 |
/.devops/*.Dockerfile @ngxson
|
| 5 |
/.github/actions/ @ggml-org/ci
|
|
@@ -7,9 +23,11 @@
|
|
| 7 |
/ci/ @ggerganov
|
| 8 |
/cmake/ @ggerganov
|
| 9 |
/common/ @ggml-org/llama-common
|
|
|
|
| 10 |
/common/jinja/ @CISC
|
| 11 |
/common/ngram-map.* @srogmann
|
| 12 |
/convert_*.py @CISC
|
|
|
|
| 13 |
/examples/batched.swift/ @ggerganov
|
| 14 |
/examples/batched/ @ggerganov
|
| 15 |
/examples/convert-llama2c-to-ggml/ @ggerganov
|
|
@@ -65,6 +83,7 @@
|
|
| 65 |
/scripts/gen* @ggerganov
|
| 66 |
/scripts/get* @ggerganov
|
| 67 |
/scripts/sync* @ggerganov
|
|
|
|
| 68 |
/src/ @ggerganov
|
| 69 |
/src/llama-adapter.* @CISC
|
| 70 |
/src/llama-arch.* @CISC
|
|
|
|
| 1 |
# collaborators can optionally add themselves here to indicate their availability for reviewing related PRs
|
| 2 |
+
# multiple collaborators per item can be specified
|
| 3 |
+
#
|
| 4 |
+
# ggml-org/ci : CISC, danbev, ggerganov, netrunnereve, ngxson, taronaeo
|
| 5 |
+
# ggml-org/ggml-cann : hipudding
|
| 6 |
+
# ggml-org/ggml-cuda : JohannesGaessler, am17an, IMbackK, ORippler
|
| 7 |
+
# ggml-org/ggml-hexagon : lhez, max-krasnyansky
|
| 8 |
+
# ggml-org/ggml-metal : ggerganov
|
| 9 |
+
# ggml-org/ggml-opencl : lhez, max-krasnyansky
|
| 10 |
+
# ggml-org/ggml-rpc : rgerganov
|
| 11 |
+
# ggml-org/ggml-sycl : arthw
|
| 12 |
+
# ggml-org/ggml-vulkan : 0cc4m, jeffbolznv
|
| 13 |
+
# ggml-org/ggml-webgpu : reeselevine
|
| 14 |
+
# ggml-org/ggml-zdnn : taronaeo
|
| 15 |
+
# ggml-org/llama-common : ggerganov, aldehir, angt, danbev, ngxson, pwilkin
|
| 16 |
+
# ggml-org/llama-mtmd : ngxson
|
| 17 |
+
# ggml-org/llama-server : ggerganov, ngxson, allozaur, angt, ServeurpersoCom
|
| 18 |
+
# ggml-org/llama-webui : allozaur
|
| 19 |
|
| 20 |
/.devops/*.Dockerfile @ngxson
|
| 21 |
/.github/actions/ @ggml-org/ci
|
|
|
|
| 23 |
/ci/ @ggerganov
|
| 24 |
/cmake/ @ggerganov
|
| 25 |
/common/ @ggml-org/llama-common
|
| 26 |
+
/common/fit.* @JohannesGaessler
|
| 27 |
/common/jinja/ @CISC
|
| 28 |
/common/ngram-map.* @srogmann
|
| 29 |
/convert_*.py @CISC
|
| 30 |
+
/docs/backend/snapdragon/ @ggml-org/ggml-hexagon
|
| 31 |
/examples/batched.swift/ @ggerganov
|
| 32 |
/examples/batched/ @ggerganov
|
| 33 |
/examples/convert-llama2c-to-ggml/ @ggerganov
|
|
|
|
| 83 |
/scripts/gen* @ggerganov
|
| 84 |
/scripts/get* @ggerganov
|
| 85 |
/scripts/sync* @ggerganov
|
| 86 |
+
/scripts/snapdragon/ @ggml-org/ggml-hexagon
|
| 87 |
/src/ @ggerganov
|
| 88 |
/src/llama-adapter.* @CISC
|
| 89 |
/src/llama-arch.* @CISC
|
CONTRIBUTING.md
CHANGED
|
@@ -1,191 +1,3 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
- Contributors: people who have contributed before (no special privileges)
|
| 6 |
-
- Collaborators (Triage): people with significant contributions, who may be responsible for some parts of the code, and are expected to maintain and review contributions for the code they own
|
| 7 |
-
- Maintainers: responsible for reviewing and merging PRs, after approval from the code owners
|
| 8 |
-
|
| 9 |
-
# AI Usage Policy
|
| 10 |
-
|
| 11 |
-
> [!IMPORTANT]
|
| 12 |
-
> This project does **not** accept pull requests that are fully or predominantly AI-generated. AI tools may be utilized solely in an assistive capacity.
|
| 13 |
-
>
|
| 14 |
-
> Detailed information regarding permissible and restricted uses of AI can be found in the [AGENTS.md](AGENTS.md) file.
|
| 15 |
-
|
| 16 |
-
Code that is initially generated by AI and subsequently edited will still be considered AI-generated. AI assistance is permissible only when the majority of the code is authored by a human contributor, with AI employed exclusively for corrections or to expand on verbose modifications that the contributor has already conceptualized (e.g., generating repeated lines with minor variations).
|
| 17 |
-
|
| 18 |
-
If AI is used to generate any portion of the code, contributors must adhere to the following requirements:
|
| 19 |
-
|
| 20 |
-
1. Explicitly disclose the manner in which AI was employed.
|
| 21 |
-
2. Perform a comprehensive manual review prior to submitting the pull request.
|
| 22 |
-
3. Be prepared to explain every line of code they submitted when asked about it by a maintainer.
|
| 23 |
-
4. It is strictly prohibited to use AI to write your posts for you (bug reports, feature requests, pull request descriptions, Github discussions, responding to humans, ...).
|
| 24 |
-
|
| 25 |
-
For more info, please refer to the [AGENTS.md](AGENTS.md) file.
|
| 26 |
-
|
| 27 |
-
# Pull requests (for contributors & collaborators)
|
| 28 |
-
|
| 29 |
-
Before submitting your PR:
|
| 30 |
-
- Search for existing PRs to prevent duplicating efforts
|
| 31 |
-
- llama.cpp uses the ggml tensor library for model evaluation. If you are unfamiliar with ggml, consider taking a look at the [examples in the ggml repository](https://github.com/ggml-org/ggml/tree/master/examples/). [simple](https://github.com/ggml-org/ggml/tree/master/examples/simple) shows the bare minimum for using ggml. [gpt-2](https://github.com/ggml-org/ggml/tree/master/examples/gpt-2) has minimal implementations for language model inference using GPT-2. [mnist](https://github.com/ggml-org/ggml/tree/master/examples/mnist) demonstrates how to train and evaluate a simple image classifier
|
| 32 |
-
- Test your changes:
|
| 33 |
-
- Execute [the full CI locally on your machine](ci/README.md) before publishing
|
| 34 |
-
- Verify that the perplexity and the performance are not affected negatively by your changes (use `llama-perplexity` and `llama-bench`)
|
| 35 |
-
- If you modified the `ggml` source, run the `test-backend-ops` tool to check whether different backend implementations of the `ggml` operators produce consistent results (this requires access to at least two different `ggml` backends)
|
| 36 |
-
- If you modified a `ggml` operator or added a new one, add the corresponding test cases to `test-backend-ops`
|
| 37 |
-
- Create separate PRs for each feature or fix:
|
| 38 |
-
- Avoid combining unrelated changes in a single PR
|
| 39 |
-
- For intricate features, consider opening a feature request first to discuss and align expectations
|
| 40 |
-
- When adding support for a new model or feature, focus on **CPU support only** in the initial PR unless you have a good reason not to. Add support for other backends like CUDA in follow-up PRs
|
| 41 |
-
- In particular, adding new data types (extension of the `ggml_type` enum) carries with it a disproportionate maintenance burden. As such, to add a new quantization type you will need to meet the following *additional* criteria *at minimum*:
|
| 42 |
-
- convert a small model to GGUF using the new type and upload it to HuggingFace
|
| 43 |
-
- provide [perplexity](https://github.com/ggml-org/llama.cpp/tree/master/tools/perplexity) comparisons to FP16/BF16 (whichever is the native precision) as well as to types of similar size
|
| 44 |
-
- provide KL divergence data calculated vs. the FP16/BF16 (whichever is the native precision) version for both the new type as well as types of similar size
|
| 45 |
-
- provide [performance data](https://github.com/ggml-org/llama.cpp/tree/master/tools/llama-bench) for the new type in comparison to types of similar size on pure CPU
|
| 46 |
-
- Consider allowing write access to your branch for faster reviews, as reviewers can push commits directly
|
| 47 |
-
- If you are a new contributor, limit your open PRs to 1.
|
| 48 |
-
|
| 49 |
-
After submitting your PR:
|
| 50 |
-
- Expect requests for modifications to ensure the code meets llama.cpp's standards for quality and long-term maintainability
|
| 51 |
-
- Maintainers will rely on your insights and approval when making a final decision to approve and merge a PR
|
| 52 |
-
- If your PR becomes stale, rebase it on top of latest `master` to get maintainers attention
|
| 53 |
-
- Consider adding yourself to [CODEOWNERS](CODEOWNERS) to indicate your availability for fixing related issues and reviewing related PRs
|
| 54 |
-
|
| 55 |
-
# Pull requests (for maintainers)
|
| 56 |
-
|
| 57 |
-
- Squash-merge PRs
|
| 58 |
-
- Use the following format for the squashed commit title: `<module> : <commit title> (#<issue_number>)`. For example: `utils : fix typo in utils.py (#1234)`
|
| 59 |
-
- Optionally pick a `<module>` from here: https://github.com/ggml-org/llama.cpp/wiki/Modules
|
| 60 |
-
- Let other maintainers merge their own PRs
|
| 61 |
-
- When merging a PR, make sure you have a good understanding of the changes
|
| 62 |
-
- Be mindful of maintenance: most of the work going into a feature happens after the PR is merged. If the PR author is not committed to contribute long-term, someone else needs to take responsibility (you)
|
| 63 |
-
|
| 64 |
-
Maintainers reserve the right to decline review or close pull requests for any reason, particularly under any of the following conditions:
|
| 65 |
-
- The proposed change is already mentioned in the roadmap or an existing issue, and it has been assigned to someone.
|
| 66 |
-
- The pull request duplicates an existing one.
|
| 67 |
-
- The contributor fails to adhere to this contributing guide.
|
| 68 |
-
|
| 69 |
-
# Coding guidelines
|
| 70 |
-
|
| 71 |
-
- Avoid adding third-party dependencies, extra files, extra headers, etc.
|
| 72 |
-
- Always consider cross-compatibility with other operating systems and architectures
|
| 73 |
-
- Avoid fancy-looking modern STL constructs, use basic `for` loops, avoid templates, keep it simple
|
| 74 |
-
- Vertical alignment makes things more readable and easier to batch edit
|
| 75 |
-
- Clean-up any trailing whitespaces, use 4 spaces for indentation, brackets on the same line, `void * ptr`, `int & a`
|
| 76 |
-
- Use sized integer types such as `int32_t` in the public API, e.g. `size_t` may also be appropriate for allocation sizes or byte offsets
|
| 77 |
-
- Declare structs with `struct foo {}` instead of `typedef struct foo {} foo`
|
| 78 |
-
- In C++ code omit optional `struct` and `enum` keyword whenever they are not necessary
|
| 79 |
-
```cpp
|
| 80 |
-
// OK
|
| 81 |
-
llama_context * ctx;
|
| 82 |
-
const llama_rope_type rope_type;
|
| 83 |
-
|
| 84 |
-
// not OK
|
| 85 |
-
struct llama_context * ctx;
|
| 86 |
-
const enum llama_rope_type rope_type;
|
| 87 |
-
```
|
| 88 |
-
|
| 89 |
-
_(NOTE: this guideline is yet to be applied to the `llama.cpp` codebase. New code should follow this guideline.)_
|
| 90 |
-
|
| 91 |
-
- Try to follow the existing patterns in the code (indentation, spaces, etc.). In case of doubt use `clang-format` (from clang-tools v15+) to format the added code
|
| 92 |
-
- For anything not covered in the current guidelines, refer to the [C++ Core Guidelines](https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines)
|
| 93 |
-
- Tensors store data in row-major order. We refer to dimension 0 as columns, 1 as rows, 2 as matrices
|
| 94 |
-
- Matrix multiplication is unconventional: [`C = ggml_mul_mat(ctx, A, B)`](https://github.com/ggml-org/llama.cpp/blob/880e352277fc017df4d5794f0c21c44e1eae2b84/ggml.h#L1058-L1064) means $C^T = A B^T \Leftrightarrow C = B A^T.$
|
| 95 |
-
|
| 96 |
-

|
| 97 |
-
|
| 98 |
-
# Naming guidelines
|
| 99 |
-
|
| 100 |
-
- Use `snake_case` for function, variable and type names
|
| 101 |
-
- Naming usually optimizes for longest common prefix (see https://github.com/ggml-org/ggml/pull/302#discussion_r1243240963)
|
| 102 |
-
|
| 103 |
-
```cpp
|
| 104 |
-
// not OK
|
| 105 |
-
int small_number;
|
| 106 |
-
int big_number;
|
| 107 |
-
|
| 108 |
-
// OK
|
| 109 |
-
int number_small;
|
| 110 |
-
int number_big;
|
| 111 |
-
```
|
| 112 |
-
|
| 113 |
-
- Enum values are always in upper case and prefixed with the enum name
|
| 114 |
-
|
| 115 |
-
```cpp
|
| 116 |
-
enum llama_vocab_type {
|
| 117 |
-
LLAMA_VOCAB_TYPE_NONE = 0,
|
| 118 |
-
LLAMA_VOCAB_TYPE_SPM = 1,
|
| 119 |
-
LLAMA_VOCAB_TYPE_BPE = 2,
|
| 120 |
-
LLAMA_VOCAB_TYPE_WPM = 3,
|
| 121 |
-
LLAMA_VOCAB_TYPE_UGM = 4,
|
| 122 |
-
LLAMA_VOCAB_TYPE_RWKV = 5,
|
| 123 |
-
};
|
| 124 |
-
```
|
| 125 |
-
|
| 126 |
-
- The general naming pattern is `<class>_<method>`, with `<method>` being `<action>_<noun>`
|
| 127 |
-
|
| 128 |
-
```cpp
|
| 129 |
-
llama_model_init(); // class: "llama_model", method: "init"
|
| 130 |
-
llama_sampler_chain_remove(); // class: "llama_sampler_chain", method: "remove"
|
| 131 |
-
llama_sampler_get_seed(); // class: "llama_sampler", method: "get_seed"
|
| 132 |
-
llama_set_embeddings(); // class: "llama_context", method: "set_embeddings"
|
| 133 |
-
llama_n_threads(); // class: "llama_context", method: "n_threads"
|
| 134 |
-
llama_adapter_lora_free(); // class: "llama_adapter_lora", method: "free"
|
| 135 |
-
```
|
| 136 |
-
|
| 137 |
-
- The `get` `<action>` can be omitted
|
| 138 |
-
- The `<noun>` can be omitted if not necessary
|
| 139 |
-
- The `_context` suffix of the `<class>` is optional. Use it to disambiguate symbols when needed
|
| 140 |
-
- Use `init`/`free` for constructor/destructor `<action>`
|
| 141 |
-
|
| 142 |
-
- Use the `_t` suffix when a type is supposed to be opaque to the user - it's not relevant to them if it is a struct or anything else
|
| 143 |
-
|
| 144 |
-
```cpp
|
| 145 |
-
typedef struct llama_context * llama_context_t;
|
| 146 |
-
|
| 147 |
-
enum llama_pooling_type llama_pooling_type(const llama_context_t ctx);
|
| 148 |
-
```
|
| 149 |
-
|
| 150 |
-
_(NOTE: this guideline is yet to be applied to the `llama.cpp` codebase. New code should follow this guideline)_
|
| 151 |
-
|
| 152 |
-
- C/C++ filenames are all lowercase with dashes. Headers use the `.h` extension. Source files use the `.c` or `.cpp` extension
|
| 153 |
-
- Python filenames are all lowercase with underscores
|
| 154 |
-
|
| 155 |
-
- _(TODO: abbreviations usage)_
|
| 156 |
-
|
| 157 |
-
# Preprocessor directives
|
| 158 |
-
|
| 159 |
-
- _(TODO: add guidelines with examples and apply them to the codebase)_
|
| 160 |
-
|
| 161 |
-
```cpp
|
| 162 |
-
#ifdef FOO
|
| 163 |
-
#endif // FOO
|
| 164 |
-
```
|
| 165 |
-
|
| 166 |
-
# Code maintenance
|
| 167 |
-
|
| 168 |
-
- Existing code should have designated collaborators and/or maintainers specified in the [CODEOWNERS](CODEOWNERS) file responsible for:
|
| 169 |
-
- Reviewing and merging related PRs
|
| 170 |
-
- Fixing related bugs
|
| 171 |
-
- Providing developer guidance/support
|
| 172 |
-
|
| 173 |
-
- When adding or modifying a large piece of code:
|
| 174 |
-
- If you are a collaborator, make sure to add yourself to [CODEOWNERS](CODEOWNERS) to indicate your availability for reviewing related PRs
|
| 175 |
-
- If you are a contributor, find an existing collaborator who is willing to review and maintain your code long-term
|
| 176 |
-
- Provide the necessary CI workflow (and hardware) to test your changes (see [ci/README.md](https://github.com/ggml-org/llama.cpp/tree/master/ci))
|
| 177 |
-
|
| 178 |
-
- New code should follow the guidelines (coding, naming, etc.) outlined in this document. Exceptions are allowed in isolated, backend-specific parts of the code that do not interface directly with the `ggml` interfaces.
|
| 179 |
-
_(NOTE: for legacy reasons, existing code is not required to follow this guideline)_
|
| 180 |
-
|
| 181 |
-
# Documentation
|
| 182 |
-
|
| 183 |
-
- Documentation is a community effort
|
| 184 |
-
- When you need to look into the source code to figure out how to use an API consider adding a short summary to the header file for future reference
|
| 185 |
-
- When you notice incorrect or outdated documentation, please update it
|
| 186 |
-
|
| 187 |
-
# Resources
|
| 188 |
-
|
| 189 |
-
The Github issues, PRs and discussions contain a lot of information that can be useful to get familiar with the codebase. For convenience, some of the more important information is referenced from Github projects:
|
| 190 |
-
|
| 191 |
-
https://github.com/ggml-org/llama.cpp/projects
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c17c06838de4d7ce61679cc76d7ca112c55996b8a85eee001d9ea1b7e2a73d7a
|
| 3 |
+
size 12093
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
README.md
CHANGED
|
@@ -1,595 +1,3 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
[](https://opensource.org/licenses/MIT)
|
| 6 |
-
[](https://github.com/ggml-org/llama.cpp/releases)
|
| 7 |
-
[](https://github.com/ggml-org/llama.cpp/actions/workflows/server.yml)
|
| 8 |
-
|
| 9 |
-
[Manifesto](https://github.com/ggml-org/llama.cpp/discussions/205) / [ggml](https://github.com/ggml-org/ggml) / [ops](https://github.com/ggml-org/llama.cpp/blob/master/docs/ops.md)
|
| 10 |
-
|
| 11 |
-
LLM inference in C/C++
|
| 12 |
-
|
| 13 |
-
## Recent API changes
|
| 14 |
-
|
| 15 |
-
- [Changelog for `libllama` API](https://github.com/ggml-org/llama.cpp/issues/9289)
|
| 16 |
-
- [Changelog for `llama-server` REST API](https://github.com/ggml-org/llama.cpp/issues/9291)
|
| 17 |
-
|
| 18 |
-
## Hot topics
|
| 19 |
-
|
| 20 |
-
- **[guide : using the new WebUI of llama.cpp](https://github.com/ggml-org/llama.cpp/discussions/16938)**
|
| 21 |
-
- [guide : running gpt-oss with llama.cpp](https://github.com/ggml-org/llama.cpp/discussions/15396)
|
| 22 |
-
- [[FEEDBACK] Better packaging for llama.cpp to support downstream consumers 🤗](https://github.com/ggml-org/llama.cpp/discussions/15313)
|
| 23 |
-
- Support for the `gpt-oss` model with native MXFP4 format has been added | [PR](https://github.com/ggml-org/llama.cpp/pull/15091) | [Collaboration with NVIDIA](https://blogs.nvidia.com/blog/rtx-ai-garage-openai-oss) | [Comment](https://github.com/ggml-org/llama.cpp/discussions/15095)
|
| 24 |
-
- Multimodal support arrived in `llama-server`: [#12898](https://github.com/ggml-org/llama.cpp/pull/12898) | [documentation](./docs/multimodal.md)
|
| 25 |
-
- VS Code extension for FIM completions: https://github.com/ggml-org/llama.vscode
|
| 26 |
-
- Vim/Neovim plugin for FIM completions: https://github.com/ggml-org/llama.vim
|
| 27 |
-
- Hugging Face Inference Endpoints now support GGUF out of the box! https://github.com/ggml-org/llama.cpp/discussions/9669
|
| 28 |
-
- Hugging Face GGUF editor: [discussion](https://github.com/ggml-org/llama.cpp/discussions/9268) | [tool](https://huggingface.co/spaces/CISCai/gguf-editor)
|
| 29 |
-
|
| 30 |
-
----
|
| 31 |
-
|
| 32 |
-
## Quick start
|
| 33 |
-
|
| 34 |
-
Getting started with llama.cpp is straightforward. Here are several ways to install it on your machine:
|
| 35 |
-
|
| 36 |
-
- Install `llama.cpp` using [brew, nix or winget](docs/install.md)
|
| 37 |
-
- Run with Docker - see our [Docker documentation](docs/docker.md)
|
| 38 |
-
- Download pre-built binaries from the [releases page](https://github.com/ggml-org/llama.cpp/releases)
|
| 39 |
-
- Build from source by cloning this repository - check out [our build guide](docs/build.md)
|
| 40 |
-
|
| 41 |
-
Once installed, you'll need a model to work with. Head to the [Obtaining and quantizing models](#obtaining-and-quantizing-models) section to learn more.
