パラメータの指定

MLflow 入門

Weston Bassler

Senior MLOps Engineer

パラメータ

コーディング

1 unsplash.com
MLflow 入門

パラメータの指定

parameter_1_name:

    type: data_type

    default: default_value

parameter_2_name:

    type: data_type

    default: default_value
MLflow 入門

パラメータの指定

データ型

  • float、string など
  • 既定は string

既定値

  • 実行時に未指定なら使われる値
  • True、False、None など
MLflow 入門

parameters ブロック

name: project_name
python_env: python_env.yaml
entry_points:
  main:

    parameters:
      parameter_1:
        type: data_type
        default: default_value
      parameter_2:
        type: data_type
        default: default_value

    command: "python train.py {parameter_1_name} {parameter_2_name}"
MLflow 入門

train_model.py

# Import libraries and modules
import mlflow
import mlflow.sklearn
import pandas as pd

import sys

from sklearn.linear_model import LinearRegression
from sklearn.model_selection import train_test_split

# Training Data
df = pd.read_csv('Salary_predict.csv')
X = df[["experience", "age", "interview_score"]]
y = df[["Salary"]]
MLflow 入門

train_model.py

# Train test split 
X_train, X_test, y_train, y_test = train_test_split(X, y, train_size=0.7,
                                                    random_state=0)

# Set Auto logging for Scikit-learn flavor
mlflow.sklearn.autolog()

# Parameters
n_jobs_param = int(sys.argv[1])
fit_intercept_param = bool(sys.argv[2])

# Train model
lr = LinearRegression(n_jobs=n_jobs_param, fit_intercept=fit_intercept_param)

lr.fit(X_train, y_train)
MLflow 入門

MLproject

name: salary_model
python_env: python_env.yaml
entry_points:
  main:

    parameters:
      n_jobs_param:
        type: int
        default: 1
      fit_intercept_param:
        type: bool
        default: True

    command: "python train_model.py {n_jobs_param} {fit_intercept_param}"
MLflow 入門

パラメータの実行

Python

mlflow.projects.run()

parameters={
        'parameter_1': data_value, 
        'parameter_2': data_value, 
    }

CLI

mlflow run

-P parameter_1=parameter_1_value

-P parameter_2=parameter_2_value
MLflow 入門

Projects の実行

# Import MLflow Module
import mlflow

# Run local Project
mlflow.projects.run(
    uri='./', entry_point='main', 
    experiment_name='Salary Model',

    parameters={
        'n_jobs_param': 2, 
        'fit_intercept_param': False

})
MLflow 入門

出力

2023/04/07 18:50:22 INFO mlflow.projects.backend.local: === Running command 'source 
/.mlflow/envs/mlflow/bin/activate && python train_model.py 2 False' in run with ID
'422a6f589c984049ac03698efec8a286' ===

2023/04/07 18:50:28 INFO mlflow.projects: === Run (ID 
'422a6f589c984049ac03698efec8a286') succeeded ===
MLflow 入門

実行コマンド

# Run main entry point from Salary Model experiment
mlflow run --entry-point main --experiment-name "Salary Model" \

    -P n_jobs_param=3 -P fit_intercept_param=True ./
MLflow 入門

出力

# Run main entry point from Salary Model experiment
mlflow run --entry-point main --experiment-name "Salary Model" \
    -P n_jobs_param=3 -P fit_intercept_param=True ./

2023/04/07 19:08:51 INFO mlflow.projects.backend.local: === Running command 'source 
/.mlflow/envs/mlflow/bin/activate && python train_model.py 3 True' in run with ID 
'7c0a8885a3c7475a9aafc43c6fcae04d' ===
MLflow 入門

Laten we oefenen!

MLflow 入門

Preparing Video For Download...