Add vLLM
This commit is contained in:
parent
d1ae090cf3
commit
f7031fda8a
114
Dockerfile
114
Dockerfile
@ -582,3 +582,117 @@ COPY /frontend/ /opt/backstory/frontend/
|
||||
ENV PATH=/opt/backstory:$PATH
|
||||
|
||||
ENTRYPOINT [ "/entrypoint.sh" ]
|
||||
|
||||
|
||||
FROM ubuntu:24.04 AS ollama-ov-server
|
||||
|
||||
SHELL ["/bin/bash", "-c"]
|
||||
|
||||
RUN apt-get update && apt install -y software-properties-common libtbb-dev
|
||||
RUN add-apt-repository ppa:deadsnakes/ppa \
|
||||
&& apt-get update \
|
||||
&& apt-get install -y python3.10 net-tools
|
||||
RUN ln -sf /usr/bin/python3.10 /usr/bin/python3
|
||||
|
||||
RUN apt-get install -y ca-certificates git wget curl gcc g++ \
|
||||
&& apt-get clean \
|
||||
&& rm -rf /var/lib/apt/lists/*
|
||||
|
||||
WORKDIR /home/ollama_ov_server
|
||||
ARG GOVERSION=1.24.1
|
||||
RUN curl -fsSL https://golang.org/dl/go${GOVERSION}.linux-$(case $(uname -m) in x86_64) echo amd64 ;; aarch64) echo arm64 ;; esac).tar.gz | tar xz -C /usr/local
|
||||
ENV PATH=/usr/local/go/bin:$PATH
|
||||
|
||||
RUN wget https://storage.openvinotoolkit.org/repositories/openvino_genai/packages/nightly/2025.2.0.0.dev20250513/openvino_genai_ubuntu24_2025.2.0.0.dev20250513_x86_64.tar.gz
|
||||
RUN tar -xzf openvino_genai_ubuntu24_2025.2.0.0.dev20250513_x86_64.tar.gz
|
||||
ENV GENAI_DIR=/home/ollama_ov_server/openvino_genai_ubuntu24_2025.2.0.0.dev20250513_x86_64
|
||||
|
||||
RUN source /home/ollama_ov_server/openvino_genai_ubuntu24_2025.2.0.0.dev20250513_x86_64/setupvars.sh
|
||||
|
||||
ENV CGO_ENABLED=1
|
||||
ENV GODEBUG=cgocheck=0
|
||||
|
||||
ENV CGO_LDFLAGS=-L$GENAI_DIR/runtime/lib/intel64
|
||||
ENV CGO_CFLAGS=-I$GENAI_DIR/runtime/include
|
||||
|
||||
WORKDIR /home/ollama_ov_server
|
||||
RUN git clone https://github.com/openvinotoolkit/openvino_contrib.git
|
||||
WORKDIR /home/ollama_ov_server/openvino_contrib/modules/ollama_openvino
|
||||
|
||||
RUN go build -o /usr/bin/ollama .
|
||||
|
||||
ENV OLLAMA_HOST=0.0.0.0:11434
|
||||
EXPOSE 11434
|
||||
|
||||
RUN apt-get update \
|
||||
&& DEBIAN_FRONTEND=noninteractive apt-get install -y \
|
||||
pip \
|
||||
&& apt-get clean \
|
||||
&& rm -rf /var/lib/apt/lists/{apt,dpkg,cache,log}
|
||||
|
||||
RUN pip install huggingface_hub modelscope
|
||||
|
||||
#ENV model=Qwen3-4B-int4-ov
|
||||
#ENV model=Qwen3-8B-int4-ov -- didn't work
|
||||
#RUN huggingface-cli download OpenVINO/${model}
|
||||
#RUN modelscope download --model OpenVINO/${model} --local_dir ./${model}
|
||||
|
||||
#RUN tar -zcvf /root/.ollama/models/${model}.tar.gz /root/.cache/hub/models--OpenVINO--${model}
|
||||
#RUN { \
|
||||
# echo "FROM ${model}.tar.gz" ; \
|
||||
# echo "ModelType 'OpenVINO'" ; \
|
||||
#} > /root/.ollama/models/Modelfile
|
||||
#
|
||||
#RUN /bin/bash -c "source /home/ollama_ov_server/openvino_genai_ubuntu24_2025.2.0.0.dev20250513_x86_64/setupvars.sh && /usr/bin/ollama create ${model}:v1 -f /root/.ollama/models/Modelfile"
|
||||
|
||||
ENTRYPOINT ["/bin/bash", "-c", "source /home/ollama_ov_server/openvino_genai_ubuntu24_2025.2.0.0.dev20250513_x86_64/setupvars.sh && /usr/bin/ollama serve"]
|
||||
|
||||
FROM python AS vllm
|
||||
|
||||
RUN apt-get update \
|
||||
&& DEBIAN_FRONTEND=noninteractive apt-get install -y \
|
||||
git \
|
||||
&& apt-get clean \
|
||||
&& rm -rf /var/lib/apt/lists/{apt,dpkg,cache,log}
|
||||
|
||||
WORKDIR /opt
|
||||
|
||||
RUN git clone https://github.com/vllm-project/vllm.git
|
||||
|
||||
WORKDIR /opt/vllm
|
||||
|
||||
RUN wget -O - https://astral.sh/uv/install.sh | sh
|
||||