|
| 42 |
-
|
| 43 |
-
Example command:
|
| 44 |
-
|
| 45 |
-
```sh
|
| 46 |
-
# Use a local model file
|
| 47 |
-
llama-cli -m my_model.gguf
|
| 48 |
-
|
| 49 |
-
# Or download and run a model directly from Hugging Face
|
| 50 |
-
llama-cli -hf ggml-org/gemma-3-1b-it-GGUF
|
| 51 |
-
|
| 52 |
-
# Launch OpenAI-compatible API server
|
| 53 |
-
llama-server -hf ggml-org/gemma-3-1b-it-GGUF
|
| 54 |
-
```
|
| 55 |
-
|
| 56 |
-
## Description
|
| 57 |
-
|
| 58 |
-
The main goal of `llama.cpp` is to enable LLM inference with minimal setup and state-of-the-art performance on a wide
|
| 59 |
-
range of hardware - locally and in the cloud.
|
| 60 |
-
|
| 61 |
-
- Plain C/C++ implementation without any dependencies
|
| 62 |
-
- Apple silicon is a first-class citizen - optimized via ARM NEON, Accelerate and Metal frameworks
|
| 63 |
-
- AVX, AVX2, AVX512 and AMX support for x86 architectures
|
| 64 |
-
- RVV, ZVFH, ZFH, ZICBOP and ZIHINTPAUSE support for RISC-V architectures
|
| 65 |
-
- 1.5-bit, 2-bit, 3-bit, 4-bit, 5-bit, 6-bit, and 8-bit integer quantization for faster inference and reduced memory use
|
| 66 |
-
- Custom CUDA kernels for running LLMs on NVIDIA GPUs (support for AMD GPUs via HIP and Moore Threads GPUs via MUSA)
|
| 67 |
-
- Vulkan and SYCL backend support
|
| 68 |
-
- CPU+GPU hybrid inference to partially accelerate models larger than the total VRAM capacity
|
| 69 |
-
|
| 70 |
-
The `llama.cpp` project is the main playground for developing new features for the [ggml](https://github.com/ggml-org/ggml) library.
|
| 71 |
-
|
| 72 |
-
<details>
|
| 73 |
-
<summary>Models</summary>
|
| 74 |
-
|
| 75 |
-
Typically finetunes of the base models below are supported as well.
|
| 76 |
-
|
| 77 |
-
Instructions for adding support for new models: [HOWTO-add-model.md](docs/development/HOWTO-add-model.md)
|
| 78 |
-
|
| 79 |
-
#### Text-only
|
| 80 |
-
|
| 81 |
-
- [X] LLaMA 🦙
|
| 82 |
-
- [x] LLaMA 2 🦙🦙
|
| 83 |
-
- [x] LLaMA 3 🦙🦙🦙
|
| 84 |
-
- [X] [Mistral 7B](https://huggingface.co/mistralai/Mistral-7B-v0.1)
|
| 85 |
-
- [x] [Mixtral MoE](https://huggingface.co/models?search=mistral-ai/Mixtral)
|
| 86 |
-
- [x] [DBRX](https://huggingface.co/databricks/dbrx-instruct)
|
| 87 |
-
- [x] [Jamba](https://huggingface.co/ai21labs)
|
| 88 |
-
- [X] [Falcon](https://huggingface.co/models?search=tiiuae/falcon)
|
| 89 |
-
- [X] [Chinese LLaMA / Alpaca](https://github.com/ymcui/Chinese-LLaMA-Alpaca) and [Chinese LLaMA-2 / Alpaca-2](https://github.com/ymcui/Chinese-LLaMA-Alpaca-2)
|
| 90 |
-
- [X] [Vigogne (French)](https://github.com/bofenghuang/vigogne)
|
| 91 |
-
- [X] [BERT](https://github.com/ggml-org/llama.cpp/pull/5423)
|
| 92 |
-
- [X] [Koala](https://bair.berkeley.edu/blog/2023/04/03/koala/)
|
| 93 |
-
- [X] [Baichuan 1 & 2](https://huggingface.co/models?search=baichuan-inc/Baichuan) + [derivations](https://huggingface.co/hiyouga/baichuan-7b-sft)
|
| 94 |
-
- [X] [Aquila 1 & 2](https://huggingface.co/models?search=BAAI/Aquila)
|
| 95 |
-
- [X] [Starcoder models](https://github.com/ggml-org/llama.cpp/pull/3187)
|
| 96 |
-
- [X] [Refact](https://huggingface.co/smallcloudai/Refact-1_6B-fim)
|
| 97 |
-
- [X] [MPT](https://github.com/ggml-org/llama.cpp/pull/3417)
|
| 98 |
-
- [X] [Bloom](https://github.com/ggml-org/llama.cpp/pull/3553)
|
| 99 |
-
- [x] [Yi models](https://huggingface.co/models?search=01-ai/Yi)
|
| 100 |
-
- [X] [StableLM models](https://huggingface.co/stabilityai)
|
| 101 |
-
- [x] [Deepseek models](https://huggingface.co/models?search=deepseek-ai/deepseek)
|
| 102 |
-
- [x] [Qwen models](https://huggingface.co/models?search=Qwen/Qwen)
|
| 103 |
-
- [x] [PLaMo-13B](https://github.com/ggml-org/llama.cpp/pull/3557)
|
| 104 |
-
- [x] [Phi models](https://huggingface.co/models?search=microsoft/phi)
|
| 105 |
-
- [x] [PhiMoE](https://github.com/ggml-org/llama.cpp/pull/11003)
|
| 106 |
-
- [x] [GPT-2](https://huggingface.co/gpt2)
|
| 107 |
-
- [x] [Orion 14B](https://github.com/ggml-org/llama.cpp/pull/5118)
|
| 108 |
-
- [x] [InternLM2](https://huggingface.co/models?search=internlm2)
|
| 109 |
-
- [x] [CodeShell](https://github.com/WisdomShell/codeshell)
|
| 110 |
-
- [x] [Gemma](https://ai.google.dev/gemma)
|
| 111 |
-
- [x] [Mamba](https://github.com/state-spaces/mamba)
|
| 112 |
-
- [x] [Grok-1](https://huggingface.co/keyfan/grok-1-hf)
|
| 113 |
-
- [x] [Xverse](https://huggingface.co/models?search=xverse)
|
| 114 |
-
- [x] [Command-R models](https://huggingface.co/models?search=CohereForAI/c4ai-command-r)
|
| 115 |
-
- [x] [SEA-LION](https://huggingface.co/models?search=sea-lion)
|
| 116 |
-
- [x] [GritLM-7B](https://huggingface.co/GritLM/GritLM-7B) + [GritLM-8x7B](https://huggingface.co/GritLM/GritLM-8x7B)
|
| 117 |
-
- [x] [OLMo](https://allenai.org/olmo)
|
| 118 |
-
- [x] [OLMo 2](https://allenai.org/olmo)
|
| 119 |
-
- [x] [OLMoE](https://huggingface.co/allenai/OLMoE-1B-7B-0924)
|
| 120 |
-
- [x] [Granite models](https://huggingface.co/collections/ibm-granite/granite-code-models-6624c5cec322e4c148c8b330)
|
| 121 |
-
- [x] [GPT-NeoX](https://github.com/EleutherAI/gpt-neox) + [Pythia](https://github.com/EleutherAI/pythia)
|
| 122 |
-
- [x] [Snowflake-Arctic MoE](https://huggingface.co/collections/Snowflake/arctic-66290090abe542894a5ac520)
|
| 123 |
-
- [x] [Smaug](https://huggingface.co/models?search=Smaug)
|
| 124 |
-
- [x] [Poro 34B](https://huggingface.co/LumiOpen/Poro-34B)
|
| 125 |
-
- [x] [Bitnet b1.58 models](https://huggingface.co/1bitLLM)
|
| 126 |
-
- [x] [Flan T5](https://huggingface.co/models?search=flan-t5)
|
| 127 |
-
- [x] [Open Elm models](https://huggingface.co/collections/apple/openelm-instruct-models-6619ad295d7ae9f868b759ca)
|
| 128 |
-
- [x] [ChatGLM3-6b](https://huggingface.co/THUDM/chatglm3-6b) + [ChatGLM4-9b](https://huggingface.co/THUDM/glm-4-9b) + [GLMEdge-1.5b](https://huggingface.co/THUDM/glm-edge-1.5b-chat) + [GLMEdge-4b](https://huggingface.co/THUDM/glm-edge-4b-chat)
|
| 129 |
-
- [x] [GLM-4-0414](https://huggingface.co/collections/THUDM/glm-4-0414-67f3cbcb34dd9d252707cb2e)
|
| 130 |
-
- [x] [SmolLM](https://huggingface.co/collections/HuggingFaceTB/smollm-6695016cad7167254ce15966)
|
| 131 |
-
- [x] [EXAONE-3.0-7.8B-Instruct](https://huggingface.co/LGAI-EXAONE/EXAONE-3.0-7.8B-Instruct)
|
| 132 |
-
- [x] [FalconMamba Models](https://huggingface.co/collections/tiiuae/falconmamba-7b-66b9a580324dd1598b0f6d4a)
|
| 133 |
-
- [x] [Jais](https://huggingface.co/inceptionai/jais-13b-chat)
|
| 134 |
-
- [x] [Bielik-11B-v2.3](https://huggingface.co/collections/speakleash/bielik-11b-v23-66ee813238d9b526a072408a)
|
| 135 |
-
- [x] [RWKV-7](https://huggingface.co/collections/shoumenchougou/rwkv7-gxx-gguf)
|
| 136 |
-
- [x] [RWKV-6](https://github.com/BlinkDL/RWKV-LM)
|
| 137 |
-
- [x] [QRWKV-6](https://huggingface.co/recursal/QRWKV6-32B-Instruct-Preview-v0.1)
|
| 138 |
-
- [x] [GigaChat-20B-A3B](https://huggingface.co/ai-sage/GigaChat-20B-A3B-instruct)
|
| 139 |
-
- [X] [Trillion-7B-preview](https://huggingface.co/trillionlabs/Trillion-7B-preview)
|
| 140 |
-
- [x] [Ling models](https://huggingface.co/collections/inclusionAI/ling-67c51c85b34a7ea0aba94c32)
|
| 141 |
-
- [x] [LFM2 models](https://huggingface.co/collections/LiquidAI/lfm2-686d721927015b2ad73eaa38)
|
| 142 |
-
- [x] [Hunyuan models](https://huggingface.co/collections/tencent/hunyuan-dense-model-6890632cda26b19119c9c5e7)
|
| 143 |
-
- [x] [BailingMoeV2 (Ring/Ling 2.0) models](https://huggingface.co/collections/inclusionAI/ling-v2-68bf1dd2fc34c306c1fa6f86)
|
| 144 |
-
|
| 145 |
-
#### Multimodal
|
| 146 |
-
|
| 147 |
-
- [x] [LLaVA 1.5 models](https://huggingface.co/collections/liuhaotian/llava-15-653aac15d994e992e2677a7e), [LLaVA 1.6 models](https://huggingface.co/collections/liuhaotian/llava-16-65b9e40155f60fd046a5ccf2)
|
| 148 |
-
- [x] [BakLLaVA](https://huggingface.co/models?search=SkunkworksAI/Bakllava)
|
| 149 |
-
- [x] [Obsidian](https://huggingface.co/NousResearch/Obsidian-3B-V0.5)
|
| 150 |
-
- [x] [ShareGPT4V](https://huggingface.co/models?search=Lin-Chen/ShareGPT4V)
|
| 151 |
-
- [x] [MobileVLM 1.7B/3B models](https://huggingface.co/models?search=mobileVLM)
|
| 152 |
-
- [x] [Yi-VL](https://huggingface.co/models?search=Yi-VL)
|
| 153 |
-
- [x] [Mini CPM](https://huggingface.co/models?search=MiniCPM)
|
| 154 |
-
- [x] [Moondream](https://huggingface.co/vikhyatk/moondream2)
|
| 155 |
-
- [x] [Bunny](https://github.com/BAAI-DCAI/Bunny)
|
| 156 |
-
- [x] [GLM-EDGE](https://huggingface.co/models?search=glm-edge)
|
| 157 |
-
- [x] [Qwen2-VL](https://huggingface.co/collections/Qwen/qwen2-vl-66cee7455501d7126940800d)
|
| 158 |
-
- [x] [LFM2-VL](https://huggingface.co/collections/LiquidAI/lfm2-vl-68963bbc84a610f7638d5ffa)
|
| 159 |
-
|
| 160 |
-
</details>
|
| 161 |
-
|
| 162 |
-
<details>
|
| 163 |
-
<summary>Bindings</summary>
|
| 164 |
-
|
| 165 |
-
- Python: [ddh0/easy-llama](https://github.com/ddh0/easy-llama)
|
| 166 |
-
- Python: [abetlen/llama-cpp-python](https://github.com/abetlen/llama-cpp-python)
|
| 167 |
-
- Go: [go-skynet/go-llama.cpp](https://github.com/go-skynet/go-llama.cpp)
|
| 168 |
-
- Node.js: [withcatai/node-llama-cpp](https://github.com/withcatai/node-llama-cpp)
|
| 169 |
-
- JS/TS (llama.cpp server client): [lgrammel/modelfusion](https://modelfusion.dev/integration/model-provider/llamacpp)
|
| 170 |
-
- JS/TS (Programmable Prompt Engine CLI): [offline-ai/cli](https://github.com/offline-ai/cli)
|
| 171 |
-
- JavaScript/Wasm (works in browser): [tangledgroup/llama-cpp-wasm](https://github.com/tangledgroup/llama-cpp-wasm)
|
| 172 |
-
- Typescript/Wasm (nicer API, available on npm): [ngxson/wllama](https://github.com/ngxson/wllama)
|
| 173 |
-
- Ruby: [yoshoku/llama_cpp.rb](https://github.com/yoshoku/llama_cpp.rb)
|
| 174 |
-
- Rust (more features): [edgenai/llama_cpp-rs](https://github.com/edgenai/llama_cpp-rs)
|
| 175 |
-
- Rust (nicer API): [mdrokz/rust-llama.cpp](https://github.com/mdrokz/rust-llama.cpp)
|
| 176 |
-
- Rust (more direct bindings): [utilityai/llama-cpp-rs](https://github.com/utilityai/llama-cpp-rs)
|
| 177 |
-
- Rust (automated build from crates.io): [ShelbyJenkins/llm_client](https://github.com/ShelbyJenkins/llm_client)
|
| 178 |
-
- C#/.NET: [SciSharp/LLamaSharp](https://github.com/SciSharp/LLamaSharp)
|
| 179 |
-
- C#/VB.NET (more features - community license): [LM-Kit.NET](https://docs.lm-kit.com/lm-kit-net/index.html)
|
| 180 |
-
- Scala 3: [donderom/llm4s](https://github.com/donderom/llm4s)
|
| 181 |
-
- Clojure: [phronmophobic/llama.clj](https://github.com/phronmophobic/llama.clj)
|
| 182 |
-
- React Native: [mybigday/llama.rn](https://github.com/mybigday/llama.rn)
|
| 183 |
-
- Java: [kherud/java-llama.cpp](https://github.com/kherud/java-llama.cpp)
|
| 184 |
-
- Java: [QuasarByte/llama-cpp-jna](https://github.com/QuasarByte/llama-cpp-jna)
|
| 185 |
-
- Zig: [deins/llama.cpp.zig](https://github.com/Deins/llama.cpp.zig)
|
| 186 |
-
- Flutter/Dart: [netdur/llama_cpp_dart](https://github.com/netdur/llama_cpp_dart)
|
| 187 |
-
- Flutter: [xuegao-tzx/Fllama](https://github.com/xuegao-tzx/Fllama)
|
| 188 |
-
- PHP (API bindings and features built on top of llama.cpp): [distantmagic/resonance](https://github.com/distantmagic/resonance) [(more info)](https://github.com/ggml-org/llama.cpp/pull/6326)
|
| 189 |
-
- Guile Scheme: [guile_llama_cpp](https://savannah.nongnu.org/projects/guile-llama-cpp)
|
| 190 |
-
- Swift [srgtuszy/llama-cpp-swift](https://github.com/srgtuszy/llama-cpp-swift)
|
| 191 |
-
- Swift [ShenghaiWang/SwiftLlama](https://github.com/ShenghaiWang/SwiftLlama)
|
| 192 |
-
- Delphi [Embarcadero/llama-cpp-delphi](https://github.com/Embarcadero/llama-cpp-delphi)
|
| 193 |
-
- Go (no CGo needed): [hybridgroup/yzma](https://github.com/hybridgroup/yzma)
|
| 194 |
-
- Android: [llama.android](/examples/llama.android)
|
| 195 |
-
|
| 196 |
-
</details>
|
| 197 |
-
|
| 198 |
-
<details>
|
| 199 |
-
<summary>UIs</summary>
|
| 200 |
-
|
| 201 |
-
*(to have a project listed here, it should clearly state that it depends on `llama.cpp`)*
|
| 202 |
-
|
| 203 |
-
- [AI Sublime Text plugin](https://github.com/yaroslavyaroslav/OpenAI-sublime-text) (MIT)
|
| 204 |
-
- [BonzAI App](https://apps.apple.com/us/app/bonzai-your-local-ai-agent/id6752847988) (proprietary)
|
| 205 |
-
- [cztomsik/ava](https://github.com/cztomsik/ava) (MIT)
|
| 206 |
-
- [Dot](https://github.com/alexpinel/Dot) (GPL)
|
| 207 |
-
- [eva](https://github.com/ylsdamxssjxxdd/eva) (MIT)
|
| 208 |
-
- [iohub/collama](https://github.com/iohub/coLLaMA) (Apache-2.0)
|
| 209 |
-
- [janhq/jan](https://github.com/janhq/jan) (AGPL)
|
| 210 |
-
- [johnbean393/Sidekick](https://github.com/johnbean393/Sidekick) (MIT)
|
| 211 |
-
- [KanTV](https://github.com/zhouwg/kantv?tab=readme-ov-file) (Apache-2.0)
|
| 212 |
-
- [KodiBot](https://github.com/firatkiral/kodibot) (GPL)
|
| 213 |
-
- [llama.vim](https://github.com/ggml-org/llama.vim) (MIT)
|
| 214 |
-
- [LARS](https://github.com/abgulati/LARS) (AGPL)
|
| 215 |
-
- [Llama Assistant](https://github.com/vietanhdev/llama-assistant) (GPL)
|
| 216 |
-
- [LlamaLib](https://github.com/undreamai/LlamaLib) (Apache-2.0)
|
| 217 |
-
- [LLMFarm](https://github.com/guinmoon/LLMFarm?tab=readme-ov-file) (MIT)
|
| 218 |
-
- [LLMUnity](https://github.com/undreamai/LLMUnity) (MIT)
|
| 219 |
-
- [LMStudio](https://lmstudio.ai/) (proprietary)
|
| 220 |
-
- [LocalAI](https://github.com/mudler/LocalAI) (MIT)
|
| 221 |
-
- [LostRuins/koboldcpp](https://github.com/LostRuins/koboldcpp) (AGPL)
|
| 222 |
-
- [MindMac](https://mindmac.app) (proprietary)
|
| 223 |
-
- [MindWorkAI/AI-Studio](https://github.com/MindWorkAI/AI-Studio) (FSL-1.1-MIT)
|
| 224 |
-
- [Mobile-Artificial-Intelligence/maid](https://github.com/Mobile-Artificial-Intelligence/maid) (MIT)
|
| 225 |
-
- [Mozilla-Ocho/llamafile](https://github.com/Mozilla-Ocho/llamafile) (Apache-2.0)
|
| 226 |
-
- [nat/openplayground](https://github.com/nat/openplayground) (MIT)
|
| 227 |
-
- [nomic-ai/gpt4all](https://github.com/nomic-ai/gpt4all) (MIT)
|
| 228 |
-
- [ollama/ollama](https://github.com/ollama/ollama) (MIT)
|
| 229 |
-
- [oobabooga/text-generation-webui](https://github.com/oobabooga/text-generation-webui) (AGPL)
|
| 230 |
-
- [PocketPal AI](https://github.com/a-ghorbani/pocketpal-ai) (MIT)
|
| 231 |
-
- [psugihara/FreeChat](https://github.com/psugihara/FreeChat) (MIT)
|
| 232 |
-
- [ptsochantaris/emeltal](https://github.com/ptsochantaris/emeltal) (MIT)
|
| 233 |
-
- [pythops/tenere](https://github.com/pythops/tenere) (AGPL)
|
| 234 |
-
- [ramalama](https://github.com/containers/ramalama) (MIT)
|
| 235 |
-
- [semperai/amica](https://github.com/semperai/amica) (MIT)
|
| 236 |
-
- [withcatai/catai](https://github.com/withcatai/catai) (MIT)
|
| 237 |
-
- [Autopen](https://github.com/blackhole89/autopen) (GPL)
|
| 238 |
-
|
| 239 |
-
</details>
|
| 240 |
-
|
| 241 |
-
<details>
|
| 242 |
-
<summary>Tools</summary>
|
| 243 |
-
|
| 244 |
-
- [akx/ggify](https://github.com/akx/ggify) – download PyTorch models from HuggingFace Hub and convert them to GGML
|
| 245 |
-
- [akx/ollama-dl](https://github.com/akx/ollama-dl) – download models from the Ollama library to be used directly with llama.cpp
|
| 246 |
-
- [crashr/gppm](https://github.com/crashr/gppm) – launch llama.cpp instances utilizing NVIDIA Tesla P40 or P100 GPUs with reduced idle power consumption
|
| 247 |
-
- [gpustack/gguf-parser](https://github.com/gpustack/gguf-parser-go/tree/main/cmd/gguf-parser) - review/check the GGUF file and estimate the memory usage
|
| 248 |
-
- [Styled Lines](https://marketplace.unity.com/packages/tools/generative-ai/styled-lines-llama-cpp-model-292902) (proprietary licensed, async wrapper of inference part for game development in Unity3d with pre-built Mobile and Web platform wrappers and a model example)
|
| 249 |
-
- [unslothai/unsloth](https://github.com/unslothai/unsloth) – 🦥 exports/saves fine-tuned and trained models to GGUF (Apache-2.0)
|
| 250 |
-
|
| 251 |
-
</details>
|
| 252 |
-
|
| 253 |
-
<details>
|
| 254 |
-
<summary>Infrastructure</summary>
|
| 255 |
-
|
| 256 |
-
- [Paddler](https://github.com/intentee/paddler) - Open-source LLMOps platform for hosting and scaling AI in your own infrastructure
|
| 257 |
-
- [GPUStack](https://github.com/gpustack/gpustack) - Manage GPU clusters for running LLMs
|
| 258 |
-
- [llama_cpp_canister](https://github.com/onicai/llama_cpp_canister) - llama.cpp as a smart contract on the Internet Computer, using WebAssembly
|
| 259 |
-
- [llama-swap](https://github.com/mostlygeek/llama-swap) - transparent proxy that adds automatic model switching with llama-server
|
| 260 |
-
- [Kalavai](https://github.com/kalavai-net/kalavai-client) - Crowdsource end to end LLM deployment at any scale
|
| 261 |
-
- [llmaz](https://github.com/InftyAI/llmaz) - ☸️ Easy, advanced inference platform for large language models on Kubernetes.
