Skip to content

Commit

Permalink
Merge branch 'master' of https://github.com/l3utterfly/llama.cpp
Browse files Browse the repository at this point in the history
  • Loading branch information
l3utterfly committed Sep 8, 2024
2 parents 32335d5 + a876861 commit d11bd3b
Show file tree
Hide file tree
Showing 199 changed files with 25,042 additions and 15,107 deletions.
25 changes: 11 additions & 14 deletions .devops/full-cuda.Dockerfile
Original file line number Diff line number Diff line change
@@ -1,18 +1,16 @@
ARG UBUNTU_VERSION=22.04

# This needs to generally match the container host's environment.
ARG CUDA_VERSION=11.7.1

ARG CUDA_VERSION=12.6.0
# Target the CUDA build image
ARG BASE_CUDA_DEV_CONTAINER=nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}

FROM ${BASE_CUDA_DEV_CONTAINER} AS build

# Unless otherwise specified, we make a fat build.
ARG CUDA_DOCKER_ARCH=all
# CUDA architecture to build for (defaults to all supported archs)
ARG CUDA_DOCKER_ARCH=default

RUN apt-get update && \
apt-get install -y build-essential python3 python3-pip git libcurl4-openssl-dev libgomp1
apt-get install -y build-essential cmake python3 python3-pip git libcurl4-openssl-dev libgomp1

COPY requirements.txt requirements.txt
COPY requirements requirements
Expand All @@ -24,13 +22,12 @@ WORKDIR /app

COPY . .

# Set nvcc architecture
ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
# Enable CUDA
ENV GGML_CUDA=1
# Enable cURL
ENV LLAMA_CURL=1

RUN make -j$(nproc)
# Use the default CUDA archs if not specified
RUN if [ "${CUDA_DOCKER_ARCH}" != "default" ]; then \
export CMAKE_ARGS="-DCMAKE_CUDA_ARCHITECTURES=${CUDA_DOCKER_ARCH}"; \
fi && \
cmake -B build -DGGML_CUDA=ON -DLLAMA_CURL=ON ${CMAKE_ARGS} -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined . && \
cmake --build build --config Release -j$(nproc) && \
cp build/bin/* .

ENTRYPOINT ["/app/.devops/tools.sh"]
44 changes: 44 additions & 0 deletions .devops/llama-cli-cann.Dockerfile
Original file line number Diff line number Diff line change
@@ -0,0 +1,44 @@
ARG ASCEND_VERSION=8.0.rc2.alpha003-910b-openeuler22.03-py3.8

FROM cosdt/cann:$ASCEND_VERSION AS build

WORKDIR /app

COPY . .

RUN yum install -y gcc g++ cmake make
ENV ASCEND_TOOLKIT_HOME=/usr/local/Ascend/ascend-toolkit/latest
ENV LIBRARY_PATH=${ASCEND_TOOLKIT_HOME}/lib64:$LIBRARY_PATH
ENV LD_LIBRARY_PATH=${ASCEND_TOOLKIT_HOME}/lib64:${ASCEND_TOOLKIT_HOME}/lib64/plugin/opskernel:${ASCEND_TOOLKIT_HOME}/lib64/plugin/nnengine:${ASCEND_TOOLKIT_HOME}/opp/built-in/op_impl/ai_core/tbe/op_tiling:${LD_LIBRARY_PATH}
ENV PYTHONPATH=${ASCEND_TOOLKIT_HOME}/python/site-packages:${ASCEND_TOOLKIT_HOME}/opp/built-in/op_impl/ai_core/tbe:${PYTHONPATH}
ENV PATH=${ASCEND_TOOLKIT_HOME}/bin:${ASCEND_TOOLKIT_HOME}/compiler/ccec_compiler/bin:${PATH}
ENV ASCEND_AICPU_PATH=${ASCEND_TOOLKIT_HOME}
ENV ASCEND_OPP_PATH=${ASCEND_TOOLKIT_HOME}/opp
ENV TOOLCHAIN_HOME=${ASCEND_TOOLKIT_HOME}/toolkit
ENV ASCEND_HOME_PATH=${ASCEND_TOOLKIT_HOME}

