How to set a tag at the experiment level in MLFlow

Question

I can see that an experiment in MLFlow can have tags (like runs can have tags). I'm able to set a run's tag using mlflow.set_tag, but how do I set it for an experiment?

I see the rest API (https://www.mlflow.org/docs/latest/rest-api.html#set-experiment-tag), i'm trying to set it from Python — Nestor, Oct 23 '21 at 16:12

Alex Ott · Accepted Answer · 2022-09-23T18:46:32.437

3

If you look into the Python API, the very first example in mlflow.tracking package that shows how to create the MLflowClient is really showing how to tag experiment using the client.set_experiment_tag function (doc):

from mlflow.tracking import MlflowClient

# Create an experiment with a name that is unique and case sensitive.
client = MlflowClient()
experiment_id = client.create_experiment("Social NLP Experiments")
client.set_experiment_tag(experiment_id, "nlp.framework", "Spark NLP")

you can also set it for model version with set_model_version_tag function, and for registered model with set_registered_model_tag.

edited Sep 23 '22 at 18:46

answered Oct 23 '21 at 18:00

Alex Ott

80,552
8
87
132

As of Sep 23, 2022, all links appear to be broken. – Harshvardhan Sep 23 '22 at 16:04
1

fixed - they moved `MLflowClient` into a separate package. Thank you for reporting – Alex Ott Sep 23 '22 at 18:51

score 0 · Answer 2 · answered Jun 06 '23 at 15:31

While the client is very flexible, you can set tags at experiments with mlflow alone.

For instance, this is how to set a description visible in UI:

mlflow.set_experiment(experiment_name=EXPERIMENT_NAME)
mlflow.set_experiment_tag('mlflow.note.content',EXPERIMENT_DESCRIPTION)

NOTE: tested under mlflow=2.3.2.

How to set a tag at the experiment level in MLFlow

2 Answers2