From 657cef8557d1080d5cb55d737a7de7e0c777ec37 Mon Sep 17 00:00:00 2001
From: icew1998 <wangdongsheng25@huawei.com>
Date: Wed, 5 Jun 2024 09:36:15 +0000
Subject: [PATCH] =?UTF-8?q?=E6=B7=BB=E5=8A=A0=E6=94=AF=E6=8C=81AI=E5=A4=A7?=
 =?UTF-8?q?=E6=A8=A1=E5=9E=8B=E5=BC=80=E7=AE=B1=E5=8D=B3=E7=94=A8=E7=9A=84?=
 =?UTF-8?q?=E5=AE=B9=E5=99=A8=E5=8C=96=E6=9C=8D=E5=8A=A1=E9=95=9C=E5=83=8F?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Signed-off-by: icew1998 <wangdongsheng25@huawei.com>
---
 llm-server/1.0.0.cpu/22.03-lts-sp3/Dockerfile | 23 ++++++++++++
 llm-server/1.0.0.gpu/22.03-lts-sp3/Dockerfile | 20 ++++++++++
 llm-server/README.md                          | 37 +++++++++++++++++++
 llm-server/meta.yml                           |  4 ++
 4 files changed, 84 insertions(+)
 create mode 100644 llm-server/1.0.0.cpu/22.03-lts-sp3/Dockerfile
 create mode 100644 llm-server/1.0.0.gpu/22.03-lts-sp3/Dockerfile
 create mode 100644 llm-server/README.md
 create mode 100644 llm-server/meta.yml

diff --git a/llm-server/1.0.0.cpu/22.03-lts-sp3/Dockerfile b/llm-server/1.0.0.cpu/22.03-lts-sp3/Dockerfile
new file mode 100644
index 00000000..41173fe3
--- /dev/null
+++ b/llm-server/1.0.0.cpu/22.03-lts-sp3/Dockerfile
@@ -0,0 +1,23 @@
+FROM openeuler/openeuler:22.03-lts-sp3
+
+RUN sed -i 's|http://repo.openeuler.org/|https://mirrors.huaweicloud.com/openeuler/|g' /etc/yum.repos.d/openEuler.repo &&\
+    yum update -y &&\
+    yum install -y python3 python3-pip shadow-utils cmake gcc g++ git make &&\
+    yum clean all
+RUN git clone https://github.com/OpenMathLib/OpenBLAS.git && \
+    cd OpenBLAS && \
+    make && \
+    make install && \
+    cd .. && \
+    rm -rf OpenBLAS
+RUN CMAKE_ARGS="-DLLAMA_BLAS=ON -DLLAMA_BLAS_VENDOR=OpenBLAS" pip3 install -i https://pypi.tuna.tsinghua.edu.cn/simple --no-cache-dir guidance llama-cpp-python[server]
+
+ENV KEY=sk-123456
+ENV MODEL=/models/model.gguf
+ENV MODEL_NAME=qwen-1.5
+ENV THREADS=8
+ENV CONTEXT=8192
+
+EXPOSE 8000
+
+ENTRYPOINT python3 -m llama_cpp.server --host 0.0.0.0 --port 8000 --api_key $KEY --model $MODEL --model_alias $MODEL_NAME --n_threads $THREADS --n_ctx $CONTEXT
\ No newline at end of file
diff --git a/llm-server/1.0.0.gpu/22.03-lts-sp3/Dockerfile b/llm-server/1.0.0.gpu/22.03-lts-sp3/Dockerfile
new file mode 100644
index 00000000..9a5485de
--- /dev/null
+++ b/llm-server/1.0.0.gpu/22.03-lts-sp3/Dockerfile
@@ -0,0 +1,20 @@
+FROM nvidia/cuda:12.1.1-devel-rockylinux9
+
+RUN sed -e 's|^mirrorlist=|#mirrorlist=|g' \
+         -e 's|^#baseurl=http://dl.rockylinux.org/$contentdir|baseurl=https://mirror.nju.edu.cn/rocky|g' \
+         -i.bak \
+         /etc/yum.repos.d/rocky-extras.repo \
+         /etc/yum.repos.d/rocky.repo && \
+    dnf -y update && \
+    dnf -y install python3 python3-devel cmake gcc g++
+
+RUN CMAKE_ARGS="-DLLAMA_CUDA=on" pip3 install --no-cache-dir -i https://pypi.tuna.tsinghua.edu.cn/simple -v guidance llama-cpp-python[server]
+
+ENV KEY=sk-123456
+ENV MODEL=/models/model.gguf
+ENV MODEL_NAME=qwen-1.5
+ENV CONTEXT=8192
+
+EXPOSE 8000
+
+ENTRYPOINT python3 -m llama_cpp.server --host 0.0.0.0 --port 8000 --api_key $KEY --model $MODEL --model_alias $MODEL_NAME --n_gpu_layers -1 --n_ctx $CONTEXT
\ No newline at end of file
diff --git a/llm-server/README.md b/llm-server/README.md
new file mode 100644
index 00000000..2434435a
--- /dev/null
+++ b/llm-server/README.md
@@ -0,0 +1,37 @@
+# 支持百川、chatglm、星火等AI大模型的容器化封装
+
+已配好相关依赖，分为CPU和GPU版本，降低使用门槛，开箱即用。
+
+### 启动方式
+docker-compose
+```
+version: '3'
+services:
+  model:
+    image: openeuler/llm-server:1.0.0.gpu-oe2203sp3   #镜像名称与Tag
+    restart: on-failure:5
+    ports:
+      - 10011:8000    #监听端口号，修改“10011”以更换端口
+    volumes:
+      - /root/models:/models
+    environment:
+      - MODEL=/models/qwen1_5-7b-chat-q4_k_m.gguf  #容器内的模型文件路径
+      - MODEL_NAME=qwen1.5  #模型名称
+      - KEY=sk-12345678  #API Key
+      - CONTEXT=8192  #上下文大小
+      - THREADS=8    #CPU线程数，仅CPU部署时需要
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              count: all
+              capabilities: [gpu]
+```
+```
+docker-compose -f docker-compose.yaml up
+```
+或者
+```
+docker run -d --gpus all --restart on-failure:5 -p 宿主机端口:容器端口(默认8000) -v 宿主机模型挂载路径:/models -e MODEL=/models/模型名(gguf格式) -e MODEL_NAME=baichuan-7b(自定义模型名称) -e KEY=sk-12345678(api key) image_name:tag
+```
\ No newline at end of file
diff --git a/llm-server/meta.yml b/llm-server/meta.yml
new file mode 100644
index 00000000..28dd2abf
--- /dev/null
+++ b/llm-server/meta.yml
@@ -0,0 +1,4 @@
+1.0.0.gpu-oe2203sp3:
+  llm-server/1.0.0.gpu/22.03-lts-sp3/Dockerfile
+1.0.0.cpu-oe2203sp3:
+  llm-server/1.0.0.cpu/22.03-lts-sp3/Dockerfile
\ No newline at end of file
-- 
Gitee