# find libascend_hal.so, because the drive hasn`t been mounted.
ENV LD_LIBRARY_PATH=${ASCEND_TOOLKIT_HOME}/runtime/lib64/stub:$LD_LIBRARY_PATH

RUN echo "Building with static libs" && \
source /usr/local/Ascend/ascend-toolkit/set_env.sh --force && \
cmake -B build -DGGML_CANN=ON -DBUILD_SHARED_LIBS=OFF && \
cmake --build build --config Release --target llama-cli

# TODO: use image with NNRT
FROM cosdt/cann:$ASCEND_VERSION AS runtime
COPY --from=build /app/build/bin/llama-cli /llama-cli

ENV LC_ALL=C.utf8

ENV ASCEND_TOOLKIT_HOME=/usr/local/Ascend/ascend-toolkit/latest
ENV LIBRARY_PATH=${ASCEND_TOOLKIT_HOME}/lib64:$LIBRARY_PATH
ENV LD_LIBRARY_PATH=${ASCEND_TOOLKIT_HOME}/lib64:${ASCEND_TOOLKIT_HOME}/lib64/plugin/opskernel:${ASCEND_TOOLKIT_HOME}/lib64/plugin/nnengine:${ASCEND_TOOLKIT_HOME}/opp/built-in/op_impl/ai_core/tbe/op_tiling:${LD_LIBRARY_PATH}
ENV PYTHONPATH=${ASCEND_TOOLKIT_HOME}/python/site-packages:${ASCEND_TOOLKIT_HOME}/opp/built-in/op_impl/ai_core/tbe:${PYTHONPATH}
ENV PATH=${ASCEND_TOOLKIT_HOME}/bin:${ASCEND_TOOLKIT_HOME}/compiler/ccec_compiler/bin:${PATH}
ENV ASCEND_AICPU_PATH=${ASCEND_TOOLKIT_HOME}
ENV ASCEND_OPP_PATH=${ASCEND_TOOLKIT_HOME}/opp
ENV TOOLCHAIN_HOME=${ASCEND_TOOLKIT_HOME}/toolkit
ENV ASCEND_HOME_PATH=${ASCEND_TOOLKIT_HOME}

ENTRYPOINT ["/llama-cli" ]
24 changes: 13 additions & 11 deletions .devops/llama-cli-cuda.Dockerfile
Original file line number Diff line number Diff line change
@@ -1,35 +1,37 @@
ARG UBUNTU_VERSION=22.04
# This needs to generally match the container host's environment.
ARG CUDA_VERSION=11.7.1
ARG CUDA_VERSION=12.6.0
# Target the CUDA build image
ARG BASE_CUDA_DEV_CONTAINER=nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}
# Target the CUDA runtime image
ARG BASE_CUDA_RUN_CONTAINER=nvidia/cuda:${CUDA_VERSION}-runtime-ubuntu${UBUNTU_VERSION}

FROM ${BASE_CUDA_DEV_CONTAINER} AS build

# Unless otherwise specified, we make a fat build.
ARG CUDA_DOCKER_ARCH=all
# CUDA architecture to build for (defaults to all supported archs)
ARG CUDA_DOCKER_ARCH=default

RUN apt-get update && \
apt-get install -y build-essential git
apt-get install -y build-essential git cmake

WORKDIR /app

COPY . .

# Set nvcc architecture
ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
# Enable CUDA
ENV GGML_CUDA=1

RUN make -j$(nproc) llama-cli
# Use the default CUDA archs if not specified
RUN if [ "${CUDA_DOCKER_ARCH}" != "default" ]; then \
export CMAKE_ARGS="-DCMAKE_CUDA_ARCHITECTURES=${CUDA_DOCKER_ARCH}"; \
fi && \
cmake -B build -DGGML_CUDA=ON ${CMAKE_ARGS} -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined . && \
cmake --build build --config Release --target llama-cli -j$(nproc)

