tf-chat-glm2-6b.yml

ROSTemplateFormatVersion: '2015-09-01'
Transform: Aliyun::Terraform-v1.5
Workspace:
  main.tf: |-
    variable "zone_id" {
      type        = string
      description = <<EOT
      {
        "AssociationProperty": "ZoneId",
        "Label": {
          "zh-cn": "可用区ID",
          "en": "Zone ID"
        }
      }
      EOT
    }

    variable "instance_type" {
      type        = string
      description = <<EOT
      {
        "Label": {
            "zh-cn": "实例类型",
            "en": "Instance Type"
        },
        "AssociationProperty": "ALIYUN::ECS::Instance::InstanceType",
        "AssociationPropertyMetadata": {
          "Constraints": {
            "Memory": [
              64
            ],
          "Architecture": "X86"
          }
        }
      }
      EOT
      default     = "ecs.hfg8i.4xlarge"
    }

    resource "alicloud_vpc" "vpc" {
      vpc_name   = "chat-glm2-6b-vpc"
      cidr_block = "192.168.0.0/16"
    }

    resource "alicloud_vswitch" "vswitch" {
      vpc_id     = alicloud_vpc.vpc.id
      zone_id    = var.zone_id
      cidr_block = "192.168.0.0/24"
    }


    resource "alicloud_security_group" "group" {
      vpc_id = alicloud_vpc.vpc.id
    }

    resource "alicloud_security_group_rule" "rule" {
      type              = "ingress"
      ip_protocol       = "tcp"
      nic_type          = "intranet"
      policy            = "accept"
      port_range        = "7860/7860"
      priority          = 1
      security_group_id = alicloud_security_group.group.id
      cidr_ip           = "0.0.0.0/0"
    }

    resource "alicloud_instance" "ecs" {
      availability_zone          = var.zone_id
      security_groups            = alicloud_security_group.group.*.id
      instance_type              = var.instance_type
      system_disk_category       = "cloud_essd"
      image_id                   = "aliyun_3_x64_20G_alibase_20240528.vhd"
      instance_name              = "Chat-GLM2-6B"
      vswitch_id                 = alicloud_vswitch.vswitch.id
      internet_max_bandwidth_out = 100
      system_disk_size = 100
    }


    locals {
      command         = <<EOF
    #!/bin/bash
    sudo dnf config-manager --add-repo=https://mirrors.aliyun.com/docker-ce/linux/centos/docker-ce.repo
    sudo dnf -y install dnf-plugin-releasever-adapter --repo alinux3-plus
    sudo dnf -y install docker-ce --nobest
    sudo systemctl start docker
    sudo systemctl enable docker
    sudo docker pull intel/xfastertransformer:1.3.1
    sudo docker run --name xFT -h xFT --privileged --shm-size=16g --network host -v /mnt:/mnt -w /mnt/xFasterTransformer intel/xfastertransformer:1.3.1 sh -c "tail -f /dev/null" &
    sleep 5
    sudo docker exec xFT sh -c '
    apt update
    apt install -y wget git git-lfs vim tmux
    cd /root/xFasterTransformer
    git lfs install
    mkdir /mnt/data
    cd /mnt/data
    git clone --depth 1 https://www.modelscope.cn/ZhipuAI/chatglm2-6b.git /mnt/data/chatglm2-6b
    python -c "import xfastertransformer as xft; xft.ChatGLM2Convert().convert(\"/mnt/data/chatglm2-6b\")"
    mkdir -p ~/.config/pip
    cat > ~/.config/pip/pip.conf <<EOX
    [global]
    index-url=http://mirrors.cloud.aliyuncs.com/pypi/simple/

    [install]
    trusted-host=mirrors.cloud.aliyuncs.com
    EOX

    cd /root/xFasterTransformer/examples/web_demo
    pip install -r requirements.txt
    cpu_count=$(lscpu | grep "^CPU(s):" | awk "{print \$NF}")
    OMP_NUM_THREADS=$((cpu_count / 2))
    export LD_PRELOAD=libiomp5.so
    export GRADIO_SERVER_NAME="0.0.0.0"
    cpu_list=$(seq -s, 0 2 $((cpu_count - 2)))
    nohup numactl -C $cpu_list -m 0 python ChatGLM2.py -t /mnt/data/chatglm2-6b -m /mnt/data/chatglm2-6b-xft -d bf16 > output.log 2>&1 &
    sleep 5
    '
    EOF
      base_64_command = base64encode(local.command)
    }


    resource "alicloud_ecs_command" "command" {
      name            = "chat-glm2-6b-command"
      command_content = local.base_64_command
      type            = "RunShellScript"
      timeout         = 3600
      working_dir = "/root"
    }

    resource "alicloud_ecs_invocation" "default" {
      command_id  = alicloud_ecs_command.command.id
      instance_id = [alicloud_instance.ecs.id]
      timeouts {
        create = "3600s"
      }
    }

    output "Url" {
      description = <<EOT
      {
        "Label": "Web 访问地址",
        "Description": "ChatGLM2-6B Chat页面访问地址."
      }
      EOT
      value = format("http://%s:7860", alicloud_instance.ecs.public_ip)
    }
Metadata:
  ALIYUN::ROS::Interface:
    TemplateTags:
    - acs:document-help:ecs:使用基于英特尔CPU的c8i实例部署ChatGLM2-6B