|
| 262 |
-
- [LLMKube](https://github.com/defilantech/llmkube) - Kubernetes operator for llama.cpp with multi-GPU and Apple Silicon Metal
|
| 263 |
-
support"
|
| 264 |
-
</details>
|
| 265 |
-
|
| 266 |
-
<details>
|
| 267 |
-
<summary>Games</summary>
|
| 268 |
-
|
| 269 |
-
- [Lucy's Labyrinth](https://github.com/MorganRO8/Lucys_Labyrinth) - A simple maze game where agents controlled by an AI model will try to trick you.
|
| 270 |
-
|
| 271 |
-
</details>
|
| 272 |
-
|
| 273 |
-
|
| 274 |
-
## Supported backends
|
| 275 |
-
|
| 276 |
-
| Backend | Target devices |
|
| 277 |
-
| --- | --- |
|
| 278 |
-
| [Metal](docs/build.md#metal-build) | Apple Silicon |
|
| 279 |
-
| [BLAS](docs/build.md#blas-build) | All |
|
| 280 |
-
| [BLIS](docs/backend/BLIS.md) | All |
|
| 281 |
-
| [SYCL](docs/backend/SYCL.md) | Intel and Nvidia GPU |
|
| 282 |
-
| [OpenVINO [In Progress]](docs/backend/OPENVINO.md) | Intel CPUs, GPUs, and NPUs |
|
| 283 |
-
| [MUSA](docs/build.md#musa) | Moore Threads GPU |
|
| 284 |
-
| [CUDA](docs/build.md#cuda) | Nvidia GPU |
|
| 285 |
-
| [HIP](docs/build.md#hip) | AMD GPU |
|
| 286 |
-
| [ZenDNN](docs/build.md#zendnn) | AMD CPU |
|
| 287 |
-
| [Vulkan](docs/build.md#vulkan) | GPU |
|
| 288 |
-
| [CANN](docs/build.md#cann) | Ascend NPU |
|
| 289 |
-
| [OpenCL](docs/backend/OPENCL.md) | Adreno GPU |
|
| 290 |
-
| [IBM zDNN](docs/backend/zDNN.md) | IBM Z & LinuxONE |
|
| 291 |
-
| [WebGPU [In Progress]](docs/build.md#webgpu) | All |
|
| 292 |
-
| [RPC](https://github.com/ggml-org/llama.cpp/tree/master/tools/rpc) | All |
|
| 293 |
-
| [Hexagon [In Progress]](docs/backend/snapdragon/README.md) | Snapdragon |
|
| 294 |
-
| [VirtGPU](docs/backend/VirtGPU.md) | VirtGPU APIR |
|
| 295 |
-
|
| 296 |
-
## Obtaining and quantizing models
|
| 297 |
-
|
| 298 |
-
The [Hugging Face](https://huggingface.co) platform hosts a [number of LLMs](https://huggingface.co/models?library=gguf&sort=trending) compatible with `llama.cpp`:
|
| 299 |
-
|
| 300 |
-
- [Trending](https://huggingface.co/models?library=gguf&sort=trending)
|
| 301 |
-
- [LLaMA](https://huggingface.co/models?sort=trending&search=llama+gguf)
|
| 302 |
-
|
| 303 |
-
You can either manually download the GGUF file or directly use any `llama.cpp`-compatible models from [Hugging Face](https://huggingface.co/) or other model hosting sites, such as [ModelScope](https://modelscope.cn/), by using this CLI argument: `-hf <user>/<model>[:quant]`. For example:
|
| 304 |
-
|
| 305 |
-
```sh
|
| 306 |
-
llama-cli -hf ggml-org/gemma-3-1b-it-GGUF
|
| 307 |
-
```
|
| 308 |
-
|
| 309 |
-
By default, the CLI would download from Hugging Face, you can switch to other options with the environment variable `MODEL_ENDPOINT`. For example, you may opt to downloading model checkpoints from ModelScope or other model sharing communities by setting the environment variable, e.g. `MODEL_ENDPOINT=https://www.modelscope.cn/`.
|
| 310 |
-
|
| 311 |
-
After downloading a model, use the CLI tools to run it locally - see below.
|
| 312 |
-
|
| 313 |
-
`llama.cpp` requires the model to be stored in the [GGUF](https://github.com/ggml-org/ggml/blob/master/docs/gguf.md) file format. Models in other data formats can be converted to GGUF using the `convert_*.py` Python scripts in this repo.
|
| 314 |
-
|
| 315 |
-
The Hugging Face platform provides a variety of online tools for converting, quantizing and hosting models with `llama.cpp`:
|
| 316 |
-
|
| 317 |
-
- Use the [GGUF-my-repo space](https://huggingface.co/spaces/ggml-org/gguf-my-repo) to convert to GGUF format and quantize model weights to smaller sizes
|
| 318 |
-
- Use the [GGUF-my-LoRA space](https://huggingface.co/spaces/ggml-org/gguf-my-lora) to convert LoRA adapters to GGUF format (more info: https://github.com/ggml-org/llama.cpp/discussions/10123)
|
| 319 |
-
- Use the [GGUF-editor space](https://huggingface.co/spaces/CISCai/gguf-editor) to edit GGUF meta data in the browser (more info: https://github.com/ggml-org/llama.cpp/discussions/9268)
|
| 320 |
-
- Use the [Inference Endpoints](https://ui.endpoints.huggingface.co/) to directly host `llama.cpp` in the cloud (more info: https://github.com/ggml-org/llama.cpp/discussions/9669)
|
| 321 |
-
|
| 322 |
-
To learn more about model quantization, [read this documentation](tools/quantize/README.md)
|
| 323 |
-
|
| 324 |
-
## [`llama-cli`](tools/cli)
|
| 325 |
-
|
| 326 |
-
#### A CLI tool for accessing and experimenting with most of `llama.cpp`'s functionality.
|
| 327 |
-
|
| 328 |
-
- <details open>
|
| 329 |
-
<summary>Run in conversation mode</summary>
|
| 330 |
-
|
| 331 |
-
Models with a built-in chat template will automatically activate conversation mode. If this doesn't occur, you can manually enable it by adding `-cnv` and specifying a suitable chat template with `--chat-template NAME`
|
| 332 |
-
|
| 333 |
-
```bash
|
| 334 |
-
llama-cli -m model.gguf
|
| 335 |
-
|
| 336 |
-
# > hi, who are you?
|
| 337 |
-
# Hi there! I'm your helpful assistant! I'm an AI-powered chatbot designed to assist and provide information to users like you. I'm here to help answer your questions, provide guidance, and offer support on a wide range of topics. I'm a friendly and knowledgeable AI, and I'm always happy to help with anything you need. What's on your mind, and how can I assist you today?
|
| 338 |
-
#
|
| 339 |
-
# > what is 1+1?
|
| 340 |
-
# Easy peasy! The answer to 1+1 is... 2!
|
| 341 |
-
```
|
| 342 |
-
|
| 343 |
-
</details>
|
| 344 |
-
|
| 345 |
-
- <details>
|
| 346 |
-
<summary>Run in conversation mode with custom chat template</summary>
|
| 347 |
-
|
| 348 |
-
```bash
|
| 349 |
-
# use the "chatml" template (use -h to see the list of supported templates)
|
| 350 |
-
llama-cli -m model.gguf -cnv --chat-template chatml
|
| 351 |
-
|
| 352 |
-
# use a custom template
|
| 353 |
-
llama-cli -m model.gguf -cnv --in-prefix 'User: ' --reverse-prompt 'User:'
|
| 354 |
-
```
|
| 355 |
-
|
| 356 |
-
</details>
|
| 357 |
-
|
| 358 |
-
- <details>
|
| 359 |
-
<summary>Constrain the output with a custom grammar</summary>
|
| 360 |
-
|
| 361 |
-
```bash
|
| 362 |
-
llama-cli -m model.gguf -n 256 --grammar-file grammars/json.gbnf -p 'Request: schedule a call at 8pm; Command:'
|
| 363 |
-
|
| 364 |
-
# {"appointmentTime": "8pm", "appointmentDetails": "schedule a a call"}
|
| 365 |
-
```
|
| 366 |
-
|
| 367 |
-
The [grammars/](grammars/) folder contains a handful of sample grammars. To write your own, check out the [GBNF Guide](grammars/README.md).
|
| 368 |
-
|
| 369 |
-
For authoring more complex JSON grammars, check out https://grammar.intrinsiclabs.ai/
|
| 370 |
-
|
| 371 |
-
</details>
|
| 372 |
-
|
| 373 |
-
|
| 374 |
-
## [`llama-server`](tools/server)
|
| 375 |
-
|
| 376 |
-
#### A lightweight, [OpenAI API](https://github.com/openai/openai-openapi) compatible, HTTP server for serving LLMs.
|
| 377 |
-
|
| 378 |
-
- <details open>
|
| 379 |
-
<summary>Start a local HTTP server with default configuration on port 8080</summary>
|
| 380 |
-
|
| 381 |
-
```bash
|
| 382 |
-
llama-server -m model.gguf --port 8080
|
| 383 |
-
|
| 384 |
-
# Basic web UI can be accessed via browser: http://localhost:8080
|
| 385 |
-
# Chat completion endpoint: http://localhost:8080/v1/chat/completions
|
| 386 |
-
```
|
| 387 |
-
|
| 388 |
-
</details>
|
| 389 |
-
|
| 390 |
-
- <details>
|
| 391 |
-
<summary>Support multiple-users and parallel decoding</summary>
|
| 392 |
-
|
| 393 |
-
```bash
|
| 394 |
-
# up to 4 concurrent requests, each with 4096 max context
|
| 395 |
-
llama-server -m model.gguf -c 16384 -np 4
|
| 396 |
-
```
|
| 397 |
-
|
| 398 |
-
</details>
|
| 399 |
-
|
| 400 |
-
- <details>
|
| 401 |
-
<summary>Enable speculative decoding</summary>
|
| 402 |
-
|
| 403 |
-
```bash
|
| 404 |
-
# the draft.gguf model should be a small variant of the target model.gguf
|
| 405 |
-
llama-server -m model.gguf -md draft.gguf
|
| 406 |
-
```
|
| 407 |
-
|
| 408 |
-
</details>
|
| 409 |
-
|
| 410 |
-
- <details>
|
| 411 |
-
<summary>Serve an embedding model</summary>
|
| 412 |
-
|
| 413 |
-
```bash
|
| 414 |
-
# use the /embedding endpoint
|
| 415 |
-
llama-server -m model.gguf --embedding --pooling cls -ub 8192
|
| 416 |
-
```
|
| 417 |
-
|
| 418 |
-
</details>
|
| 419 |
-
|
| 420 |
-
- <details>
|
| 421 |
-
<summary>Serve a reranking model</summary>
|
| 422 |
-
|
| 423 |
-
```bash
|
| 424 |
-
# use the /reranking endpoint
|
| 425 |
-
llama-server -m model.gguf --reranking
|
| 426 |
-
```
|
| 427 |
-
|
| 428 |
-
</details>
|
| 429 |
-
|
| 430 |
-
- <details>
|
| 431 |
-
<summary>Constrain all outputs with a grammar</summary>
|
| 432 |
-
|
| 433 |
-
```bash
|
| 434 |
-
# custom grammar
|
| 435 |
-
llama-server -m model.gguf --grammar-file grammar.gbnf
|
| 436 |
-
|
| 437 |
-
# JSON
|
| 438 |
-
llama-server -m model.gguf --grammar-file grammars/json.gbnf
|
| 439 |
-
```
|
| 440 |
-
|
| 441 |
-
</details>
|
| 442 |
-
|
| 443 |
-
|
| 444 |
-
## [`llama-perplexity`](tools/perplexity)
|
| 445 |
-
|
| 446 |
-
#### A tool for measuring the [perplexity](tools/perplexity/README.md) [^1] (and other quality metrics) of a model over a given text.
|
| 447 |
-
|
| 448 |
-
- <details open>
|
| 449 |
-
<summary>Measure the perplexity over a text file</summary>
|
| 450 |
-
|
| 451 |
-
```bash
|
| 452 |
-
llama-perplexity -m model.gguf -f file.txt
|
| 453 |
-
|
| 454 |
-
# [1]15.2701,[2]5.4007,[3]5.3073,[4]6.2965,[5]5.8940,[6]5.6096,[7]5.7942,[8]4.9297, ...
|
| 455 |
-
# Final estimate: PPL = 5.4007 +/- 0.67339
|
| 456 |
-
```
|
| 457 |
-
|
| 458 |
-
</details>
|
| 459 |
-
|
| 460 |
-
- <details>
|
| 461 |
-
<summary>Measure KL divergence</summary>
|
| 462 |
-
|
| 463 |
-
```bash
|
| 464 |
-
# TODO
|
| 465 |
-
```
|
| 466 |
-
|
| 467 |
-
</details>
|
| 468 |
-
|
| 469 |
-
[^1]: [https://huggingface.co/docs/transformers/perplexity](https://huggingface.co/docs/transformers/perplexity)
|
| 470 |
-
|
| 471 |
-
## [`llama-bench`](tools/llama-bench)
|
| 472 |
-
|
| 473 |
-
#### Benchmark the performance of the inference for various parameters.
|
| 474 |
-
|
| 475 |
-
- <details open>
|
| 476 |
-
<summary>Run default benchmark</summary>
|
| 477 |
-
|
| 478 |
-
```bash
|
| 479 |
-
llama-bench -m model.gguf
|
| 480 |
-
|
| 481 |
-
# Output:
|
| 482 |
-
# | model | size | params | backend | threads | test | t/s |
|
| 483 |
-
# | ------------------- | ---------: | ---------: | ---------- | ------: | ------------: | -------------------: |
|
| 484 |
-
# | qwen2 1.5B Q4_0 | 885.97 MiB | 1.54 B | Metal,BLAS | 16 | pp512 | 5765.41 ± 20.55 |
|
| 485 |
-
# | qwen2 1.5B Q4_0 | 885.97 MiB | 1.54 B | Metal,BLAS | 16 | tg128 | 197.71 ± 0.81 |
|
| 486 |
-
#
|
| 487 |
-
# build: 3e0ba0e60 (4229)
|
| 488 |
-
```
|
| 489 |
-
|
| 490 |
-
</details>
|
| 491 |
-
|
| 492 |
-
## [`llama-simple`](examples/simple)
|
| 493 |
-
|
| 494 |
-
#### A minimal example for implementing apps with `llama.cpp`. Useful for developers.
|
| 495 |
-
|
| 496 |
-
- <details>
|
| 497 |
-
<summary>Basic text completion</summary>
|
| 498 |
-
|
| 499 |
-
```bash
|
| 500 |
-
llama-simple -m model.gguf
|
| 501 |
-
|
| 502 |
-
# Hello my name is Kaitlyn and I am a 16 year old girl. I am a junior in high school and I am currently taking a class called "The Art of
|
| 503 |
-
```
|
| 504 |
-
|
| 505 |
-
</details>
|
| 506 |
-
|
| 507 |
-
|
| 508 |
-
## Contributing
|
| 509 |
-
|
| 510 |
-
- Contributors can open PRs
|
| 511 |
-
- Collaborators will be invited based on contributions
|
| 512 |
-
- Maintainers can push to branches in the `llama.cpp` repo and merge PRs into the `master` branch
|
| 513 |
-
- Any help with managing issues, PRs and projects is very appreciated!
|
| 514 |
-
- See [good first issues](https://github.com/ggml-org/llama.cpp/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22) for tasks suitable for first contributions
|
| 515 |
-
- Read the [CONTRIBUTING.md](CONTRIBUTING.md) for more information
|
| 516 |
-
- Make sure to read this: [Inference at the edge](https://github.com/ggml-org/llama.cpp/discussions/205)
|
| 517 |
-
- A bit of backstory for those who are interested: [Changelog podcast](https://changelog.com/podcast/532)
|
| 518 |
-
|
| 519 |
-
## Other documentation
|
| 520 |
-
|
| 521 |
-
- [cli](tools/cli/README.md)
|
| 522 |
-
- [completion](tools/completion/README.md)
|
| 523 |
-
- [server](tools/server/README.md)
|
| 524 |
-
- [GBNF grammars](grammars/README.md)
|
| 525 |
-
|
| 526 |
-
#### Development documentation
|
| 527 |
-
|
| 528 |
-
- [How to build](docs/build.md)
|
| 529 |
-
- [Running on Docker](docs/docker.md)
|
| 530 |
-
- [Build on Android](docs/android.md)
|
| 531 |
-
- [Performance troubleshooting](docs/development/token_generation_performance_tips.md)
|
| 532 |
-
- [GGML tips & tricks](https://github.com/ggml-org/llama.cpp/wiki/GGML-Tips-&-Tricks)
|
| 533 |
-
|
| 534 |
-
#### Seminal papers and background on the models
|
| 535 |
-
|
| 536 |
-
If your issue is with model generation quality, then please at least scan the following links and papers to understand the limitations of LLaMA models. This is especially important when choosing an appropriate model size and appreciating both the significant and subtle differences between LLaMA models and ChatGPT:
|
| 537 |
-
- LLaMA:
|
| 538 |
-
- [Introducing LLaMA: A foundational, 65-billion-parameter large language model](https://ai.facebook.com/blog/large-language-model-llama-meta-ai/)
|
| 539 |
-
- [LLaMA: Open and Efficient Foundation Language Models](https://arxiv.org/abs/2302.13971)
|
| 540 |
-
- GPT-3
|
| 541 |
-
- [Language Models are Few-Shot Learners](https://arxiv.org/abs/2005.14165)
|
| 542 |
-
- GPT-3.5 / InstructGPT / ChatGPT:
|
| 543 |
-
- [Aligning language models to follow instructions](https://openai.com/research/instruction-following)
|
| 544 |
-
- [Training language models to follow instructions with human feedback](https://arxiv.org/abs/2203.02155)
|
| 545 |
-
|
| 546 |
-
## XCFramework
|
| 547 |
-
The XCFramework is a precompiled version of the library for iOS, visionOS, tvOS,
|
| 548 |
-
and macOS. It can be used in Swift projects without the need to compile the
|
| 549 |
-
library from source. For example:
|
| 550 |
-
```swift
|
| 551 |
-
// swift-tools-version: 5.10
|
| 552 |
-
// The swift-tools-version declares the minimum version of Swift required to build this package.
