Skip to content

Commit c9ecf62

Browse files
compiladecompilade
committed
Merge branch 'master' into compilade/mamba2
2 parents 1ee6c48 + a800ae4 commit c9ecf62

File tree

702 files changed

+103784
-67948
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

702 files changed

+103784
-67948
lines changed

‎.clang-tidy

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -17,8 +17,10 @@ Checks: >
1717
-clang-analyzer-security.insecureAPI.DeprecatedOrUnsafeBufferHandling,
1818
performance-*,
1919
portability-*,
20+
-portability-simd-intrinsics,
2021
misc-*,
2122
-misc-const-correctness,
2223
-misc-non-private-member-variables-in-classes,
2324
-misc-no-recursion,
25+
-misc-use-anonymous-namespace,
2426
FormatStyle: none

‎.devops/cpu.Dockerfile

Lines changed: 92 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,92 @@
1+
ARG UBUNTU_VERSION=22.04
2+
3+
FROM ubuntu:$UBUNTU_VERSION AS build
4+
5+
ARG TARGETARCH
6+
7+
ARG GGML_CPU_ARM_ARCH=armv8-a
8+
9+
RUN apt-get update && \
10+
apt-get install -y build-essential git cmake libcurl4-openssl-dev
11+
12+
WORKDIR /app
13+
14+
COPY . .
15+
16+
RUN if [ "$TARGETARCH" = "amd64" ]; then \
17+
cmake -S . -B build -DCMAKE_BUILD_TYPE=Release -DLLAMA_CURL=ON -DGGML_NATIVE=OFF -DGGML_BACKEND_DL=ON -DGGML_CPU_ALL_VARIANTS=ON; \
18+
elif [ "$TARGETARCH" = "arm64" ]; then \
19+
cmake -S . -B build -DCMAKE_BUILD_TYPE=Release -DLLAMA_CURL=ON -DGGML_NATIVE=OFF -DGGML_CPU_ARM_ARCH=${GGML_CPU_ARM_ARCH}; \
20+
else \
21+
echo "Unsupported architecture"; \
22+
exit 1; \
23+
fi && \
24+
cmake --build build -j $(nproc)
25+
26+
RUN mkdir -p /app/lib && \
27+
find build -name "*.so" -exec cp {} /app/lib \;
28+
29+
RUN mkdir -p /app/full \
30+
&& cp build/bin/* /app/full \
31+
&& cp *.py /app/full \
32+
&& cp -r gguf-py /app/full \
33+
&& cp -r requirements /app/full \
34+
&& cp requirements.txt /app/full \
35+
&& cp .devops/tools.sh /app/full/tools.sh
36+
37+
## Base image
38+
FROM ubuntu:$UBUNTU_VERSION AS base
39+
40+
RUN apt-get update \
41+
&& apt-get install -y libgomp1 curl\
42+
&& apt autoremove -y \
43+
&& apt clean -y \
44+
&& rm -rf /tmp/* /var/tmp/* \
45+
&& find /var/cache/apt/archives /var/lib/apt/lists -not -name lock -type f -delete \
46+
&& find /var/cache -type f -delete
47+
48+
COPY --from=build /app/lib/ /app
49+
50+
### Full
51+
FROM base AS full
52+
53+
COPY --from=build /app/full /app
54+
55+
WORKDIR /app
56+
57+
RUN apt-get update \
58+
&& apt-get install -y \
59+
git \
60+
python3 \
61+
python3-pip \
62+
&& pip install --upgrade pip setuptools wheel \
63+
&& pip install -r requirements.txt \
64+
&& apt autoremove -y \
65+
&& apt clean -y \
66+
&& rm -rf /tmp/* /var/tmp/* \
67+
&& find /var/cache/apt/archives /var/lib/apt/lists -not -name lock -type f -delete \
68+
&& find /var/cache -type f -delete
69+
70+
ENTRYPOINT ["/app/tools.sh"]
71+
72+
### Light, CLI only
73+
FROM base AS light
74+
75+
COPY --from=build /app/full/llama-cli /app
76+
77+
WORKDIR /app
78+
79+
ENTRYPOINT [ "/app/llama-cli" ]
80+
81+
### Server, Server only
82+
FROM base AS server
83+
84+
ENV LLAMA_ARG_HOST=0.0.0.0
85+
86+
COPY --from=build /app/full/llama-server /app
87+
88+
WORKDIR /app
89+
90+
HEALTHCHECK CMD [ "curl", "-f", "http://localhost:8080/health" ]
91+
92+
ENTRYPOINT [ "/app/llama-server" ]

‎.devops/cuda.Dockerfile

Lines changed: 94 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,94 @@
1+
ARG UBUNTU_VERSION=22.04
2+
# This needs to generally match the container host's environment.
3+
ARG CUDA_VERSION=12.4.0
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 ${BASE_CUDA_DEV_CONTAINER} AS build
10+
11+
# CUDA architecture to build for (defaults to all supported archs)
12+
ARG CUDA_DOCKER_ARCH=default
13+
14+
RUN apt-get update && \
15+
apt-get install -y build-essential cmake python3 python3-pip git libcurl4-openssl-dev libgomp1
16+
17+
WORKDIR /app
18+
19+
COPY . .
20+
21+
RUN if [ "${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 -DLLAMA_CURL=ON ${CMAKE_ARGS} -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined . && \
25+
cmake --build build --config Release -j$(nproc)
26+
27+
RUN mkdir -p /app/lib && \
28+
find build -name "*.so" -exec cp {} /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+
&& pip install --upgrade pip setuptools wheel \
64+
&& pip install -r requirements.txt \
65+
&& apt autoremove -y \
66+
&& apt clean -y \
67+
&& rm -rf /tmp/* /var/tmp/* \
68+
&& find /var/cache/apt/archives /var/lib/apt/lists -not -name lock -type f -delete \
69+
&& find /var/cache -type f -delete
70+
71+
72+
ENTRYPOINT ["/app/tools.sh"]
73+
74+
### Light, CLI only
75+
FROM base AS light
76+
77+
COPY --from=build /app/full/llama-cli /app
78+
79+
WORKDIR /app
80+
81+
ENTRYPOINT [ "/app/llama-cli" ]
82+
83+
### Server, Server only
84+
FROM base AS server
85+
86+
ENV LLAMA_ARG_HOST=0.0.0.0
87+
88+
COPY --from=build /app/full/llama-server /app
89+
90+
WORKDIR /app
91+
92+
HEALTHCHECK CMD [ "curl", "-f", "http://localhost:8080/health" ]
93+
94+
ENTRYPOINT [ "/app/llama-server" ]

‎.devops/full-cuda.Dockerfile

Lines changed: 0 additions & 33 deletions
This file was deleted.

‎.devops/full-musa.Dockerfile

Lines changed: 0 additions & 26 deletions
This file was deleted.

‎.devops/full-rocm.Dockerfile

Lines changed: 0 additions & 50 deletions
This file was deleted.

‎.devops/full.Dockerfile

Lines changed: 0 additions & 25 deletions
This file was deleted.

0 commit comments

Comments
 (0)