MLflow 入门
Weston Bassler
Senior MLOps Engineer
保存
记录
加载
# 将模型保存到本地文件系统 mlflow.sklearn.save_model(model, path)
# 将模型作为 artifact 记录到 MLflow Tracking mlflow.sklearn.log_model(model, artifact_path)
# 从本地或 MLflow Tracking 加载模型 mlflow.sklearn.load_model(model_uri)
本地文件系统 - relative/path/to/local/model 或 /Users/me/path/to/local/model
relative/path/to/local/model
/Users/me/path/to/local/model
MLflow Tracking - runs:/<mlflow_run_id>/run-relative/path/to/model
runs:/<mlflow_run_id>/run-relative/path/to/model
支持 S3 - s3://my_bucket/path/to/model
s3://my_bucket/path/to/model
# 模型 lr = LogisticRegression() lr.fit(X, y) # 本地保存模型 mlflow.sklearn.save_model(lr, "local_path")
# 模型 lr = LogisticRegression() lr.fit(X, y)
# 本地保存模型 mlflow.sklearn.save_model(lr, "local_path")
ls local_path/
MLmodel model.pkl requirements.txt python_env.yaml
# 从本地路径加载模型 model = mlflow.sklearn.load_model("local_path") # 显示模型 model
# 从本地路径加载模型 model = mlflow.sklearn.load_model("local_path")
# 显示模型 model
LogisticRegression()
# 模型 lr = LogisticRegression(n_jobs=n_jobs) lr.fit(X, y) # 记录模型 mlflow.sklearn.log_model(lr, "tracking_path")
# 模型 lr = LogisticRegression(n_jobs=n_jobs) lr.fit(X, y)
# 记录模型 mlflow.sklearn.log_model(lr, "tracking_path")
# runs 的格式 runs:/<mlflow_run_id>/run-relative/path/to/model
# 获取最后的活动运行 run = mlflow.last_active_run() run
# 获取最后的活动运行 run = mlflow.last_active_run()
run
<Run: data=<RunData: metrics={}, params={}, tags={'mlflow.runName': 'run_name'}>, info=<RunInfo: artifact_uri='uri', end_time='end_time', experiment_id='0', lifecycle_stage='active', run_id='run_id', run_name='name', run_uuid='run_uuid', start_time=start_time, status='FINISHED', user_id='user_id'>>
# 显示最后一次运行的 run_id run.info.run_id
'8c2061731caf447e805a2ac65630e70c'
# 获取最后的活动运行 run = mlflow.last_active_run() # 设置 run_id 变量 run_id = run.info.run_id run_id
# 设置 run_id 变量 run_id = run.info.run_id
run_id
# 将 run_id 作为 f-string 文字传入 model = mlflow.sklearn.load_model(f"runs:/{run_id}/tracking_path") # 显示模型 model
# 将 run_id 作为 f-string 文字传入 model = mlflow.sklearn.load_model(f"runs:/{run_id}/tracking_path")
Preparing Video For Download...