|
| 553 |
-
|
| 554 |
-
import PackageDescription
|
| 555 |
-
|
| 556 |
-
let package = Package(
|
| 557 |
-
name: "MyLlamaPackage",
|
| 558 |
-
targets: [
|
| 559 |
-
.executableTarget(
|
| 560 |
-
name: "MyLlamaPackage",
|
| 561 |
-
dependencies: [
|
| 562 |
-
"LlamaFramework"
|
| 563 |
-
]),
|
| 564 |
-
.binaryTarget(
|
| 565 |
-
name: "LlamaFramework",
|
| 566 |
-
url: "https://github.com/ggml-org/llama.cpp/releases/download/b5046/llama-b5046-xcframework.zip",
|
| 567 |
-
checksum: "c19be78b5f00d8d29a25da41042cb7afa094cbf6280a225abe614b03b20029ab"
|
| 568 |
-
)
|
| 569 |
-
]
|
| 570 |
-
)
|
| 571 |
-
```
|
| 572 |
-
The above example is using an intermediate build `b5046` of the library. This can be modified
|
| 573 |
-
to use a different version by changing the URL and checksum.
|
| 574 |
-
|
| 575 |
-
## Completions
|
| 576 |
-
Command-line completion is available for some environments.
|
| 577 |
-
|
| 578 |
-
#### Bash Completion
|
| 579 |
-
```bash
|
| 580 |
-
$ build/bin/llama-cli --completion-bash > ~/.llama-completion.bash
|
| 581 |
-
$ source ~/.llama-completion.bash
|
| 582 |
-
```
|
| 583 |
-
Optionally this can be added to your `.bashrc` or `.bash_profile` to load it
|
| 584 |
-
automatically. For example:
|
| 585 |
-
```console
|
| 586 |
-
$ echo "source ~/.llama-completion.bash" >> ~/.bashrc
|
| 587 |
-
```
|
| 588 |
-
|
| 589 |
-
## Dependencies
|
| 590 |
-
|
| 591 |
-
- [yhirose/cpp-httplib](https://github.com/yhirose/cpp-httplib) - Single-header HTTP server, used by `llama-server` - MIT license
|
| 592 |
-
- [stb-image](https://github.com/nothings/stb) - Single-header image format decoder, used by multimodal subsystem - Public domain
|
| 593 |
-
- [nlohmann/json](https://github.com/nlohmann/json) - Single-header JSON library, used by various tools/examples - MIT License
|
| 594 |
-
- [miniaudio.h](https://github.com/mackron/miniaudio) - Single-header audio format decoder, used by multimodal subsystem - Public domain
|
| 595 |
-
- [subprocess.h](https://github.com/sheredom/subprocess.h) - Single-header process launching solution for C and C++ - Public domain
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cfc44b7ba25614df70e6b65e3341cae0310163bd32fd31a6b928a542df433faf
|
| 3 |
+
size 30786
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ci/run.sh
CHANGED
|
@@ -57,6 +57,13 @@ SRC=`pwd`
|
|
| 57 |
CMAKE_EXTRA="-DLLAMA_FATAL_WARNINGS=${LLAMA_FATAL_WARNINGS:-ON} -DLLAMA_OPENSSL=OFF -DGGML_SCHED_NO_REALLOC=ON"
|
| 58 |
CTEST_EXTRA=""
|
| 59 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 60 |
if [ ! -z ${GG_BUILD_METAL} ]; then
|
| 61 |
CMAKE_EXTRA="${CMAKE_EXTRA} -DGGML_METAL=ON"
|
| 62 |
fi
|
|
@@ -112,6 +119,11 @@ if [ ! -z ${GG_BUILD_VULKAN} ]; then
|
|
| 112 |
CMAKE_EXTRA="${CMAKE_EXTRA} -DGGML_METAL=OFF -DGGML_BLAS=OFF"
|
| 113 |
fi
|
| 114 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 115 |
fi
|
| 116 |
|
| 117 |
if [ ! -z ${GG_BUILD_WEBGPU} ]; then
|
|
@@ -144,35 +156,7 @@ fi
|
|
| 144 |
|
| 145 |
if [ -n "${GG_BUILD_KLEIDIAI}" ]; then
|
| 146 |
echo ">>===== Enabling KleidiAI support"
|
| 147 |
-
|
| 148 |
-
CANDIDATES=(
|
| 149 |
-
"armv9-a+dotprod+i8mm+sve2"
|
| 150 |
-
"armv9-a+dotprod+i8mm"
|
| 151 |
-
"armv8.6-a+dotprod+i8mm"
|
| 152 |
-
"armv8.2-a+dotprod"
|
| 153 |
-
)
|
| 154 |
-
CPU=""
|
| 155 |
-
|
| 156 |
-
for cpu in "${CANDIDATES[@]}"; do
|
| 157 |
-
if echo 'int main(){}' | ${CXX:-c++} -march="$cpu" -x c++ - -c -o /dev/null >/dev/null 2>&1; then
|
| 158 |
-
CPU="$cpu"
|
| 159 |
-
break
|
| 160 |
-
fi
|
| 161 |
-
done
|
| 162 |
-
|
| 163 |
-
if [ -z "$CPU" ]; then
|
| 164 |
-
echo "ERROR: None of the required ARM baselines (armv9/armv8.6/armv8.2 + dotprod) are supported by this compiler."
|
| 165 |
-
exit 1
|
| 166 |
-
fi
|
| 167 |
-
|
| 168 |
-
echo ">>===== Using ARM baseline: ${CPU}"
|
| 169 |
-
|
| 170 |
-
CMAKE_EXTRA="${CMAKE_EXTRA:+$CMAKE_EXTRA } \
|
| 171 |
-
-DGGML_NATIVE=OFF \
|
| 172 |
-
-DGGML_CPU_KLEIDIAI=ON \
|
| 173 |
-
-DGGML_CPU_AARCH64=ON \
|
| 174 |
-
-DGGML_CPU_ARM_ARCH=${CPU} \
|
| 175 |
-
-DBUILD_SHARED_LIBS=OFF"
|
| 176 |
fi
|
| 177 |
|
| 178 |
if [ ! -z ${GG_BUILD_BLAS} ]; then
|
|
@@ -242,13 +226,13 @@ function gg_run_ctest_debug {
|
|
| 242 |
|
| 243 |
set -e
|
| 244 |
|
| 245 |
-
# Check
|
| 246 |
gg_check_build_requirements
|
| 247 |
|
| 248 |
-
(
|
| 249 |
-
(time
|
| 250 |
|
| 251 |
-
(time ctest --output-on-failure -L main -E "test-opt|test-backend-ops" ${CTEST_EXTRA}) 2>&1 | tee -a $OUT/${ci}-ctest.log
|
| 252 |
|
| 253 |
set +e
|
| 254 |
}
|
|
@@ -273,16 +257,16 @@ function gg_run_ctest_release {
|
|
| 273 |
|
| 274 |
set -e
|
| 275 |
|
| 276 |
-
# Check
|
| 277 |
gg_check_build_requirements
|
| 278 |
|
| 279 |
-
(
|
| 280 |
-
(time
|
| 281 |
|
| 282 |
if [ -z ${GG_BUILD_LOW_PERF} ]; then
|
| 283 |
-
(time ctest --output-on-failure -L 'main|python' ${CTEST_EXTRA}) 2>&1 | tee -a $OUT/${ci}-ctest.log
|
| 284 |
else
|
| 285 |
-
(time ctest --output-on-failure -L main -E test-opt ${CTEST_EXTRA}) 2>&1 | tee -a $OUT/${ci}-ctest.log
|
| 286 |
fi
|
| 287 |
|
| 288 |
set +e
|
|
@@ -340,7 +324,7 @@ function gg_run_ctest_with_model_debug {
|
|
| 340 |
cd build-ci-debug
|
| 341 |
set -e
|
| 342 |
|
| 343 |
-
(LLAMACPP_TEST_MODELFILE="$model" time ctest --output-on-failure -L model) 2>&1 | tee -a $OUT/${ci}-ctest.log
|
| 344 |
|
| 345 |
set +e
|
| 346 |
cd ..
|
|
@@ -353,7 +337,7 @@ function gg_run_ctest_with_model_release {
|
|
| 353 |
cd build-ci-release
|
| 354 |
set -e
|
| 355 |
|
| 356 |
-
(LLAMACPP_TEST_MODELFILE="$model" time ctest --output-on-failure -L model) 2>&1 | tee -a $OUT/${ci}-ctest.log
|
| 357 |
|
| 358 |
# test memory leaks
|
| 359 |
#if [[ ! -z ${GG_BUILD_METAL} ]]; then
|
|
@@ -407,8 +391,8 @@ function gg_run_qwen3_0_6b {
|
|
| 407 |
|
| 408 |
set -e
|
| 409 |
|
| 410 |
-
(
|
| 411 |
-
(time
|
| 412 |
|
| 413 |
python3 ../convert_hf_to_gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf --outtype f16
|
| 414 |
python3 ../convert_hf_to_gguf.py ${path_models} --outfile ${path_models}/ggml-model-bf16.gguf --outtype bf16
|
|
@@ -556,8 +540,8 @@ function gg_run_embd_bge_small {
|
|
| 556 |
|
| 557 |
set -e
|
| 558 |
|
| 559 |
-
(
|
| 560 |
-
(time
|
| 561 |
|
| 562 |
python3 ../convert_hf_to_gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf
|
| 563 |
|
|
@@ -601,8 +585,8 @@ function gg_run_rerank_tiny {
|
|
| 601 |
|
| 602 |
set -e
|
| 603 |
|
| 604 |
-
(
|
| 605 |
-
(time
|
| 606 |
|
| 607 |
python3 ../convert_hf_to_gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf
|
| 608 |
|
|
@@ -648,12 +632,36 @@ function gg_sum_rerank_tiny {
|
|
| 648 |
}
|
| 649 |
|
| 650 |
function gg_check_build_requirements {
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 651 |
if ! command -v cmake &> /dev/null; then
|
| 652 |
gg_printf 'cmake not found, please install'
|
| 653 |
fi
|
| 654 |
|
| 655 |
-
if ! command -v
|
| 656 |
-
gg_printf '
|
| 657 |
fi
|
| 658 |
|
| 659 |
if ! command -v ctest &> /dev/null; then
|
|
|
|
| 57 |
CMAKE_EXTRA="-DLLAMA_FATAL_WARNINGS=${LLAMA_FATAL_WARNINGS:-ON} -DLLAMA_OPENSSL=OFF -DGGML_SCHED_NO_REALLOC=ON"
|
| 58 |
CTEST_EXTRA=""
|
| 59 |
|
| 60 |
+
# Default to use make unless specified for compatibility
|
| 61 |
+
CMAKE_GENERATOR="Unix Makefiles"
|
| 62 |
+
|
| 63 |
+
if [ ! -z "${GG_BUILD_NINJA}" ]; then
|
| 64 |
+
CMAKE_GENERATOR="Ninja"
|
| 65 |
+
fi
|
| 66 |
+
|
| 67 |
if [ ! -z ${GG_BUILD_METAL} ]; then
|
| 68 |
CMAKE_EXTRA="${CMAKE_EXTRA} -DGGML_METAL=ON"
|
| 69 |
fi
|
|
|
|
| 119 |
CMAKE_EXTRA="${CMAKE_EXTRA} -DGGML_METAL=OFF -DGGML_BLAS=OFF"
|
| 120 |
fi
|
| 121 |
|
| 122 |
+
# Build shared libs on Windows
|
| 123 |
+
# to reduce binary size and avoid errors in library loading unit tests
|
| 124 |
+
if uname -s | grep -qi nt; then
|
| 125 |
+
CMAKE_EXTRA="${CMAKE_EXTRA} -DBUILD_SHARED_LIBS=ON"
|
| 126 |
+
fi
|
| 127 |
fi
|
| 128 |
|
| 129 |
if [ ! -z ${GG_BUILD_WEBGPU} ]; then
|
|
|
|
| 156 |
|
| 157 |
if [ -n "${GG_BUILD_KLEIDIAI}" ]; then
|
| 158 |
echo ">>===== Enabling KleidiAI support"
|
| 159 |
+
CMAKE_EXTRA="${CMAKE_EXTRA:+$CMAKE_EXTRA } -DGGML_CPU_KLEIDIAI=ON"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 160 |
fi
|
| 161 |
|
| 162 |
if [ ! -z ${GG_BUILD_BLAS} ]; then
|
|
|
|
| 226 |
|
| 227 |
set -e
|
| 228 |
|
| 229 |
+
# Check required binaries are installed
|
| 230 |
gg_check_build_requirements
|
| 231 |
|
| 232 |
+
(cmake -G "${CMAKE_GENERATOR}" -DCMAKE_BUILD_TYPE=Debug ${CMAKE_EXTRA} .. ) 2>&1 | tee -a $OUT/${ci}-cmake.log
|
| 233 |
+
(time cmake --build . --config Debug -j$(nproc)) 2>&1 | tee -a $OUT/${ci}-make.log
|
| 234 |
|
| 235 |
+
(time ctest -C Debug --output-on-failure -L main -E "test-opt|test-backend-ops" ${CTEST_EXTRA}) 2>&1 | tee -a $OUT/${ci}-ctest.log
|
| 236 |
|
| 237 |
set +e
|
| 238 |
}
|
|
|
|
| 257 |
|
| 258 |
set -e
|
| 259 |
|
| 260 |
+
# Check required binaries are installed
|
| 261 |
gg_check_build_requirements
|
| 262 |
|
| 263 |
+
(cmake -G "${CMAKE_GENERATOR}" -DCMAKE_BUILD_TYPE=Release ${CMAKE_EXTRA} .. ) 2>&1 | tee -a $OUT/${ci}-cmake.log
|
| 264 |
+
(time cmake --build . --config Release -j$(nproc)) 2>&1 | tee -a $OUT/${ci}-make.log
|
| 265 |
|
| 266 |
if [ -z ${GG_BUILD_LOW_PERF} ]; then
|
| 267 |
+
(time ctest -C Release --output-on-failure -L 'main|python' ${CTEST_EXTRA}) 2>&1 | tee -a $OUT/${ci}-ctest.log
|
| 268 |
else
|
| 269 |
+
(time ctest -C Release --output-on-failure -L main -E test-opt ${CTEST_EXTRA}) 2>&1 | tee -a $OUT/${ci}-ctest.log
|
| 270 |
fi
|
| 271 |
|
| 272 |
set +e
|
|
|
|
| 324 |
cd build-ci-debug
|
| 325 |
set -e
|
| 326 |
|
| 327 |
+
(LLAMACPP_TEST_MODELFILE="$model" time ctest -C Debug --output-on-failure -L model) 2>&1 | tee -a $OUT/${ci}-ctest.log
|
| 328 |
|
| 329 |
set +e
|
| 330 |
cd ..
|
|
|
|
| 337 |
cd build-ci-release
|
| 338 |
set -e
|
| 339 |
|
| 340 |
+
(LLAMACPP_TEST_MODELFILE="$model" time ctest -C Release --output-on-failure -L model) 2>&1 | tee -a $OUT/${ci}-ctest.log
|
| 341 |
|
| 342 |
# test memory leaks
|
| 343 |
#if [[ ! -z ${GG_BUILD_METAL} ]]; then
|
|
|
|
| 391 |
|
| 392 |
set -e
|
| 393 |
|
| 394 |
+
(cmake -G "${CMAKE_GENERATOR}" -DCMAKE_BUILD_TYPE=Release ${CMAKE_EXTRA} .. ) 2>&1 | tee -a $OUT/${ci}-cmake.log
|
| 395 |
+
(time cmake --build . --config Release -j$(nproc)) 2>&1 | tee -a $OUT/${ci}-make.log
|
| 396 |
|
| 397 |
python3 ../convert_hf_to_gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf --outtype f16
|
| 398 |
python3 ../convert_hf_to_gguf.py ${path_models} --outfile ${path_models}/ggml-model-bf16.gguf --outtype bf16
|
|
|
|
| 540 |
|
| 541 |
set -e
|
| 542 |
|
| 543 |
+
(cmake -G "${CMAKE_GENERATOR}" -DCMAKE_BUILD_TYPE=Release ${CMAKE_EXTRA} .. ) 2>&1 | tee -a $OUT/${ci}-cmake.log
|
| 544 |
+
(time cmake --build . --config Release -j$(nproc)) 2>&1 | tee -a $OUT/${ci}-make.log
|
| 545 |
|
| 546 |
python3 ../convert_hf_to_gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf
|
| 547 |
|
|
|
|
| 585 |
|
| 586 |
set -e
|
| 587 |
|
| 588 |
+
(cmake -G "${CMAKE_GENERATOR}" -DCMAKE_BUILD_TYPE=Release ${CMAKE_EXTRA} .. ) 2>&1 | tee -a $OUT/${ci}-cmake.log
|
| 589 |
+
(time cmake --build . --config Release -j$(nproc)) 2>&1 | tee -a $OUT/${ci}-make.log
|
| 590 |
|
| 591 |
python3 ../convert_hf_to_gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf
|
| 592 |
|
|
|
|
| 632 |
}
|
| 633 |
|
| 634 |
function gg_check_build_requirements {
|
| 635 |
+
if ! command -v git &> /dev/null; then
|
| 636 |
+
gg_printf 'git not found, please install'
|
| 637 |
+
fi
|
| 638 |
+
|
| 639 |
+
if ! command -v git-lfs &> /dev/null; then
|
| 640 |
+
gg_printf 'git-lfs not found, please install'
|
| 641 |
+
fi
|
| 642 |
+
|
| 643 |
+
if ! command -v wget &> /dev/null; then
|
| 644 |
+
gg_printf 'wget not found, please install'
|
| 645 |
+
fi
|
| 646 |
+
|
| 647 |
+
if ! command -v python3 &> /dev/null; then
|
| 648 |
+
gg_printf 'python3 not found, please install'
|
| 649 |
+
fi
|
| 650 |
+
|
| 651 |
+
if ! command -v pip3 &> /dev/null; then
|
| 652 |
+
gg_printf 'pip3 not found, please install'
|
| 653 |
+
fi
|
| 654 |
+
|
| 655 |
+
if ! python3 -m ensurepip --help &> /dev/null; then
|
| 656 |
+
gg_printf 'ensurepip not found, please install python3-venv package'
|
| 657 |
+
fi
|
| 658 |
+
|
| 659 |
if ! command -v cmake &> /dev/null; then
|
| 660 |
gg_printf 'cmake not found, please install'
|
| 661 |
fi
|
| 662 |
|
| 663 |
+
if ! command -v ccache &> /dev/null; then
|
| 664 |
+
gg_printf 'ccache not found, please consider installing for faster builds'
|
| 665 |
fi
|
| 666 |
|
| 667 |
if ! command -v ctest &> /dev/null; then
|
cmake/arm64-linux-clang.cmake
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
set( CMAKE_SYSTEM_NAME Linux )
|
| 2 |
+
set( CMAKE_SYSTEM_PROCESSOR arm64 )
|
| 3 |
+
|
| 4 |
+
set( target aarch64-linux-gnu )
|
| 5 |
+
|
| 6 |
+
set( CMAKE_C_COMPILER clang )
|
| 7 |
+
set( CMAKE_CXX_COMPILER clang++ )
|
| 8 |
+
|
| 9 |
+
set( CMAKE_C_COMPILER_TARGET ${target} )
|
| 10 |
+
set( CMAKE_CXX_COMPILER_TARGET ${target} )
|
| 11 |
+
|
| 12 |
+
set( arch_c_flags "-march=armv8.7-a -fvectorize -ffp-model=fast -fno-finite-math-only" )
|
| 13 |
+
set( warn_c_flags "-Wno-format -Wno-unused-variable -Wno-unused-function -Wno-gnu-zero-variadic-macro-arguments" )
|
| 14 |
+
|
| 15 |
+
set( CMAKE_C_FLAGS_INIT "${arch_c_flags} ${warn_c_flags}" )
|
| 16 |
+
set( CMAKE_CXX_FLAGS_INIT "${arch_c_flags} ${warn_c_flags}" )
|
| 17 |
+
|
common/CMakeLists.txt
CHANGED
|
@@ -1,151 +1,3 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
llama_add_compile_flags()
|
| 6 |
-
|
| 7 |
-
# Build info header
|
| 8 |
-
|
| 9 |
-
if(EXISTS "${PROJECT_SOURCE_DIR}/.git")
|
| 10 |
-
set(GIT_DIR "${PROJECT_SOURCE_DIR}/.git")
|
| 11 |
-
|
| 12 |
-
# Is git submodule
|
| 13 |
-
if(NOT IS_DIRECTORY "${GIT_DIR}")
|
| 14 |
-
file(READ ${GIT_DIR} REAL_GIT_DIR_LINK)
|
| 15 |
-
string(REGEX REPLACE "gitdir: (.*)\n$" "\\1" REAL_GIT_DIR ${REAL_GIT_DIR_LINK})
|
| 16 |
-
string(FIND "${REAL_GIT_DIR}" "/" SLASH_POS)
|
| 17 |
-
if (SLASH_POS EQUAL 0)
|
| 18 |
-
set(GIT_DIR "${REAL_GIT_DIR}")
|
| 19 |
-
else()
|
| 20 |
-
set(GIT_DIR "${PROJECT_SOURCE_DIR}/${REAL_GIT_DIR}")
|
| 21 |
-
endif()
|
| 22 |
-
endif()
|
| 23 |
-
|
| 24 |
-
if(EXISTS "${GIT_DIR}/index")
|
| 25 |
-
# For build-info.cpp below
|
| 26 |
-
set_property(DIRECTORY APPEND PROPERTY CMAKE_CONFIGURE_DEPENDS "${GIT_DIR}/index")
|
| 27 |
-
else()
|
| 28 |
-
message(WARNING "Git index not found in git repository.")
