https://kedro.org/ logo
#plugins-integrations
Title
# plugins-integrations
t

Tomás Rojas

03/15/2023, 5:34 AM
Hi team! I am looking at the
kedro mlflow
documentation and there is this thing called
pipeline_ml_factory
which is said to be used inside a
pipeline.py
file which I cannot find is this a kedro version issue?. Do you know any repository that uses this plugin with pytorch lightning? so I can see a more concrete example of its usage. Here is the page I've been looking
m

marrrcin

03/15/2023, 8:21 AM
1. For Kedro mlflow and `pipeline_ml_factory`there is a great official tutorial here: https://github.com/Galileo-Galilei/kedro-mlflow-tutorial 2. For Kedro + PyTorch Lightning - https://getindata.com/blog/deep-learning-with-azure-pytorch-distributed-training-done-right-kedro/ (there is a GH repo linked there too). You don’t have to use Azure ML. Adding kedro-mlflow is simple, as PyTorch-Lightning has MLflowLogger built-in, you just need to pass proper config:
Copy code
mlflow_logger = MLFlowLogger(
        experiment_name,
        tracking_uri,
    )
    mlflow_logger._run_id = mlflow.active_run().info.run_id
which can be injected e.g. from Kedro hooks and then just plug it in:
Copy code
pl.Trainer(logger=mlflow_logger, ...)
y

Yolan Honoré-Rougé

03/15/2023, 11:14 AM
This is indeed a typo in the documentation: in recent versions, ``pipeline.py`` has been renamed ``pipeline_registry.py`` . Thanks @marrrcin for pointing out the tutorial!
t

Tomás Rojas

03/24/2023, 1:31 PM
Thank you very much! I'll take a look
m

marrrcin

03/24/2023, 3:15 PM
@Tomás Rojas for PyTorch + Kedro, this also might be of your interest: https://kedro-org.slack.com/archives/C03QPDQTLBG/p1678891373264199
t

Tomás Rojas

03/26/2023, 7:58 PM
Thanks!