FROM ${BASE_CUDA_RUN_CONTAINER} AS runtime

RUN apt-get update && \
apt-get install -y libgomp1

COPY --from=build /app/llama-cli /llama-cli
COPY --from=build /app/build/ggml/src/libggml.so /libggml.so
COPY --from=build /app/build/src/libllama.so /libllama.so
COPY --from=build /app/build/bin/llama-cli /llama-cli

ENTRYPOINT [ "/llama-cli" ]
29 changes: 16 additions & 13 deletions .devops/llama-server-cuda.Dockerfile
Original file line number Diff line number Diff line change
@@ -1,38 +1,41 @@
ARG UBUNTU_VERSION=22.04
# This needs to generally match the container host's environment.
ARG CUDA_VERSION=11.7.1
ARG CUDA_VERSION=12.6.0
# Target the CUDA build image
ARG BASE_CUDA_DEV_CONTAINER=nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}
# Target the CUDA runtime image
ARG BASE_CUDA_RUN_CONTAINER=nvidia/cuda:${CUDA_VERSION}-runtime-ubuntu${UBUNTU_VERSION}

FROM ${BASE_CUDA_DEV_CONTAINER} AS build

# Unless otherwise specified, we make a fat build.
ARG CUDA_DOCKER_ARCH=all
# CUDA architecture to build for (defaults to all supported archs)
ARG CUDA_DOCKER_ARCH=default

RUN apt-get update && \
apt-get install -y build-essential git libcurl4-openssl-dev
apt-get install -y build-essential git cmake libcurl4-openssl-dev

WORKDIR /app

COPY . .

# Set nvcc architecture
ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
# Enable CUDA
ENV GGML_CUDA=1
# Enable cURL
ENV LLAMA_CURL=1

RUN make -j$(nproc) llama-server
# Use the default CUDA archs if not specified
RUN if [ "${CUDA_DOCKER_ARCH}" != "default" ]; then \
export CMAKE_ARGS="-DCMAKE_CUDA_ARCHITECTURES=${CUDA_DOCKER_ARCH}"; \
fi && \
cmake -B build -DGGML_CUDA=ON -DLLAMA_CURL=ON ${CMAKE_ARGS} -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined . && \
cmake --build build --config Release --target llama-server -j$(nproc)

FROM ${BASE_CUDA_RUN_CONTAINER} AS runtime

RUN apt-get update && \
apt-get install -y libcurl4-openssl-dev libgomp1 curl

COPY --from=build /app/llama-server /llama-server
COPY --from=build /app/build/ggml/src/libggml.so /libggml.so
COPY --from=build /app/build/src/libllama.so /libllama.so
COPY --from=build /app/build/bin/llama-server /llama-server

# Must be set to 0.0.0.0 so it can listen to requests from host machine
ENV LLAMA_ARG_HOST=0.0.0.0

HEALTHCHECK CMD [ "curl", "-f", "http://localhost:8080/health" ]

Expand Down
2 changes: 2 additions & 0 deletions .devops/llama-server-intel.Dockerfile
Original file line number Diff line number Diff line change
Expand Up @@ -26,6 +26,8 @@ RUN apt-get update && \
COPY --from=build /app/build/bin/llama-server /llama-server

ENV LC_ALL=C.utf8
# Must be set to 0.0.0.0 so it can listen to requests from host machine
ENV LLAMA_ARG_HOST=0.0.0.0

HEALTHCHECK CMD [ "curl", "-f", "http://localhost:8080/health" ]

Expand Down
2 changes: 2 additions & 0 deletions .devops/llama-server-rocm.Dockerfile
Original file line number Diff line number Diff line change
Expand Up @@ -39,6 +39,8 @@ ENV GPU_TARGETS=${ROCM_DOCKER_ARCH}
ENV GGML_HIPBLAS=1
ENV CC=/opt/rocm/llvm/bin/clang
ENV CXX=/opt/rocm/llvm/bin/clang++
# Must be set to 0.0.0.0 so it can listen to requests from host machine
ENV LLAMA_ARG_HOST=0.0.0.0