|
| 29 |
-
endif()
|
| 30 |
-
else()
|
| 31 |
-
message(WARNING "Git repository not found; to enable automatic generation of build info, make sure Git is installed and the project is a Git repository.")
|
| 32 |
-
endif()
|
| 33 |
-
|
| 34 |
-
set(TEMPLATE_FILE "${CMAKE_CURRENT_SOURCE_DIR}/build-info.cpp.in")
|
| 35 |
-
set(OUTPUT_FILE "${CMAKE_CURRENT_BINARY_DIR}/build-info.cpp")
|
| 36 |
-
configure_file(${TEMPLATE_FILE} ${OUTPUT_FILE})
|
| 37 |
-
|
| 38 |
-
set(TARGET build_info)
|
| 39 |
-
add_library(${TARGET} OBJECT ${OUTPUT_FILE})
|
| 40 |
-
if (BUILD_SHARED_LIBS)
|
| 41 |
-
set_target_properties(${TARGET} PROPERTIES POSITION_INDEPENDENT_CODE ON)
|
| 42 |
-
endif()
|
| 43 |
-
|
| 44 |
-
set(TARGET common)
|
| 45 |
-
|
| 46 |
-
add_library(${TARGET} STATIC
|
| 47 |
-
arg.cpp
|
| 48 |
-
arg.h
|
| 49 |
-
base64.hpp
|
| 50 |
-
chat-auto-parser-generator.cpp
|
| 51 |
-
chat-auto-parser-helpers.cpp
|
| 52 |
-
chat-auto-parser.h
|
| 53 |
-
chat-diff-analyzer.cpp
|
| 54 |
-
chat-peg-parser.cpp
|
| 55 |
-
chat-peg-parser.h
|
| 56 |
-
chat.cpp
|
| 57 |
-
chat.h
|
| 58 |
-
common.cpp
|
| 59 |
-
common.h
|
| 60 |
-
console.cpp
|
| 61 |
-
console.h
|
| 62 |
-
debug.cpp
|
| 63 |
-
debug.h
|
| 64 |
-
download.cpp
|
| 65 |
-
download.h
|
| 66 |
-
http.h
|
| 67 |
-
json-partial.cpp
|
| 68 |
-
json-partial.h
|
| 69 |
-
json-schema-to-grammar.cpp
|
| 70 |
-
llguidance.cpp
|
| 71 |
-
log.cpp
|
| 72 |
-
log.h
|
| 73 |
-
ngram-cache.cpp
|
| 74 |
-
ngram-cache.h
|
| 75 |
-
ngram-map.cpp
|
| 76 |
-
ngram-map.h
|
| 77 |
-
ngram-mod.cpp
|
| 78 |
-
ngram-mod.h
|
| 79 |
-
peg-parser.cpp
|
| 80 |
-
peg-parser.h
|
| 81 |
-
preset.cpp
|
| 82 |
-
preset.h
|
| 83 |
-
regex-partial.cpp
|
| 84 |
-
reasoning-budget.cpp
|
| 85 |
-
reasoning-budget.h
|
| 86 |
-
regex-partial.h
|
| 87 |
-
sampling.cpp
|
| 88 |
-
sampling.h
|
| 89 |
-
speculative.cpp
|
| 90 |
-
speculative.h
|
| 91 |
-
unicode.cpp
|
| 92 |
-
unicode.h
|
| 93 |
-
jinja/lexer.cpp
|
| 94 |
-
jinja/lexer.h
|
| 95 |
-
jinja/parser.cpp
|
| 96 |
-
jinja/parser.h
|
| 97 |
-
jinja/runtime.cpp
|
| 98 |
-
jinja/runtime.h
|
| 99 |
-
jinja/value.cpp
|
| 100 |
-
jinja/value.h
|
| 101 |
-
jinja/string.cpp
|
| 102 |
-
jinja/string.h
|
| 103 |
-
jinja/caps.cpp
|
| 104 |
-
jinja/caps.h
|
| 105 |
-
)
|
| 106 |
-
|
| 107 |
-
target_include_directories(${TARGET} PUBLIC . ../vendor)
|
| 108 |
-
target_compile_features (${TARGET} PUBLIC cxx_std_17)
|
| 109 |
-
|
| 110 |
-
if (BUILD_SHARED_LIBS)
|
| 111 |
-
set_target_properties(${TARGET} PROPERTIES POSITION_INDEPENDENT_CODE ON)
|
| 112 |
-
endif()
|
| 113 |
-
|
| 114 |
-
target_link_libraries(${TARGET} PRIVATE
|
| 115 |
-
build_info
|
| 116 |
-
cpp-httplib
|
| 117 |
-
)
|
| 118 |
-
|
| 119 |
-
if (LLAMA_LLGUIDANCE)
|
| 120 |
-
include(ExternalProject)
|
| 121 |
-
set(LLGUIDANCE_SRC ${CMAKE_BINARY_DIR}/llguidance/source)
|
| 122 |
-
set(LLGUIDANCE_PATH ${LLGUIDANCE_SRC}/target/release)
|
| 123 |
-
set(LLGUIDANCE_LIB_NAME "${CMAKE_STATIC_LIBRARY_PREFIX}llguidance${CMAKE_STATIC_LIBRARY_SUFFIX}")
|
| 124 |
-
|
| 125 |
-
ExternalProject_Add(llguidance_ext
|
| 126 |
-
GIT_REPOSITORY https://github.com/guidance-ai/llguidance
|
| 127 |
-
# v1.0.1:
|
| 128 |
-
GIT_TAG d795912fedc7d393de740177ea9ea761e7905774
|
| 129 |
-
PREFIX ${CMAKE_BINARY_DIR}/llguidance
|
| 130 |
-
SOURCE_DIR ${LLGUIDANCE_SRC}
|
| 131 |
-
BUILD_IN_SOURCE TRUE
|
| 132 |
-
CONFIGURE_COMMAND ""
|
| 133 |
-
BUILD_COMMAND cargo build --release --package llguidance
|
| 134 |
-
INSTALL_COMMAND ""
|
| 135 |
-
BUILD_BYPRODUCTS ${LLGUIDANCE_PATH}/${LLGUIDANCE_LIB_NAME} ${LLGUIDANCE_PATH}/llguidance.h
|
| 136 |
-
UPDATE_COMMAND ""
|
| 137 |
-
)
|
| 138 |
-
target_compile_definitions(${TARGET} PUBLIC LLAMA_USE_LLGUIDANCE)
|
| 139 |
-
|
| 140 |
-
add_library(llguidance STATIC IMPORTED)
|
| 141 |
-
set_target_properties(llguidance PROPERTIES IMPORTED_LOCATION ${LLGUIDANCE_PATH}/${LLGUIDANCE_LIB_NAME})
|
| 142 |
-
add_dependencies(llguidance llguidance_ext)
|
| 143 |
-
|
| 144 |
-
target_include_directories(${TARGET} PRIVATE ${LLGUIDANCE_PATH})
|
| 145 |
-
target_link_libraries(${TARGET} PRIVATE llguidance)
|
| 146 |
-
if (WIN32)
|
| 147 |
-
target_link_libraries(${TARGET} PRIVATE ws2_32 userenv ntdll bcrypt)
|
| 148 |
-
endif()
|
| 149 |
-
endif()
|
| 150 |
-
|
| 151 |
-
target_link_libraries(${TARGET} PUBLIC llama Threads::Threads)
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a4fe95e36bff8944120f212aaeef45c24c483038ddc30bffc5694102fdff30f0
|
| 3 |
+
size 4864
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
common/arg.cpp
CHANGED
|
@@ -1,8 +1,10 @@
|
|
| 1 |
#include "arg.h"
|
| 2 |
|
|
|
|
| 3 |
#include "chat.h"
|
| 4 |
#include "common.h"
|
| 5 |
#include "download.h"
|
|
|
|
| 6 |
#include "json-schema-to-grammar.h"
|
| 7 |
#include "log.h"
|
| 8 |
#include "sampling.h"
|
|
@@ -290,14 +292,16 @@ static bool common_params_handle_remote_preset(common_params & params, llama_exa
|
|
| 290 |
hf_tag = "default";
|
| 291 |
}
|
| 292 |
|
| 293 |
-
|
| 294 |
-
std::string model_endpoint = get_model_endpoint();
|
| 295 |
auto preset_url = model_endpoint + hf_repo + "/resolve/main/preset.ini";
|
| 296 |
|
| 297 |
// prepare local path for caching
|
| 298 |
auto preset_fname = clean_file_name(hf_repo + "_preset.ini");
|
| 299 |
auto preset_path = fs_get_cache_file(preset_fname);
|
| 300 |
-
|
|
|
|
|
|
|
|
|
|
| 301 |
const bool has_preset = status >= 200 && status < 400;
|
| 302 |
|
| 303 |
// remote preset is optional, so we don't error out if not found
|
|
@@ -326,60 +330,49 @@ struct handle_model_result {
|
|
| 326 |
common_params_model mmproj;
|
| 327 |
};
|
| 328 |
|
| 329 |
-
static handle_model_result common_params_handle_model(
|
| 330 |
-
|
| 331 |
-
|
| 332 |
-
bool offline) {
|
| 333 |
handle_model_result result;
|
| 334 |
-
// handle pre-fill default model path and url based on hf_repo and hf_file
|
| 335 |
-
{
|
| 336 |
-
if (!model.docker_repo.empty()) { // Handle Docker URLs by resolving them to local paths
|
| 337 |
-
model.path = common_docker_resolve_model(model.docker_repo);
|
| 338 |
-
model.name = model.docker_repo; // set name for consistency
|
| 339 |
-
} else if (!model.hf_repo.empty()) {
|
| 340 |
-
// short-hand to avoid specifying --hf-file -> default it to --model
|
| 341 |
-
if (model.hf_file.empty()) {
|
| 342 |
-
if (model.path.empty()) {
|
| 343 |
-
auto auto_detected = common_get_hf_file(model.hf_repo, bearer_token, offline);
|
| 344 |
-
if (auto_detected.repo.empty() || auto_detected.ggufFile.empty()) {
|
| 345 |
-
exit(1); // error message already printed
|
| 346 |
-
}
|
| 347 |
-
model.name = model.hf_repo; // repo name with tag
|
| 348 |
-
model.hf_repo = auto_detected.repo; // repo name without tag
|
| 349 |
-
model.hf_file = auto_detected.ggufFile;
|
| 350 |
-
if (!auto_detected.mmprojFile.empty()) {
|
| 351 |
-
result.found_mmproj = true;
|
| 352 |
-
result.mmproj.hf_repo = model.hf_repo;
|
| 353 |
-
result.mmproj.hf_file = auto_detected.mmprojFile;
|
| 354 |
-
}
|
| 355 |
-
} else {
|
| 356 |
-
model.hf_file = model.path;
|
| 357 |
-
}
|
| 358 |
-
}
|
| 359 |
|
| 360 |
-
|
| 361 |
-
|
| 362 |
-
|
| 363 |
-
|
| 364 |
-
|
| 365 |
-
|
| 366 |
-
|
| 367 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 368 |
|
| 369 |
-
|
| 370 |
-
|
| 371 |
-
auto f = string_split<std::string>(model.url, '#').front();
|
| 372 |
-
f = string_split<std::string>(f, '?').front();
|
| 373 |
-
model.path = fs_get_cache_file(string_split<std::string>(f, '/').back());
|
| 374 |
-
}
|
| 375 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 376 |
}
|
| 377 |
-
}
|
| 378 |
|
| 379 |
-
|
| 380 |
-
|
| 381 |
-
|
| 382 |
-
|
|
|
|
| 383 |
LOG_ERR("error: failed to download model from %s\n", model.url.c_str());
|
| 384 |
exit(1);
|
| 385 |
}
|
|
@@ -434,6 +427,9 @@ static bool parse_bool_value(const std::string & value) {
|
|
| 434 |
static bool common_params_parse_ex(int argc, char ** argv, common_params_context & ctx_arg) {
|
| 435 |
common_params & params = ctx_arg.params;
|
| 436 |
|
|
|
|
|
|
|
|
|
|
| 437 |
std::unordered_map<std::string, std::pair<common_arg *, bool>> arg_to_options;
|
| 438 |
for (auto & opt : ctx_arg.options) {
|
| 439 |
for (const auto & arg : opt.args) {
|
|
@@ -539,8 +535,17 @@ static bool common_params_parse_ex(int argc, char ** argv, common_params_context
|
|
| 539 |
// parse the first time to get -hf option (used for remote preset)
|
| 540 |
parse_cli_args();
|
| 541 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 542 |
// maybe handle remote preset
|
| 543 |
-
if (!params.model.hf_repo.empty()) {
|
| 544 |
std::string cli_hf_repo = params.model.hf_repo;
|
| 545 |
bool has_preset = common_params_handle_remote_preset(params, ctx_arg.ex);
|
| 546 |
|
|
@@ -571,7 +576,7 @@ static bool common_params_parse_ex(int argc, char ** argv, common_params_context
|
|
| 571 |
}
|
| 572 |
|
| 573 |
// handle model and download
|
| 574 |
-
{
|
| 575 |
auto res = common_params_handle_model(params.model, params.hf_token, params.offline);
|
| 576 |
if (params.no_mmproj) {
|
| 577 |
params.mmproj = {};
|
|
@@ -592,7 +597,7 @@ static bool common_params_parse_ex(int argc, char ** argv, common_params_context
|
|
| 592 |
|
| 593 |
// model is required (except for server)
|
| 594 |
// TODO @ngxson : maybe show a list of available models in CLI in this case
|
| 595 |
-
if (params.model.path.empty() && ctx_arg.ex != LLAMA_EXAMPLE_SERVER && !params.usage && !params.completion) {
|
| 596 |
throw std::invalid_argument("error: --model is required\n");
|
| 597 |
}
|
| 598 |
|
|
@@ -635,8 +640,6 @@ static bool common_params_parse_ex(int argc, char ** argv, common_params_context
|
|
| 635 |
));
|
| 636 |
}
|
| 637 |
|
| 638 |
-
common_log_set_verbosity_thold(params.verbosity);
|
| 639 |
-
|
| 640 |
return true;
|
| 641 |
}
|
| 642 |
|
|
@@ -1042,8 +1045,8 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
|
|
| 1042 |
{"--version"},
|
| 1043 |
"show version and build info",
|
| 1044 |
[](common_params &) {
|
| 1045 |
-
fprintf(stderr, "version: %d (%s)\n",
|
| 1046 |
-
fprintf(stderr, "built with %s for %s\n",
|
| 1047 |
exit(0);
|
| 1048 |
}
|
| 1049 |
));
|
|
@@ -1061,12 +1064,10 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
|
|
| 1061 |
{"-cl", "--cache-list"},
|
| 1062 |
"show list of models in cache",
|
| 1063 |
[](common_params &) {
|
| 1064 |
-
printf("model cache directory: %s\n", fs_get_cache_directory().c_str());
|
| 1065 |
auto models = common_list_cached_models();
|
| 1066 |
printf("number of models in cache: %zu\n", models.size());
|
| 1067 |
for (size_t i = 0; i < models.size(); i++) {
|
| 1068 |
-
|
| 1069 |
-
printf("%4d. %s\n", (int) i + 1, model.to_string().c_str());
|
| 1070 |
}
|
| 1071 |
exit(0);
|
| 1072 |
}
|
|
@@ -1084,7 +1085,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
|
|
| 1084 |
[](common_params & params) {
|
| 1085 |
params.verbose_prompt = true;
|
| 1086 |
}
|
| 1087 |
-
));
|
| 1088 |
add_opt(common_arg(
|
| 1089 |
{"--display-prompt"},
|
| 1090 |
{"--no-display-prompt"},
|
|
@@ -1314,6 +1315,14 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
|
|
| 1314 |
params.kv_unified = value;
|
| 1315 |
}
|
| 1316 |
).set_env("LLAMA_ARG_KV_UNIFIED").set_examples({LLAMA_EXAMPLE_SERVER, LLAMA_EXAMPLE_PERPLEXITY, LLAMA_EXAMPLE_BATCHED, LLAMA_EXAMPLE_BENCH, LLAMA_EXAMPLE_PARALLEL}));
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1317 |
add_opt(common_arg(
|
| 1318 |
{"--context-shift"},
|
| 1319 |
{"--no-context-shift"},
|
|
@@ -1830,23 +1839,23 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
|
|
| 1830 |
).set_sparam());
|
| 1831 |
add_opt(common_arg(
|
| 1832 |
{"--grammar"}, "GRAMMAR",
|
| 1833 |
-
|
| 1834 |
[](common_params & params, const std::string & value) {
|
| 1835 |
-
params.sampling.grammar = value;
|
| 1836 |
}
|
| 1837 |
).set_sparam());
|
| 1838 |
add_opt(common_arg(
|
| 1839 |
{"--grammar-file"}, "FNAME",
|
| 1840 |
"file to read grammar from",
|
| 1841 |
[](common_params & params, const std::string & value) {
|
| 1842 |
-
params.sampling.grammar = read_file(value);
|
| 1843 |
}
|
| 1844 |
).set_sparam());
|
| 1845 |
add_opt(common_arg(
|
| 1846 |
{"-j", "--json-schema"}, "SCHEMA",
|
| 1847 |
"JSON schema to constrain generations (https://json-schema.org/), e.g. `{}` for any JSON object\nFor schemas w/ external $refs, use --grammar + example/json_schema_to_grammar.py instead",
|
| 1848 |
[](common_params & params, const std::string & value) {
|
| 1849 |
-
params.sampling.grammar = json_schema_to_grammar(json::parse(value));
|
| 1850 |
}
|
| 1851 |
).set_sparam());
|
| 1852 |
add_opt(common_arg(
|
|
@@ -1863,7 +1872,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
|
|
| 1863 |
std::istreambuf_iterator<char>(),
|
| 1864 |
std::back_inserter(schema)
|
| 1865 |
);
|
| 1866 |
-
params.sampling.grammar = json_schema_to_grammar(json::parse(schema));
|
| 1867 |
}
|
| 1868 |
).set_sparam());
|
| 1869 |
add_opt(common_arg(
|
|
@@ -2343,19 +2352,21 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
|
|
| 2343 |
}
|
| 2344 |
).set_env("LLAMA_ARG_N_GPU_LAYERS"));
|
| 2345 |
add_opt(common_arg(
|
| 2346 |
-
{"-sm", "--split-mode"}, "{none,layer,row}",
|
| 2347 |
"how to split the model across multiple GPUs, one of:\n"
|
| 2348 |
"- none: use one GPU only\n"
|
| 2349 |
-
"- layer (default): split layers and KV across GPUs\n"
|
| 2350 |
-
"- row: split
|
|
|
|
| 2351 |
[](common_params & params, const std::string & value) {
|
| 2352 |
-
|
| 2353 |
-
if (arg_next == "none") {
|
| 2354 |
params.split_mode = LLAMA_SPLIT_MODE_NONE;
|
| 2355 |
-
} else if (
|
| 2356 |
params.split_mode = LLAMA_SPLIT_MODE_LAYER;
|
| 2357 |
-
} else if (
|
| 2358 |
params.split_mode = LLAMA_SPLIT_MODE_ROW;
|
|
|
|
|
|
|
| 2359 |
} else {
|
| 2360 |
throw std::invalid_argument("invalid value");
|
| 2361 |
}
|
|
@@ -2415,6 +2426,20 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
|
|
| 2415 |
}
|
| 2416 |
}
|
| 2417 |
).set_env("LLAMA_ARG_FIT"));
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2418 |
add_opt(common_arg(
|
| 2419 |
{ "-fitt", "--fit-target" }, "MiB0,MiB1,MiB2,...",
|
| 2420 |
string_format("target margin per device for --fit, comma-separated list of values, "
|
|
@@ -2583,7 +2608,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
|
|
| 2583 |
{"-hf", "-hfr", "--hf-repo"}, "<user>/<model>[:quant]",
|
| 2584 |
"Hugging Face model repository; quant is optional, case-insensitive, default to Q4_K_M, or falls back to the first file in the repo if Q4_K_M doesn't exist.\n"
|
| 2585 |
"mmproj is also downloaded automatically if available. to disable, add --no-mmproj\n"
|
| 2586 |
-
"example:
|
| 2587 |
"(default: unused)",
|
| 2588 |
[](common_params & params, const std::string & value) {
