# 定义流水线要使用的 Docker 镜像
image: image.sourcefind.cn:5000/dcu/admin/base/vllm:0.8.5-ubuntu22.04-dtk25.04.1-rc5-das1.6-py3.10-20250724

# 定义流水线的所有阶段
stages:
  - prepare
  - install
  - test

# === 阶段 1: 准备阶段 ===
download_assets:
  stage: prepare
  tags:
    - demos
  script:
    - echo "--- 1. Preparing assets ---"
    - pip install modelscope
    - nohup modelscope download --model "AI-ModelScope/sdxl-vae-fp16-fix" --local_dir "./sdxl-vae-fp16-fix" > downloadsdxl-vae-fp16-fix.log 2>&1 &
    - nohup modelscope download --model "stabilityai/stable-diffusion-xl-base-1.0" --local_dir "./stable-diffusion-xl-base-1.0" > downloadsdxl-stable-diffusion-xl-base-1.0.log 2>&1 &
    - curl -f -C - -o "rocblas-install-0626-bug.tar.gz" "https://wuzh01.hpccube.com:65015/efile/s/d/amVycnJycnk=/a6a7342d017b1748"
    - curl -f -C - -o "package-miopen-dev-0801-ubuntu20.tar.gz" "https://ksefile.hpccube.com:65241/efile/s/d/amVycnJycnk=/9c2334cb9cdb8b39"
    - echo "--- Asset download finished ---"
  cache:
    key: "assets-cache-${CI_COMMIT_REF_SLUG}"
    paths:
      - "sdxl-vae-fp16-fix/"
      - "stable-diffusion-xl-base-1.0/"
      - "rocblas-install-0626-bug.tar.gz"
      - "package-miopen-dev-0801-ubuntu20.tar.gz"

# === 阶段 2: 安装阶段 ===
setup_environment:
  stage: install
  tags:
    - demos
  script:
    - echo "--- 2. Setting up environment ---"
    - tar -xzvf "rocblas-install-0626-bug.tar.gz"
    - tar -xzvf "package-miopen-dev-0801-ubuntu20.tar.gz"
    - cd "stable-diffusion-xl-base-1.0"
    - mv vae vae_bak || true
    - mkdir -p vae
    - cp ../sdxl-vae-fp16-fix/c* ./vae/
    - cp ../sdxl-vae-fp16-fix/s* ./vae/
    - cp ../sdxl-vae-fp16-fix/d* ./vae/
    - cd ..
    - curl -f -C - -o "diffusers-0.33.1-py3-none-any.whl" "https://ksefile.hpccube.com:65241/efile/s/d/amVycnJycnk=/63ec0d10ce960f90"
    - curl -f -C - -o "lightop-0.5.0+das.dtk25041-cp310-cp310-linux_x86_64.whl" "https://ksefile.hpccube.com:65241/efile/s/d/amVycnJycnk=/da522f7e175bb092"
    - curl -f -C - -o "litserve-0.2.15-py3-none-any.whl" "https://ksefile.hpccube.com:65241/efile/s/d/amVycnJycnk=/36943fb1ca62ac3b"
    - pip install "diffusers-0.33.1-py3-none-any.whl" "lightop-0.5.0+das.dtk25041-cp310-cp310-linux_x86_64.whl" "litserve-0.2.15-py3-none-any.whl"
    - curl -f -C - -o "transformers.tar.gz" "https://ksefile.hpccube.com:65241/efile/s/d/amVycnJycnk=/f587d939a37d1727"
    - tar -xzvf "transformers.tar.gz"
    - echo "--- Environment setup finished ---"
  cache:
    key: "assets-cache-${CI_COMMIT_REF_SLUG}"
    paths:
      - "sdxl-vae-fp16-fix/"
      - "stable-diffusion-xl-base-1.0/"
      - "rocblas-install/"
      - "package-miopen-dev-0801-ubuntu20/"
      - "transformers/"
    policy: pull-push

# === 阶段 3: 测试阶段 ===
benchmark_sdxl:
  stage: test
  tags:
    - demos
  script:
    - echo "--- 3. Running SDXL benchmark ---"
    # 设置环境变量
    - export LD_LIBRARY_PATH="$CI_PROJECT_DIR/rocblas-install/lib/:$LD_LIBRARY_PATH"
    - export LD_LIBRARY_PATH="$CI_PROJECT_DIR/package-miopen-dev-0801-ubuntu20/lib/:$LD_LIBRARY_PATH"
    - export PYTORCH_MIOPEN_SUGGEST_NHWC=1
    - export PYTHONPATH="$CI_PROJECT_DIR/transformers:$PYTHONPATH"
    - export MODEL_PATH="$CI_PROJECT_DIR/stable-diffusion-xl-base-1.0"
    - echo "Starting benchmark script, full log will be saved to benchmark.log..."
    # 将 Python 脚本的所有输出（包括错误）都重定向到 benchmark.log 文件
    - python test.py 
  artifacts:
    name: "$CI_JOB_NAME-$CI_COMMIT_REF_NAME"
    paths:
      - "*.png"
      - "results.json"
      - "benchmark.log"  # <-- 关键：将日志文件也作为产物保存
    expire_in: 1 week
  cache:
    key: "assets-cache-${CI_COMMIT_REF_SLUG}"
    policy: pull