PythonでMLflowからmetricの履歴を取得する - Qiita

import pandas as pd import mlflow def get_metric_history(run_id, metric): client = mlflow.tracking.MlflowClient() history = client.get_metric_history(run_id, metric) history = [dict(key=m.key, value=m.value, timestamp=m.timestamp, step=m.step) for m in history] history = pd.DataFrame(history).sort_values("step") history.timestamp = pd.to_datetime(history.timestamp, unit="ms") return history train_