Hello Metaflowers! I have a generic question. When you want to save some kind of a trained model, not necessarily a classifier/regressor, (say a fit KMeans clusterer on a huge dataset or something else that takes time to compute), and reuse it across different runs, do you store it as a Metaflow data artifact or do you use any specialized tool?

you can store them as Metaflow artifacts so you get your models versioned and registered conveniently. Find examples here https://outerbounds.com/docs/use-keras-with-metaflow

Woot. And if you'd like to skip the training if the model has already been trained, would you put a check there to inspect the "latest_successful_run" and see if there's a readily available fitted model?

@adventurous-australia-32236 Yep - that's a very valid pattern.