diff --git a/AI/vllm-cpu/0.10.1/22.03-lts-sp4/Dockerfile b/AI/vllm-cpu/0.10.1/22.03-lts-sp4/Dockerfile new file mode 100644 index 0000000000000000000000000000000000000000..b40c0d93fa486ab21e1d3d73d88b9053dabf510f --- /dev/null +++ b/AI/vllm-cpu/0.10.1/22.03-lts-sp4/Dockerfile @@ -0,0 +1,54 @@ +ARG BASE=openeuler/openeuler:22.03-lts-sp4 +ARG VERSION=v0.10.1 + +FROM ${BASE} AS base + +ARG VERSION + +RUN yum update -y && \ + yum install -y \ + ccache \ + git \ + curl \ + wget \ + ca-certificates \ + gcc-toolset-12-gcc \ + gcc-toolset-12-gcc-c++ \ + python \ + python-pip \ + python3-devel \ + git \ + ffmpeg \ + libSM \ + libXext \ + jq \ + lsof \ + numactl-devel && \ + yum clean all + +ENV PATH=/opt/openEuler/gcc-toolset-12/root/usr/bin/:$PATH +ENV LD_LIBRARY_PATH=/opt/openEuler/gcc-toolset-12/root/usr/lib64/:$LD_LIBRARY_PATH + +WORKDIR /workspace + +RUN git clone -b ${VERSION} https://github.com/vllm-project/vllm.git + +WORKDIR /workspace/vllm + +RUN pip install -r requirements/cpu.txt + +FROM base AS build + +RUN pip install -r requirements/cpu-build.txt + +RUN VLLM_TARGET_DEVICE=cpu python3 setup.py bdist_wheel + +FROM base AS release + +COPY --from=build /workspace/vllm/dist /workspace/vllm/dist + +RUN pip install dist/*.whl + +WORKDIR /workspace + +ENTRYPOINT ["python", "-m", "vllm.entrypoints.openai.api_server"] \ No newline at end of file diff --git a/AI/vllm-cpu/0.10.1/24.03-lts/Dockerfile b/AI/vllm-cpu/0.10.1/24.03-lts/Dockerfile new file mode 100644 index 0000000000000000000000000000000000000000..d81feb542db6506da29119f7ba3415cae0e65dd3 --- /dev/null +++ b/AI/vllm-cpu/0.10.1/24.03-lts/Dockerfile @@ -0,0 +1,51 @@ +ARG BASE=openeuler/openeuler:24.03-lts +ARG VERSION=v0.10.1 + +FROM ${BASE} AS base + +ARG VERSION + +RUN yum update -y && \ + yum install -y \ + ccache \ + git \ + curl \ + wget \ + ca-certificates \ + gcc \ + gcc-c++ \ + python \ + python-pip \ + python3-devel \ + git \ + ffmpeg \ + libSM \ + libXext \ + jq \ + lsof \ + numactl-devel && \ + yum clean all + +WORKDIR /workspace + +RUN git clone -b ${VERSION} https://github.com/vllm-project/vllm.git + +WORKDIR /workspace/vllm + +RUN pip install -r requirements/cpu.txt + +FROM base AS build + +RUN pip install -r requirements/cpu-build.txt + +RUN VLLM_TARGET_DEVICE=cpu python3 setup.py bdist_wheel + +FROM base AS release + +COPY --from=build /workspace/vllm/dist /workspace/vllm/dist + +RUN pip install dist/*.whl + +WORKDIR /workspace + +ENTRYPOINT ["python", "-m", "vllm.entrypoints.openai.api_server"] diff --git a/AI/vllm-cpu/README.md b/AI/vllm-cpu/README.md index 0f2b7e30435df5ea4a42fb4b5fa7166a94187ad0..9661ea1e9a66981577d6c79cb7a828438eba47ca 100644 --- a/AI/vllm-cpu/README.md +++ b/AI/vllm-cpu/README.md @@ -44,7 +44,8 @@ The tag of each vLLM docker image is consist of the version of vLLM and the vers |[0.9.0-oe2403lts](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.9.0/24.03-lts/Dockerfile)| vLLM 0.9.0 on openEuler 24.03-LTS | amd64, arm64 | |[0.9.1-oe2203sp4](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.9.1/22.03-lts-sp4/Dockerfile)| vLLM 0.9.1 on openEuler 22.03-LTS-SP4 | amd64, arm64 | |[0.9.1-oe2403lts](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.9.1/24.03-lts/Dockerfile)| vLLM 0.9.1 on openEuler 24.03-LTS | amd64, arm64 | - +|[0.10.1-oe2203sp4](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.10.1/22.03-lts-sp4/Dockerfile)| vLLM 0.10.1 on openEuler 22.03-LTS-SP4 | amd64, arm64 | +|[0.10.1-oe2403lts](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.10.1/24.03-lts/Dockerfile)| vLLM 0.10.1 on openEuler 24.03-LTS | amd64, arm64 | # Usage diff --git a/AI/vllm-cpu/doc/image-info.yml b/AI/vllm-cpu/doc/image-info.yml new file mode 100644 index 0000000000000000000000000000000000000000..ffdfd48a1e351ff32142141682963afe8a498bab --- /dev/null +++ b/AI/vllm-cpu/doc/image-info.yml @@ -0,0 +1,85 @@ +name: vllm-cpu +category: ai +description: vLLM是一个高性能的、可扩展的、易于使用的用于大语言模型（LLM）推理的库。它支持CPU和GPU推理，并且可以与流行的深度学习框架（如PyTorch和TensorFlow）无缝集成。 +environment: | + 本应用在Docker环境中运行，安装Docker执行如下命令 + ``` + yum install -y docker + ``` +tags: | + vLLM镜像的Tag由其版本信息和基础镜像版本信息组成，详细内容如下 + + | Tags | Currently | Architectures| + |--|--|--| + |[0.6.3-oe2403lts](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.6.3/24.03-lts/Dockerfile)| vLLM 0.6.3 on openEuler 24.03-LTS | amd64 | + |[0.8.3-oe2203sp4](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.8.3/22.03-lts-sp4/Dockerfile)| vLLM 0.8.3 on openEuler 22.03-LTS-SP4 | amd64, arm64 | + |[0.8.3-oe2403lts](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.8.3/24.03-lts/Dockerfile)| vLLM 0.8.3 on openEuler 24.03-LTS | amd64, arm64 | + |[0.8.4-oe2203sp4](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.8.4/22.03-lts-sp4/Dockerfile)| vLLM 0.8.4 on openEuler 22.03-LTS-SP4 | amd64 | + |[0.8.4-oe2403lts](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.8.4/24.03-lts/Dockerfile)| vLLM 0.8.4 on openEuler 24.03-LTS | amd64 | + |[0.8.5-oe2203sp4](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.8.5/22.03-lts-sp4/Dockerfile)| vLLM 0.8.5 on openEuler 22.03-LTS-SP4 | amd64, arm64 | + |[0.8.5-oe2403lts](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.8.5/24.03-lts/Dockerfile)| vLLM 0.8.5 on openEuler 24.03-LTS | amd64, arm64 | + |[0.9.0-oe2203sp4](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.9.0/22.03-lts-sp4/Dockerfile)| vLLM 0.9.0 on openEuler 22.03-LTS-SP4 | amd64, arm64 | + |[0.9.0-oe2403lts](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.9.0/24.03-lts/Dockerfile)| vLLM 0.9.0 on openEuler 24.03-LTS | amd64, arm64 | + |[0.9.1-oe2203sp4](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.9.1/22.03-lts-sp4/Dockerfile)| vLLM 0.9.1 on openEuler 22.03-LTS-SP4 | amd64, arm64 | + |[0.9.1-oe2403lts](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.9.1/24.03-lts/Dockerfile)| vLLM 0.9.1 on openEuler 24.03-LTS | amd64, arm64 | + |[0.10.1-oe2203sp4](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.10.1/22.03-lts-sp4/Dockerfile)| vLLM 0.10.1 on openEuler 22.03-LTS-SP4 | amd64, arm64 | + |[0.10.1-oe2403lts](https://gitee.com/openeuler/openeuler-docker-images/blob/master/AI/vllm-cpu/0.10.1/24.03-lts/Dockerfile)| vLLM 0.10.1 on openEuler 24.03-LTS | amd64, arm64 | + +download: | + 拉取镜像到本地 + ``` + docker pull openeuler/vllm-cpu:{Tag} + ``` + +usage: | + - 启动容器 + ``` + docker run \ + --name my-vllm \ + -it openeuler/vllm-cpu:{Tag} bash + ``` + 用户可根据自身需求选择对应的硬件设备{device}、对应版本的{Tag}以及容器启动的选项。 + + - 参数说明 + | 配置项 | 描述 | + |--|--| + | `--name my-vllm` | 容器名称。| + | `-it` | 以交互模式启动容器。 | + | `openeuler/vllm-cpu:{Tag}` | 指定要运行的镜像为 `openeuler/vllm-cpu`，其中` {Tag}` 是需要替换的镜像标签。 | + + - 容器测试 + + 查看运行日志 + ``` + docker logs -f my-vllm + ``` + + 使用shell交互 + ``` + docker exec -it my-vllm /bin/bash + ``` + +license: BSD 3-Clause license +similar_packages: + - Ollama: Ollama是一个用于在本地运行大型语言模型（LLM）的工具，旨在简化模型的部署和管理。它支持多种LLM，并提供了一个统一的接口来与这些模型进行交互。 + - ONNX Runtime: ONNX Runtime是一个高性能的推理引擎，用于运行ONNX（Open Neural Network Exchange）格式的机器学习模型。ONNX是一个开放的生态系统，旨在促进不同深度学习框架之间的互操作性。 +dependency: + - transformers + - datasets + - torch + - sentencepiece + - protobuf + - numpy + - psutil + - aiohttp + - aiofiles + - fastapi + - uvicorn + +homepage: https://github.com/vllm-project/vllm +upstream: + version_url: vllm-project/vllm + backend: GitHub + version_scheme: RPM + version_filter: rc + version_prefix: v \ No newline at end of file diff --git a/AI/vllm-cpu/doc/picture/logo.png b/AI/vllm-cpu/doc/picture/logo.png new file mode 100644 index 0000000000000000000000000000000000000000..e1c94e01947825f454087061c81d7a068f8dfe6a Binary files /dev/null and b/AI/vllm-cpu/doc/picture/logo.png differ diff --git a/AI/vllm-cpu/meta.yml b/AI/vllm-cpu/meta.yml index a37dd739662de91525d1bc8dbed3886b4666666a..d5fb3998bc81cfd8c6eda78ca09973a831256663 100644 --- a/AI/vllm-cpu/meta.yml +++ b/AI/vllm-cpu/meta.yml @@ -29,3 +29,9 @@ 0.9.1-oe2403lts: path: 0.9.1/24.03-lts/Dockerfile + +0.10.1-oe2203sp4: + path: 0.10.1/22.03-lts-sp4/Dockerfile + +0.10.1-oe2403lts: + path: 0.10.1/24.03-lts/Dockerfile \ No newline at end of file