vincent-antaki

I'm not sure I see what's the point of having the context implement such TruncableStep-like interface. To allow reusage of a step at different point within the pipeline? If so,...

Somewhat linked: #316 Same issue: #211

Just looked at the opened issues and #254 seems to be the exact same suggestion.

update: the in memory observer/subscriber exists since at least 0.5.6. The OnDisk is not.

Extra information : So far, we've let the multiprocess library performs its default behaviour (which is a deep copy if i remember correctly). And, as far as step copying goes,...

> Related to #485: Services could have savers as well, not only cloners. It would be interesting to allow to save services. But do we have this use-case of saving...

@guillaume-chevalier I don't mind dependency inversion of Trainer/AutoML. However, your solution does not address point 2 and 3 of the consideration list. I'll implement it anyway because it is sufficient...

Prettier training curves plot mainly :)

I have no preference between step and component. As for higher order component, I think it makes less clear then wrapper.

As for alternatives we could have the service assertion wrappers be step instead. We'd then either loose the step.assert_has_service() syntactic sugar or have that function return the two steps and...