ENV PATH=~/.local/bin:$PATH
|
||||
|
||||
RUN { \
|
||||
echo '#!/bin/bash' ; \
|
||||
echo 'source /opt/vllm/.venv/bin/activate' ; \
|
||||
echo 'if [[ "${1}" != "" ]]; then bash -c "${@}"; else bash -i; fi' ; \
|
||||
} > /opt/vllm/shell ; \
|
||||
chmod +x /opt/vllm/shell
|
||||
|
||||
RUN uv venv --python 3.12 --seed
|
||||
|
||||
SHELL [ "/opt/vllm/shell" ]
|
||||
|
||||
RUN pip install --upgrade pip ; \
|
||||
pip install -v -r requirements/xpu.txt
|
||||
RUN VLLM_TARGET_DEVICE=xpu python setup.py install
|
||||
|
||||
SHELL [ "/bin/bash", "-c" ]
|
||||
|
||||
RUN { \
|
||||
echo '#!/bin/bash'; \
|
||||
echo 'echo "Container: vLLM"'; \
|
||||
echo 'set -e'; \
|
||||
echo 'python -m vllm.entrypoints.openai.api_server \'; \
|
||||
echo ' --model=facebook/opt-13b \' ; \
|
||||
echo ' --dtype=bfloat16 \' ; \
|
||||
echo ' --max_model_len=1024 \' ; \
|
||||
echo ' --distributed-executor-backend=ray \' ; \
|
||||
echo ' --pipeline-parallel-size=2 \' ; \
|
||||
echo ' -tp=8' ; \
|
||||
} > /entrypoint.sh \
|
||||
&& chmod +x /entrypoint.sh
|
||||
|
||||
ENTRYPOINT [ "/entrypoint.sh" ]
|
||||
@ -18,9 +18,13 @@ services:
|
||||
# To use Anthropic, uncomment the following lines and comment out the OpenAI lines
|
||||
# - DEFAULT_LLM_PROVIDER=anthropic
|
||||
# - MODEL_NAME=claude-3-5-haiku-latest
|
||||
- DEFAULT_LLM_PROVIDER=openai
|
||||
- DEFAULT_LLM_PROVIDER=ollama
|
||||
- MODEL_NAME=${MODEL_NAME:-qwen2.5:7b}
|
||||
- OPENAI_URL=http://ollama:11434
|
||||
- OLLAMA_HOST=http://ollama:11434
|
||||
# Test with OpenVINO; it doesn't work though
|
||||
# - MODEL_NAME=Qwen3-4B-int4-ov:v1
|
||||
# - OLLAMA_HOST=http://ollama-ov-server:11434
|
||||
|
||||
devices:
|
||||
- /dev/dri:/dev/dri
|
||||
depends_on:
|
||||
@ -60,9 +64,9 @@ services:
|
||||
- REDIS_URL=redis://redis:6379
|
||||
- REDIS_DB=1
|
||||
- SSL_ENABLED=false
|
||||
- DEFAULT_LLM_PROVIDER=openai
|
||||
- DEFAULT_LLM_PROVIDER=ollama
|
||||
- MODEL_NAME=${MODEL_NAME:-qwen2.5:7b}
|
||||
- OPENAI_URL=http://ollama:11434
|
||||
- OLLAMA_HOST=http://ollama:11434
|
||||
devices:
|
||||
- /dev/dri:/dev/dri
|
||||
depends_on:
|
||||
@ -184,6 +188,50 @@ services:
|
||||
- CAP_PERFMON # Access to perf_events (vs. overloaded CAP_SYS_ADMIN)
|
||||
- CAP_SYS_PTRACE # PTRACE_MODE_READ_REALCREDS ptrace access mode check
|
||||
|
||||
ollama-ov-server:
|
||||
build:
|
||||
context: .
|
||||
dockerfile: Dockerfile
|
||||
target: ollama-ov-server
|
||||
container_name: ollama-ov-server
|
||||
restart: "no"
|
||||
env_file:
|
||||
- .env
|
||||
environment:
|
||||
- OLLAMA_HOST=0.0.0.0
|
||||
- ONEAPI_DEVICE_SELECTOR=level_zero:0
|
||||
devices:
|
||||
- /dev/dri:/dev/dri
|
||||
ports:
|
||||
- 11435:11434 # ollama serve port
|
||||
networks:
|
||||
- internal
|
||||
volumes:
|
||||
- ./cache:/root/.cache # Cache hub models and neo_compiler_cache
|
||||
- ./ollama:/root/.ollama # Cache the ollama models
|
||||
|
||||
vllm:
|
||||
build:
|
||||
context: .
|
||||
dockerfile: Dockerfile
|
||||
target: vllm
|
||||
container_name: vllm
|
||||
restart: "always"
|
||||
env_file:
|
||||
- .env
|
||||
environment:
|
||||
- OLLAMA_HOST=0.0.0.0
|
||||
# - ONEAPI_DEVICE_SELECTOR=level_zero:0
|
||||
devices:
|
||||
- /dev/dri:/dev/dri
|
||||
ports:
|
||||
- 11438:8000 # ollama serve port
|
||||
networks:
|
||||
- internal
|
||||
volumes:
|
||||
- ./cache:/root/.cache # Cache hub models and neo_compiler_cache
|
||||
- ./ollama:/root/.ollama # Cache the ollama models
|
||||
|
||||
jupyter:
|
||||
build:
|
||||
context: .
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user