how to save the trained autoML object or all models in it?

[feihongloveworld]

how to save the trained autoML object or all models in it?

[sebhrusen]

Hi @feihongloveworld, this is addressed by #6437
I can't remember why this PR is on standby, we'll try to reactivate it (cc: @tomasfryda )

Right now, you can't save an AutoML object, the only thing you can do is to loop through the trained models and save them individually.

[tomasfryda]

I think we put it on hold until the pipelines are merged so we don't create functionality that would be disabled the next release due to new functionality. And also due to other priorities.

I have some prototype implementation but there are missing things - if we load the automl object and do h2o.get_leaderboard(loaded_automl, extra_colums="ALL") it can fail since we use some (I think training) frame available during training to calculate the predict_time_per_row_ms and we probably don't want to persist that frame along with the automl. We can of course precalculate metrics like this but it could be misleading as it would be relevant only to the training machine (who'd want to know the inference speed on a machine they would not use for inference?).