|
| 2589 |
params.model.hf_repo = value;
|
|
@@ -2812,6 +2837,13 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
|
|
| 2812 |
params.port = value;
|
| 2813 |
}
|
| 2814 |
).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_PORT"));
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2815 |
add_opt(common_arg(
|
| 2816 |
{"--path"}, "PATH",
|
| 2817 |
string_format("path to serve static files from (default: %s)", params.public_path.c_str()),
|
|
@@ -2848,6 +2880,15 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
|
|
| 2848 |
params.webui_mcp_proxy = value;
|
| 2849 |
}
|
| 2850 |
).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_WEBUI_MCP_PROXY"));
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2851 |
add_opt(common_arg(
|
| 2852 |
{"--webui"},
|
| 2853 |
{"--no-webui"},
|
|
@@ -3250,6 +3291,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
|
|
| 3250 |
"Set verbosity level to infinity (i.e. log all messages, useful for debugging)",
|
| 3251 |
[](common_params & params) {
|
| 3252 |
params.verbosity = INT_MAX;
|
|
|
|
| 3253 |
}
|
| 3254 |
));
|
| 3255 |
add_opt(common_arg(
|
|
@@ -3270,6 +3312,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
|
|
| 3270 |
"(default: %d)\n", params.verbosity),
|
| 3271 |
[](common_params & params, int value) {
|
| 3272 |
params.verbosity = value;
|
|
|
|
| 3273 |
}
|
| 3274 |
).set_env("LLAMA_LOG_VERBOSITY"));
|
| 3275 |
add_opt(common_arg(
|
|
@@ -3494,7 +3537,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
|
|
| 3494 |
throw std::invalid_argument("unknown speculative decoding type without draft model");
|
| 3495 |
}
|
| 3496 |
}
|
| 3497 |
-
).set_examples({LLAMA_EXAMPLE_SERVER}));
|
| 3498 |
add_opt(common_arg(
|
| 3499 |
{"--spec-ngram-size-n"}, "N",
|
| 3500 |
string_format("ngram size N for ngram-simple/ngram-map speculative decoding, length of lookup n-gram (default: %d)", params.speculative.ngram_size_n),
|
|
|
|
| 1 |
#include "arg.h"
|
| 2 |
|
| 3 |
+
#include "build-info.h"
|
| 4 |
#include "chat.h"
|
| 5 |
#include "common.h"
|
| 6 |
#include "download.h"
|
| 7 |
+
#include "hf-cache.h"
|
| 8 |
#include "json-schema-to-grammar.h"
|
| 9 |
#include "log.h"
|
| 10 |
#include "sampling.h"
|
|
|
|
| 292 |
hf_tag = "default";
|
| 293 |
}
|
| 294 |
|
| 295 |
+
std::string model_endpoint = common_get_model_endpoint();
|
|
|
|
| 296 |
auto preset_url = model_endpoint + hf_repo + "/resolve/main/preset.ini";
|
| 297 |
|
| 298 |
// prepare local path for caching
|
| 299 |
auto preset_fname = clean_file_name(hf_repo + "_preset.ini");
|
| 300 |
auto preset_path = fs_get_cache_file(preset_fname);
|
| 301 |
+
common_download_opts opts;
|
| 302 |
+
opts.bearer_token = params.hf_token;
|
| 303 |
+
opts.offline = params.offline;
|
| 304 |
+
const int status = common_download_file_single(preset_url, preset_path, opts);
|
| 305 |
const bool has_preset = status >= 200 && status < 400;
|
| 306 |
|
| 307 |
// remote preset is optional, so we don't error out if not found
|
|
|
|
| 330 |
common_params_model mmproj;
|
| 331 |
};
|
| 332 |
|
| 333 |
+
static handle_model_result common_params_handle_model(struct common_params_model & model,
|
| 334 |
+
const std::string & bearer_token,
|
| 335 |
+
bool offline) {
|
|
|
|
| 336 |
handle_model_result result;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 337 |
|
| 338 |
+
if (!model.docker_repo.empty()) {
|
| 339 |
+
model.path = common_docker_resolve_model(model.docker_repo);
|
| 340 |
+
model.name = model.docker_repo;
|
| 341 |
+
} else if (!model.hf_repo.empty()) {
|
| 342 |
+
// If -m was used with -hf, treat the model "path" as the hf_file to download
|
| 343 |
+
if (model.hf_file.empty() && !model.path.empty()) {
|
| 344 |
+
model.hf_file = model.path;
|
| 345 |
+
model.path = "";
|
| 346 |
+
}
|
| 347 |
+
common_download_opts opts;
|
| 348 |
+
opts.bearer_token = bearer_token;
|
| 349 |
+
opts.offline = offline;
|
| 350 |
+
auto download_result = common_download_model(model, opts, true);
|
| 351 |
+
|
| 352 |
+
if (download_result.model_path.empty()) {
|
| 353 |
+
LOG_ERR("error: failed to download model from Hugging Face\n");
|
| 354 |
+
exit(1);
|
| 355 |
+
}
|
| 356 |
|
| 357 |
+
model.name = model.hf_repo;
|
| 358 |
+
model.path = download_result.model_path;
|
|
|
|
|
|
|
|
|
|
|
|
|
| 359 |
|
| 360 |
+
if (!download_result.mmproj_path.empty()) {
|
| 361 |
+
result.found_mmproj = true;
|
| 362 |
+
result.mmproj.path = download_result.mmproj_path;
|
| 363 |
+
}
|
| 364 |
+
} else if (!model.url.empty()) {
|
| 365 |
+
if (model.path.empty()) {
|
| 366 |
+
auto f = string_split<std::string>(model.url, '#').front();
|
| 367 |
+
f = string_split<std::string>(f, '?').front();
|
| 368 |
+
model.path = fs_get_cache_file(string_split<std::string>(f, '/').back());
|
| 369 |
}
|
|
|
|
| 370 |
|
| 371 |
+
common_download_opts opts;
|
| 372 |
+
opts.bearer_token = bearer_token;
|
| 373 |
+
opts.offline = offline;
|
| 374 |
+
auto download_result = common_download_model(model, opts);
|
| 375 |
+
if (download_result.model_path.empty()) {
|
| 376 |
LOG_ERR("error: failed to download model from %s\n", model.url.c_str());
|
| 377 |
exit(1);
|
| 378 |
}
|
|
|
|
| 427 |
static bool common_params_parse_ex(int argc, char ** argv, common_params_context & ctx_arg) {
|
| 428 |
common_params & params = ctx_arg.params;
|
| 429 |
|
| 430 |
+
// setup log directly from params.verbosity: see tools/cli/cli.cpp
|
| 431 |
+
common_log_set_verbosity_thold(params.verbosity);
|
| 432 |
+
|
| 433 |
std::unordered_map<std::string, std::pair<common_arg *, bool>> arg_to_options;
|
| 434 |
for (auto & opt : ctx_arg.options) {
|
| 435 |
for (const auto & arg : opt.args) {
|
|
|
|
| 535 |
// parse the first time to get -hf option (used for remote preset)
|
| 536 |
parse_cli_args();
|
| 537 |
|
| 538 |
+
// TODO: Remove later
|
| 539 |
+
try {
|
| 540 |
+
hf_cache::migrate_old_cache_to_hf_cache(params.hf_token, params.offline);
|
| 541 |
+
} catch (const std::exception & e) {
|
| 542 |
+
LOG_WRN("HF cache migration failed: %s\n", e.what());
|
| 543 |
+
}
|
| 544 |
+
// export_graph_ops loads only metadata
|
| 545 |
+
const bool skip_model_download = ctx_arg.ex == LLAMA_EXAMPLE_EXPORT_GRAPH_OPS;
|
| 546 |
+
|
| 547 |
// maybe handle remote preset
|
| 548 |
+
if (!params.model.hf_repo.empty() && !skip_model_download) {
|
| 549 |
std::string cli_hf_repo = params.model.hf_repo;
|
| 550 |
bool has_preset = common_params_handle_remote_preset(params, ctx_arg.ex);
|
| 551 |
|
|
|
|
| 576 |
}
|
| 577 |
|
| 578 |
// handle model and download
|
| 579 |
+
if (!skip_model_download) {
|
| 580 |
auto res = common_params_handle_model(params.model, params.hf_token, params.offline);
|
| 581 |
if (params.no_mmproj) {
|
| 582 |
params.mmproj = {};
|
|
|
|
| 597 |
|
| 598 |
// model is required (except for server)
|
| 599 |
// TODO @ngxson : maybe show a list of available models in CLI in this case
|
| 600 |
+
if (params.model.path.empty() && ctx_arg.ex != LLAMA_EXAMPLE_SERVER && !skip_model_download && !params.usage && !params.completion) {
|
| 601 |
throw std::invalid_argument("error: --model is required\n");
|
| 602 |
}
|
| 603 |
|
|
|
|
| 640 |
));
|
| 641 |
}
|
| 642 |
|
|
|
|
|
|
|
| 643 |
return true;
|
| 644 |
}
|
| 645 |
|
|
|
|
| 1045 |
{"--version"},
|
| 1046 |
"show version and build info",
|
| 1047 |
[](common_params &) {
|
| 1048 |
+
fprintf(stderr, "version: %d (%s)\n", llama_build_number(), llama_commit());
|
| 1049 |
+
fprintf(stderr, "built with %s for %s\n", llama_compiler(), llama_build_target());
|
| 1050 |
exit(0);
|
| 1051 |
}
|
| 1052 |
));
|
|
|
|
| 1064 |
{"-cl", "--cache-list"},
|
| 1065 |
"show list of models in cache",
|
| 1066 |
[](common_params &) {
|
|
|
|
| 1067 |
auto models = common_list_cached_models();
|
| 1068 |
printf("number of models in cache: %zu\n", models.size());
|
| 1069 |
for (size_t i = 0; i < models.size(); i++) {
|
| 1070 |
+
printf("%4zu. %s\n", i + 1, models[i].to_string().c_str());
|
|
|
|
| 1071 |
}
|
| 1072 |
exit(0);
|
| 1073 |
}
|
|
|
|
| 1085 |
[](common_params & params) {
|
| 1086 |
params.verbose_prompt = true;
|
| 1087 |
}
|
| 1088 |
+
).set_examples({LLAMA_EXAMPLE_COMPLETION, LLAMA_EXAMPLE_CLI, LLAMA_EXAMPLE_EMBEDDING, LLAMA_EXAMPLE_RETRIEVAL}));
|
| 1089 |
add_opt(common_arg(
|
| 1090 |
{"--display-prompt"},
|
| 1091 |
{"--no-display-prompt"},
|
|
|
|
| 1315 |
params.kv_unified = value;
|
| 1316 |
}
|
| 1317 |
).set_env("LLAMA_ARG_KV_UNIFIED").set_examples({LLAMA_EXAMPLE_SERVER, LLAMA_EXAMPLE_PERPLEXITY, LLAMA_EXAMPLE_BATCHED, LLAMA_EXAMPLE_BENCH, LLAMA_EXAMPLE_PARALLEL}));
|
| 1318 |
+
add_opt(common_arg(
|
| 1319 |
+
{"--cache-idle-slots"},
|
| 1320 |
+
{"--no-cache-idle-slots"},
|
| 1321 |
+
"save and clear idle slots on new task (default: enabled, requires unified KV and cache-ram)",
|
| 1322 |
+
[](common_params & params, bool value) {
|
| 1323 |
+
params.cache_idle_slots = value;
|
| 1324 |
+
}
|
| 1325 |
+
).set_env("LLAMA_ARG_CACHE_IDLE_SLOTS").set_examples({LLAMA_EXAMPLE_SERVER}));
|
| 1326 |
add_opt(common_arg(
|
| 1327 |
{"--context-shift"},
|
| 1328 |
{"--no-context-shift"},
|
|
|
|
| 1839 |
).set_sparam());
|
| 1840 |
add_opt(common_arg(
|
| 1841 |
{"--grammar"}, "GRAMMAR",
|
| 1842 |
+
"BNF-like grammar to constrain generations (see samples in grammars/ dir)",
|
| 1843 |
[](common_params & params, const std::string & value) {
|
| 1844 |
+
params.sampling.grammar = {COMMON_GRAMMAR_TYPE_USER, value};
|
| 1845 |
}
|
| 1846 |
).set_sparam());
|
| 1847 |
add_opt(common_arg(
|
| 1848 |
{"--grammar-file"}, "FNAME",
|
| 1849 |
"file to read grammar from",
|
| 1850 |
[](common_params & params, const std::string & value) {
|
| 1851 |
+
params.sampling.grammar = {COMMON_GRAMMAR_TYPE_USER, read_file(value)};
|
| 1852 |
}
|
| 1853 |
).set_sparam());
|
| 1854 |
add_opt(common_arg(
|
| 1855 |
{"-j", "--json-schema"}, "SCHEMA",
|
| 1856 |
"JSON schema to constrain generations (https://json-schema.org/), e.g. `{}` for any JSON object\nFor schemas w/ external $refs, use --grammar + example/json_schema_to_grammar.py instead",
|
| 1857 |
[](common_params & params, const std::string & value) {
|
| 1858 |
+
params.sampling.grammar = {COMMON_GRAMMAR_TYPE_OUTPUT_FORMAT, json_schema_to_grammar(json::parse(value))};
|
| 1859 |
}
|
| 1860 |
).set_sparam());
|
| 1861 |
add_opt(common_arg(
|
|
|
|
| 1872 |
std::istreambuf_iterator<char>(),
|
| 1873 |
std::back_inserter(schema)
|
| 1874 |
);
|
| 1875 |
+
params.sampling.grammar = {COMMON_GRAMMAR_TYPE_OUTPUT_FORMAT, json_schema_to_grammar(json::parse(schema))};
|
| 1876 |
}
|
| 1877 |
).set_sparam());
|
| 1878 |
add_opt(common_arg(
|
|
|
|
| 2352 |
}
|
| 2353 |
).set_env("LLAMA_ARG_N_GPU_LAYERS"));
|
| 2354 |
add_opt(common_arg(
|
| 2355 |
+
{"-sm", "--split-mode"}, "{none,layer,row,tensor}",
|
| 2356 |
"how to split the model across multiple GPUs, one of:\n"
|
| 2357 |
"- none: use one GPU only\n"
|
| 2358 |
+
"- layer (default): split layers and KV across GPUs (pipelined)\n"
|
| 2359 |
+
"- row: split weight across GPUs by rows (parallelized)\n"
|
| 2360 |
+
"- tensor: split weights and KV across GPUs (parallelized, EXPERIMENTAL)",
|
| 2361 |
[](common_params & params, const std::string & value) {
|
| 2362 |
+
if (value == "none") {
|
|
|
|
| 2363 |
params.split_mode = LLAMA_SPLIT_MODE_NONE;
|
| 2364 |
+
} else if (value == "layer") {
|
| 2365 |
params.split_mode = LLAMA_SPLIT_MODE_LAYER;
|
| 2366 |
+
} else if (value == "row") {
|
| 2367 |
params.split_mode = LLAMA_SPLIT_MODE_ROW;
|
| 2368 |
+
} else if (value == "tensor") {
|
| 2369 |
+
params.split_mode = LLAMA_SPLIT_MODE_TENSOR;
|
| 2370 |
} else {
|
| 2371 |
throw std::invalid_argument("invalid value");
|
| 2372 |
}
|
|
|
|
| 2426 |
}
|
| 2427 |
}
|
| 2428 |
).set_env("LLAMA_ARG_FIT"));
|
| 2429 |
+
add_opt(common_arg(
|
| 2430 |
+
{ "-fitp", "--fit-print" }, "[on|off]",
|
| 2431 |
+
string_format("print the estimated required memory ('on' or 'off', default: '%s')", params.fit_params_print ? "on" : "off"),
|
| 2432 |
+
[](common_params & params, const std::string & value) {
|
| 2433 |
+
if (is_truthy(value)) {
|
| 2434 |
+
params.fit_params_print = true;
|
| 2435 |
+
} else if (is_falsey(value)) {
|
| 2436 |
+
params.fit_params_print = false;
|
| 2437 |
+
} else {
|
| 2438 |
+
throw std::runtime_error(
|
| 2439 |
+
string_format("error: unknown value for --fit-print: '%s'\n", value.c_str()));
|
| 2440 |
+
}
|
| 2441 |
+
}
|
| 2442 |
+
).set_examples({LLAMA_EXAMPLE_FIT_PARAMS}).set_env("LLAMA_ARG_FIT_ESTIMATE"));
|
| 2443 |
add_opt(common_arg(
|
| 2444 |
{ "-fitt", "--fit-target" }, "MiB0,MiB1,MiB2,...",
|
| 2445 |
string_format("target margin per device for --fit, comma-separated list of values, "
|
|
|
|
| 2608 |
{"-hf", "-hfr", "--hf-repo"}, "<user>/<model>[:quant]",
|
| 2609 |
"Hugging Face model repository; quant is optional, case-insensitive, default to Q4_K_M, or falls back to the first file in the repo if Q4_K_M doesn't exist.\n"
|
| 2610 |
"mmproj is also downloaded automatically if available. to disable, add --no-mmproj\n"
|
| 2611 |
+
"example: ggml-org/GLM-4.7-Flash-GGUF:Q4_K_M\n"
|
| 2612 |
"(default: unused)",
|
| 2613 |
[](common_params & params, const std::string & value) {
|
| 2614 |
params.model.hf_repo = value;
|
|
|
|
| 2837 |
params.port = value;
|
| 2838 |
}
|
| 2839 |
).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_PORT"));
|
| 2840 |
+
add_opt(common_arg(
|
| 2841 |
+
{"--reuse-port"},
|
| 2842 |
+
string_format("allow multiple sockets to bind to the same port (default: %s)", params.reuse_port ? "enabled" : "disabled"),
|
| 2843 |
+
[](common_params & params) {
|
| 2844 |
+
params.reuse_port = true;
|
| 2845 |
+
}
|
| 2846 |
+
).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_REUSE_PORT"));
|
| 2847 |
add_opt(common_arg(
|
| 2848 |
{"--path"}, "PATH",
|
| 2849 |
string_format("path to serve static files from (default: %s)", params.public_path.c_str()),
|
|
|
|
| 2880 |
params.webui_mcp_proxy = value;
|
| 2881 |
}
|
| 2882 |
).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_WEBUI_MCP_PROXY"));
|
| 2883 |
+
add_opt(common_arg(
|
| 2884 |
+
{"--tools"}, "TOOL1,TOOL2,...",
|
| 2885 |
+
"experimental: whether to enable built-in tools for AI agents - do not enable in untrusted environments (default: no tools)\n"
|
| 2886 |
+
"specify \"all\" to enable all tools\n"
|
| 2887 |
+
"available tools: read_file, file_glob_search, grep_search, exec_shell_command, write_file, edit_file, apply_diff",
|
| 2888 |
+
[](common_params & params, const std::string & value) {
|
| 2889 |
+
params.server_tools = parse_csv_row(value);
|
| 2890 |
+
}
|
| 2891 |
+
).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_TOOLS"));
