MLflow: find model version with best metric using python code

Question

I am trying to use API workflow (python code) to find a model version that has the best metric (for instance, “accuracy”) among several model versions. I understand we can use web UI to do so, but I would love to write python code to achieve this. Could someone help me?

user_13980177 · Answer

import mlflow 
client = mlflow.tracking.MlflowClient()
runs = client.search_runs("my_experiment_id", "", order_by=["metrics.rmse DESC"], max_results=1)
best_run = runs[0]

https://mlflow.org/docs/latest/python_api/mlflow.tracking.html#mlflow.tracking.MlflowClient.search_runs

MLflow: find model version with best metric using python code

1 Answers