# Databricks开源项目MLflow入门学习

**Published by:** [liyao](https://paragraph.com/@liyao/)
**Published on:** 2022-02-11
**URL:** https://paragraph.com/@liyao/databricks-mlflow

## Content

An open source platform for the machine learning lifecycle https://mlflow.org/上下文为开发者提供类似谷歌TFX、Facebook FBLearner Flow等平台类似好处可以支持任何工具和算法项目架构MLflow Tracking记录和查询实验：代码、数据、配置和结果https://www.mlflow.org/docs/latest/tracking.htmlMLflow Projects可在任何平台上重复运行的打包格式https://www.mlflow.org/docs/latest/projects.htmlMLflow Models将模型发送到各种部署工具的通用格式https://www.mlflow.org/docs/latest/models.htmlModel Registry中央存储库：存储、注释、发现和管理模型https://mlflow.org/docs/latest/model-registry.html项目使用Anaconda环境安装，可以便捷获取包且对包能够进行管理，同时对环境可以统一管理的发行版本mlflow安装UI启动examples/mlflow_tracking.pycd examples mlflow ui 模型运行# 模型生成 python3 sklearn_logistic_regression/train.py # API启动 mlflow models serve -m runs:/<Model_id>/model --port 1234 可能报错：mlflow.utils.process.ShellCommandException: Non-zero exitcode: 1 mlflow models serve -m runs:/<Model_id>/model --port 1234 --no-conda 流处理Demohttps://github.com/mlflow/mlflow/tree/master/examples/multistep_workflow思考官方待引入（MLflow Monitoring）MLFlow现阶段还是类似Pipeline的辅助工具，定义了Pipeline工具和标准，缺少AI平台概念如何与Spark、Docker做集成That's all!

## Publication Information

- [liyao](https://paragraph.com/@liyao/): Publication homepage
- [All Posts](https://paragraph.com/@liyao/): More posts from this publication
- [RSS Feed](https://api.paragraph.com/blogs/rss/@liyao): Subscribe to updates