# Enable cURL
ENV LLAMA_CURL=1
Expand Down
2 changes: 2 additions & 0 deletions .devops/llama-server-vulkan.Dockerfile
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,8 @@ RUN cp /app/build/bin/llama-server /llama-server && \
rm -rf /app

ENV LC_ALL=C.utf8
# Must be set to 0.0.0.0 so it can listen to requests from host machine
ENV LLAMA_ARG_HOST=0.0.0.0

HEALTHCHECK CMD [ "curl", "-f", "http://localhost:8080/health" ]

Expand Down
2 changes: 2 additions & 0 deletions .devops/llama-server.Dockerfile
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,8 @@ RUN apt-get update && \
COPY --from=build /app/llama-server /llama-server

ENV LC_ALL=C.utf8
# Must be set to 0.0.0.0 so it can listen to requests from host machine
ENV LLAMA_ARG_HOST=0.0.0.0

HEALTHCHECK CMD [ "curl", "-f", "http://localhost:8080/health" ]

Expand Down
53 changes: 46 additions & 7 deletions .devops/nix/devshells.nix
Original file line number Diff line number Diff line change
@@ -1,13 +1,52 @@
{ inputs, ... }:

{
perSystem =
{ config, lib, ... }:
{
config,
lib,
system,
...
}:
{
devShells =
lib.concatMapAttrs
(name: package: {
${name} = package.passthru.shell;
${name + "-extra"} = package.passthru.shell-extra;
})
config.packages;
let
pkgs = import inputs.nixpkgs { inherit system; };
stdenv = pkgs.stdenv;
scripts = config.packages.python-scripts;
in
lib.pipe (config.packages) [
(lib.concatMapAttrs (
name: package: {
${name} = pkgs.mkShell {
name = "${name}";
inputsFrom = [ package ];
shellHook = ''
echo "Entering ${name} devShell"
'';
};
"${name}-extra" =
if (name == "python-scripts") then
null
else
pkgs.mkShell {
name = "${name}-extra";
inputsFrom = [
package
scripts
];
# Extra packages that *may* be used by some scripts
packages = [
pkgs.python3Packages.tiktoken
];
shellHook = ''
echo "Entering ${name} devShell"
addToSearchPath "LD_LIBRARY_PATH" "${lib.getLib stdenv.cc.cc}/lib"
'';
};
}
))
(lib.filterAttrs (name: value: value != null))
];
};
}
18 changes: 8 additions & 10 deletions .devops/nix/nixpkgs-instances.nix
Original file line number Diff line number Diff line change
Expand Up @@ -26,16 +26,14 @@
config.cudaSupport = true;
config.allowUnfreePredicate =
p:
builtins.all
(
license:
license.free
|| builtins.elem license.shortName [
"CUDA EULA"
"cuDNN EULA"
]
)
(p.meta.licenses or [ p.meta.license ]);
builtins.all (
license:
license.free
|| builtins.elem license.shortName [
"CUDA EULA"
"cuDNN EULA"
]
) (p.meta.licenses or [ p.meta.license ]);
};
# Ensure dependencies use ROCm consistently
pkgsRocm = import inputs.nixpkgs {
Expand Down
36 changes: 36 additions & 0 deletions .devops/nix/package-gguf-py.nix
Original file line number Diff line number Diff line change
@@ -0,0 +1,36 @@
{
lib,
llamaVersion,
numpy,
tqdm,
sentencepiece,
pyyaml,
poetry-core,
buildPythonPackage,
pytestCheckHook,
}:

buildPythonPackage {
pname = "gguf";
version = llamaVersion;
pyproject = true;
nativeBuildInputs = [ poetry-core ];
propagatedBuildInputs = [
numpy
tqdm
sentencepiece
pyyaml
];
src = lib.cleanSource ../../gguf-py;
pythonImportsCheck = [
"numpy"
"gguf"
];
nativeCheckInputs = [ pytestCheckHook ];
doCheck = true;
meta = with lib; {
description = "Python package for writing binary files in the GGUF format";
license = licenses.mit;
maintainers = [ maintainers.ditsuke ];
};
}
Loading

0 comments on commit d11bd3b

Please sign in to comment.