模型服务

MLflow 入门

Weston Bassler

Senior MLOps Engineer

MLflow 模型

标准化模型打包
记录模型
模型评估

模型部署

机器学习生命周期

¹ datacamp.com

REST API

/ping - 健康检查
/health - 健康检查
/version - 获取 MLflow 版本
/invocations - 模型打分
端口 5000

Invocations 端点

/invocations

No,Name,Subject
1,Bill Johnson,English
2,Gary Valentine,Mathematics

Content-Type:application/json 或 application/csv

{
    "1": {
        "No": "1",
        "Name": "Bill Johnson",
        "Subject": "English"
    },
    "2": {
        "No": "2",
        "Name": "Gary Valentine",
        "Subject": "Mathematics"
    }
}

CSV 格式

Pandas DataFrame
pandas_df.to_csv()

JSON 格式

dataframe_split - 分裂（split）方向的 pandas DataFrame
dataframe_records - 记录（records）方向的 pandas DataFrame

DataFrame（split）

# Dataframe split orientation
{
  "dataframe_split": {
      "columns": ["sex", "age", "weight"],
      "data": [["male", 23, 160], ["female", 33, 120]]
  }
}

服务模型

# MLflow 服务命令
mlflow models serve --help
用法: mlflow models serve [OPTIONS]

Serve URI

# 本地文件系统
mlflow models serve -m relative/path/to/local/model

# 运行 ID
mlflow models serve -m runs:/<mlflow_run_id>/artifacts/model

# AWS S3
mlflow models serve -m s3://my_bucket/path/to/model

服务示例

# 从运行中提供模型
mlflow models serve -m runs:/e84a122920de4bdeaedb54146deeb429/artifacts/model

2023/03/12 16:28:28 INFO mlflow.models.flavor_backend_registry: 
Selected backend for flavor 'python_function'
2023/03/12 16:28:28 INFO mlflow.pyfunc.backend: === Running command 
'exec gunicorn --timeout=60 -b 127.0.0.1:5000 -w 1 ${GUNICORN_CMD_ARGS} -- 
mlflow.pyfunc.scoring_server.wsgi:app'
[2023-03-12 16:28:29 -0400] [48431] [INFO] Starting gunicorn 20.1.0
[2023-03-12 16:28:29 -0400] [48431] [INFO] Listening at: http://127.0.0.1:5000 
(48431)
[2023-03-12 16:28:29 -0400] [48431] [INFO] Using worker: sync
[2023-03-12 16:28:29 -0400] [48432] [INFO] Booting worker with pid: 48432

Invocations 请求

# 向 MLflow 发送 dataframe_split 方向的载荷
curl http://127.0.0.1:5000/invocations -H 'Content-Type: application/json' -d '{
  "dataframe_split": {
      "columns": ["sex", "age", "weight"],
      "data": [["male", 23, 160], ["female", 33, 120]]
  }
}'

{"predictions": [1, 0]}

Passons à la pratique !

MLflow 入门