|
| 2892 |
add_opt(common_arg(
|
| 2893 |
{"--webui"},
|
| 2894 |
{"--no-webui"},
|
|
|
|
| 3291 |
"Set verbosity level to infinity (i.e. log all messages, useful for debugging)",
|
| 3292 |
[](common_params & params) {
|
| 3293 |
params.verbosity = INT_MAX;
|
| 3294 |
+
common_log_set_verbosity_thold(INT_MAX);
|
| 3295 |
}
|
| 3296 |
));
|
| 3297 |
add_opt(common_arg(
|
|
|
|
| 3312 |
"(default: %d)\n", params.verbosity),
|
| 3313 |
[](common_params & params, int value) {
|
| 3314 |
params.verbosity = value;
|
| 3315 |
+
common_log_set_verbosity_thold(value);
|
| 3316 |
}
|
| 3317 |
).set_env("LLAMA_LOG_VERBOSITY"));
|
| 3318 |
add_opt(common_arg(
|
|
|
|
| 3537 |
throw std::invalid_argument("unknown speculative decoding type without draft model");
|
| 3538 |
}
|
| 3539 |
}
|
| 3540 |
+
).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_SPEC_TYPE"));
|
| 3541 |
add_opt(common_arg(
|
| 3542 |
{"--spec-ngram-size-n"}, "N",
|
| 3543 |
string_format("ngram size N for ngram-simple/ngram-map speculative decoding, length of lookup n-gram (default: %d)", params.speculative.ngram_size_n),
|
common/build-info.cpp.in
CHANGED
|
@@ -1,4 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
int LLAMA_BUILD_NUMBER = @LLAMA_BUILD_NUMBER@;
|
| 2 |
-
char const *LLAMA_COMMIT = "@LLAMA_BUILD_COMMIT@";
|
| 3 |
-
char const *LLAMA_COMPILER = "@BUILD_COMPILER@";
|
| 4 |
-
char const *LLAMA_BUILD_TARGET = "@BUILD_TARGET@";
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#include "build-info.h"
|
| 2 |
+
|
| 3 |
+
#include <cstdio>
|
| 4 |
+
#include <string>
|
| 5 |
+
|
| 6 |
int LLAMA_BUILD_NUMBER = @LLAMA_BUILD_NUMBER@;
|
| 7 |
+
char const * LLAMA_COMMIT = "@LLAMA_BUILD_COMMIT@";
|
| 8 |
+
char const * LLAMA_COMPILER = "@BUILD_COMPILER@";
|
| 9 |
+
char const * LLAMA_BUILD_TARGET = "@BUILD_TARGET@";
|
| 10 |
+
|
| 11 |
+
int llama_build_number(void) {
|
| 12 |
+
return LLAMA_BUILD_NUMBER;
|
| 13 |
+
}
|
| 14 |
+
|
| 15 |
+
const char * llama_commit(void) {
|
| 16 |
+
return LLAMA_COMMIT;
|
| 17 |
+
}
|
| 18 |
+
|
| 19 |
+
const char * llama_compiler(void) {
|
| 20 |
+
return LLAMA_COMPILER;
|
| 21 |
+
}
|
| 22 |
+
|
| 23 |
+
const char * llama_build_target(void) {
|
| 24 |
+
return LLAMA_BUILD_TARGET;
|
| 25 |
+
}
|
| 26 |
+
|
| 27 |
+
const char * llama_build_info(void) {
|
| 28 |
+
static std::string s = "b" + std::to_string(LLAMA_BUILD_NUMBER) + "-" + LLAMA_COMMIT;
|
| 29 |
+
return s.c_str();
|
| 30 |
+
}
|
| 31 |
+
|
| 32 |
+
void llama_print_build_info(void) {
|
| 33 |
+
fprintf(stderr, "%s: build = %d (%s)\n", __func__, llama_build_number(), llama_commit());
|
| 34 |
+
fprintf(stderr, "%s: built with %s for %s\n", __func__, llama_compiler(), llama_build_target());
|
| 35 |
+
}
|
common/build-info.h
ADDED
|
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#pragma once
|
| 2 |
+
|
| 3 |
+
int llama_build_number(void);
|
| 4 |
+
|
| 5 |
+
const char * llama_commit(void);
|
| 6 |
+
const char * llama_compiler(void);
|
| 7 |
+
|
| 8 |
+
const char * llama_build_target(void);
|
| 9 |
+
const char * llama_build_info(void);
|
| 10 |
+
|
| 11 |
+
void llama_print_build_info(void);
|
common/chat-auto-parser-generator.cpp
CHANGED
|
@@ -1,3 +1,4 @@
|
|
|
|
|
| 1 |
#include "chat-auto-parser.h"
|
| 2 |
#include "chat-peg-parser.h"
|
| 3 |
#include "chat.h"
|
|
@@ -5,6 +6,7 @@
|
|
| 5 |
#include "json-schema-to-grammar.h"
|
| 6 |
#include "log.h"
|
| 7 |
#include "nlohmann/json.hpp"
|
|
|
|
| 8 |
|
| 9 |
#include <stdexcept>
|
| 10 |
#include <string>
|
|
@@ -23,13 +25,13 @@ static void foreach_function(const json & tools, const std::function<void(const
|
|
| 23 |
|
| 24 |
namespace autoparser {
|
| 25 |
|
| 26 |
-
parser_build_context::parser_build_context(common_chat_peg_builder & p, const
|
| 27 |
p(p),
|
| 28 |
inputs(inputs),
|
| 29 |
reasoning_parser(p.eps()) {}
|
| 30 |
|
| 31 |
common_chat_params peg_generator::generate_parser(const common_chat_template & tmpl,
|
| 32 |
-
const struct
|
| 33 |
// Run differential analysis to extract template structure
|
| 34 |
struct autoparser autoparser;
|
| 35 |
autoparser.analyze_template(tmpl);
|
|
@@ -37,17 +39,16 @@ common_chat_params peg_generator::generate_parser(const common_chat_template &
|
|
| 37 |
}
|
| 38 |
|
| 39 |
common_chat_params peg_generator::generate_parser(const common_chat_template & tmpl,
|
| 40 |
-
const struct
|
| 41 |
const autoparser & autoparser) {
|
| 42 |
-
// Build the parser using the analysis results
|
| 43 |
-
auto parser = autoparser.build_parser(inputs);
|
| 44 |
-
|
| 45 |
// Create the result structure
|
| 46 |
common_chat_params data;
|
| 47 |
data.prompt = common_chat_template_direct_apply(tmpl, inputs);
|
| 48 |
data.format = COMMON_CHAT_FORMAT_PEG_NATIVE;
|
| 49 |
data.preserved_tokens = autoparser.preserved_tokens;
|
| 50 |
-
|
|
|
|
|
|
|
| 51 |
|
| 52 |
// Build grammar if tools are present
|
| 53 |
bool has_tools =
|
|
@@ -65,9 +66,13 @@ common_chat_params peg_generator::generate_parser(const common_chat_template &
|
|
| 65 |
data.grammar = build_grammar([&](const common_grammar_builder & builder) {
|
| 66 |
foreach_function(inputs.tools, [&](const json & tool) {
|
| 67 |
const auto & function = tool.at("function");
|
| 68 |
-
auto schema = function.at("parameters");
|
| 69 |
builder.resolve_refs(schema);
|
| 70 |
});
|
|
|
|
|
|
|
|
|
|
|
|
|
| 71 |
parser.build_grammar(builder, data.grammar_lazy);
|
| 72 |
});
|
| 73 |
|
|
@@ -82,44 +87,41 @@ common_chat_params peg_generator::generate_parser(const common_chat_template &
|
|
| 82 |
return data;
|
| 83 |
}
|
| 84 |
|
| 85 |
-
common_peg_arena autoparser::build_parser(const
|
| 86 |
if (!analysis_complete) {
|
| 87 |
throw std::invalid_argument("Cannot call build_parser on autoparser without performing analysis first, call analyze_template(...)");
|
| 88 |
}
|
| 89 |
return build_chat_peg_parser([&](common_chat_peg_builder & p) {
|
| 90 |
-
// If the template uses Python dict format (single-quoted strings in JSON structures),
|
| 91 |
-
// pre-register a json-string rule that accepts both quote styles. This must happen
|
| 92 |
-
// before any call to p.json() so that all JSON parsing inherits the flexible rule.
|
| 93 |
-
if (tools.format.uses_python_dicts) {
|
| 94 |
-
p.rule("json-string", p.quoted_string());
|
| 95 |
-
}
|
| 96 |
-
|
| 97 |
parser_build_context ctx(p, inputs);
|
| 98 |
bool extract_reasoning = inputs.reasoning_format != COMMON_REASONING_FORMAT_NONE;
|
| 99 |
-
bool enable_thinking = inputs.enable_thinking;
|
| 100 |
|
| 101 |
-
ctx.extracting_reasoning = extract_reasoning &&
|
| 102 |
ctx.content = &content;
|
|
|
|
| 103 |
|
| 104 |
// Build reasoning parser
|
| 105 |
ctx.reasoning_parser = reasoning.build_parser(ctx);
|
| 106 |
|
|
|
|
|
|
|
| 107 |
bool has_tools = inputs.tools.is_array() && !inputs.tools.empty();
|
| 108 |
bool has_response_format = inputs.json_schema.is_object() && !inputs.json_schema.empty();
|
|
|
|
| 109 |
|
| 110 |
if (has_response_format) {
|
| 111 |
auto response_format = p.rule("response-format", p.content(p.schema(p.json(), "response-format-schema", inputs.json_schema)));
|
| 112 |
-
|
| 113 |
p.literal("```json") + p.space() + response_format + p.space() + p.literal("```"),
|
| 114 |
response_format
|
| 115 |
}) + p.end();
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 116 |
}
|
| 117 |
-
|
| 118 |
-
if (has_tools && inputs.tool_choice != COMMON_CHAT_TOOL_CHOICE_NONE && jinja_caps.supports_tool_calls) {
|
| 119 |
-
return tools.build_parser(ctx);
|
| 120 |
-
}
|
| 121 |
-
|
| 122 |
-
return content.build_parser(ctx);
|
| 123 |
});
|
| 124 |
}
|
| 125 |
|
|
@@ -130,24 +132,15 @@ common_peg_parser analyze_reasoning::build_parser(parser_build_context & ctx) co
|
|
| 130 |
return p.eps();
|
| 131 |
}
|
| 132 |
|
| 133 |
-
bool thinking_forced_open = (mode == reasoning_mode::FORCED_OPEN);
|
| 134 |
-
bool thinking_forced_closed = (mode == reasoning_mode::FORCED_CLOSED);
|
| 135 |
-
|
| 136 |
-
if (thinking_forced_open || thinking_forced_closed) {
|
| 137 |
-
// Thinking is forced open OR forced closed with enable_thinking=true
|
| 138 |
-
// In both cases, expect only the closing tag (opening was in template)
|
| 139 |
-
// However, since we might have incorrectly detected the open/close pattern,
|
| 140 |
-
// we admit an optional starting marker
|
| 141 |
-
return p.optional(p.literal(start)) + p.reasoning(p.until(end)) + end;
|
| 142 |
-
}
|
| 143 |
if (mode == reasoning_mode::TAG_BASED || mode == reasoning_mode::TOOLS_ONLY) {
|
| 144 |
-
|
| 145 |
-
|
| 146 |
-
|
| 147 |
-
|
|
|
|
|
|
|
|
|
|
| 148 |
}
|
| 149 |
-
} else if (mode == reasoning_mode::DELIMITER) {
|
| 150 |
-
return p.optional(p.reasoning(p.until(end)) + end);
|
| 151 |
}
|
| 152 |
|
| 153 |
return p.eps();
|
|
@@ -205,10 +198,19 @@ common_peg_parser analyze_tools::build_tool_parser_json_native(parser_build_cont
|
|
| 205 |
args_field = format.function_field + "." + args_field;
|
| 206 |
}
|
| 207 |
|
| 208 |
-
auto tools_parser = p.
|
| 209 |
-
|
| 210 |
-
|
| 211 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 212 |
|
| 213 |
// Handle content wrappers if present
|
| 214 |
if (ctx.content && ctx.content->is_always_wrapped()) {
|
|
@@ -227,6 +229,44 @@ common_peg_parser analyze_tools::build_tool_parser_json_native(parser_build_cont
|
|
| 227 |
p.end();
|
| 228 |
}
|
| 229 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 230 |
common_peg_parser analyze_tools::build_tool_parser_tag_json(parser_build_context & ctx) const {
|
| 231 |
auto & p = ctx.p;
|
| 232 |
const auto & inputs = ctx.inputs;
|
|
@@ -237,20 +277,30 @@ common_peg_parser analyze_tools::build_tool_parser_tag_json(parser_build_context
|
|
| 237 |
foreach_function(inputs.tools, [&](const json & tool) {
|
| 238 |
const auto & func = tool.at("function");
|
| 239 |
std::string name = func.at("name");
|
| 240 |
-
const auto & schema = func.at("parameters");
|
| 241 |
|
| 242 |
// Build call_id parser based on position (if supported)
|
|
|
|
| 243 |
common_peg_parser call_id_section = p.eps();
|
| 244 |
if (call_id.pos == call_id_position::BETWEEN_FUNC_AND_ARGS && !call_id.prefix.empty() &&
|
| 245 |
-
!call_id.suffix.empty()) {
|
| 246 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 247 |
}
|
| 248 |
-
|
| 249 |
-
|
| 250 |
-
|
| 251 |
-
|
| 252 |
-
|
|
|
|
| 253 |
}
|
|
|
|
|
|
|
|
|
|
| 254 |
tool_choice |= p.rule("tool-" + name, func_parser);
|
| 255 |
});
|
| 256 |
|
|
@@ -295,49 +345,42 @@ common_peg_parser analyze_tools::build_tool_parser_tag_tagged(parser_build_conte
|
|
| 295 |
const auto & inputs = ctx.inputs;
|
| 296 |
bool force_tools = inputs.tool_choice == COMMON_CHAT_TOOL_CHOICE_REQUIRED;
|
| 297 |
|
|
|
|
|
|
|
| 298 |
common_peg_parser tool_choice = p.choice();
|
| 299 |
|
| 300 |
foreach_function(inputs.tools, [&](const json & tool) {
|
| 301 |
-
const auto &
|
| 302 |
-
std::string
|
| 303 |
-
|
| 304 |
-
|
| 305 |
-
if (!params.contains("properties") || !params.at("properties").is_object()) {
|
| 306 |
-
return;
|
| 307 |
-
}
|
| 308 |
|
| 309 |
-
const auto & properties = params.at("properties");
|
| 310 |
std::set<std::string> required;
|
| 311 |
-
if (params.contains("required")
|
| 312 |
params.at("required").get_to(required);
|
| 313 |
}
|
| 314 |
|
|
|
|
|
|
|
|
|
|
| 315 |
// Build parser for each argument, separating required and optional
|
| 316 |
std::vector<common_peg_parser> required_parsers;
|
| 317 |
std::vector<common_peg_parser> optional_parsers;
|
| 318 |
for (const auto & [param_name, param_schema] : properties.items()) {
|
| 319 |
-
bool
|
| 320 |
-
|
| 321 |
-
auto
|
| 322 |
-
|
| 323 |
-
|
| 324 |
-
|
| 325 |
-
|
| 326 |
-
|
| 327 |
-
|
| 328 |
-
|
| 329 |
-
|
| 330 |
-
|
| 331 |
-
|
| 332 |
-
|
| 333 |
-
arguments.value_prefix +
|
| 334 |
-
(type == "string" ? p.tool_arg_string_value(p.schema(p.until(arguments.value_suffix),
|
| 335 |
-
"tool-" + name + "-arg-" + param_name + "-schema",
|
| 336 |
-
param_schema, true)) :
|
| 337 |
-
p.tool_arg_json_value(p.schema(
|
| 338 |
-
p.json(), "tool-" + name + "-arg-" + param_name + "-schema", param_schema, format.uses_python_dicts)) +
|
| 339 |
-
p.space()) +
|
| 340 |
-
p.tool_arg_close(p.literal(arguments.value_suffix)));
|
| 341 |
|
| 342 |
auto named_arg = p.rule("tool-" + name + "-arg-" + param_name, arg);
|
| 343 |
if (is_required) {
|
|
@@ -362,53 +405,34 @@ common_peg_parser analyze_tools::build_tool_parser_tag_tagged(parser_build_conte
|
|
| 362 |
for (const auto & opt : optional_parsers) {
|
| 363 |
any_opt |= opt;
|
| 364 |
}
|
| 365 |
-
args_seq = args_seq + p.repeat(p.space() + any_opt, 0,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 366 |
}
|
| 367 |
|
| 368 |
// Build call_id parser based on position (if supported)
|
| 369 |
common_peg_parser call_id_section = p.eps();
|
| 370 |
bool have_call_id = false;
|
| 371 |
if (call_id.pos == call_id_position::BETWEEN_FUNC_AND_ARGS && !call_id.prefix.empty() &&
|
| 372 |
-
!call_id.suffix.empty()) {
|
| 373 |
have_call_id = true;
|
| 374 |
-
|
| 375 |
-
|
| 376 |
-
|
| 377 |
-
|
| 378 |
-
|
| 379 |
-
if (!function.name_suffix.empty()) {
|
| 380 |
-
func_parser = p.tool_open(function.name_prefix + p.tool_name(p.literal(name)) + function.name_suffix) +
|
| 381 |
-
call_id_section + p.space() + args_seq;
|
| 382 |
-
matched_atomic = true;
|
| 383 |
-
} else if (have_call_id) {
|
| 384 |
-
func_parser = p.atomic(p.tool_open(function.name_prefix + p.tool_name(p.literal(name)) + function.name_suffix) +
|
| 385 |
-
call_id_section) + p.space() + args_seq;
|
| 386 |
-
matched_atomic = true;
|
| 387 |
-
} else if (!arguments.name_prefix.empty() && properties.size() > 0) {
|
| 388 |
-
func_parser = p.atomic(p.tool_open(function.name_prefix + p.tool_name(p.literal(name)) + function.name_suffix) +
|
| 389 |
-
call_id_section + p.space() + p.peek(p.literal(arguments.name_prefix))) + args_seq;
|
| 390 |
-
matched_atomic = true;
|
| 391 |
-
} else {
|
| 392 |
-
func_parser = p.tool_open(function.name_prefix + p.tool_name(p.literal(name)) + function.name_suffix) +
|
| 393 |
-
call_id_section + p.space() + args_seq;
|
| 394 |
-
}
|
| 395 |
-
|
| 396 |
-
if (!function.close.empty()) {
|
| 397 |
-
func_parser = func_parser + p.space() + p.tool_close(p.literal(function.close));
|
| 398 |
-
} else if (!format.per_call_end.empty()) {
|
| 399 |
-
// When there's no func_close but there is a per_call_end marker, use peek() to ensure
|
| 400 |
-
// we only emit tool_close when we can actually see the closing marker. This prevents
|
| 401 |
-
// premature closing during partial parsing when we've seen e.g. "</" which could be
|
| 402 |
-
// either "</tool_call>" (end) or "<arg_key>" prefix that failed to match.
