请问如何只部署一个bge-large-zh

# 把 chatglm3 删除即可。

version: '3.5'

services:

##################### 使用fastchat部署大模型 #####################
  fastchat-controller:
    restart: always
    container_name: fastchat-controller
    image: fastchat-cpu:v1.0
    build:
      context: .
      dockerfile: docker/fastchat/DockerfileCpu
    # 获取当前ip地址
    command: python3 -m fastchat.serve.controller --host 0.0.0.0 --port 21001

fastchat-api:
    restart: always
    container_name: fastchat-api
    image: fastchat-cpu:v1.0
    build:
      context: .
      dockerfile: docker/fastchat/DockerfileCpu
    # 获取当前ip地址 执行模型。
    ports:
      - "8000:8000"
    command: 
        bash -c "python3 -m fastchat.serve.openai_api_server --controller-address http://fastchat-controller:21001 \
         --host 0.0.0.0 --port 8000 "
    depends_on:
      - "fastchat-controller"
  fastchat-worker-bge-large-zh:
    restart: always
    container_name: fastchat-worker-bge-large-zh
    image: fastchat-gpu:v1.0
    volumes:
      - ./models/bge-large-zh:/data/models/bge-large-zh
    build:
      context: .
      dockerfile: docker/fastchat/DockerfileGpu
    # 是以 GPU 资源。
    deploy:
        resources:
            reservations:
                devices:
                  - driver: "nvidia"
                    count: "all"
                    capabilities: ["gpu"]
    # 获取当前ip地址 执行模型。
    command: 
        bash -c "python3 -m fastchat.serve.model_worker --model-names bge-large-zh,text-embedding-ada-002 \
         --model-path /data/models/bge-large-zh --controller-address http://fastchat-controller:21001 \
         --worker-address http://fastchat-worker-bge-large-zh:8001 --host 0.0.0.0 --port 8001 "
    depends_on:
      - "fastchat-controller"

networks:
  default:
    name: knowledge-network

fly-llm/dify-docker-compose

内容风险标识

评论 (1)

fly-llm/dify-docker-compose .gitee-modal { width: 500px !important; }

内容风险标识