|
| 403 |
-
func_parser = func_parser + p.tool_close(p.peek(p.literal(format.per_call_end)));
|
| 404 |
-
} else {
|
| 405 |
-
func_parser =
|
| 406 |
-
func_parser + p.tool_close(p.space()); // force this to process tool closing callbacks in mapper
|
| 407 |
-
}
|
| 408 |
-
if (!matched_atomic) {
|
| 409 |
-
func_parser = p.atomic(func_parser);
|
| 410 |
}
|
| 411 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 412 |
tool_choice |= p.rule("tool-" + name, func_parser);
|
| 413 |
});
|
| 414 |
|
|
@@ -419,14 +443,14 @@ common_peg_parser analyze_tools::build_tool_parser_tag_tagged(parser_build_conte
|
|
| 419 |
if (!format.per_call_start.empty()) {
|
| 420 |
auto wrapped_call = format.per_call_start + p.space() + tool_choice + p.space() + format.per_call_end;
|
| 421 |
if (inputs.parallel_tool_calls) {
|
| 422 |
-
tool_calls = p.trigger_rule("tool-call", wrapped_call + p.zero_or_more(p.space() + wrapped_call));
|
| 423 |
} else {
|
| 424 |
-
tool_calls = p.trigger_rule("tool-call", wrapped_call);
|
| 425 |
}
|
| 426 |
if (!format.section_start.empty()) {
|
| 427 |
tool_calls = p.trigger_rule("tool-calls",
|
| 428 |
p.literal(format.section_start) + p.space() + tool_calls + p.space() +
|
| 429 |
-
(format.section_end.empty() ? p.end() : p.literal(format.section_end)));
|
| 430 |
}
|
| 431 |
} else {
|
| 432 |
std::string separator = ", "; // Default
|
|
|
|
| 1 |
+
#include "chat-auto-parser-helpers.h"
|
| 2 |
#include "chat-auto-parser.h"
|
| 3 |
#include "chat-peg-parser.h"
|
| 4 |
#include "chat.h"
|
|
|
|
| 6 |
#include "json-schema-to-grammar.h"
|
| 7 |
#include "log.h"
|
| 8 |
#include "nlohmann/json.hpp"
|
| 9 |
+
#include "peg-parser.h"
|
| 10 |
|
| 11 |
#include <stdexcept>
|
| 12 |
#include <string>
|
|
|
|
| 25 |
|
| 26 |
namespace autoparser {
|
| 27 |
|
| 28 |
+
parser_build_context::parser_build_context(common_chat_peg_builder & p, const generation_params & inputs) :
|
| 29 |
p(p),
|
| 30 |
inputs(inputs),
|
| 31 |
reasoning_parser(p.eps()) {}
|
| 32 |
|
| 33 |
common_chat_params peg_generator::generate_parser(const common_chat_template & tmpl,
|
| 34 |
+
const struct generation_params & inputs) {
|
| 35 |
// Run differential analysis to extract template structure
|
| 36 |
struct autoparser autoparser;
|
| 37 |
autoparser.analyze_template(tmpl);
|
|
|
|
| 39 |
}
|
| 40 |
|
| 41 |
common_chat_params peg_generator::generate_parser(const common_chat_template & tmpl,
|
| 42 |
+
const struct generation_params & inputs,
|
| 43 |
const autoparser & autoparser) {
|
|
|
|
|
|
|
|
|
|
| 44 |
// Create the result structure
|
| 45 |
common_chat_params data;
|
| 46 |
data.prompt = common_chat_template_direct_apply(tmpl, inputs);
|
| 47 |
data.format = COMMON_CHAT_FORMAT_PEG_NATIVE;
|
| 48 |
data.preserved_tokens = autoparser.preserved_tokens;
|
| 49 |
+
|
| 50 |
+
auto parser = autoparser.build_parser(inputs);
|
| 51 |
+
data.parser = parser.save();
|
| 52 |
|
| 53 |
// Build grammar if tools are present
|
| 54 |
bool has_tools =
|
|
|
|
| 66 |
data.grammar = build_grammar([&](const common_grammar_builder & builder) {
|
| 67 |
foreach_function(inputs.tools, [&](const json & tool) {
|
| 68 |
const auto & function = tool.at("function");
|
| 69 |
+
auto schema = function.contains("parameters") ? function.at("parameters") : json::object();
|
| 70 |
builder.resolve_refs(schema);
|
| 71 |
});
|
| 72 |
+
if (has_response_format) {
|
| 73 |
+
auto schema = inputs.json_schema;
|
| 74 |
+
builder.resolve_refs(schema);
|
| 75 |
+
}
|
| 76 |
parser.build_grammar(builder, data.grammar_lazy);
|
| 77 |
});
|
| 78 |
|
|
|
|
| 87 |
return data;
|
| 88 |
}
|
| 89 |
|
| 90 |
+
common_peg_arena autoparser::build_parser(const generation_params & inputs) const {
|
| 91 |
if (!analysis_complete) {
|
| 92 |
throw std::invalid_argument("Cannot call build_parser on autoparser without performing analysis first, call analyze_template(...)");
|
| 93 |
}
|
| 94 |
return build_chat_peg_parser([&](common_chat_peg_builder & p) {
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 95 |
parser_build_context ctx(p, inputs);
|
| 96 |
bool extract_reasoning = inputs.reasoning_format != COMMON_REASONING_FORMAT_NONE;
|
|
|
|
| 97 |
|
| 98 |
+
ctx.extracting_reasoning = extract_reasoning && reasoning.mode != reasoning_mode::NONE;
|
| 99 |
ctx.content = &content;
|
| 100 |
+
ctx.reasoning = &reasoning;
|
| 101 |
|
| 102 |
// Build reasoning parser
|
| 103 |
ctx.reasoning_parser = reasoning.build_parser(ctx);
|
| 104 |
|
| 105 |
+
auto parser = p.eps();
|
| 106 |
+
|
| 107 |
bool has_tools = inputs.tools.is_array() && !inputs.tools.empty();
|
| 108 |
bool has_response_format = inputs.json_schema.is_object() && !inputs.json_schema.empty();
|
| 109 |
+
bool pure_content = reasoning.mode == reasoning_mode::NONE;
|
| 110 |
|
| 111 |
if (has_response_format) {
|
| 112 |
auto response_format = p.rule("response-format", p.content(p.schema(p.json(), "response-format-schema", inputs.json_schema)));
|
| 113 |
+
parser = ctx.reasoning_parser + p.space() + p.choice({
|
| 114 |
p.literal("```json") + p.space() + response_format + p.space() + p.literal("```"),
|
| 115 |
response_format
|
| 116 |
}) + p.end();
|
| 117 |
+
pure_content = false;
|
| 118 |
+
} else if (has_tools && inputs.tool_choice != COMMON_CHAT_TOOL_CHOICE_NONE && jinja_caps.supports_tool_calls) {
|
| 119 |
+
parser = tools.build_parser(ctx);
|
| 120 |
+
pure_content = false;
|
| 121 |
+
} else {
|
| 122 |
+
parser = content.build_parser(ctx);
|
| 123 |
}
|
| 124 |
+
return pure_content ? p.prefix(inputs.generation_prompt, reasoning.start) + parser : p.prefix(inputs.generation_prompt, reasoning.start) << parser;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 125 |
});
|
| 126 |
}
|
| 127 |
|
|
|
|
| 132 |
return p.eps();
|
| 133 |
}
|
| 134 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 135 |
if (mode == reasoning_mode::TAG_BASED || mode == reasoning_mode::TOOLS_ONLY) {
|
| 136 |
+
if (!end.empty()) {
|
| 137 |
+
if (!start.empty()) {
|
| 138 |
+
// Standard tag-based: optional(<think>reasoning</think>)
|
| 139 |
+
return p.optional(start + p.reasoning(p.until(end)) + end + p.space());
|
| 140 |
+
}
|
| 141 |
+
// Delimiter-style (empty start)
|
| 142 |
+
return p.optional(p.reasoning(p.until(end)) + end + p.space());
|
| 143 |
}
|
|
|
|
|
|
|
| 144 |
}
|
| 145 |
|
| 146 |
return p.eps();
|
|
|
|
| 198 |
args_field = format.function_field + "." + args_field;
|
| 199 |
}
|
| 200 |
|
| 201 |
+
auto tools_parser = p.eps();
|
| 202 |
+
if (format.section_start.empty() && !format.per_call_start.empty()) {
|
| 203 |
+
auto single_tool_parser = p.standard_json_tools(
|
| 204 |
+
format.per_call_start, format.per_call_end, inputs.tools, inputs.parallel_tool_calls,
|
| 205 |
+
inputs.tool_choice == COMMON_CHAT_TOOL_CHOICE_REQUIRED, name_field, args_field, format.tools_array_wrapped,
|
| 206 |
+
format.fun_name_is_key, format.id_field, format.gen_id_field, format.parameter_order);
|
| 207 |
+
tools_parser = p.trigger_rule("tool-calls", p.one_or_more(single_tool_parser + p.space()));
|
| 208 |
+
} else {
|
| 209 |
+
tools_parser = p.standard_json_tools(
|
| 210 |
+
format.section_start, format.section_end, inputs.tools, inputs.parallel_tool_calls,
|
| 211 |
+
inputs.tool_choice == COMMON_CHAT_TOOL_CHOICE_REQUIRED, name_field, args_field, format.tools_array_wrapped,
|
| 212 |
+
format.fun_name_is_key, format.id_field, format.gen_id_field, format.parameter_order);
|
| 213 |
+
}
|
| 214 |
|
| 215 |
// Handle content wrappers if present
|
| 216 |
if (ctx.content && ctx.content->is_always_wrapped()) {
|
|
|
|
| 229 |
p.end();
|
| 230 |
}
|
| 231 |
|
| 232 |
+
common_peg_parser analyze_tools::build_func_parser(common_chat_peg_builder & p, const std::string & name,
|
| 233 |
+
const common_peg_parser & call_id_section, bool have_call_id,
|
| 234 |
+
const common_peg_parser & args,
|
| 235 |
+
std::optional<common_peg_parser> atomic_peek) const {
|
| 236 |
+
auto open = p.tool_open(function.name_prefix + p.tool_name(p.literal(name)) + function.name_suffix);
|
| 237 |
+
bool matched_atomic = false;
|
| 238 |
+
common_peg_parser func_parser = p.eps();
|
| 239 |
+
|
| 240 |
+
if (!function.name_suffix.empty()) {
|
| 241 |
+
func_parser = open + call_id_section + p.space() + args;
|
| 242 |
+
matched_atomic = true;
|
| 243 |
+
} else if (have_call_id) {
|
| 244 |
+
func_parser = p.atomic(open + call_id_section) + p.space() + args;
|
| 245 |
+
matched_atomic = true;
|
| 246 |
+
} else if (atomic_peek.has_value()) {
|
| 247 |
+
func_parser = p.atomic(open + call_id_section + p.space() + *atomic_peek) + args;
|
| 248 |
+
matched_atomic = true;
|
| 249 |
+
} else {
|
| 250 |
+
func_parser = open + call_id_section + p.space() + args;
|
| 251 |
+
}
|
| 252 |
+
|
| 253 |
+
if (!function.close.empty()) {
|
| 254 |
+
func_parser = func_parser + p.space() + p.tool_close(p.literal(function.close));
|
| 255 |
+
} else if (!format.per_call_end.empty()) {
|
| 256 |
+
// When there's no func_close but there is a per_call_end marker, use peek() to ensure
|
| 257 |
+
// we only emit tool_close when we can actually see the closing marker. This prevents
|
| 258 |
+
// premature closing during partial parsing when we've seen e.g. "</" which could be
|
| 259 |
+
// either "</tool_call>" (end) or "<arg_key>" prefix that failed to match.
|
| 260 |
+
func_parser = func_parser + p.tool_close(p.peek(p.literal(format.per_call_end)));
|
| 261 |
+
} else {
|
| 262 |
+
func_parser = func_parser + p.tool_close(p.space()); // force this to process tool closing callbacks in mapper
|
| 263 |
+
}
|
| 264 |
+
if (!matched_atomic) {
|
| 265 |
+
func_parser = p.atomic(func_parser);
|
| 266 |
+
}
|
| 267 |
+
return func_parser;
|
| 268 |
+
}
|
| 269 |
+
|
| 270 |
common_peg_parser analyze_tools::build_tool_parser_tag_json(parser_build_context & ctx) const {
|
| 271 |
auto & p = ctx.p;
|
| 272 |
const auto & inputs = ctx.inputs;
|
|
|
|
| 277 |
foreach_function(inputs.tools, [&](const json & tool) {
|
| 278 |
const auto & func = tool.at("function");
|
| 279 |
std::string name = func.at("name");
|
| 280 |
+
const auto & schema = func.contains("parameters") ? func.at("parameters") : json::object();
|
| 281 |
|
| 282 |
// Build call_id parser based on position (if supported)
|
| 283 |
+
bool have_call_id = false;
|
| 284 |
common_peg_parser call_id_section = p.eps();
|
| 285 |
if (call_id.pos == call_id_position::BETWEEN_FUNC_AND_ARGS && !call_id.prefix.empty() &&
|
| 286 |
+
(!call_id.suffix.empty() || !arguments.start.empty())) {
|
| 287 |
+
if (!call_id.suffix.empty()) {
|
| 288 |
+
call_id_section = p.optional(call_id.prefix + p.tool_id(p.until(call_id.suffix))) + call_id.suffix;
|
| 289 |
+
} else {
|
| 290 |
+
call_id_section = p.optional(call_id.prefix + p.tool_id(p.until(arguments.start)));
|
| 291 |
+
}
|
| 292 |
+
have_call_id = true;
|
| 293 |
}
|
| 294 |
+
auto args_parser = p.tool_args(p.schema(p.json(), "tool-" + name + "-schema", schema));
|
| 295 |
+
if (!arguments.start.empty()) {
|
| 296 |
+
args_parser = p.literal(arguments.start) + args_parser;
|
| 297 |
+
}
|
| 298 |
+
if (!arguments.end.empty()) {
|
| 299 |
+
args_parser = args_parser + p.literal(arguments.end);
|
| 300 |
}
|
| 301 |
+
|
| 302 |
+
auto atomic_peek = !arguments.start.empty() ? std::optional(p.peek(p.literal(arguments.start))) : std::nullopt;
|
| 303 |
+
auto func_parser = build_func_parser(p, name, call_id_section, have_call_id, args_parser, atomic_peek);
|
| 304 |
tool_choice |= p.rule("tool-" + name, func_parser);
|
| 305 |
});
|
| 306 |
|
|
|
|
| 345 |
const auto & inputs = ctx.inputs;
|
| 346 |
bool force_tools = inputs.tool_choice == COMMON_CHAT_TOOL_CHOICE_REQUIRED;
|
| 347 |
|
| 348 |
+
auto until_suffix = p.rule("until-suffix", p.until(arguments.value_suffix));
|
| 349 |
+
|
| 350 |
common_peg_parser tool_choice = p.choice();
|
| 351 |
|
| 352 |
foreach_function(inputs.tools, [&](const json & tool) {
|
| 353 |
+
const auto & func = tool.at("function");
|
| 354 |
+
std::string name = func.at("name");
|
| 355 |
+
auto params = func.contains("parameters") ? func.at("parameters") : json::object();
|
| 356 |
+
const auto & properties = params.contains("properties") ? params.at("properties") : json::object();
|
|
|
|
|
|
|
|
|
|
| 357 |
|
|
|
|
| 358 |
std::set<std::string> required;
|
| 359 |
+
if (params.contains("required")) {
|
| 360 |
params.at("required").get_to(required);
|
| 361 |
}
|
| 362 |
|
| 363 |
+
auto schema_info = common_schema_info();
|
| 364 |
+
schema_info.resolve_refs(params);
|
| 365 |
+
|
| 366 |
// Build parser for each argument, separating required and optional
|
| 367 |
std::vector<common_peg_parser> required_parsers;
|
| 368 |
std::vector<common_peg_parser> optional_parsers;
|
| 369 |
for (const auto & [param_name, param_schema] : properties.items()) {
|
| 370 |
+
bool is_required = required.find(param_name) != required.end();
|
| 371 |
+
|
| 372 |
+
auto arg =
|
| 373 |
+
p.tool_arg(p.tool_arg_open(arguments.name_prefix + p.tool_arg_name(p.literal(param_name)) +
|
| 374 |
+
arguments.name_suffix) +
|
| 375 |
+
arguments.value_prefix +
|
| 376 |
+
(schema_info.resolves_to_string(param_schema) ?
|
| 377 |
+
p.tool_arg_string_value(p.schema(until_suffix,
|
| 378 |
+
"tool-" + name + "-arg-" + param_name + "-schema",
|
| 379 |
+
param_schema, true)) :
|
| 380 |
+
p.tool_arg_json_value(p.schema(
|
| 381 |
+
p.json(), "tool-" + name + "-arg-" + param_name + "-schema", param_schema, false)) +
|
| 382 |
+
p.space()) +
|
| 383 |
+
p.tool_arg_close(p.literal(arguments.value_suffix)));
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 384 |
|
| 385 |
auto named_arg = p.rule("tool-" + name + "-arg-" + param_name, arg);
|
| 386 |
if (is_required) {
|
|
|
|
| 405 |
for (const auto & opt : optional_parsers) {
|
| 406 |
any_opt |= opt;
|
| 407 |
}
|
| 408 |
+
args_seq = args_seq + p.repeat(p.space() + any_opt, 0, -1);
|
| 409 |
+
}
|
| 410 |
+
|
| 411 |
+
if (!arguments.start.empty()) {
|
| 412 |
+
args_seq = p.literal(arguments.start) + args_seq;
|
| 413 |
+
}
|
| 414 |
+
if (!arguments.end.empty()) {
|
| 415 |
+
args_seq = args_seq + p.literal(arguments.end);
|
| 416 |
}
|
| 417 |
|
| 418 |
// Build call_id parser based on position (if supported)
|
| 419 |
common_peg_parser call_id_section = p.eps();
|
| 420 |
bool have_call_id = false;
|
| 421 |
if (call_id.pos == call_id_position::BETWEEN_FUNC_AND_ARGS && !call_id.prefix.empty() &&
|
| 422 |
+
(!call_id.suffix.empty() || !arguments.start.empty())) {
|
| 423 |
have_call_id = true;
|
| 424 |
+
if (!call_id.suffix.empty()) {
|
| 425 |
+
call_id_section = p.optional(call_id.prefix + p.tool_id(p.until(call_id.suffix)) + call_id.suffix);
|
| 426 |
+
} else {
|
| 427 |
+
call_id_section = p.optional(call_id.prefix + p.tool_id(p.until(arguments.start)));
|
| 428 |
+
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 429 |
}
|
| 430 |
|
| 431 |
+
// Only peek for an arg tag when there are required args that must follow.
|
| 432 |
+
// When all args are optional, the model may emit no arg tags at all (#20650).
|
| 433 |
+
auto atomic_peek = (!arguments.name_prefix.empty() && !required_parsers.empty()) ?
|
| 434 |
+
std::optional(p.peek(p.literal(arguments.name_prefix))) : std::nullopt;
|
| 435 |
+
auto func_parser = build_func_parser(p, name, call_id_section, have_call_id, args_seq, atomic_peek);
|
| 436 |
tool_choice |= p.rule("tool-" + name, func_parser);
|
| 437 |
});
|
| 438 |
|
|
|
|
| 443 |
if (!format.per_call_start.empty()) {
|
| 444 |
auto wrapped_call = format.per_call_start + p.space() + tool_choice + p.space() + format.per_call_end;
|
| 445 |
if (inputs.parallel_tool_calls) {
|
| 446 |
+
tool_calls = p.trigger_rule("tool-call", wrapped_call + p.zero_or_more(p.space() + wrapped_call) + p.space());
|
| 447 |
} else {
|
| 448 |
+
tool_calls = p.trigger_rule("tool-call", wrapped_call + p.space());
|
| 449 |
}
|
| 450 |
if (!format.section_start.empty()) {
|
| 451 |
tool_calls = p.trigger_rule("tool-calls",
|
| 452 |
p.literal(format.section_start) + p.space() + tool_calls + p.space() +
|
| 453 |
+
(format.section_end.empty() ? p.end() : p.literal(format.section_end) + p.space()));
|
| 454 |
}
|
| 455 |
} else {
|
| 456 |
std::string separator = ", "; // Default
|
common/chat-auto-parser-helpers.cpp
CHANGED
|
@@ -1,9 +1,11 @@
|
|
| 1 |
#include "chat-auto-parser-helpers.h"
|
| 2 |
|
| 3 |
#include "chat-auto-parser.h"
|
|
|
|
| 4 |
#include "chat.h"
|
| 5 |
#include "log.h"
|
| 6 |
#include "nlohmann/json.hpp"
|
|
|
|
| 7 |
|
| 8 |
#include <cctype>
|
| 9 |
#include <numeric>
|
|
@@ -186,6 +188,21 @@ diff_split calculate_diff_split(const std::string & left, const std::string & ri
|
|
| 186 |
result.suffix = "";
|
| 187 |
// pick prefix = all as representation
|
| 188 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 189 |
return result;
|
| 190 |
}
|
| 191 |
|
|
@@ -294,7 +311,7 @@ std::vector<segment> prune_whitespace_segments(const std::vector<segment> & segm
|
|
| 294 |
namespace autoparser {
|
| 295 |
|
| 296 |
std::string apply_template(const common_chat_template & tmpl, const template_params & params) {
|
| 297 |
-
|
| 298 |
tmpl_params.messages = params.messages;
|
| 299 |
tmpl_params.tools = params.tools;
|
| 300 |
tmpl_params.add_generation_prompt = params.add_generation_prompt;
|
|
|
|
| 1 |
#include "chat-auto-parser-helpers.h"
|
| 2 |
|
| 3 |
#include "chat-auto-parser.h"
|
| 4 |
+
#include "chat-peg-parser.h"
|
| 5 |
#include "chat.h"
|
| 6 |
#include "log.h"
|
| 7 |
#include "nlohmann/json.hpp"
|
| 8 |
+
#include "peg-parser.h"
|
| 9 |
|
| 10 |
#include <cctype>
|
| 11 |
#include <numeric>
|
|
|
|
| 188 |
result.suffix = "";
|
| 189 |
// pick prefix = all as representation
|
| 190 |
}
|
| 191 |
+
|
| 192 |
+
// When left has no unique content (result.left is empty), left is entirely
|
| 193 |
+
// shared with right. The simultaneous prefix/suffix segment matching can
|
| 194 |
+
// incorrectly consume trailing segments of left as suffix when those same
|
| 195 |
+
// segments also appear at the end of right (e.g. "\n" at the end of both
|
| 196 |
+
// the shared content and the generation prompt). This rotates the diff.
|
| 197 |
+
// Fix: if left is a prefix of right, enforce that directly.
|
| 198 |
+
if (result.left.empty() && !result.right.empty() &&
|
| 199 |
+
left.size() <= right.size() &&
|
| 200 |
+
right.substr(0, left.size()) == left) {
|
| 201 |
+
result.prefix = left;
|
| 202 |
+
result.suffix = "";
|
| 203 |
+
result.right = right.substr(left.size());
|
| 204 |
+
}
|
| 205 |
+
|
| 206 |
return result;
|
| 207 |
}
|
| 208 |
|
|
|
|
| 311 |
namespace autoparser {
|
| 312 |
|
| 313 |
std::string apply_template(const common_chat_template & tmpl, const template_params & params) {
|
| 314 |
+
generation_params tmpl_params;
|
| 315 |
tmpl_params.messages = params.messages;
|
| 316 |
tmpl_params.tools = params.tools;
|
| 317 |
tmpl_params.add_generation_prompt = params.add_generation_prompt;
|